STP 1561-2012

Selected Technical Papers STP1561
Flammability and Sensitivity

of Materials in Oxygen-Enriched
Atmospheres: 13th Volume
Editors:
Samuel Edgar Davis
Theodore A. Steinberg
ASTM International
100 Barr Harbor Drive
PO Box C700
West Conshohocken, PA 19428-2959
Printed in the U.S.A.
ASTM Stock #: STP1561

Library of Congress Cataloging-in-Publication Data
ISBN: 978-0-8031-7547-1
ISSN: 0899-6652
Copyright © 2012 ASTM INTERNATIONAL, West Conshohocken, PA. All rights reserved.
This material may not be reproduced or copied, in whole or in part, in any printed, mechanical,
electronic, film, or other distribution and storage media, without the written consent of the
publisher.
Photocopy Rights
Authorization to photocopy items for internal, personal, or educational classroom use, or the
internal, personal, or educational classroom use of specific clients, is granted by ASTM
International provided that the appropriate fee is paid to ASTM International, 100 Barr Harbor
Drive, P.O. Box C700, West Conshohocken, PA 19428-2959, Tel: 610-832-9634; online:
http://www.astm.org/copyright.
The Society is not responsible, as a body, for the statements and opinions expressed in this
publication. ASTM International does not endorse any products represented in this publication.
Peer Review Policy
Each paper published in this volume was evaluated by two peer reviewers and at least one editor.
The authors addressed all of the reviewers’ comments to the satisfaction of both the technical
editor(s) and the ASTM International Committee on Publications.
The quality of the papers in this publication reflects not only the obvious efforts of the authors
and the technical editor(s), but also the work of the peer reviewers. In keeping with long-standing
publication practices, ASTM International maintains the anonymity of the peer reviewers. The
ASTM International Committee on Publications acknowledges with appreciation their dedication
and contribution of time and effort on behalf of ASTM International.
Citation of Papers
When citing papers from this publication, the appropriate citation includes the paper authors,
“paper title”, J. ASTM Intl., volume and number, Paper doi, ASTM International, West
Conshohocken, PA, Paper, year listed in the footnote of the paper. A citation is provided as a
footnote on page one of each paper.
Printed in Bay Shore, NY

November, 2012
Foreword
THIS COMPILATION OF Selected Technical Papers, STP1561, Flammability
and Sensitivity of Materials in Oxygen-Enriched Atmospheres: 13th Volume,
contains peer-reviewed papers that were presented at a symposium held
September 19–21, 2012 in Montreal, Quebec, Canada. The symposium was
sponsored by ASTM International Committee G04 on Compatibility and
Sensitivity of Materials in Oxygen Enriched Atmospheres.
The Symposium Co-Chairpersons and STP Editors are Samuel Edgar Davis,
NASA, George C. Marshall Space Flight Center, Huntsville, AL, USA, and
Theodore A. Steinberg, Queensland University of Technology, Australia.
Contents
Overview ............................................................ vii
An Elementary Overview of the Selection of Materials for Service in Oxygen-

enriched Environments
S. E. Davis ......................................................... 1
Development of a Liquid Oxygen Facility for Rocket Engine Testing
H. W. Mulkey and D. M. Lineberry ........................................ 28
Rudimentary Cleaning Compared to Level 300A
C. Y. Piña Arpin and J. Stolzfus .......................................... 47
Flow Friction or Spontaneous Ignition?
J. M. Stoltzfus, T. D. Gallus, and K. Sparks .................................. 62
Ignition Testing by Fracture of Oil- and Particle-Contaminated Aluminum Pipes
in High Pressure Oxygen at Ambient and Elevated Temperature
M. Meilinger........................................................ 81
Fire Without Cause of an Oxygen Component With an Aluminum Seal
......................
T. Tillack, C. Binder, T. Brock, N. Treisch, and P. Dielforder 105
Pressure Dependence of Aluminum Ignition in Gaseous Oxygen and Possible
Ignition Mechanisms in Brazed Aluminum Heat Exchangers
F. Crayssac, J.-C. Rostaing, E. Werlen, L. Sun, P. Houghton, W. Kleinberg,
and F. Coste ........................................................ 117
Auto-Ignition and Combustion Properties of Iron/Steel Micro-Particles in Oxygen
Atmospheres Heated by Rapid Compression
V. V. Leschevich, O. G. Penyazkov, and J.-C. Rostaing ......................... 147
Another Dangerous Practice in the SCUBA Diving Community
W. Ciscato, C. Binder, O. Hesse, S. Lehné, and T. Tillack........................ 162
Safe for Oxygen but Not for Oxygen-Enriched Gas Mixtures in SCUBA Diving
T. Tillack, C. Binder, O. Hesse, S. Lehné, and W. Ciscato........................ 172
Burning of CP Titanium (Grade 2) in Oxygen-Enriched Atmospheres
J. M. Stoltzfus, N. Jeffers, and T. D. Gallus .................................. 182
Flammability Evaluation of Metals in Piping Configurations Under Flowing
Conditions: Hollow-Vessel Promoted Ignition Test
M. Ridlova, J.-C. Rostaing, A. Colson, and O. Longuet ......................... 198
Situational Non-Flammable Control Valves for Oxygen Service
C. Dobinson, P. Nelson, and T. Steinberg ................................... 225
Combustion Products of Bulk Aluminum Rods Burning in High-Pressure Oxygen
.....................
O. Plagens, D. Lynn, M. Castillo, T. Paulos, and T. Steinberg 233
A Comparison of Fire Reaction Effects for Halogenated Versus Non-Halogenated
Non-Metals in High-Pressure Oxygen Components
B. S. Forsyth, G. J. A. Chiffoleau, G. A. Odom and B. E. Newton.................. 245
Risk Evaluation Approach for Polymers in Oxygen Breathing Systems
...................
G. J. Chiffoleau, B. E. Newton, B. S. Forsyth, and E. T. Forsyth 262
Development of Burn Curves to Assist With Metals Selection in Oxygen
E. T. Forsyth, N. Linley, G. J. A. Chiffoleau, J. Hooser, and B. E. Newton ............. 274
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
Overview
STP 1561 is the thirteenth Special Technical Publication (STP) originating from
the ASTM Committee G04 focusing on the Flammability and Sensitivity of Ma-
terials in Oxygen-Enriched Atmospheres. The twelve previous STP volumes origi-
nating from the ASTM G04 committee are: 812, 910, 986, 1040, 1111, 1197, 1267,
1319, 1395, 1454, 1479, and 1522. Copies of these STP volumes are available
from ASTM, International.
The ASTM, International, G04 committee on the Flammability and Sen-
sitivity of Materials in Oxygen-Enriched Atmospheres continues to grow in its
international appeal. The thirteenth symposium was attended by a number of
professionals representing several countries. These included the United States,
Australia, Germany, Canada, France, Switzerland and Belarus. A number of pro-
fessionals from other nations also attended the symposium and shared impor-
tant information in person even though they were unable to submit a formal
paper for publication.
As with the past STPs, the thirteenth volume expands upon the objectives
that have been carried forward since the first ASTM Committee G04 STP was
published in 1983. These objectives include:
• Review the current research on polymers and metals ignition and
combustion;
• Overview principles of oxygen systems design and issues related to mate-
rials compatibility with oxygen; contribute to the knowledge on the most
current risk management concepts, practices, approaches, and procedures
used by individuals and organization involved in the design, use, retrofit-
ting, maintenance, and cleaning of oxygen systems;
• Review of accident/incident case studies related to oxygen systems and
oxygen handling procedures;
• Provide the most current data related to the flammability and sensitivity
of materials in oxygen-enriched atmospheres to designers, users, manufac-
turers and maintainers of oxygen components and systems and to support
Committee G04’s Technical and Professional Training Course on Fire
Hazards in Oxygen Systems.
• Discuss enhancement, development, and use of standards sponsored by
ASTM Committee G04 on Compatibility and Sensitivity of Materials in
Oxygen Enriched Atmospheres;
• Provide a readily accessible reference addressing oxygen compatibility.
vii
The thirteenth volume consists of a group of peer-reviewed papers that were
presented at the Committee G04’s Thirteenth International Symposium held in
Montreal, Quebec, Canada in September 2012. The volume consists of seventeen
papers on topics related to ignition and combustion of non-metals, ignition and
combustion of metals, oxygen compatibility of components and systems, analysis
of ignition and combustion, failure analysis and safety, and includes aerospace,
military, scuba diving, and industrial oxygen applications.
The papers presented in the thirteenth volume are arranged into six groups
that offer a variety of valuable information. The first group of papers consists
of three overview papers related to the development of safe oxygen systems.
The first paper, submitted by NASA George C. Marshall Space Flight Center,
is aimed at new users of oxygen systems and presents a concise, elementary
discussion of the selection of materials for producing safe oxygen systems. The
second paper, submitted by University of Alabama in Huntsville, discusses the
effort and decisions that were required by a university to build the oxygen sys-
tem for a rocket engine test facility. The third paper, submitted by NASA White
Sands Test Facility, focuses on the effectiveness of different methods for field
cleaning large oxygen system hardware.
The second group consists of five papers focusing on ignition mechanisms
within oxygen systems and how to avoid them. The first paper, submitted by
NASA White Sands Test Facility, discusses the elusive ignition mechanism called
flow friction and if other ignition mechanisms may actually be mistakenly called
flow friction. The second paper, submitted by Linde Engineering, evaluates the
aluminum alloys common for air separation equipment and analyses the effects
of contamination and elevated pressures on the most common aluminum alloys.
The third paper, submitted by the BAM Federal Institute for Materials Research
and Testing in Germany, discusses the severe ignitions that have occurred in
their all-metal oxygen components which contain aluminum seals. The fourth
paper, submitted by Air Liquide Corporation of France, discusses their testing of
ignitions with aluminum using a laser igniter to vary the energy input, with a
comparison of pressures and flow rates. The fifth paper, submitted by the Bela-
rus National Academy of Sciences, examined iron particle contamination within
a system and demonstrated how little energy was required to ignite these iron
particles using only rapid pressurization as the ignition source.
The third group of papers consists of two manuscripts related to the specific
hazards that exist with the oxygen mixture breathed by divers in the scuba indus-
try. The first, submitted by Gentoo Divers International of Switzerland, evaluated
the issue of unsafe conditions in scuba diving because a few of the suppliers of
scuba equipment are making material changes without realizing the dangers that
the high pressure mixture of oxygen and nitrogen pose, and that these should be
treated as oxygen systems instead of air systems. The other paper, submitted by
the BAM Federal Institute for Materials Research and Testing in Germany, dis-
cusses the dangers of allowing only visual inspection for any detrimental effects
that testing may have presented on scuba equipment seal materials.
viii
The fourth group consists of four papers that discuss issues related to oxygen
system level safety. The first, submitted by NASA White Sands Test Facility, in-
vestigates the factors influencing the fire hazards in commercially pure titanium,
focusing on the effects of oxygen concentration and specimen diameter. The second,
submitted by Air Liquide Corporation of France, investigates the special hazards
involved with flowing oxygen gas and a hollow material, especially focusing on how
much more severely this configuration burns when compared to a standard solid
test specimen in a stagnant atmosphere. The third paper, submitted by Oxycheck
Pty Ltd of Australia, investigates the benefits of building hardware by situational
non-flammability of the components, a condition where a component is built using
only nonflammable materials. The fourth, submitted by Queensland University of
Technology in Australia, investigates the special hazards presented by the com-
bustion products that are given off by materials burning in an oxygen fire.
The fifth group consists of two papers that cover issues related to oxygen safe-
ty in breathing systems. Both of these papers were submitted by Wendell Hull
and Associates. The first compares the hazards present when using halogenated
versus non-halogenated polymers as seals for breathing systems, and the con-
cerns presented by toxic combustion products. The second details taking a risk
assessment approach to determining the safety of a breathing system, and the
successful technique for this assessment.
The last manuscript, submitted by Wendell Hull and Associates, provides de-
tailed investigations and discussions related to the burn curves that have been
generated for metals that are burning in a standard test fixture.
The thirteenth volume of Flammability and Sensitivity of Materials in
Oxygen-Enriched Atmospheres provides a diverse source of new information
to air separation industries, oxygen manufacturers, manufacturers of compo-
nents for oxygen and other industrial gases service, manufacturers of mate-
rials intended for oxygen service, and users of oxygen and oxygen-enriched
atmospheres, including aerospace, medical, industrial gases, chemical process-
ing, steel and metals refining, as well as to military, commercial or recreational
diving.
Samuel Edgar Davis
Materials and Processes Laboratory
National Aeronautics and Space Administration
George C. Marshall Space Flight Center
Huntsville, Alabama
Theodore A. Steinberg
School of Engineering Systems
Faculty of Built Environment and Engineering
Queensland University of Technology
Brisbane, Queensland, Australia
ix
Flammability and Sensitivity of Materials in Oxygen-Enriched Atmospheres
STP 1561, 2012
Available online at www.astm.org
DOI:10.1520/STP20120001
Samuel Edgar Davis1
An Elementary Overview of the Selection of

Materials for Service in Oxygen-enriched
Environments
REFERENCE: Davis, Samuel Edgar, “An Elementary Overview of the Selec-

tion of Materials for Service in Oxygen-enriched Environments,” Flammability
and Sensitivity of Materials in Oxygen-Enriched Atmospheres on September
19–21, 2012 in Montreal, QC; STP 1561, Samuel Edgar Davis and Theodore
A. Steinberg, Editors, pp. 1–27, doi:10.1520/STP20120001, ASTM Interna-
tional, West Conshohocken, PA 2012.
ABSTRACT: The process of selecting materials for use in oxygen or oxygen-
enriched environments is one that continues to be investigated by many
industries because of the importance to those industries of oxygen systems.
There are several excellent resources available to assist oxygen systems
design engineers and end-users, with the most comprehensive being the
second edition of ASTM MNL-36, Safe Use of Oxygen and Oxygen Systems:
Handbook for Design, Operation and Maintenance. ASTM also makes avail-
able several standards for oxygen systems. However, the ASTM publications
are extremely detailed and typically are designed for professionals who
already possess a working knowledge of oxygen systems. No notable
resource exists, whether an ASTM publication or that of some other organi-
zation, that can be used to educate engineers or technicians who have no
prior knowledge of the nuances of oxygen system design and safety. This
paper will fill the void for information needed by organizations that design or
operate oxygen systems. The information in this paper is not new, but it pro-
vides a concise and easily understood summary of the selection of materials
for oxygen systems. This paper will serve well as an employee’s first
introduction to oxygen system materials selection, and probably as the
employee’s first introduction to ASTM.
Manuscript received January 20, 2012; accepted for publication May 2, 2012; published online
November 2012.
1
Materials and Processes Laboratory, George C. Marshall Space Flight Center, NASA,
Huntsville, AL 35812.
Copyright V
C 2012 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA
19428-2959.
1
2 STP 1561 ON FLAMMABILITY AND SENSITIVITY OF MATERIALS
Introduction
Oxygen is a highly reactive nonmetallic element that reacts with many other
elements to form oxides through oxidation reactions. Oxygen constitutes
20.9 % of Earth’s atmosphere and exists primarily in the form of O2 molecules.
O2 is a pale blue odorless and tasteless gas that can also exist in a liquid state at
temperatures below 297 F (183 C) and as a solid below 362 F
(219 C). Oxygen in its liquid state is important because it provides a very
large number of oxygen molecules per unit volume of containment. Liquid
oxygen is easily stored, is easily transferred from one tank to another, and is
readily converted to gaseous oxygen.
The highly reactive nature of oxygen requires that special care be taken
when it is being used in order to ensure the safety of personnel and equipment.
Materials that do not burn in normal atmosphere can burn violently, sometimes
to the point of explosion, in oxygen. Also, liquid oxygen is very cold and can
do severe damage to human skin on contact. Therefore, the use of oxygen or
oxygen-enriched media requires special safety precautions.
Hazards Inherent in Oxygen-enriched Environments

The hazards inherent in oxygen systems arise primarily from the risk of fires.
The standard fire triangle demonstrates that a fire requires three separate and
independent components, represented as legs of the triangle. The three compo-
nent legs are fuel, oxidizer, and ignition source. The oxygen system obviously
has the oxidizer leg, but what about a fuel leg? In an oxygen system, the hard-
ware components actually can become the fuel!
Oxygen fires are mysterious and not well understood. Laboratory experi-
ments with oxygen and materials have shown that oxygen fires are infrequent,
even if the experimental conditions maximize the probability of a fire. Also,
fires created in a laboratory setting are difficult to duplicate. It is very important
to understand oxygen fire hazards, however, because oxygen fire events are
often catastrophic.
One special note: oxygen-rich systems must always be treated as if they
were pure oxygen systems. Many personnel injuries and deaths, and much
major equipment damage, have resulted from the assumption that oxygen-rich
systems can be treated as if they were air systems.
Oxygen and Materials

Oxygen is a powerful oxidizer, and as such it completes one leg of the fire tri-
angle (see Fig. 1). A fuel can be almost any material, even the system hard-
ware, under the right conditions. Ignition sources are more common than most
engineers believe. These three legs of the fire triangle can come together in the
right circumstances to provide a very exothermic chemical reaction. The
DAVIS, doi:10.1520/STP20120001 3
FIG. 1—The fire triangle.
oxygen system designer must always remember that fires and explosions occur
readily in oxygen systems, even when the systems are not under extreme condi-
tions. In fact, all industries that use oxygen systems have experienced fires at
one time or another. Many of these have resulted in a loss of life.
Factors Significantly Influencing Combustion Probability and Severity of

Materials in Oxygen Systems
The ignition and combustion properties of materials are not the same from one
oxygen system to another. Several factors determine ignition probability and
fire severity. The three most important factors that significantly increase both
the probability of ignition of a material and the severity of the fire once ignition
has occurred are increasing oxygen concentration, increasing system pressure,
and increasing system temperature.
Oxygen Concentration Effects—Combustible materials, with very few

exceptions, burn far more readily and rapidly in environments that have
higher concentrations of oxygen. The oxygen concentration within a system
should be minimized to the lowest allowable level. Table 1 shows the
effects of increasing oxygen concentrations on the fire hazard status of materi-
als. This table provides data for materials that were tested in different oxygen
concentrations, with measurements for burn length and burn rate. The tested
material specimens were 12 in. (305 mm) long, 2.5 in. (64 mm) wide, and at a
thickness that was consistent for each material. The data were generated via
testing per Test 1 of NASA-STD-6001 [1] at the National Aeronautics and
Space Administration’s (NASA’s) George C. Marshall Space Flight Center
(MSFC).
TABLE 1—Nonmetal flammability versus oxygen.
Oxygen Burn Length, Burn Rate,

Material Concentration, % in. (mm) in./s (mm/s)
Polyester-based foam 23 1.3 (33) 0.07 (1.8)
25 12 (305) 0.67 (17)
Polyurethane and epoxy 21 1.0 (25.4) 0.05 (1.3)
26 12 (305) 0.33 (8.4)
Polyurethane foam 21 1.0 (25.4) 0.05 (1.3)
23 9.0 (228) 0.27 (6.9)
Polyisocyanurate foam 21 2.0 (50.8) 0.07 (1.8)
30 12 (305) 0.13 (3.3)
Silicone RTV 21 0.5 (12.7) 0.01 (0.3)
(Room Temperature Vulcanizing) 24 2.6 (66.0) 0.02 (0.5)
30 12 (305) 0.02 (0.5)
Note: Data are for oxygen concentration effect comparison only and are not to be considered standard values
for listed materials.
Pressure Effects—Combustible materials, with few exceptions, burn far

more readily and rapidly in oxygen systems with elevated pressures. The use of
pressure within an oxygen system should be minimized to the lowest allowable
level. Table 2 shows the effects of increasing pressures on the fire hazard status
of metals. The table provides data for metals that were tested in oxygen at dif-
ferent pressures. The metal specimens were rods 12 in. (305 mm) long and
0.125 in. (3.2 mm) in diameter. These data were generated via testing per
ASTM G124 [2] at NASA’s MSFC.
Temperature Effects—Temperature effects on combustible materials, with

very few exceptions, are analogous to pressure effects in that materials burn far
TABLE 2—Metal flammability versus pressure.
Pressure, Burn Length, Burn Rate,

Material psi (kPa) in. (mm) in./s (mm/s)
Bronze/aluminum mixture 50 (345) 0.0 (0.0) 0.0 (0.0)
100 (689) 12 (305) 0.3 (7.6)
Aluminum 4043 25 (172) 4.4 (112) 0.6 (15.2)
50 (345) 12 (305) 1.1 (27.9)
316 stainless steel 500 (3447) 1.7 (43) 0.3 (7.6)
1000 (6895) 12 (305) 0.4 (10.2)
304 stainless steel 500 (3447) 3.8 (97) 0.3 (7.6)
1000 (6895) 12 (305) 0.4 (10.2)
316L stainless steel 250 (1724) 2.6 (66) 0.2 (5.1)
1000 (6895) 12 (305) 0.4 (10.2)
Data are for pressure effects comparison only and are not to be considered standard values for listed materials.
DAVIS, doi:10.1520/STP20120001 5
more readily and rapidly in a gaseous oxygen environment as the system tem-
perature is increased. Designers of gaseous oxygen systems should minimize
the system temperature, with ambient temperature being ideal. Note that tem-
peratures can increase substantially above the initial temperature during normal
system operations because of factors such as the effectiveness of the insulation
and the frictional heating that occurs during normal system operation. Table 3
shows the effects of increasing initial temperatures on the fire hazard status of
metals. The table provides data for materials that were tested in oxygen at dif-
ferent initial test temperatures, with specimens heated by an induction field.
The tested metals were rods that were 12 in. (305 mm) long and 0.125 in.
(3.2 mm) in diameter. These data were generated via testing per ASTM G124
[2] at NASA’s MSFC.
The data presented in the tables demonstrate that fire hazards are far
greater in systems with increased oxygen concentrations, higher pressures, and
increased temperatures than in normal atmospheric conditions. These hazards
must be minimized through improved system designs.
Generalizations on Effects—The ignition resistance and burn inhibition prop-

erties inherent in a material are determined by several factors. The following list
provides generalized rules on the fire safety of materials in oxygen systems:
(1) The higher the oxygen concentration within the system, the more eas-
ily the system materials ignite, and the more rapidly they burn once
ignited. So, the lowest usable oxygen concentration is preferred.
(2) The higher the pressure the material is subjected to, the more easily
the material ignites, and the more rapidly it burns once ignited. So,
the lowest usable pressure is preferred.
(3) The higher the temperature the material is exposed to, the more easily
the material ignites, and the more rapidly it burns once ignited. So,
the lowest workable temperature is preferred.
TABLE 3—Metal flammability versus temperature.
Pressure, Temperature, Burn Length, Burn Rate,

Material psi (kPa) F (K) in. (mm) in./s (mm/s)
304L stainless steel 500 (3447) 75 (297) 4.0 (102) 0.35 (8.9)
500 (3447) 1000 (811) 12 (305) 0.43 (10.9)
15-5 PH stainless steel 500 (3447) 75 (297) 2.9 (74) 0.33 (8.4)
500 (3447) 1000 (811) 12 (305) 0.35 (8.9)
InconelTM 718 750 (5171) 75 (297) 4.2 (107) 0.35 (8.9)
750 (5171) 1700 (1200) 12 (305) 0.60 (15.2)
316 stainless steel 500 (3447) 75 (297) 5.6 (142) 0.34 (8.6)
500 (3447) 1000 (811) 12 (305) 0.37 (9.4)
Note: Data are for temperature effects comparison only and are not to be considered standard values for listed
materials.
(4) The smaller the thickness of a material, the more easily the material
ignites, and the more rapidly it burns once ignited. So, thicker materi-
als are preferable to thinner ones.
(5) The greater the flow rate of oxygen passing over a material, the more
easily the material ignites, and the more rapidly it burns in the direc-
tion of the flow once ignited. So, flow rates within the system should
be kept as low as functionally allowed.
Safe Oxygen Systems—Safety must be a primary goal for any oxygen sys-
tem design. There are four ground rules for designing safer oxygen systems, as
described in the ASTM Technical and Professional Training (TPT) course Fire
Hazards in Oxygen Systems. These are as follows:
Conservation: Conserve both oxygen and materials within the system.
Maximization: Maximize the use of the most burn-resistant materials
available.
Minimization: Minimize the potential ignition sources that are inherent in
all systems.
Utilization: Utilize good system design practices.
Conserve Oxygen and Materials—The first step in designing safer oxygen

systems is the conservation of oxygen and materials. The safest systems will
conserve, or minimize, the amount of oxygen in the system. Less oxygen
within a system minimizes the likelihood of ignition and allows for quenching
in case a fire does begin. The amount of oxygen within a system can be mini-
mized in three ways: (1) minimizing the system pressure by pressurizing the
system no more than is absolutely necessary, (2) using the lowest concentration
of oxygen permitted for the application, and (3) designing the system so that it
uses the shortest possible flow path of oxygen.
The safest oxygen systems will also minimize the amount of materials
within the system. Less material within a system provides less fuel to burn.
However, this factor can be misleading. In general, thicker materials are less
likely to ignite than thinner ones. Therefore, the ignition hazard of a material
must be weighed against the safety provided by utilizing less fuel within a
system.
Maximize Good Materials—Poor material choices can greatly increase the

likelihood of a fire occurring in an oxygen system. Some materials are more
difficult to ignite than others and, when ignited, are more resistant to sustained
burning. Materials also vary in the amount of energy released when they burn.
Therefore, careful selection of materials can lessen the chances of ignition and
enhance the burn resistance of a system, thereby limiting the amount of dam-
age resulting from a fire. Despite the fact that heat sources can be inherent in
DAVIS, doi:10.1520/STP20120001 7
an oxygen system or its surroundings, design elements can limit the amount of,
or dissipate altogether, the heat within the system.
Several factors enter into the decision process when selecting the materials
to be used in an oxygen system. However, the issue of the compatibility of the
materials with the oxygen fluid is the only issue considered in this paper. Mate-
rials compatibility is an important factor for the safety of oxygen systems.
Materials compatibility not only covers the potential chemical breakdown of a
material with oxygen, but also encompasses the resistance of the material to
ignition and burning.
The most compatible materials for oxygen environments possess the fol-
lowing properties:
1. Difficult to ignite
2. Burn with low heat of combustion
3. Self-extinguish quickly once ignited
4. Burn slowly
The above-listed properties are all determined via testing. Oxygen systems
are not new, so extensive test data generated by the government and private
industries already exist. However, newer materials, and even some older ones,
have no test data. Therefore, it is important to generate new data for desired
materials. Caution: The compatibility of untested materials is unknown. It can-
not be determined by a material’s similarity to other materials with known
properties, whether the similarity is in composition, physical properties, etc.
Slight differences can drastically change the oxygen compatibility of a
material.
Testing, and not evaluation, is the only method by which safe materials
can be chosen. However, there is no decisive test or evaluation method that
will clearly indicate the best materials for use in oxygen. Several test techni-
ques have been developed to help determine which materials will probably be
safer for specific applications in oxygen systems. These tests demonstrate ei-
ther the ease of ignition or the tendency for sustained burning of the material.
The Ambient Pressure Liquid Oxygen Mechanical Impact Test, per ASTM
D2512 [3], is one of the oldest and is still used today. This test involves drop-
ping a metal plummet onto a disk of the test material that is immersed in liquid
oxygen. Materials that are very incompatible with oxygen will burn when the
energy from the falling plummet is transferred to the sample. This burn, or
reaction, is witnessed by the test operator. The reaction might be a visible flash,
an audible report, the appearance of charring on the surface of the sample, or a
combination of these. Figure 2 shows a diagram of the section of the test fixture
where the plummet will strike the material sample. Figure 3 shows the tester in
operation at NASA’s MSFC. Figure 4 shows a severe reaction that can happen
with a material that is not compatible with liquid oxygen, which clearly dem-
onstrates the importance of testing.
After several years of oxygen systems failure investigations, this one test
proved to be insufficient for determining whether a material is safe for use in
every application within an oxygen system. This Ambient Pressure Mechanical
Impact Test, in fact, did not adequately address elevated pressure scenarios
because several materials that passed the test were burning in high pressure ox-
ygen systems. The tester was later modified to perform the mechanical impacts
in elevated pressure oxygen. The new tester, the High Pressure Liquid and Gas-
eous Oxygen Mechanical Impact Tester, which is operated per ASTM G86 [4],
performs impact testing at pressures of up to 10 000 psi (68 948 kPa). The tester
works in basically the same way as the ambient pressure tester, except that this
test fixture seals a head onto a base in order to maintain the elevated test pres-
sure. The tester was again modified to test materials in gaseous oxygen and at
elevated temperatures. The test setup can be seen in Fig. 5.
The Mechanical Impact Test has also, over time, proven to be insufficient.
The test provides useful data for applications in which actual impacts occur
within the system, such as a valve material shutting against a valve seat. The
test does not, however, provide useful information for other ignition scenarios.
The Mechanical Impact Test also does not differentiate well between metals
FIG. 2—Ambient mechanical impact test fixture (from NASA-STD-6001 [1]).

DAVIS, doi:10.1520/STP20120001 9
FIG. 3—Ambient mechanical impact tester in operation.
that are very compatible with oxygen and those that are minimally compatible.
In order to provide more useful data, several new test methods have been
developed.
The Promoted Ignition Test, or Combustion Behavior of Metallic Materials in
Oxygen, per ASTM G124 [2], has become the new baseline standard for determin-
ing whether a material is safe in an oxygen environment. This test uses a 1=8 in.
(3.2 mm) diameter rod of the candidate material, which is hung vertically in a test
chamber. The rod is surrounded by oxygen and ignited at the bottom by an alumi-
num or magnesium promoter. The test demonstrates the behavior of a material
once burning is initiated and is used to determine which materials do not ignite or,
if ignited, will or will not exhibit self-sustained burning.
FIG. 4—Ambient mechanical impact tester showing a severe reaction.
The ASTM G124 [2] Promoted Ignition Test requires a rod of at least 4 in.
(102 mm) in length with a diameter of 1=8 in. (3.2 mm). The promoter, con-
nected to the bottom of the rod, is initiated, and the sample rod is observed for
its burn characteristics, primarily its burn length. Self-sustained combustion, as
research has confirmed, occurs if the sample burns more than 1.2 in. (30 mm).
Samples that burn more than this are considered unacceptable for unlimited
use in an oxygen system at the test pressure. However, limited use of that mate-
rial can be allowed under certain circumstances, as is discussed later. A dia-
gram of the Ambient Temperature Promoted Ignition Tester is shown in Fig. 6.
A photograph of a sample being loaded into MSFC’s Elevated Temperature
Promoted Ignition Tester is shown in Fig. 7.
The ASTM G124 [2] Promoted Ignition Test, due to its harshness, is still
not a perfect test for materials in oxygen environments. Nonmetals that are
essential seal materials for oxygen systems will not meet the acceptance crite-
rion of this test. Also, the majority of the industry standard metals for oxygen
DAVIS, doi:10.1520/STP20120001 11
FIG. 5—High pressure mechanical impact tester; the head is bolted to the
base to sustain pressure.
systems, such as stainless steel alloys, do not meet this criterion at their typical
use pressures. The harshness of the test has necessitated the development of a
more realistic hazards evaluation process that will allow the use of materials
that do not meet the acceptance criteria. This process is called an Oxygen
Compatibility Assessment and is discussed later.
The Mechanical Impact Test and the Promoted Ignition Test provide the
pressures at which a given material will ignite or burn. However, the two tests
provide significantly different values for the lowest pressures at which a mate-
rial will burn in oxygen. This is a result of the differences between an ignition
test and a burning test. The determination of which test to use depends upon
the use application planned for the material. The Mechanical Impact Test is
more valuable for seal or seat materials that will actually be subjected to
impact loads during normal system operation. The Promoted Ignition Test is
more valuable for materials that will form the basic structure of the system.
FIG. 6—Promoted ignition test fixture (from NASA-STD-6001 [1]).
Tables 4 and 5 provide summaries of data generated at NASA’s MSFC for

selected materials in both of these tests. The data clearly show that the pressure
required in order for a given material to exhibit self-sustained burning is much
lower than the pressure that is required in order to ignite that material by means
of a mechanical impact.
Selection of Metals versus Nonmetals

The selection of both metals and nonmetals is important, and each is important
for its specific application. Given their inherent structural strength, the vast ma-
jority of the materials in oxygen systems will be metals. However, nonmetals
are important for applications such as valve seals and seats, lubricants, or appli-
cations in which rigid materials cannot be used. Metals are generally more
compatible with oxygen than nonmetals, but metals tend to burn much more
violently, for longer durations, and at much higher temperatures. Both metals
DAVIS, doi:10.1520/STP20120001 13
FIG. 7—Promoted ignition tester with sample rod.
and nonmetals become more incompatible with oxygen as the pressure or tem-
perature within the system increases.
An oxygen system designer must choose the best materials by considering
all of the factors involved, especially compatibility under the use conditions.
The metals that tend to be the most compatible include nickel alloys, copper,
brass, and bronze. These metals are more difficult to ignite and, once ignited,
tend to burn with the lowest heats of combustion. The metals that are more eas-
ily ignited, and which thus should be avoided, include magnesium, titanium,
and many aluminum alloys. The maximum use pressure within the system will
drive the material choices.
The most commonly used nonmetals in oxygen systems are elastomers,
lubricants, ceramics, and carbon fiber materials. Each of these has its own
application. Elastomers are important for providing seals within a pressurized
system. The seals must withstand the system pressure and be compatible with
TABLE 4—Minimum pressures required in order to ignite or burn metals.
Metal Oxygen Condition Promoted Ignition Test Mechanical Impact Test

Aluminum 2024 GOX <14.7 psi (<101 kPa) 1500 psi (10 342 kPa)
LOX 1500 psi (10 342 kPa)
Aluminum 2090 GOX <14.7 psi (<101 kPa) 500 psi (3447 kPa)
LOX 500 psi (3447 kPa)
Aluminum 2219 GOX <14.7 psi (<101 kPa) 1500 psi (10 342 kPa)
Aluminum 6061 GOX <14.7 psi (<101 kPa) <14.7 psi (<101 kPa)
LOX <14.7 psi (<101 kPa)
Brass GOX 10 000 psi (68 948 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Copper 12200 GOX 10 000 psi (68 948 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Haynes 214 GOX 1000 psi (6895 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Inconel 718 GOX 500 psi (3447 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Magnesium GOX <14.7 psi (<101 kPa) <14.7 psi (<101 kPa)
LOX <14.7 psi (<101 kPa)
Monel K-400=K-500 GOX 10 000 psi (68 948 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Nickel GOX 10 000 psi (68 948 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Stainless steel 17-4 GOX 400 psi (2758 kPa) 5000 psi (34 474 kPa)
LOX 5000 psi (34 474 kPa)
Stainless steel 304L GOX 250 psi (1724 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Stainless steel 316 GOX 400 psi (2758 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Stainless steel 420 GOX 750 psi (5171 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Stainless steel 440C GOX 3000 psi (20 684 kPa) 10 000 psi (68 948 kPa)
LOX 10 000 psi (68 948 kPa)
Titanium GOX <14.7 psi (<101 kPa) <14.7 psi (<101 kPa)
LOX <14.7 psi (<101 kPa)
Note: Data are for comparison purposes only and are not to be considered standard values for listed materials.
LOX ¼ Liquid Oxygen; GOX ¼ Gaseous Oxygen.
oxygen at the highest use pressure. The best elastomeric compounds are typi-
cally fluorinated polymers, which, as shown in Table 4, tend to be more
difficult to ignite and burn with lower heats of combustion.
Lubricants are sometimes used in oxygen systems for pipe threads and in
other places where metal meets metal. However, lubricant use must follow two
DAVIS, doi:10.1520/STP20120001 15
TABLE 5—Minimum pressures required in order to ignite nonmetals.
Nonmetal Oxygen Condition Minimum Ignition Pressure

Acrylic sheet GOX 5000 psi (34 474 kPa)
LOX 5000 psi (34 474 kPa)
Butyl rubber GOX 40 psi (276 kPa)
LOX <14.7 psi (<101 kPa)
Carbon fiber=epoxy (graphite epoxy) GOX 10 000 psi (68 948 kPa)a
LOX 5000 psi (34 474 kPa)a
Polytetrafluoroethylene GOX 3500 psi (24 132 kPa)
LOX 1500 psi (10 342 kPa)
Fiberglass GOX 5000 psi (34 474 kPa)
LOX 5000 psi (34 474 kPa)
Fluorinated lubricants (typical) GOX 10 000 psi (68 948 kPa)
LOX 8000 psi (55 158 kPa)
Molybdenum disulfide GOX 4000 psi (27 579 kPa)
LOX 4000 psi (27 579 kPa)
Neoprene rubber GOX 4000 psi (27 579 kPa)
LOX 4000 psi (27 579 kPa)
Nitrile rubber GOX <150 psi (<1034 kPa)
LOX <150 psi (<1034 kPa)
Polyimide (KaptonTM, etc.) GOX 200 psi (1379 kPa)
LOX <14.7 psi (<101 kPa)
Polyvinylchloride LOX <14.7 psi (<101 kPa)
Silicon carbide GOX 250 psi (1724 kPa)
Silicone rubber sheet LOX <14.7 psi (<101 kPa)
Note: Data are for comparison purposes only and are not to be considered standard values for listed materials.
LOX ¼ Liquid Oxygen; GOX ¼ Gaseous Oxygen.
a
Composite materials made from carbon fibers and epoxy binder are dependent upon the specific binder and
the layup of the sheet.
basic rules: (1) use only when it is absolutely necessary, and (2) use only the
smallest workable amount. Lubricants pose several risks within a system
because many are flammable in oxygen, and lubricants become contaminants if
they enter into the oxygen stream. If a lubricant is absolutely necessary, then
fluorinated lubricants are currently the best choices.
Ceramic materials are generally oxidized compounds and tend to be com-
patible with oxygen. However, the use of ceramics is limited because they are
typically very fragile. Most oxygen systems need to be able to withstand inter-
nal temperature fluctuations and external temperature and humidity changes as
part of their normal operation. Ceramics do not undergo changes well, and so
their usefulness is limited.
Carbon fiber composite materials, also known as graphite epoxies, have
traditionally been shunned from oxygen systems because the binder, or matrix,
materials are typically incompatible with oxygen. However, laboratory testing

has demonstrated that some of these can be used in oxygen systems. Utilizing
graphite epoxy composites is advantageous because of their low weight and
high strength. Unfortunately, the factors that make graphite epoxy materials ac-
ceptable for use in oxygen systems are not well understood. Many variables in-
herent in the use of composites have strong influences on composites’ oxygen
compatibility, namely, binder material, layup, cure, shape, etc.
Materials that are not compatible with oxygen can be used effectively as
secondary materials for oxygen service, thus reducing the cost or weight of the
system. These secondary materials are not directly in contact with the oxygen
fluid, i.e., not oxygen wetted, but they can support the materials that are. Com-
posites and ceramic materials have successfully been used as structural sup-
ports for metals that cannot withstand the system pressures at their use
pressures. A good example of this is a composite overwrapped oxygen tank.
Special care must be taken with nonmetals if they are used within a breath-
ing oxygen system because some materials can produce toxic products that can
enter the oxygen stream and thus get into the lungs of the user. Fluorinated
elastomers and lubricants can produce toxic fluorinated products, especially if
any small burn within the system allows combustion byproducts to enter the
gas stream. It should be noted, however, that the fluorinated compounds have
lower heats of combustion if ignited. Therefore, fluorinated polymers and lubri-
cants tend to be the most preferable for use in oxygen systems.
The test data provided in Tables 4 and 5 clearly show that several classes
of materials work well in oxygen systems, whereas several other classes of
materials should be avoided. The data in these tables can be used for determin-
ing which materials are best able to support the use conditions.
Table 4 provides the minimum pressures required in order to ignite or burn
common metals in the promoted ignition test and the mechanical impact test.
The data in the table provide, for various metals, the lowest pressure at which
the metal will exhibit self-sustained combustion in the promoted ignition test
and the lowest pressure at which the metal will be ignited by mechanical
impact. The standard sample size for the promoted ignition sample is a rod 12
in. (305 mm) long and 0.125 in. (3.2 mm) in diameter. This test is conducted
per ASTM G124 [2]. The sample size for the mechanical impact test is a disk
with a 0.63 in. (16 mm) diameter and a thickness of 0.06 in. (1.5 mm). The me-
chanical impact test is conducted according to the method described in ASTM
G86 [4]. These data were generated via testing at NASA’s MSFC.
Table 5 provides the minimum pressures required in order to ignite com-
mon nonmetals in the mechanical impact test. Promoted ignition testing is typi-
cally not performed on nonmetals at higher pressures because a standard
sample rod will typically burn completely. The data in the table provide the av-
erage lowest pressure at which the nonmetal will be ignited by mechanical
impact. The standard sample size for the mechanical impact test is a disk with
DAVIS, doi:10.1520/STP20120001 17
a 0.63 in. (16 mm) diameter and a thickness of roughly 0.1 in. (2.5 mm). The
mechanical impact test is conducted per ASTM G86 [4]. These data were gen-
erated via testing at NASA’s MSFC.
Several newer tests are being used to provide supplemental information on
materials for use in oxygen service. These tests include the Oxygen Index Test
(ASTM G125 [5]), the Autogenous Ignition Temperature Test (ASTM G72
[6]), the Gaseous Pneumatic Impact Test (ASTM G74 [7]), and the Elevated
Temperature Promoted Ignition Test (ASTM G124, revisions after 2009 [2]).
The values of these tests are uncertain at present, and the data generated are
being scrutinized in order to determine whether any strong correlations can be
found between test results and the successful utilization of materials in oxygen
environments.
Configuration Dependence
The energy required in order to ignite a material or sustain its burning is not a
fixed value and is influenced by several factors, whether dealing with metals or
nonmetals. The amount of energy required in order to ignite a material, based
upon the minimum temperature and pressure at which it will ignite, is deter-
mined by the thickness of the material, the shape of the part produced by the
material, and the surface configuration. Metals are specifically sensitive to
composition variations and heat treatments. Nonmetals are sensitive to cure
conditions, batch=lot variability, and the age of the material.
The flammability of a metal is strongly influenced by its configuration, or
the size and shape of the part. Solid metals are the most resistant to ignition
and burn the least if ignited. Metal parts with greater surface areas will burn
more easily than those with less surface area. If a metal tube and a metal rod of
the same material and diameter are tested, the tube will ignite more easily.
Metal mesh materials, such as filter materials, are the easiest of all configura-
tions to ignite. For example, a 316 stainless steel rod will not ignite until the
pressure is generally more than 400 psi (2758 kPa), but a 316 stainless steel
mesh filter material shaped into a rod of the same dimensions will ignite in ox-
ygen below atmospheric pressure.
The variability in the flammability of a material in different configurations
is the primary reason that stainless steels are used extensively in high pressure
oxygen systems even though they are flammable under the use pressures. Oxy-
gen system components for high pressure applications typically have thick
walls designed to withstand the stresses. This thickness is also advantageous
for ignition resistance.
Minimize Ignition Sources and Mechanisms

An ideal oxygen system will have nothing inside it that could create an ignition
potential. There will be no contaminant, no floating debris, no metal shavings
from pipe threading, and only pure oxygen entering into the system. This is an
ideal scenario that does not exist in the real world. Therefore, steps must be
taken when designing an oxygen system in order to minimize the effects of
ignition sources and mechanisms that will certainly find their way into the
system.
Ignition sources and mechanisms are materials or processes that can them-
selves ignite or cause ignition of other materials within the system. The most
common ignition sources are contaminants in the system, and the most com-
mon ignition mechanisms are the result of improper system designs that allow
the materials to be subjected to sufficient energy to ignite them during
operations.
The safest oxygen systems are designed considering the following issues:
(1) Which ignition sources or mechanisms might be present in the system? (2)
What is the severity of the hazards produced? (3) What hazards might arise
from the worst-case scenario, such as any danger to people or equipment?
Several ignition sources and mechanisms have been determined, after
many years of failure investigations, for oxygen systems. The list provided
below is not complete but does represent the most common ignition sources
and mechanisms, or at least the ones that are currently understood.
(1) Contamination
(2) Particle impact
(3) Rapid pressurization
(4) Mechanical impact
(5) Friction
(6) Static discharge
(7) Electrical arc
(8) External heat
(9) External hazards
Contamination—A contaminant, commonly called foreign object debris

(FOD), is any unwanted object or particle of material that enters into the sys-
tem. FOD is the most common ignition source. As dangerous as it is, contami-
nation should be considered unavoidable, and mitigation plans should be
developed.
Sources of contamination include (1) dirt or debris entering the system dur-
ing assembly; (2) excess lubricants that leak from threading; (3) metal flakes
that fall from the pipe threading during assembly; (4) particles brought into the
system by impure oxygen; (5) particulates that arise from within the system,
such as through a breakdown of seals, flaking of vessel plating, valve friction
shearing metal pieces, etc.; and (6) items that have been dropped into the sys-
tem by workers. Contamination is a significant factor in each ignition
mechanism.
DAVIS, doi:10.1520/STP20120001 19
Particle Impact—The particle impact ignition mechanism involves heat

being generated by a particle’s striking the surface of another material with a
velocity that will generate sufficient heat to ignite the particle or the material.
This particle might be introduced during the assembly of the system or released
during system operation. This ignition mechanism generally requires a metal
particle and metal surface for it to strike. Nonmetals are generally considered
unable to be ignited by means of particle impact. Hard plastics have been
ignited, although rarely, via this mechanism. Figure 8 shows one particle
impact ignition scenario.
The amount of laboratory research that has been conducted on this ignition
mechanism is limited, and new research might change some of the assump-
tions. Current beliefs about oxygen dictate that four factors must be present
before particle impact ignition can occur: (1) the system must be a gaseous
oxygen system; (2) the particle must be one of the more flammable metals,
such as aluminum, titanium, etc.; (3) the flow rate of the oxygen gas must be
FIG. 8—Particle impact ignition can result from a particle striking a flamma-
ble part of a valve. Note the orifice accelerating the particle, increasing the
ignition probability (from NASA TM-213740 [8]).
greater than 45 m=s (150 ft=s); and (4) the particle must impact the surface of
the material at full force and not with a glancing blow.
The particle impact ignition hazard can be mitigated by (1) cleaning sys-
tem components specifically for oxygen service; (2) threading pipes carefully
so as not to release particles; (3) utilizing internal filter elements for trapping
particles; (4) regulating the oxygen flow rate to the lowest tolerable level; (5)
designing systems to produce flow paths that have minimal points where
particles can strike at severe angles; and (6) minimizing components that can
generate particles, such as rotating valves, sliding parts, etc.
Rapid Pressurization—The rapid pressurization ignition mechanism, also

known as the heat of compression or adiabatic compression, involves extreme
heat generated by the oxygen gas undergoing pressurization. The ideal gas law
and thermodynamic equations for an adiabatic process (i.e., one with no heat
loss) demonstrate that if the oxygen gas pressure increases rapidly, then the gas
temperature increases rapidly. The formula for determining the final tempera-
ture achieved in this process is
ðc1Þ=c
Tf ¼ Ti Pf =Pi
in which c represents the adiabatic index of the gas for O2 ¼ 1.40.

If the mechanism of rapid pressurization is not carefully controlled, the
temperature of the gas and surrounding materials might increase beyond the
ignition point of the materials. Figure 9 illustrates this mechanism in a system.
Rapid pressurization ignition is not possible unless the following three
conditions are present:
(1) A significant pressure spike increases the pressure from near ambient
to at least 500 psi (3447 kPa).
FIG. 9—Rapid pressurization ignition results from the rapid compression of

oxygen (from NASA TM-213740 [8]).
DAVIS, doi:10.1520/STP20120001 21
(2) Rapid pressurization occurs within a fraction of a second.

(3) A flammable nonmetal is present close to the highest temperature gas
generated.
The final temperature resulting from rapid pressurization is very high. For
example, starting from ambient pressure, the final temperature of a system rap-
idly pressurized to 2000 psi (13 790 kPa) is 1688 F (1193 K), and rapidly
pressurizing to 5000 psi (34 474 kPa) produces a temperature of 2330 F
(1550 K). Nonmetals are easy to ignite at these temperatures.
The rapid pressurization ignition potential must be minimized in oxygen
systems via good design practices, including (1) limiting pressurization rates;
(2) minimizing the amount of nonmetals, including contaminants, in areas
affected by pressurization; (3) burying nonmetals behind metal parts in the
flow path; and (4) compressing oxygen slowly in the vicinity of soft goods.
Mechanical Impact—The mechanical impact ignition mechanism involves

the heat energy generated when two objects collide. Numerous collisions occur
with normal operations, such as the closing of valves, that are designed into the
system. However, collisions sometimes occur with large pieces of contaminant
debris or as a result of parts of the system breaking free. These sources of
mechanical impact must be mitigated. Figure 10 shows a mechanical impact
ignition hazard that has resulted in a number of fires during an oxygen transfer
process.
FIG. 10—Mechanical impact ignition results from a wrench falling onto an

asphalt pad covered in liquid oxygen (from NASA TM-213740 [8]).
FIG. 11—Frictional ignition occurs when damaged or worn soft goods result
in metal-to-metal rubbing (from NASA TM-213740 [8]).
The mechanical impact ignition mechanism is most commonly witnessed

during the opening and closing of valves, in which heat is generated from the
frequent slamming together of parts. A metal part impacting a polymer seat
can produce sufficient heat to ignite the seat. Two conditions are necessary for
mechanical impact ignition: (1) a nonmetal or highly flammable metal must be
impacted by another material such that its temperature rises above its auto-
ignition temperature, and (2) the impact must deliver a significant amount of
energy, either through multiple impacts occurring rapidly or through one single
strong impact. Valve materials are commonly subjected to multiple impacts,
especially if polymer seal chatter occurs. Large foreign objects in the flow
stream, however, typically lead to one forceful strike.
Friction—The friction ignition mechanism involves the heat energy gener-

ated when two objects rub together. Frictional rubbing of two materials in oxy-
gen can provide sufficient heat to ignite one of the materials in the friction
load. In addition, heat generated in the friction process can ignite debris near
the rubbing surfaces. A subset, and more severe case, of friction is galling
between two metal parts. Figure 11 illustrates a condition in which friction
ignition can occur in an oxygen system.
Friction ignition can occur if the following three factors are present:
(1) Two metals are rubbing together.
(2) Rubbing must be at a high speed in order to generate enough heat to
ignite one of the metals.
DAVIS, doi:10.1520/STP20120001 23
(3) The metals must undergo a high load, i.e., press together with enough
force to make the rubbing severe.
Some components, such as check valves, regulators, and relief valves,

might become unstable and chatter during use. Chattering can result in rapid
oscillation of the moving parts within these components, creating a friction
ignition hazard. This hazard should be minimized by avoiding the rubbing to-
gether of two metal parts or by slowing the rate at which they rub.
Less Common Ignition Sources and Mechanisms

Several ignition mechanisms are rarely witnessed in oxygen systems but should
not be ignored. These include static discharge, lightning strikes, explosions or
tank ruptures in nearby systems, open flame fires that get close to the system,
welding, and severe weather damage such as that caused by a hurricane, tor-
nado, earthquake, flood, etc. Special considerations might be necessary if the
oxygen system will be located in a hazard-prone area.
Ignition Happens but Nobody Knows Why; Formerly Known as “Flow

Friction”
Oxygen system failure investigators sometimes really do not know why an
ignition has occurred, but the failure investigation clearly shows that something
went wrong and one or more materials ignited. One historical theory stated that
the rapid flow of gas across a polymer could heat the polymer to the point of
ignition, which might create a kindling chain of burning that could lead to a
catastrophic failure. This ignition mechanism became known as the flow fric-
tion ignition mechanism.
The flow friction ignition theory has now been determined to be a highly
unlikely ignition mechanism. This ignition mechanism is discussed because this
mechanism is frequently listed in the literature. To date, flow friction has never led
to the ignition of a material in a laboratory experiment, even after exhaustive labo-
ratory testing under conditions ideal for creating an ignition by means of flow fric-
tion. It is now believed that these ignitions were actually caused by one or more
mechanisms that could not be determined or are currently not understood.
The flow friction ignition mechanism assumed that the following three
conditions were required:
(1) Oxygen must be at a pressure greater than 500 psi (3447 kPa).
(2) Nonmetals must be present to ignite.
(3) Rapid flow causes erosion, vibration, friction, or a similar mechanism
for heating a nonmetal.
Further research will be necessary in order to fully elucidate the actual

ignition mechanism that has been called flow friction.
Importance of Appropriate Cleaning

An important factor in the safe use of an oxygen system is appropriate clean-
ing, which provides contamination avoidance. Contaminant particles can accu-
mulate to dangerous levels that are easily ignited. The following are keys to
clean oxygen systems:
(1) Use only oxygen system approved cleaning agents and techniques.
(2) Clean all parts prior to system assembly, if possible.
(3) Minimize the number of different materials used for assembly, such
as seal materials, lubricants, etc. Users are more likely to notice an
incorrect material when few are used.
(4) Visually inspect all components before use and reject any compo-
nent that does not appear to be 100 % clean.
(5) Avoid using vendor labeled “oxygen compatible” parts. The vendor
claim must be independently verified, or the part must be rebuilt
using only oxygen compatible materials.
(6) Protect clean parts from contamination by keeping them covered
until assembly and assembling as many parts as possible in a clean
area.
(7) Carefully assemble so that two surfaces rubbing together will not
generate particulates that can enter into the system.
(8) Use lubricants only when absolutely necessary, and ensure that the
lubricants are oxygen compatible. Use the smallest amount of lubri-
cant allowable.
(9) Ensure that verification testing of the cleanliness level has been per-
formed and that the cleanliness level is within the allowable
tolerance.
(10) Ensure that vent lines guard against contaminants’ entering the
system.
(11) Use an inert gas to blow through the system in order to remove any
stray particles prior to the introduction of oxygen.
Trivial Overview of an ASTM Oxygen Compatibility Assessment

Students who attend the ASTM TPT class Fire Hazards in Oxygen Systems
learn that an oxygen compatibility assessment (OCA) is the applications part
of determining whether hardware is acceptable for use in contact with oxygen.
An OCA is best described as a formal hazards analysis for the safety of specific
materials and hardware in a defined oxygen environment. All of the informa-
tion that has been presented up to this point is the education needed in order to
determine what is safe and what is unsafe for oxygen systems. The OCA is the
point where the “rubber meets the road,” where issues, solutions, and recom-
mendations are formally documented for use.
DAVIS, doi:10.1520/STP20120001 25
Many organizations, including ASTM, have over the years generated a

number of standards and procedures that produce safer oxygen systems. These
standards and procedures are used during the OCA to enhance system safety.
The best education available on performing an OCA is obtained by taking part
in the ASTM course mentioned earlier.
An OCA is a tool that leads the user through a step-by-step procedure for
assessing the risks associated with an oxygen system. The OCA utilizes the
system materials list, drawings of components, and system operating condi-
tions to proceed through a structured risk assessment. The structural approach
of the OCA helps prevent some areas from being overlooked while providing a
structural approach to necessary remedies. An oxygen system designer will
find that the OCA helps outline the known hazards that might be present, high-
lights areas that need further review, and provides safety improvement
strategies.
An OCA is not necessary in one specific case: that in which all of the mate-
rials used within the system are oxygen compatible at the maximum use tem-
peratures and pressures of the system. This approach, however, allows very
few materials to be used, and almost all of these are expensive to purchase.
The desire to allow a wider selection of materials, especially lighter weight and
less expensive materials, necessitates the use of the OCA tool.
The OCA involves several specific steps during the formal process. These
steps are provided here with, for brevity, little information. This explanation is
provided simply to inform the reader of the existence of the OCA process and
the information that the assessment will provide. Anyone needing to produce
an OCA is strongly encouraged to complete the ASTM TPT course.
An OCA is accomplished by following these steps for an oxygen system in
which all of the conditions present are known:
(1) Determine the worst case operating conditions at each point within
the system,
(2) Assess the flammability of the oxygen-wetted materials at their “worst
case” conditions,
(3) Evaluate potential ignition mechanisms and determine the probability
of their occurrence,
(4) Evaluate any kindling chain within the system and determine the most
severe possible effects,
(5) Determine the reaction effects, i.e., the severity of the worst potential
outcome of a hazard, including human casualty, equipment destruc-
tion, etc.,
(6) Compile a list of hazards and determine methods of correcting them,
and
(7) Document the results for each hazard or component.
New information becomes available almost every day concerning oxygen
systems hazards. This new information has necessitated updating of the current
ASTM TPT course, and this course update was in progress as of the writing of
this paper. The last course update, from 2009, stopped at step 7 above. How-
ever, NASA’s MSFC has incorporated three additional steps into its OCA pro-
cess that will most likely be added to the next edition of the ASTM TPT
course. These additional steps are as follows:
(8) Provide recommendations and required modifications for system
safety improvements,
(9) Provide the limitations of the provided OCA, and
(10) Determine the safety of the hardware or system if it is built exactly
as has been proposed.
Conclusions
Oxygen systems are inherently dangerous and must be designed only by pro-
fessionals who have a working knowledge of the intricacies of these sys-
tems. The information provided in this paper serves to introduce neophytes
to important safety concerns regarding oxygen systems. This paper does not
provide sufficient information for the reader to design a safe oxygen system,
or even to determine the safety of an existing one. It does serve to provide
introductory information that will give the reader a basic understanding of
oxygen systems, and background information that will allow an understand-
ing of the more detailed standards and procedures available on the subject
[9–17].
References
[1] NASA-STD-6001, “Flammability, Offgassing, and Compatibility

Requirements and Test Procedures,” National Aeronautics and Space
Administration, Washington, DC, USA, 2011.
[2] ASTM G124, 2010, “Standard Test Method for Determining the Burning
Behavior of Metallic Materials in Oxygen-Enriched Atmospheres,” Annual
Book of ASTM Standards, Vol. 14.04, ASTM International, West Consho-
hocken, PA.
[3] ASTM D2512, 2011, “Standard Test Method for Compatibility of Materi-
als with Liquid Oxygen,” Annual Book of ASTM Standards, Vol. 14.04,
ASTM International, West Conshohocken, PA.
[4] ASTM G86, 2011, “Standard Test Method for Determining Ignition Sen-
sitivity of Materials to Mechanical Impact in Ambient Liquid Oxygen and
Pressurized Liquid and Gaseous Oxygen Environments,” Annual Book of
ASTM Standards, Vol. 14.04, ASTM International, West Conshohocken,
PA.
DAVIS, doi:10.1520/STP20120001 27
[5] ASTM G125, 2008, “Standard Test Method for Measuring Liquid and
Solid Material Fire Limits in Gaseous Oxidants,” Annual Book of ASTM
Standards, Vol. 14.04, ASTM International, West Conshohocken, PA.
[6] ASTM G72, 2009, “Standard Test Method for Autogenous Ignition Tem-
perature of Liquids and Solids in a High-Pressure Oxygen-Enriched Envi-
ronment,” Annual Book of ASTM Standards, Vol. 14.04, ASTM
International, West Conshohocken, PA.
[7] ASTM G74, 2008, “Standard Test Method for Ignition Sensitivity of
Materials to Gaseous Fluid Impact,” Annual Book of ASTM Standards,
Vol. 14.04, ASTM International, West Conshohocken, PA.
[8] NASA TM-213740, 2007, “Guide for Oxygen Compatibility Assessments
on Oxygen Components and Systems.”
[9] ASTM G128, 2008, “Standard Guide for Control of Hazards and Risks in
Oxygen Enriched Systems,” Annual Book of ASTM Standards, Vol.
14.04, ASTM International, West Conshohocken, PA.
[10] ASTM G145, 2008, “Standard Guide for Studying Fire Incidents in Oxy-
gen Systems,” Annual Book of ASTM Standards, Vol. 14.04, ASTM Inter-
national, West Conshohocken, PA.
[11] Beeson, H., Smith, S., and Stewart, W., Safe Use of Oxygen and Oxygen
Systems: Handbook for Design, Operation, and Maintenance, 2nd. ed.,
ASTM Manual 36 (MNL36), ASTM International, West Conshohocken,
PA, 2007.
[12] CGA G-4, 2008, “Oxygen,” Compressed Gas Association, Inc., Chantilly,
VA.
[13] CGA G-4.1, 2009, “Cleaning Equipment for Oxygen Service,” Com-
pressed Gas Association, Inc., Chantilly, VA.
[14] ISO 14624-1, 2003, “Space Systems—Safety and Compatibility of Mate-
rials—Part 1: Determination of Upward Flammability of Materials,” ISO
Copyright Office, Geneva.
[15] ISO 14624-4, 2004, “Space Systems—Safety and Compatibility of Mate-
rials—Part 4: Determination of Upward Flammability of Materials in
Pressurized Gaseous Oxygen or Oxygen-Enriched Environments,” ISO
Copyright Office, Geneva.
[16] NFPA 53, 2011, “Recommended Practice on Materials, Equipment, and
Systems Used in Oxygen-Enriched Atmospheres,” National Fire Protec-
tion Association, Quincy, MA.
[17] NASA, 2012, “Materials and Processes Technical Information System
(MAPTIS),” http://maptis.nasa.gov
STP 1561, 2012
DOI:10.1520/STP20120003
H. W. Mulkey1 and D. M. Lineberry2
Development of a Liquid Oxygen Facility

for Rocket Engine Testing
REFERENCE: Mulkey, H. W. and Lineberry, D. M., “Development of a Liquid

Oxygen Facility for Rocket Engine Testing,” Flammability and Sensitivity of
Materials in Oxygen-Enriched Atmospheres on September 19–21, 2012 in
Montreal, QC; STP 1561, Samuel Edgar Davis and Theodore A. Steinberg,
Editors, pp. 28–46, doi:10.1520/STP20120003, ASTM International, West
Conshohocken, PA 2012.
ABSTRACT: The Propulsion Research Center at the University of Alabama
Huntsville has developed a liquid oxygen propellant feed system that has
advanced into operational maturity. In support of safe system functional de-
velopment, an oxygen compatibility and hazards assessment was completed
in order to identify and minimize operational risks. The objective of this effort
was to minimize potential risks that might have been overlooked in the initial
design phase or resulted from a discontinuity between design and fabrica-
tion. The effort involved an internal review of the operating procedures and
the system configuration, as well as an overall review by an oxygen safety
engineering consultant. This paper outlines the review procedures applied to
the oxygen propellant systems, summarizes the significant findings, and
details configuration changes made to the system. The review process identi-
fied several areas in which reliance on procedural control, without compro-
mising safe system operation, could be reduced or removed through design
changes. Such changes were implemented so as to create a more fault toler-
ant system and minimize risks to test equipment, system hardware, and test
personnel. Major upgrades included initiatives to manage the oxygen risk
and mitigate specific oxygen ignition mechanisms. Oxygen compatible mate-
rials and facility configuration changes enabled safer experimentation in the
demonstration testing of the cryogenic propellant facility. The findings pre-
sented in this paper represent the system changes made in an effort to
Manuscript received January 30, 2012; accepted for publication April 27, 2012; published online
October2012.
1
Research Assistant, Mechanical and Aerospace Engineering, Propulsion Research Center,
University of Alabama Huntsville, Huntsville, AL 35899.
2
Research Engineer, Mechanical and Aerospace Engineering, Propulsion Research Center,
University of Alabama Huntsville, Huntsville, AL 35899.
Copyright V
19428-2959.
28
MULKEY AND LINEBERRY, doi:10.1520/STP20120003 29
increase safety and decrease the risk of fire in order to achieve a low proba-
bility of ignition and a low consequence of ignition.
KEYWORDS: liquid oxygen, liquid rocket engine testing, university of ala-
bama huntsville propulsion research center, propulsion testing facility, oxygen
hazards
Introduction
The Propulsion Research Center (PRC) has a high-pressure combustion facility
for small-scale liquid, gaseous, and solid rocket testing located on the Univer-
sity of Alabama Huntsville campus. In this environment, students conduct
research and operate facilities containing inherent hazards. Rocket engine test-
ing can be unpredictable and necessitates proper realization of the potential
hazards. The high personnel turnover rate associated with student-related uni-
versity research further emphasizes the need for a fault minimized testing facil-
ity. Thus, the objective from a facility design standpoint is to minimize failure
associated with operator error. This objective is especially important in an aca-
demic setting and requires a comprehensive training program on the opera-
tional risks, as well as uniformity in standard operating procedures, in order to
maintain personnel safety and long-term system functionality.
In addition to the functional requirements, new test stand capability devel-
opment requires appropriate investigation of the associated hazards and meth-
ods for mitigating these hazards through design, development, and operation.
In general, oxygen systems should be reviewed by knowledgeable personnel
with training in various fire hazards and design principles essential to the safe
use of such systems. Thus, an internal review was conducted and applied to the
operational gaseous oxygen (GOX) igniter system and liquid oxygen (LOX) sys-
tem. As part of this process, a formal oxygen compatibility assessment (OCA)
and hazards analysis were developed and documented for the existing LOX facil-
ity in accordance with the standards [1,2] for the safe use of oxygen systems.
Rocket Engine Test Facility

The test facility consists of two independent systems that make rocket engine
operation possible: a main propellant system and an igniter system. The main
propellant system supplies the fuel and oxidizer for the primary investigation.
To enhance the research capability, a LOX system was designed and incorpo-
rated into the main propellant oxidizer branch. The PRC LOX facility system
was designed for a maximum flow rate of 1.4 kg/s (3 lbm/s), flow velocities of
4.6 to 9.1 m/s (15 to 30 ft/s), and a maximum system pressure of 13.8 MPa
(2000 psi). The LOX main feed tank volume has a capacity of 87.1 l (23 gal)
and is cryogenically chilled using a liquid nitrogen (LN2) outer jacket. LN2 is
bled off of the outer jacket and flows through Teflon tubing, jacketing the LOX
delivery lines to reduce heat loss from the oxygen main propellant flow. The
igniter system supplies a secondary fuel and oxidizer during the ignition phase
to start combustion of the main propellants. In the PRC test stand, ignition pro-
pellants consist of GOX and gaseous hydrogen delivered from single standard
K-bottle gas cylinders. Mass flow for each system is regulated through critical
orifices. The test stand also has a supply of gaseous nitrogen (GN2) for use as
both a pressurant and a purge gas. Fuel and oxidizer supplies are separated by a
reinforced concrete wall, and additionally the rocket engine is separated from
the feed system components by a bulkhead.
Liquid Oxygen System Assessment Strategy

The PRC implemented a three-part approach to evaluating the LOX capability
enhancement in the test stand. The approach consisted of personnel training, an
internal system assessment, and an external review. In addition to the Red
Cross first aid and cardiopulmonary resuscitation–automated external defibril-
lator training required of all PRC personnel, the ASTM “Oxygen Systems
Operation and Maintenance” and “Fire Hazards in Oxygen Systems” training
courses were provided to system designers and operators. These courses pro-
vided insight into the fundamental principles of the safe use of oxygen systems.
The training familiarized PRC personnel with oxygen specific hazards, poten-
tial ignition mechanisms, and best practices for mitigating the risks associated
with oxygen systems. The training also provided guidelines for system- and
component-level evaluations.
The internal system assessment followed guidelines outlined by ASTM for
managing oxygen hazards. PRC specific oxygen system– and component-level
compatibility assessment procedures were developed based on the ASTM
guidelines representing the best practices for oxygen systems. The procedure
used the ASTM recommendations for design, materials selection, maintenance,
and oxygen compatibility testing of materials and components. The systems
were evaluated in both nominal and any off nominal situations. This attempt to
reduce the reaction effects of individual components represents an efficient
way to ensure successful operation in oxygen systems, making the system the
most fault tolerant to the unpredictable nature of the reactivity of oxygen. The
system OCA also serves as a tool that one can use to better recognize and docu-
ment the chosen components and potential configuration hazards of the existing
system.
After the internal system assessment was completed, an external review
was conducted. Oxygen safety engineering consultants reviewed the docu-
ments from the internal OCA and inspected the facility. After the evaluation, a
set of recommendations was provided for implementation. The recommenda-
tions from the audit led to an operational facility tactic to eliminate flammable
materials and ignition mechanisms, provide procedural control, or implement
barriers to reduce risks in the PRC oxygen systems.
Oxygen Compatibility Assessment Procedure

In accordance with ASTM methodology, PRC personnel completed and docu-
mented an OCA for both GOX and LOX systems of the hot-fire test facility.
The specific focus of the OCA was system-level components in the current
configuration of the LOX main propellant system and the GOX igniter system.
The thought process throughout the OCA development as applied to the LOX
system focused on realizing, identifying, and managing the associated hazards.
In addition, the OCA process provided a means to identify and address compo-
nent- and configuration-based hazards. Figure 1 shows the flow chart used by
the PRC to manage the oxygen hazard.
The OCA procedure requires that the worst case operating conditions be
established in order to assess the flammability of the oxygen wetted materials,
evaluate the presence and probability of ignition mechanisms, and determine
reaction effects. The OCA documents this analysis for the specific oxygen sys-
tem components. For each component, worst case operating conditions were
established in conjunction with a single point system failure resulting in unpre-
dictable or catastrophic system operation [1]. This most severe condition
addressed the extreme fire hazards associated with oxygen concentration, tem-
perature, pressure, flow rate, and cleanliness level.
FIG. 1—Oxygen system risk management.

All oxygen wetted LOX and GOX system components were identified, and
individual component flow path cross sections were obtained. The materials of
construction for each component were evaluated for oxygen compatibility.
This is the first step in managing the oxygen fire hazard risk related to the reac-
tivity of material in oxygen environments. Materials were judged based on
flammability, how easily the material can be ignited, the heat of combustion or
heat release per unit mass of the burning material, the minimum oxygen con-
centration required in order for the material to react, and the autoignition tem-
perature (i.e., the minimum temperature at which a material will ignite).
ASTM oxygen system standards [1,2] provide a diverse collection of engineer-
ing material test data relevant to all of these compatibility factors. These data
were applied using technical judgment in conjunction with an evaluation of
configuration dependences. Oxygen system components were evaluated in
terms of potential oxygen ignition mechanisms and the relevant characteristic
elements.
The kindling chain and reaction effect of specific LOX and GOX system
components were considered in the next part of the OCA. This part of the OCA
identifies the result of a fire in a system component. The effects of an ignition
were assessed to determine whether a fire could propagate and burn out the
component, leading to system failure, a non-functional system, or simply a
leaky valve that could be replaced. The reaction effect of a specific component
supplied a measure of the consequence of fire in that component relating to test
objectives, equipment damage, and test personnel. The final step of the OCA
process was to assess the consequences of component failure and determine
whether corrective action was required. Any failure that could endanger per-
sonnel required corrective action. If the failure did not endanger personnel, the
decision for corrective action was based on the likelihood of an occurrence, the
consequence to the system or to the test objective, and the level of risk deemed
acceptable. When selecting the corrective action, a four tiered approach con-
sistent with the review audit methodology was used. The first approach was to
replace flammable material with more burn resistant oxygen compatible mate-
rial. This could be as simple as replacing the soft goods in a valve port connec-
tion, or it might involve a more intensive repair, such as replacing the valve
itself. The second approach was to eliminate the ignition mechanism from the
system. Possible approaches for resolving these mechanisms include adding fil-
ters to remove potential particulate, replacing fast opening valves with multi-
turn valves, and adding distance volume pieces to absorb compression heating,
among others. The third approach was to eliminate the possibility and/or
reduce the severity of failure through procedural control. This approach
involves ensuring that no personnel are exposed to components during situa-
tions that might initiate a failure, as well as following test procedures that can
eliminate ignition mechanisms. The fourth and final approach was to apply
appropriate barriers to protect personnel and equipment from potential failures.
FIG. 2—OCA decision tree.
Barrier protection was considered a last resort, but it can be an effective means
to reduce risks in oxygen systems. Figure 2 outlines a decision tree relating to
the iteration followed in this process.
Oxygen Compatibility Assessment Documentation

A specific component OCA example from the PRC’s LOX system is provided
in Table 1. The LOX main valve was chosen to illustrate the OCA procedure.
The LOX main valve is a pneumatic globe valve (PGV) located immediately
downstream of the LOX main propellant feed tank. The valve is used to isolate
the LOX supply tank from the remainder of the system. Figure 3 illustrates the
LOX main valve oxygen wetted cross section [3] with material callouts. Table
1 provides a sample assessment of the LOX main PGV under worst case oper-
ating conditions in the PRC facility. Footnotes are provided to capture the
thought process at the time of the OCA analysis. Relevant documentation pro-
viding the type of valve, the manufacturer, and the manufacturer part number
and corresponding facility label was included for comprehensive facility docu-
mentation. All components of the LOX and GOX systems were analyzed in
this manner.
TABLE 1—LOX main PGV OCA.
Nomenclature Description Manufacturer Part Number Facility Label
LOX Main Valve Pneumatic Globe Valve, 0.5 in. Ports, Ignition Mechanism Kindling Reaction
with Fail Closed Actuator, Cv ¼ 1.7 Ratings (0–4) Chain? Effect
Yes/No (A–D)
Worst Case Material Particle Component Friction/ Mechanical Electrical Flow Other/
Operating Condition Flammable? Impact Heating Galling Impact Arc/Spark Friction Chatter
Yes/No
Component Material Temperature, F Pressure, psi
Body 316 stainless Yesc

steel
Stem 316 stainless Yes
steel
Stem seal Teflon Yesm
a b d e f g h i j k l
Poppet 316 stainless Ambient 2500 Yes 1 0 0 1 0 2 0 Yes C
steel
Seat 316 stainless Yes
steel
Port seal Teflon Yes
a
Worst case operating sustained temperature from a single point failure.
b
Worst case operating pressure from single point failure.
c
Stainless steel must be considered flammable above 111 psi in 100% oxygen (Table 3.1 in Ref 1).
d
Particle impact is not a credible ignition source in LOX; however, during filling and chill, GOX flow could impact the LOX main valve seat housing, presenting a particle impact ignition hazard. Severe impact
geometries inside the valve in the given flow direction are present. LN2 jackets are employed to chill the LOX main flow lines before LOX is pressurized past the main valve. The LOX is additionally filtered to
remove particulate before it enters the tank.
e
Heat of compression is not a credible ignition source in LOX. Rapid pressurization is controlled by operation; the GOX/LOX filling the system is cold and of a sufficient volume to dissipate any heat of compres-
sion that might result upon pressurization. The cold oxygen and its sufficient volume result in a low possibility of ignition.
f
Friction/galling is not possible because there are not two or more rubbing surfaces.
g
In mechanical impact tests, Teflon experienced 3/40 reactions/tests at 2500 psi (Table 3.16 in Ref 1). This ignition mechanism is possible but not probable.
h
The component is not electrically powered.
i
Flow friction is not a credible ignition source in LOX; however, during chill down, GOX flow could leak past the Teflon sealing surface or the Teflon port connection seal, presenting all the characteristic elements
of flow friction ignition. Leaks must be monitored and corrected in order to address flow friction ignition mechanisms in this component.
j
Chatter does not occur in this component.
k
A kindling chain exists if the Teflon ignites as a result of flow friction and the flame propagates to the stem and body. The probability of ignition by flow friction is low, but a kindling chain does exist. Mechanical
impact ignition of the Teflon also is possible, but not probable. Particle impact could also be a source of ignition during LOX fill.
l
The valve possesses flammable components and one possible and two remotely possible ignition mechanisms, as well as a kindling chain; however, the valve is remotely operated, minimizing the effect on
personnel. Because of the importance of the LOX main valve in relation to isolation to the LOX tank, the effect of fire on the system objective and functional capability increases. The reaction effect is critical.
m
Teflon is flammable in 95% to 100% oxygen at ambient pressure (Table 3.12 in Ref 1).
FIG. 3—LOX main pneumatic globe valve.

Results
This section discusses some of the corrective actions implemented in order to
minimize potential oxygen hazards in the PRC’s LOX and GOX propellant
systems. The LOX system contained hardware specific hazards that necessi-
tated facility changes. These changes were related to the protection of both
hardware and personnel. The system was designed safely to procedurally con-
trol the operator risks through remote access of the severe components. The
OCA process identified practical engineering fixes for the oxygen propellant
facilities that would ensure safe system operation.
System Cleanliness
One of the primary concerns with the PRC oxygen facilities was the lack of
traceability to a cleanliness level consistent with oxygen systems. The LOX
and GOX main facility components were obtained from reputable manufactur-
ing sources familiar with oxygen cleaning practices. Multiple cleaning proc-
esses and clean assembly practices were undertaken to ensure a clean system
for safe and functional oxygen system operation. However, this approach did
not guarantee a clean system, and post-assembly verification had not been per-
formed on either system. This provided no traceable indication of component
fittings or tubing cleanliness for oxygen service. Furthermore, any part of a sys-
tem connected to an oxygen flow line, such as the GN2 purge and pressuriza-
tion lines, must also be cleaned for oxygen service, because these connected
lines could introduce hazardous material into an oxygen flow line. In order to
establish a baseline of cleanliness, a particulate and nonvolatile residue (NVR)
analysis was performed on multiple sections of each system. Stainless steel
flow lines from the LOX and GOX systems were removed from low points or
sumps, which would inherently collect the most residue and represent the most
severe concentrations.
The GOX igniter system cleanliness verification analysis established ac-
ceptable results for oxygen system operation. The oxygen cleanliness levels of
the LOX system propellant flow lines sampled were below acceptable stand-
ards. The first LOX propellant flow line that was sampled contained green par-
ticulate matter. The largest particle was 3750 lm (0.148 in.), which is larger
than the LOX propellant flow control orifice diameter. This produced obvious
concerns over the system cleanliness level. Large particulates gave rise to con-
cerns regarding damage to flow component valve seats and plugging of the ori-
fice. The source of the particulate was speculated as being, and later proven to
be, a green Teflon coated seal located at the top of the main LOX tank. The
seal had been installed in this particular tank several years earlier, although the
main tank had been in storage. The deterioration of this seal resulted in erosion
of the coating material. These particles collected in the LOX propellant flow
line. The second LOX propellant flow line sampled contained particulate as
well. This section contained less abundant particles than the former but was
still below acceptable standards for oxygen system operation. The NVR level
in both sampled LOX propellant lines was higher than nominal but well within
the bounds of oxygen cleanliness operational levels. The seal was replaced and
the LOX propellant system was thoroughly re-cleaned. These two propellant
flow lines were then sampled again. The second analysis revealed even lower
NVR levels and particulate counts, resulting in a cleanliness level appropriate
for oxygen system operation. As part of the second analysis, green chips were
taken from the deteriorated seal and analyzed. These were verified as having
the same chemical composition as the green particulates identified in the first
LOX propellant flow line oxygen cleanliness analysis. The NVR and particu-
late analysis records were added to the OCA documentation so as to provide
long term traceability.
The hazards associated with an unclean system in oxygen service are
avoidable. Cleaning, verifying that the system component is at the required ox-
ygen service cleanliness level, maintaining system cleanliness in line with sys-
tem maintenance best practices, and traceability through documentation can
mitigate the hazards associated with an unclean system. The best approach is
to start clean, assemble clean, verify clean, and maintain clean [1].
Liquid Oxygen Facility

The original LOX vent lines were 1.27 cm (0.5 in.) stainless steel and contained
a multidirectional flow path from the LOX tank main vent valve to the ambient
atmosphere away from the test area. The vent flow path contained several 90
and 45 bends, elbows, and tees, resulting in a complex flow path with numer-
ous flow direction changes. This configuration provided numerous impact loca-
tions for system particulate, which, in conjunction with high velocity gas flow,
could result in a particle impact ignition. Because stainless steel is considered a
flammable component in enriched oxygen environments [1], once ignited, a
fire could be sustained. Even if a system is filtered properly and assembled oxy-
gen clean, particulate can be generated through general maintenance of an oxy-
gen system. Particulate can additionally be generated by flow components such
as chattering flow through operation of the non-return check valve (NRCV)
downstream of the LOX main tank vent valve. The most notable hazard in the
original LOX main tank vent configuration was a stainless steel tee directly
downstream of a stainless steel burst disc. This is a highly probable location
for a particle impact ignition mechanism. The burst disc is a stainless steel ma-
terial and, if ruptured, could generate particulate traveling at high velocity
directly into a flammable stainless steel tee.
The LOX line vent configuration included many of the potential compo-
nent hardware hazards indentified in the LOX main tank vent configuration. In
the LOX line vent original component arrangement, a stainless steel flammable
tee was placed directly in front of the LOX line vent valve. There was also an
NRCV positioned directly downstream of the tee. This configuration increases
the potential for a particle ignition mechanism to interact with a flammable
component and ignite the remaining flammable materials. The LOX line vent
ties directly into the LOX main tank vent lines. Potential particulate generated
from the LOX line vent NRCV follows the venting flow path through the direc-
tional changes to the ambient atmosphere. This represents the same particle tra-
jectory impingement as seen in the LOX main tank vent configuration flow
path with multiple flammable targets.
The LOX main vent and LOX line vent configurations were restructured so
as to reduce potential ignition sources. The LOX facility main vent lines were
redesigned with larger diameter 3.81 cm (1.5 in.) copper vent lines in order to
reduce the pressure of the venting oxygen flow and, additionally, maximize the
use of more burn resistant copper material. Straighter flow path transitions
from the stainless steel flow lines into the copper vent lines minimize
the potential risks and reduce the number of potential particle impact
locations. More burn resistant copper impact locations reduce the probability
of ignition from a particle impact. All stainless steel tubing connecting the
main tank relief to the vent lines was given an appropriate flow recovery length
of ten diameters and smooth bend radii to accomplish changes in the direction
of oxygen flow.
GN2 pressure testing of the LOX system in which certain sections of the fa-
cility were filled and held under pressure kept revealing undetectable leaks.
The burst disc (BD) utilized in the system weakened in routine system pressure
checks, leading to small ruptures. This represents one common BD fatigue fail-
ure mode. The original LOX facility BD was sized for a low temperature oper-
ation. This low temperature design in the BD lowered the burst pressure in
ambient temperatures. The BD was replaced with an ambient temperature
design. This change prevented disc rupture under pressure in ambient
conditions.
In the original configuration, Viton O-rings were used in all situations to
seal ports connecting the LOX main tank and system flow component valves to
the seamless stainless steel tubing. Viton is not generally serviceable at low
temperatures. Concern existed about O-ring material embrittlement leading to
leaks through the seals that in turn could result in a potential flow friction igni-
tion source. The Viton O-ring seals were replaced with Teflon for better mate-
rial compatibility and structural integrity at low cryogenic temperatures.
In situations in which particle impact ignition mechanisms were prevalent,
stainless steel components were replaced with drop in burn resistant materials.
Tees and fittings made of Monel, which is a burn resistant material [1], were
used to replace stainless steel ones as a means of addressing this ignition mech-
anism. The area directly underneath the LOX main propellant feed tank is one
in which such measures were taken. The particle impact ignition mechanism is
not a credible ignition source in LOX because of the required particle or target
ignition impact velocities. LN2 pre-chill, shown in Fig. 3(b), was utilized in
operational procedures to maintain the oxygen fill flow in the liquid state. How-
ever, during the LOX chill down period, when filling the LOX main tank,
GOX flow is present in the lines. Replacing the stainless steel tee, which is
flammable, with a more burn resistant material removes the flammable-target-
characteristic element of particle impact. This simple engineering fix mini-
mized the potential ignition source for the particulate generated from the
NRCV directly upstream, as seen in Fig. 4.
The LOX main PGV illustrated in Fig. 3(a) contained points at which
severe impact could occur, given the flow direction from over the seat to under
the seat, should particulate be accelerated across the valve at high gas veloc-
ities. Even with LN2 outer jackets to chill the LOX main PGV prior to the
opening of this valve, it is probable that some LOX to GOX flashing will occur
across the valve seat, creating potential high velocity gas flow that could accel-
erate particulates. A potential particulate generating source is directly upstream
in the fill line if the LOX fill line NRCV chatters. Gravity could pull any gener-
ated particulate down into the flow line from the LOX tank to the LOX main
PGV. At fill pressures, the probability of chatter is low; however, this hazard is
identified for better comprehension of all risks. In addition, this is the section
that was found to contain numerous green particles in the cleanliness
verification.
The LOX fill PGV contains the same impact points discussed for the LOX
main PGV. The LOX fill PGV is not pre-chilled, representing an even higher
risk of LOX to GOX flashing, which increases the particle impact ignition
mechanism risk. In order to effectively control this ignition risk, two points of
filtration were implemented in the LOX facility, as shown in Fig. 5. The LOX
FIG. 4—LOX main tank fill.

FIG. 5—LOX facility filtration.
is filtered before entering the LOX fill PGV, flowing through the LOX main
NRCV and up into the tank. This filter prevents particulate from entering the
system during the LOX fill process. The LOX is filtered again before entering
the LOX main PGV and traveling into the main propellant flow lines. In this
situation, the LOX is filtered before it enters the system and then again down-
stream of a potential particle generation source. The filters are thickset brass
housings with a sintered bronze 220 lm (0.0087 in.) filter element. Both brass
and bronze are burn resistant materials and represent a good selection for oxy-
gen system compatibility in the PRC facility.
Gaseous Oxygen Igniter System

The PRC high pressure GOX igniter feed system contained a higher degree of
hazard potential than the LOX system because of increased pressures and flow
induced ignition probability. This system was most susceptible to particle
impact and heat of compression ignition sources. The igniter system was sized
to ensure flow velocities below the recommended maximums, which have
caused particle ignition events in oxygen tested components. However, during
pressurization, a high differential pressure between the source (K-bottle gas
cylinder) and the system (atmospheric) can result in high velocities and present
adiabatic compression ignition source potential.
The OCA identified multiple potential hazards associated with the pressur-
ization practices for the GOX ignition system. Standard gas storage cylinder
valves are inherently fast opening components and are incapable of slowly
pressurizing closely coupled downstream components. Typical cylinder valves

contain no mechanical stop other than the nylon seat. Over time, the rotating
stem valve can wear down and coin the nylon seat, which can lead to leaks and
the generation of nylon fibers. Two ignition mechanisms were associated with
the cylinder valves. A rapid pressurization time is significant in the potential
adiabatic compression ignition probability concerning Teflon lined stainless
steel flex hoses [4]. The coined nylon seat can lead to leaks, which increase the
probability of a flow friction ignition mechanism. The nylon seat material has a
high heat of combustion and, if ignited, could burn into downstream system
components. This represents a potential hazard with a severe reaction effect,
because the valve is manually operated. The PRC standard practice of cracking
a cylinder valve to bleed in pressure was not a sufficient method for slow pres-
surization of the GOX igniter system. Because the condition of the cylinder
valve seat is not known at the time of operation, there also exists potential for
nylon fibers resulting from the deterioration of the cylinder valve nylon seat to
exist in the flow path. If the cylinder valve is cracked, GOX flow will seep
through a small flow area where high surface area nylon fibers could be pres-
ent. The heat generated as this GOX flows across the nylon could ignite the ny-
lon fibers. If the seat is ignited, nylon’s high heat of combustion could further
propagate ignition and couple with GOX igniter system flammable
components.
Consider the fill section shown in Fig. 6. In the original fill configuration,
the GOX isolation manual ball valve (MBV), GOX vent MBV, and GOX main
pneumatic ball valve (PBV) are stainless steel bodied valves. Even though
stainless steel is an alloy that is commonly and successfully employed in oxy-
gen service, it is flammable in high pressure, oxygen enriched environments
[1]. Although there are internal few impact points in ball valves if they are fully
open, all characteristic elements of particle impact ignition are present because
of the high velocity gas flow that accompanies opening of the cylinder valve,
the potential particulate entrained in the flow, and a flammable target in the
stainless steel valve bodies for this particulate to impact. In historical GOX ig-
niter system operation, the particle impact ignition mechanism has been con-
trolled via the use of filters to reduce or remove particulate from the system.
The filters are implemented at connection points between the GOX supply and
the system interface.
In past operation, the heat of compression ignition source has been proce-
durally controlled. Before pressurizing the system, the GOX vent MBV was
closed, the GOX ignition isolation MBV and the GOX PBV were opened, and
the GOX cylinder valve was opened in order to pressurize the system. This
sequence provided a sufficient flow line volume to safely absorb the heat gen-
erated from rapid pressurization.
If either the GOX isolation MBV or the GOX main PBV were closed when
the oxygen cylinder valve was opened, rapid pressurization of the small
FIG. 6—GOX igniter fill configuration (original).
volume contained in the stainless steel Teflon lined flex hose could result in an
adiabatic compression ignition of the exposed Teflon near the dead end closed
valve flow path. The maximum theoretical oxygen gas temperature from isen-
tropic compression is 1010 C (1850 F). This represents a final temperature
that is well above the autoignition temperature of Teflon, which is 435 C
(814 F) [1].
In order to remove the potential for both of these ignition mechanisms in
the GOX igniter system, the standard GOX cylinder valve was fitted with a
brass regulator designed for oxygen service. This is depicted in the redesigned
GOX fill configuration shown in Fig. 7. Using a regulator directly connected to
the cylinder valve controls the pressurization time of the system. Slower pres-
surization prevents the adiabatic compression heating ignition mechanism, as
well as the high velocity gas flow associated with particle impact ignition
mechanisms. These regulators are constructed of a more burn resistant thick
walled brass material. It is important to note that this is still a procedural fix for
these potential ignition sources. For safe operation, the stainless steel GOX fill
FIG. 7—GOX igniter fill configuration (redesign).
MBV and GOX main PBV must be opened before system pressurization. Fur-
ther hazard identification was established using signage to remind operators
that the GOX fill MBV and GOX main PBV must be open prior to the GOX
cylinder valve. As depicted in Fig. 7, these measures are an attempt to make
test operators more aware of the hazards when operating the system.
In order to achieve a higher level of confidence, the GOX ignition filter ele-
ment was reduced from the original 90 lm (0.0035 in.) to 40 lm (0.0016 in.),
and the stainless steel sintered filter element was replaced with burn resistant
nickel 200. Inline sintered bronze filters in the commercial gas association 540
connection fitting to the GOX cylinder K-bottle were utilized to prevent parti-
cle impact downstream, especially impact with the stainless steel GOX ignition
filter housing. This situation ensures that the GOX flow entering the system
will be filtered in two locations—once directly at the GOX cylinder valve
before the flow enters the regulator, and again before the entrance of the GOX
fill MBV.
The original GOX vent MBV shown in Fig. 6 was a stainless steel compo-
nent and historically had been used as a high velocity component opened under
full pressure differential to exhaust the small volume of GOX from the igniter
system. This venting procedure is a manual process that increases the reaction
effect considerably. The GOX fill MBV and GOX main PBV were chosen so
as to facilitate higher flow capacities in the system. This functional capability
was not required in this venting situation. The original GOX vent MBV was
replaced with a multi-turn needle valve (MTNV) constructed of burn resistant
brass. This valve contains a metal to metal seat, removing the exposed soft
goods in the flow path essential for heat of compression ignition, and the brass
body removes the flammable target requisite for particle impact ignition. A
flow path to exhaust the oxygen gas away from the operator was built into the
GOX vent configuration. The stainless steel tubing that connects the GOX vent
MTNV seen in Fig. 7 was given smooth radii to enable straighter flow path
transitions from the stainless steel flow lines into the 1.27 cm (0.5 in.) newly
implemented GOX ignition copper vent lines, maximizing the use of more
burn resistant material.
Summary
The successful demonstration of the LOX propellant feed facility is illustrated in
Fig. 8 [5]. This testing was possible because of the safe system development.
The LOX and GOX propellant system OCA documentation procedure supplied a
FIG. 8—LOX gaseous CH4 rocket engine operation (2 s).

learning environment enabling more thorough realization of the ignition mecha-

nisms prevalent in fires in oxygen systems. Despite the best efforts in the design
and fabrication stages of the facility’s development, the review process identified
several areas in which the system could be improved so as to achieve greater
fault tolerance. The identified risks significant to these findings involved rela-
tively simple changes that would decrease the probability of ignition but not sys-
tem functionality. As an extra measure of protection, an independent oxygen
safety engineering consultant reviewed the findings of this work and performed a
system-level analysis. The safety review provided an opportunity to engage pro-
gram management, designers, technicians, graduate and undergraduate students,
and operators in creating a safe system that has a low probability of ignition and
reduced reliance on procedures to ensure safe operation. The PRC system spe-
cific design changes identified represent some, but not all, of the changes that
increase the overall system safety. No way exists to remove all operational risks
inherent in oxygen systems. Training and education are the most notable defense
against these hazards. The corrective actions outlined in this paper are specific to
the existing PRC facility, and it must be stressed that although the described ini-
tiatives are potential fixes in the illustrated situation, all oxygen systems are
unique. Every oxygen system must be evaluated in order to establish safe opera-
tions for each individual system, and the actions taken in the PRC facility might
not be applicable at other facilities. The practice specifically undertaken in this
process was to reduce the chance of operator induced errors relevant to off nomi-
nal operation. This results in increased fault tolerance of the systems and reduced
severity of reaction effects.
Acknowledgments
The writers would like to thank Huu Trinh and David Stephenson of NASA
Marshall Space Flight Center, as well as the entirety of the PRC faculty, staff,
and students; great appreciation is expressed for their supportive efforts and
patience. Mike Shoffstall of NASA White Sands Test Facility, who instructed
the PRC personnel in the more inclusive ASTM training, deserves great thanks
for the assistance provided during the OCA procedural documentation. Elliot
Forsyth of Wendell Hull and Associates is thanked for assistance. In addition,
the writers would like to express appreciation to Dr. Robert Frederick, Dr.
Hugh Coleman, Dr. Marlow Moser, Dr. Madhan Bala, Dr. John Horrack, and
Mr. Tony Hall for their support in the system review process.
References
[1] Beeson, H. D., Smith, S. R., and Stewart, W. F., Safe Use of Oxygen and
Oxygen Systems, 2nd ed., ASTM International, West Conshohocken, PA,
2007, Chaps. 1–8.
[2] ASTM G88-05, 2005, “Standard Guide for Designing Oxygen Systems,”
Annual Book of ASTM Standards, Vol. 14.04, ASTM International, West
Conshohocken, PA, pp. 1–27.
[3] Circle Seal Controls CMV/CES 12 & 60 Series, “Pneumatically Operated
Shutoff Valve Specification Sheet,” www.circle-seal.com.
[4] Janoff, D., Bamford, L. J., Newton, B. E., and Bryan, C. J., “Ignition of
PTFE-Lined Flexible Hoses by Rapid Pressurization with Oxygen,”
Svmposium on Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres, ASTM STP 1040, Vol. 4, ASTM International,
West Conshohocken, PA, 1989, pp 288–308.
[5] Mulkey, H. W., 2010, “Development of a Liquid Oxygen Facility for
Rocket Engine Injector Performance Testing,” M.S.E. thesis, University
of Alabama, Huntsville, AL.
STP 1561, 2012
DOI:10.1520/STP20120015
Christina Y. Piña Arpin1 and Joel Stoltzfus2
Rudimentary Cleaning Compared to Level

300A
REFERENCE: Piña Arpin, Christina Y. and Stoltzfus, Joel, “Rudimentary

Cleaning Compared to Level 300A,” Flammability and Sensitivity of Materials
in Oxygen-Enriched Atmospheres on September 19–21, 2012 in Montreal,
QC; STP 1561, Samuel Edgar Davis and Theodore A. Steinberg, Editors, pp.
47–61, doi:10.1520/STP20120015, ASTM International, West Conshohocken,
PA 2012.
ABSTRACT: A study was performed to characterize the cleanliness level
achievable when using a rudimentary cleaning process, and results were com-
pared to JPR 5322.1G Level 300A. While it is not ideal to clean in a shop envi-
ronment, some situations (e.g., field combat operations) require oxygen
system hardware to be maintained and cleaned to prevent a fire hazard, even
though it cannot be sent back to a precision cleaning facility. This study meas-
ured the effectiveness of basic shop cleaning. Initially, three items representing
parts of an oxygen system with maximum operating pressure of 2000 psi were
contaminated: a metal plate, valve body, and metal oxygen bottle. The contam-
inants chosen were representative of materials that could contaminate the
system during normal use: oil, lubricant, metal shavings/powder, sand, finger-
prints, tape, lip balm, and hand lotion. The cleaning process used hot water,
soap, various brushes, gaseous nitrogen, a water nozzle, plastic trays, scour-
ing pads, and a controlled shop environment. Test subjects were classified into
three groups: technical professionals having an appreciation for oxygen haz-
ards; professional precision cleaners; and a group with no previous professio-
nal knowledge of oxygen or precision cleaning. Three test subjects were in
each group, and each was provided with standard cleaning equipment, a
cleaning procedure, and one of each of the three test items to clean. The
results indicated that the achievable cleanliness level was independent of the
technical knowledge or proficiency of the personnel cleaning the items.
Manuscript received February 14, 2012; accepted for publication June 22, 2012; published
online November 2012.
1
Component Services Project Engineer, Technical Services Office, NASA Johnson Space Center
White Sands Test Facility, Las Cruces, New Mexico 88012.
2
Oxygen Group Project Manager, Materials and Components Laboratories Office, NASA
Johnson Space Center White Sands Test Facility, Las Cruces, New Mexico 88012.
Copyright V
19428-2959.
47
Results also showed that achieving a Level 300 particle count was more
difficult than achieving a Level A nonvolatile residue amount.
KEYWORDS: precision cleaning, rudimentary cleaning, Level 300, oxygen
systems, hardware cleaning procedures, process, non-cleanroom, field shop
environment
Introduction
Contamination in oxygen enriched environments can act as both a fuel and/or an
ignition source. Contamination control is necessary to mitigate risk associated
with oxygen systems. One form of contamination control is precision cleaning,
in which a value for the level of cleanliness is assigned to specific hardware. In
aerospace use, a risk assessment is typically performed to determine the level of
cleanliness required to operate a system or hardware safely. Furthermore, unlike
many other consumers/users of oxygen equipment, aerospace operations are typ-
ically able to clean and assemble hardware in a certified clean room. The need
arose in which equipment used in combat would need to be maintained and
cleaned but could not be sent to a precision cleaning facility. While developing
cleaning procedures that could be used in the field, two important questions
came up: What effect does the personnel cleaning the equipment have on the
hardware’s end cleanliness level, and what level of cleanliness could be realisti-
cally achieved? This study compares three groups of people using rudimentary
cleaning techniques to find out what level of cleanliness can be reasonably
achieved in a controlled shop environment (an access controlled room with high
air flow and sticky mats at each entrance). This study does not compare how
different environments will affect the outcome.
Precision cleaning, as typically performed by NASA White Sands Test
Facility (WSTF) in compliance with both ASTM G93 [1] and NASA JPR5322.1G
[2], is a three part process. The first part of the process is pre-cleaning, followed
by visual inspection, and ending with cleanliness verification. Pre-cleaning is a
detailed cleaning performed in a controlled shop environment and utilizes a series
of ultrasonic baths which creates implosions that agitate contaminants while the
hardware is immersed. Each bath contains either a surfactant, mild alkaline, or
caustic acid solution. The type of solution used depends on its compatibility with
the hardware. Between baths, a technician manually agitates the contaminants on
the hardware. Once the hardware is cleaned, rinsed, and dried using a nitrogen
purged oven, it is passed on to another individual for visual inspection. The hard-
ware in the visual inspection area is kept under a downflow unit to maintain its
cleanliness. A third party, trained to detect any hardware anomalies, verifies the
cleanliness of the hardware using an otoscope and a magnifying fluorescent
inspection lamp with a 5-diopter lens. To pass visual inspection, no particles or
fibers can be visible, no signs of previous contamination can be detected, and no
signs of rust, discoloration, or water spots can be seen.
PIÑA ARPIN AND STOLTZFUS, doi:10.1520/STP20120015 49
Once the hardware passes visual inspection, it is sent to the ISO (Interna-
tional Organization for Standardization) class 5 clean room for final cleaning
and cleanliness verification. This area requires full clean room garments, con-
sisting of coveralls, hood head cover, boot covers, and gloves, to be worn over
street clothes at all times. Final cleaning is performed by rinsing the hardware
with a solvent. To verify cleanliness of the hardware, a measured amount of
the rinse solvent is captured for analysis in accordance with ASTM G93-03
Type II Quantitative Tests (solvent extraction tests 1 and 2). The rinse solvent
is then passed through a membrane filter, which is used to perform the particle
count in accordance with ASTM F312-08 [3]. The solvent is then evaporated,
and the remaining nonvolatile residue (NVR), which commonly consists of re-
sidual oils or greases, is gravimetrically measured in accordance with ASTM
F331-05 [4]. Precision cleanliness levels are a quantitative measurement,
where the count and size of particles are indicated by a number (the smaller
the number the cleaner the part) and the amount of NVR per square meter of
hardware is indicated by a letter (e.g., “A” is cleaner than “B”). The chart in
Table 1 describes the cleanliness levels as specified in NASA JPR 5322.1G.
Methodology
Test Articles
Three identical sets of hardware were assembled, each containing three items
chosen to represent parts of a typical oxygen system: an aluminum plate
6 6 0.125 in. (15 15 0.3 cm), a cast stainless steel 0.25-in. (0.6-cm)
ball valve body, and a small oxygen storage bottle (Fig. 1).
TABLE 1—Cleanliness levels (per NASA JPR 5322.1G).
What is Comparable
Measured Units to: Level Requirements to Pass:
300 Range (lm) <100 100250 250300 >300
Max allow count unlimited 93 3 0
Number of
Particles &
particles & Human hair 200 Range (lm) <50 50100 101200 >200
micrometers
size of ¼ 50 to 120 l
(lm) Max allow count unlimited 154 16 0
particles
100 Range (lm) <25 2550 51100 >100
Max allow count unlimited 68 11 0
A <1 mg=m2
Nonvolatile B <2 mg=m2
milligrams Feather
residue C <3 mg=m2
(mg) 1 mg
(oils, grease)
D <4 mg=m2
FIG. 1—Test articles representing parts of the oxygen system.
Contaminants
The contaminants chosen were those representative of materials that could con-
taminate the system during use or be found in a field shop: oil, lubricant, metal
shavings/powder, sand, fingerprints, tape, lip balm, and hand lotion (Fig. 2).
Test Groups
This study grouped people into one trial group and three test groups of three
people each, assembled on a basis of their knowledge and experience regarding
oxygen and contamination control (Table 2). The trial group was comprised of
three engineers from the WSTF Oxygen Team. For discussion throughout this
FIG. 2—Contaminants were chosen for likelihood to be used in the field.

TABLE 2—Test groups.
Group Member
Group name description and ID # Group Description
Engineer 1
Trial Group Followed a procedure #1 and performed
Engineer 2
cleaning in a non-controlled environment.
Engineer 3
Cleaning technician 1
Test Group 1 Cleaning technician 2
Cleaning technician 3
Co-op student 1
Test Group 2 Co-op student 2 Followed a procedure #2 and performed cleaning in a
controlled shop environment.
Co-op student 3
Engineer 1
Test Group 3 Engineer2
Engineer 4
paper, the trial group is not considered one of the test groups, due to changes in
the written procedure and changes in test location.
The three test groups represented three sets of knowledge that were of im-
portance in this study. Group 1 was comprised of three clean room technicians
with significant experience performing precision cleaning on hardware intended
for various applications including oxygen service. Group 2 was comprised of
three co-op students who had recently graduated high school; two of them
worked as assistants in the WSTF Administrative Dept. Group 2 had no prior
knowledge of the importance of cleaning hardware for use in oxygen systems.
Group 3 was comprised of three engineers from the WSTF Oxygen Team with
detailed technical knowledge of the importance of cleaning hardware for oxy-
gen service. Two of the three engineers in this group performed the cleaning
process twice, because they had also participated in the Trial Group.
Prior to their cleaning effort, each group was briefed in the procedure and
the importance of cleaning hardware for oxygen service. The briefing was simi-
lar to that provided to field personnel who would be handling the equipment in
question. Each group received a briefing on the cleaning procedure they would
be performing.
Equipment
One important requirement that determined what equipment/tools would be
used in the rudimentary cleaning process was the common availability of such
equipment to an oxygen field shop. Another consideration was that use of the
tools should require minimum training. Groups 1–3 were supplied with stand-
ard cleaning equipment consisting of the following: a set of written procedures;
FIG. 3—Cleaning supplies and tools.
various sizes [diameters of 0.75, 0.5, and 0.375 in. (1.9, 1.3, and 1.0 cm)] of
stiff black nylon-bristle stainless-steel handle cleaning brushes; a 3-row nylon-
bristle DelrinV3 handle cleaning brush; scouring pads; NitriliteV4 gloves;
R R
KimwipesV5 (lint free wipes); an otoscope; a fluorescent inspection lamp with

R
a 5-diopter lens; a spray nozzle for 125 F (52 C) hot deionized (DI) water; fil-
tered gaseous nitrogen (GN2) at 65 psig (0.45 mPa); and three plastic basins
filled with Simple GreenV6 (aqueous degreaser), water solution, and rinse
R
water. Figure 3 represents some of the equipment supplied to the groups. The
trial group had a similar set of supplies with a few substitutions that included
tap water in place of DI water, canned air instead of GN2, and a magnifying
glass and a flashlight in place of the visual examination tools. HFE 7100V7 was
R
3 R
DelrinV is a registered trademark of E. I. du Pont de Nemours and company, Wilmington, DE.
4 R
NitriliteV is a registered trademark of Ansell Limited, Richmond, Australia.
5 R
KimwipesV is a registered trademark of Kimberly-Clark Corporation, Neenah, WI.
6 R
Simple GreenV is a registered trademark of Sunshine Makers, Incorporated, Huntington Beach,
CA.
7 R
HFE 7100V is a registered trademark of 3M Company, St. Paul, MN.
used in all cases as the verification solvent to determine the level of cleanliness
achieved on the hardware using the rudimentary process.
Environment
The Trial Group used a preliminary version of the procedure for which they set
up their own cleaning process in a room equipped only with a stainless steel
sink, domestic hot water, and a table-top surface on which to work. For the
actual testing, Groups 13 were set up on a stainless steel work top in the
WSTF pre-clean room, which is a controlled shop environment with a higher
than normal exchange of air and sticky mats at each entrance; only a smock is
worn over street clothes in this area.
Test Article Contamination

Before the study began, each type of test article was uniformly contaminated
as described in the Experimental Procedure section of this paper.
Recommended Rudimentary Procedures

Identical recommended cleaning procedures, presented as Table 3, were
supplied to the three test groups. (Note that any mention of inspection is not an
official visual inspection by a third party.)
Experimental Procedures
Once the test articles and contaminants were selected, the test articles were vis-
ually examined to ensure they were at a similar level of cleanliness and to note
the initial condition of the test articles. (This visual examination should not be
confused as an official visual inspection as identified earlier in the precision
cleaning area of this paper.) The test articles were then contaminated. Each pi-
ece from the three hardware sets was contaminated using the process outlined
in Table 4, so that each test group started with hardware that was similarly
soiled when they began the cleaning process.
The three individuals in each test group were provided with a set of identical
cleaning supplies and a set of contaminated test articles. Each individual cleaned
their three items following the recommended cleaning process and using only
the supplied set of tools. Once the individual declared their test articles to be
cleaned, the test articles were sampled using HFE 7100 by a WSTF clean room
technician (not in a test group) to measure the amount of particulate and NVR.
This process was repeated for each individual in the three test groups.
Sample Collection
To determine the cleanliness level of the hardware, a verification process simi-
lar to what is performed in the precision cleaning process was performed in the
TABLE 3—Recommended rudimentary procedures.
Equipment
Water— Hot DI water, with a high pressure wand or nozzle. If DI water
is not available, use the cleanest available water supply.
Note: In all cases, warmer and cleaner water is better. If gloves
can be worn to enable the water temperature to be higher, that
is a favorable practice. Warm or hot water cuts grease and dis-
solves oil deposits far better than cool or cold water.
Pans— Metal pans to contain cleaning solutions and rinse water. If
metal pans are not available, plastic pans may be used.
Brushes— Clean brushes in a variety of sizes and shapes to fit small crevi-
ces and reach all ends (e.g., spiral brush, end brush, bottle
brushes, toothbrush-style brush), nonmetallic scouring pads.
Warning: Ensure that the brushes will not damage the hard-
ware to be cleaned.
Warning: Do not use metallic scouring pads because they shed
particles that can either rust or become imbedded in softer
materials, or both.
Note: Brush or tool extensions can be fabricated, if necessary,
to enable physical scrubbing of parts that are otherwise
inaccessible.
Soap— Soap with no moisturizers or scents added (e.g., Simple
Green). If unavailable, use dish soap.
Gloves— Nylon or Nitrilite gloves that tend not to shed particulate. If
nonshedding gloves are not available, use rubber gloves.
Rags— Lint free cloth or paper wipes (e.g., Kimwipes), cotton cloth.
Inspection Tools— Magnifying glass, flashlight, otoscope (can be used for looking
inside vessels and components), fluorescent lamp with magni-
fying glass, bright light source.
Work Surface and Washing Station— Work surface and washing station that are located near one
another.
Metal surface that can be wiped clean or a visually clean plas-
tic or paper sheet to cover work surface.
Note: The work surface will be used to locate items that must
be kept clean during the process such as clean brushes, wipes,
bags, and gloves, as well as provide a location that clean items
can be located while preparing for bagging after the cleaning
process is complete. Finally, the work surface will be used for
disassembly and re-assembly of components.
General Procedure
Prepare Workplace— Choose a work area that is somewhat controlled to avoid open
access and to minimize contamination. Get all supplies and
tools together before beginning the cleaning process to avoid
delays once the cleaning process is underway.
TABLE 3—Continued
General Procedure
Wipe down the metal work surface or place a clean plastic
sheet or clean paper on the work surface.
Ensure that the work surface remains visually clean during
the disassembly and cleaning processes.
If necessary, wipe down the work surface to ensure that con-
tamination from one item does not get transferred to other
items.
Set up a series of wash=rinse pans.
Wash pans should progress from less to more clean.
For example, the first pan might contain hot, soapy water for
pre-cleaning and soaking dirtier items. The second pan might
contain hot, soapy water for scrubbing cleaner items. The
third pan might be for initial rinsing, and the fourth pan
might contain clean, hot water for final rinsing.
Change out the water in each pan as required when it
becomes dirty.
Disassemble— Completely disassemble the component so that all crevices and
internal surfaces are as accessible as possible.
Note: Each part must be cleaned separately, in a disassembled
configuration. If they are not, the effectiveness of the cleaning
procedure will be significantly reduced. Parts that are not
removed provide crevices and cracks that will retain contami-
nants and it will not be possible to clean the adjoining surfaces
of the parts.
Pre-clean— Remove excessive contamination.
Wipe the component.
Discard the soiled wipes.
Continue this process until as much visible contamination is
removed as possible.
Wash— Soak item in hot (the hotter, the better) soapy water as needed
to loosen contaminant from surfaces.
If item is a tank or a bottle, fill it with soapy water until it is
overflowing, then let it soak.
Scrub=agitate the part thoroughly.
Use brushes, wetted Kimwipes, or cotton cloth.
Pay special attention to threads, crevices, and hidden surfaces
that, because of their configuration, can trap contaminants.
If internal surfaces cannot be scrubbed mechanically, then
spray the inaccessible areas. If scrubbing or spraying are not
possible, fill the item with hot, soapy water, plug the holes,
and shake it vigorously.
Note: Ensure that many suds are formed during the washing
process; if there are few suds, then add more water and soap. If
suds don’t form, it indicates that the all the oil contamination
is not captured by the surfactant.
TABLE 3—Continued
Change the wash water, add soap, and repeat the washing
process twice for a total of three washes.
Note: This procedure amounts to very rigorous dishwashing
with very careful attention given to detail and precision. The
hotter and soapier the water the better. The repetitions ensure
that the surface is thoroughly cleaned and that layered contam-
ination is wetted, agitated, and removed.
Rinse— Rinse item thoroughly to remove soap and contaminant.
Dunk and spray item with clean hot water as needed to
remove all visible residues.
If item is a bottle, then fill, shake, and empty the water until
no suds are left and the water coming out is as clear as water
going in.
Note: Gloves should be rinsed often to remove contaminants
and particles obtained by touching contaminated parts. If a par-
ticle can be transferred to a cleaned part by physical contact, it
is very likely that it can be rinsed off. Rinse it off prior to
touching the parts.
Inspect rinse water.
Catch the last rinse water in clear container.
Look for particles or grease in the water using a magnify-
ing glass and by shining a bright white light through the
water
If no particles are present or grease is observed on the surface
(as a colorful sheen) and the water is as clear as it was from
the source, move on to drying; if not, repeat the wash and
rinse steps.
Note: When washing carbon steel components like the oxygen
bottles, perform the final rinse with cold water and then dry im-
mediately. The combined effects of cooling the carbon steel
surface and drying it quickly will inhibit as much as possible
the formation of rust.
Dry— Blow dry

Use oil-free, filtered GN2 or air if available.
If filtered gasses are not available, use filtered compressed or
canned air.
Note: It is preferable to dry parts quickly, removing all water
droplets, to avoid deposits of water-borne contaminants (water
spots).
Air dry.
Set aside in a clean, isolated area to air dry.
Wipe dry.
If water droplets remain, they can be carefully removed using
clean, folded Kimwipes.
TABLE 3—Continued
Inspect— Inspect item to verify cleanliness.
Visually inspect, using magnification if available, all surfaces
using the most intense light available.
Take care to inspect small crevasses, threads, and holes.
If any discoloration, water deposits, soap residue, or particu-
late are observed, remove it. If necessary, repeat the wash
and rinse steps.
WSTF pre-clean room using a 40- to 60-psig (0.3- to 0.4-mPa) rated pressure
vessel with an attached spray wand to rinse the components. The pre-clean
room location was chosen due to concerns about contaminating the WSTF
ISO class 5 clean room by taking in hardware that had only gone through a
rudimentary cleaning process. A clean room technician rinsed each test arti-
cle with 100 mL of HFE 7100, which was collected and filtered through
a 0.45-lm membrane filter. The membrane and the remaining HFE 7100 sam-
ple were sent to the WSTF chemistry lab for particle count and NVR
analysis.
Analysis of Data
Each component cleaned was evaluated on a pass/fail basis for two parameters,
particulate cleanliness and NVR cleanliness, against the target 300A criteria as
defined by NASA JPR 5322.1G and described in Table 1. Both of these cleanli-
ness parameters are adequate to reduce the fire hazards in the subject oxygen
system. For example, if the cleanliness level of hardware was measured to be a
200A, it passed the two cleanliness parameters for that test article. If, however,
the verification resulted in a 300B or >300A, both cases would have passed
only one of the two parameters: the level 300B would have passed the particle
count but failed the NVR; and the level >300A would have failed the particle
count but passed the NVR.
Due to the complexity of measuring and comparing cleanliness of hard-
ware, this study mainly focused on the quantitative analysis of cleanliness. The
visual inspection pass/fail parameters normal for precision cleaning were not
considered in this study. The performance of a test group was ranked by the
rate of passing results. For example, if Group 1 had three people who each
cleaned three parts (nine parts total), and the parts cleaned by Group 1 achieved
Level 300 three times, the success rate was 3 out of 9, or a 33 % passing rate.
Likewise, if Group 1 achieved NVR Level A eight times, the success rate was
8 out of 9, or an 89 % passing rate.
TABLE 4—Contamination process.
Test Article Steps taken for Contamination of Test Articles

Initial condition: Shear-break edges, no special protection, vis-
ually similar
1 Fingerprints: Flip & grip 10 times per each of 5 people
2 Lip balm: Smear on ( one long swipe on both sides near edge A)
R
3 KrytoxV lubricant: Dab & smear ( one long swipe on both sides
Plates: near edge B)
4 Lotion: Small dab & smear (one long swipe on both sides near
edge C)
5 Oil: 3 drops & smear (one long swipe on both sides near edge D)
6 Metal powder: Dust (like salt) entire surface
7 Sand: Dust ( like salt) entire surface
8 Bag (unsealed)
Initial condition: As received from customer, uncapped, no

gross contamination, oil residue on threads. After first cleaning
some rust was seen in bottle
Oxygen 1
R
KrytoxV lubricant: Small dab on threads, insert=remove fitting
Bottle 2 Oil: 5 drops inside (from last thread to drip into bottle)
3 Metal powder: Drop 1 spatula of powder inside & shake
4 Sand: Drop 1 spatula of sand inside and shake
5 Bag (unsealed)
Initial condition: Cleaned 200A, opened, disassembled

1 Fingerprints: Flip & grip 10 times per each of 5 people
R
2 KrytoxV lubricant: Lube initial parts, reinsert and remove
3 Oil: 3 drops and shake
Valve 4 Metal powder: Drop 2 mL of powder inside and shake
5 Sand: Drop 2 mL of sand inside and shake
R R
6 TeflonV tape: Wrap threads with TeflonV tape, insert and
remove fitting
7 Bag (unsealed)
Results and Discussion

Table 5 shows the level of cleanliness reached by each individual in each of
the groups. Each of the groups achieved similar success rates (see Table 6).
This suggests that the results obtained are not dependent upon the age, experi-
ence, technical training, or prior knowledge of oxygen hazards. The results
indicate that the instructions are adequate to produce a very consistent result,
and special training does not produce improved cleaning results.
It is also apparent that the NVR levels were met much more frequently
than the particulate levels. The percentage of parameters met for the plate,
valve body, and bottle NVR levels were 100 %, 100 %, and 77:7%
TABLE 5—Test results showing particulate and NVR levels achieved.
Plate Valve Body Oxygen Bottle
Group Particle NVR Particle NVR Particle NVR

Trial Group >300 B >300 A 200 >D
200 A 100 A >300 A
300 A 200 A >300 A
Test Group 1 >300 A 200 A 300 A

>300 A >300 A >300 >D
>300 A >300 A 300 A
Test Group 2 >300 A >300 A 300 A

>300 A 300 A >300 A
>300 A >300 A 300 A
Test Group 3 >300 A 300 A >300 A

200 A >300 A >300 A
300 A >300 A >300 B
respectively, whereas the percentage of parameters met for the particle levels
were 22: 2%, 33:3%, and 44:4%, respectively. Only 33 % of the particulate
samples passed Level 300, compared to nearly 93 % of the NVR samples. This
implies that it is much easier to remove the NVR (which includes substances
like oil, lip balm, KrytoxV8 lubricant, and fingerprints) than it is to remove and
R
TABLE 6—Percentage of parameters passed to achieve Level 300A cleanliness.
Plate Valve Body Oxygen Bottle
Group Particle NVR Particle NVR Particle NVR Group Ratio

Trial Group (Engineers) 66:6% 66:6% 66:6% 100% 33:3% 66:6% 66:6%
Test Group 1 0% 100% 33:3% 100% 66:6% 66:6% 61:1%
(Cleaning technicians)
Test Group 2 0% 100% 33:3% 100% 66:6% 100% 66:6%
(Co-op Students)
Test Group 3 66:6% 100% 33:3% 100% 0% 66:6% 61:1%
(Engineers)
All Groups: 33:3% 91:6 % 41:6 % 100% 41:6 % 75% —
Test Groups only: 22:2% 100% 33:3% 100% 44:4% 77:7% —
8 R
KrytoxV is a registered trademark of E. I. du Pont de Nemours and Company, Wilmington, DE.
control the particulate contaminant (which includes items like particles and
fibers).
The tendency to meet the NVR requirement but not meet the particulate
requirement was consistent between groups, suggesting that the efficacy of the
procedure is relatively independent of increased prior knowledge of oxygen
hazards, cleaning experience, and advanced education of the person doing the
cleaning.
Due to their availability, the brushes used for cleaning were clean room
brushes. These brushes are specially designed to shed few particles and bristles,
and it is not known what the effect of other (non-clean room) brushes would be
on the recommended rudimentary field procedure.
The oxygen bottle was the most difficult item from which to remove the
NVR. Compared to the plate and the valve body, for which 91:6% and 100 %
of the parameters passed the Level A specification, the vessel passed less fre-
quently (75 %). This leads to the conclusion that special training emphasis
should be placed on how to clean the inside of the bottle. The vessel also
showed signs of rust after the trial group cleaned it, and the other groups as
well. This was not a surprise, since water was used to clean the vessel and
water would cause flash rusting. However, it does lead to concerns as to
whether or not the vessel can be considered to have passed the 300A level
cleanliness parameters, because even if it met the analytical parameters, the
vessel would not pass a visual inspection—which is a precursor to any analyti-
cal verification of cleanliness. If this vessel had gone through the standard pre-
cision cleaning process, the rust in the vessel would have needed to be
mitigated before any verification sample was collected. It is worth noting that
the solvent used to verify cleanliness was HFE 7100, which is not a very
aggressive solvent and not normally used to validate a cleaning processes. HFE
7100 is typically used as part of a cleaning process to verify the cleanliness of
the parts once the process as a whole has been validated. A more aggressive
solvent such as AK225V9 is recommended to validate a cleaning process; how-
R
ever, HFE 7100 was chosen due to its accessibility and the equipment configu-
rations that were available.
The possible lack of control of the particulate contaminant and the pres-
ence of rust increase the possibility of a particle impact ignition mechanism
being present in such an oxygen system. An evaluation of the possibility of par-
ticle impact ignition in the subject oxygen system was recommended to ensure
that a credible fire hazard did not exist. The cleaning process tested in this pa-
per could be used to maintain oxygen systems if it were determined that this
process produced particles too few or too small to create a credible particle
impact ignition source.
9 R
AK225V is a registered trademark of Asahi Glass Company, LTD, Tokyo, Japan.
Conclusion
With proper cleaning procedures that incorporate contamination control princi-
ples, a cleanliness level of 300A is possible to quantitatively achieve in a field
shop environment. However, the consistency of cleanliness parameters met
cannot compare to those achieved in a production precision cleaning facility.
With the joint understanding of oxygen and contamination control, one can
determine if such field cleaning can meet the requirements to maintain a safe
oxygen system.
References
[1] ASTM G93, 2003, “Standard Practice for Cleaning Methods and Cleanli-
ness Levels for Material and Equipment Used in Oxygen-Enriched Envi-
ronments,” Annual Book of ASTM Standards, Vol. 14.04, ASTM
International, West Conshohocken, PA, pp. 1–22.
[2] NASA JPR5322.1G, Contamination Control Requirements Manual,
NASA Johnson Space Center Procedural Requirements, Houston, TX,
2009.
[3] ASTM F312, 2008, “Standard Test Methods for Microscopical Sizing and
Counting Particles from Aerospace Fluids on Membrane Filters,” Annual
Book of ASTM Standards, Vol. 14.04, ASTM International, West Consho-
hocken, PA, pp. 1–4.
[4] ASTM F331, 2005, “Standard Test Method for Nonvolatile Residue of
Solvent Extract from Aerospace Components (Using Flash Evaporator),”
Annual Book of ASTM Standards, Vol. 14.04, ASTM International, West
Conshohocken, PA, pp. 1–3.
STP 1561, 2012
DOI:10.1520/STP20120016
Joel M. Stoltzfus,1 Timothy D. Gallus,2 and Kyle Sparks3
Flow Friction or Spontaneous Ignition?
REFERENCE: Stoltzfus, Joel M., Gallus, Timothy D., and Sparks, Kyle, “Flow
Friction or Spontaneous Ignition?,” Flammability and Sensitivity of Materials
62–80, doi:10.1520/STP20120016, ASTM International, West Conshohocken,
PA 2012.
ABSTRACT: “Flow friction,” a proposed ignition mechanism in oxygen sys-
tems, has proved elusive in attempts at experimental verification. In this pa-
per, the literature regarding flow friction is reviewed and the experimental
verification attempts are briefly discussed. Another ignition mechanism, a
form of spontaneous combustion, is proposed as an explanation for at least
some of the fire events that have been attributed to flow friction in the litera-
ture. In addition, the results of a failure analysis performed at NASA Johnson
Space Center White Sands Test Facility are presented, and the observations
indicate that spontaneous combustion was the most likely cause of the fire in
this 2000 psig (14 MPa) oxygen-enriched system.
KEYWORDS: flow friction, spontaneous combustion, ignition mechanism,
pressurized oxygen-enriched systems
Introduction
The concept of “flow friction” as an ignition mechanism in pressurized
oxygen-enriched systems has been discussed in the literature [1,2]. Several
fires in oxygen systems ranging in pressure up to 9000 psi (62 MPa) have been
reported which have been attributed to the unintentional flow of oxygen past
Manuscript received February 14, 2012; accepted for publication April 27, 2012; published
1
Project Manager, Materials and Components Laboratories Office, NASA Johnson Space Center
White Sands Test Facility, Las Cruces, NM, 88004.
2
Engineer, NASA Test and Evaluation Contract, NASA Johnson Space Center White Sands Test
Facility, Las Cruces, NM.
3
Engineer, NASA Test and Evaluation Contract, NASA Johnson Space Center White Sands Test
Facility, Las Cruces, NM.
Copyright V
19428-2959.
62
STOLTZFUS ET AL., doi:10.1520/STP20120016 63
polymer valve seats or system seals [3–6]. In many of these cases other ignition
mechanisms such as particle impact, heat of pressurization, electrical arc, me-
chanical impact, and mechanical friction were eliminated as possible causes of
the fire events. Because leaks had been observed in the location of the fire
events, it was concluded that the cause of the fires was associated with the oxy-
gen gas leaking past a polymer seat or seal.
Experimental evidence related to “flow friction” has been difficult to come
by. Attempts to simulate leaks involving polymer seats and seals at pressures
up to 10 000 psi (69 MPa), and thereby cause ignition, have failed. This inabil-
ity to reproduce fire events in a controlled laboratory environment has led to
justifiable skepticism regarding the reality of “flow friction” as an ignition
source in real oxygen systems. This paper presents another mysterious fire
event that lacks the more verifiable ignition mechanisms mentioned above and,
yet again, points to a known oxygen-enriched, 2000 psi (14 MPa) gas leak as a
contributor to the cause, but with a twist. In the event described herein, a poten-
tial heat-generating causal factor exists that has been heretofore undiscussed in
this context: spontaneous heating and ignition.
Spontaneous Heating and Ignition in Air

The U.S. Dept. of Energy defines spontaneous combustion as the ignition of a
combustible material caused by the accumulation of heat from oxidation reactions
and indicates that fires caused by spontaneous combustion are due to spontaneous
heating, pyrophoricity, and hypergolic reactions [7]. Spontaneous heating is the
slow oxidation of an element or compound which causes its bulk temperature to
rise without the addition of an external heat source. It may occur because of the
oxidation of hydrocarbons (e.g., oils, coal, and solvents) or because of the action
of microorganisms in organic materials. Pyrophoricity relates to the ignition of a
substance upon exposure to the atmospheric oxygen in air. Although there are
some pyrophoric liquids and gases, most pyrophoric materials are very finely
divided metals. A hypergolic reaction describes a material’s ability to spontane-
ously ignite or explode upon contact with any oxidizing agent.
The spontaneous combustion of coal is a well-known phenomenon, espe-
cially with coal that is extremely friable and breaks down into smaller particles.
Hossfeld and Hatt [8] report that moist, volatile subbituminous coal will not
only smolder and catch fire while in storage piles at power plants and coal ter-
minals, but has been known to be delivered to a power plant with the rail car or
barge partially on fire.
Tsuchiya and Sumi [9] indicate that spontaneous ignition is a complex phe-
nomenon of combustible material ignited by its own heat of reaction without
external heat or other source of ignition. The factors contributing to spontane-
ous heating and ignition are heat generation and heat dissipation. If heat
generates faster than it dissipates, it accumulates (i.e., the temperature

increases). The authors state:
One of the best known examples of spontaneous ignition is that of a drying
oil such as linseed oil absorbed in cotton waste. Linseed oil slowly takes oxy-
gen from the air to form a skin of solid material, a process caused by an oxida-
tion reaction that produces heat. If the oil is absorbed in cotton waste, the heat
cannot escape rapidly and the temperature of the waste increases. This acceler-
ates the rate of oxygen absorption and results in further temperature increase.
If this process continues, the temperature of the oil-soaked waste may gradu-
ally rise until ignition occurs spontaneously [9].
Tsuchiya and Sumi further state that, in general, fibrous and finely divided
materials are more susceptible to spontaneous ignition due to their lower ther-
mal conductivity than solid materials and their increased surface area [9]. Pow-
dered metals are much more easily ignited than solid metals; for example.
When finely divided materials form a large pile, self-heating is a common
problem. The larger the pile, the easier it is for self-heating and ignition to
occur. This is because heat generation is proportional to the volume of the pile
by the third power of the radius; the heat loss, however, is proportional to the
surface area of the pile by the second power of the radius. The critical size,
above which spontaneous ignition can occur and below which spontaneous
ignition does not occur, is called the “critical diameter” or “critical radius.”
Tsuchiay and Sumi also discuss the effect of ambient temperature [9].
High surrounding temperature increases self-heating and restricts heat loss.
The critical temperature, above which spontaneous ignition can occur and
below which it does not occur, is called the “critical surface temperature.” A
theoretical relation between the critical ambient temperature and critical radius
of various materials is shown in Fig. 1. The higher the ambient temperature,
the smaller the critical radius. If the ambient temperature is below the critical
temperature, the material self-heats but does not ignite. When it is slightly
higher, spontaneous ignition can occur after a long period. The time to ignition
is called the “induction period”; the higher the ambient temperature above the
critical temperature, the shorter the induction period.
Spontaneous Ignition in Oxygen-Enriched Atmospheres

As the preceding paragraphs indicate, good data exist in support of spontane-
ous heating and ignition in ambient air conditions which is, in some measure,
related to the critical radius of the accumulation of high-surface-to-volume-
ratio fuel particles. The authors propose here that those processes leading to
fire events in oxygen-enriched, pressurized, flowing (such as would occur in a
leak) environments will occur in dramatically smaller accumulations of fuel
particles, especially if the particles are very small. It is proposed that three
FIG. 1—Surface temperature versus critical radius in spontaneous ignition [9].
factors will reduce the critical radius at which spontaneous heating and ignition
will occur: (1) oxygen purity and pressure, (2) forced convection, and (3) parti-
cle size. If it can be shown that the additive effects of these three factors can
reduce the critical radius of white pine sawdust; for example, from 15 in.
(38 cm) at 177 F (81 C) to 0.063 in. (0.16 cm) at 177 F (81 C), then spontane-
ous heating and ignition would be a possible ignition mechanism in small, leak-
ing, oxygen-enriched components such as those used in laboratory-scale
oxygen systems.
A Possible Real-Life Spontaneous Ignition Fire Event

The NASA White Sands Test Facility (WSTF) Oxygen Group was requested
to perform a failure analysis on an oxygen-enriched system which had under-
gone a fire event. The following is a description of the fire event hardware, the
FIG. 2—Damaged three-way valve shown with valve handle removed.
salient features of the failure analysis findings, and a possible ignition scenario
involving spontaneous heating and ignition.
Description of Fire Event Hardware

The system in which the fire event occurred was a portable pressure-swing-
absorption oxygen concentration system which supplied approximately 95 %
oxygen for medical purposes. The maximum system operating pressure was
2250 psi (15.5 MPa). The fire event started in a three-way valve that directed
the flow of pressurized product gas either to storage tanks or to vent (Fig. 2).
Figure 2 shows the product gas inlet on the left, the burned outlet fitting on
the bottom of the valve which connected to the storage tanks and product out-
put lines, and the burned fitting and vent line on the right.
Figure 3 shows the damaged three-way valve with the valve stem/PTFE
seat assembly removed and located on top of the valve body in the same orien-
tation it had inside the valve when the fire occurred.
FIG. 3—Damaged three-way valve with valve stem/PTFE seat assembly

removed and located on top of the valve body.
FIG. 4—Schematic representation of three-way vent valve. (See Fig. 5 for

cross-section A-A.)
Figure 4 is a schematic representation of the valve indicating the product

gas inlet line from the compressor on the left, the outlet on the bottom (this
flow path supplied product gas to the storage tanks and provided a reverse flow
option to vent the storage tanks), and the vent line on the right. The cross-
sectional view A-A, indicated in Fig. 4, is shown in Fig. 5.
Failure Analysis Findings
Improper Valve Configuration—Upon disassembly of the damaged valve,

it was determined that the valve had been improperly configured. Figures 5 and
6 show the internal parts of the damaged three-way vent valve. Figure 6(a)
illustrates the proper configuration of the seat of the ball valve as indicated in
the vendor’s catalog. The seat was a PTFE capsule which was molded over the
stainless steel ball of the valve and was therefore unable to be removed without
damage. Since the valve could not be fully disassembled, it was not possible to
clean the individual valve parts to a cleanliness level appropriate for use in
oxygen-enriched environments. The factory-installed lubricant between the
stainless steel ball and the PTFE seat was silicone, a material not recommended
for use in oxygen-enriched environments.
At the inlet or OPEN port position, through which the inlet product gas
would flow, a side ring was supposed to be positioned [Fig. 6(a)]. At the
CLOSE position, a side disc was intended to be positioned; and at the VENT
FIG. 5—Section A-A of damaged three-way vent valve showing expanded view
of improperly configured valve and possible leak path.
port position, a second side ring was to be positioned. These rings, fabricated
from metal alloy, were intended to provide support the to the PTFE seat mate-
rial to keep it in the position and configuration necessary to maintain a proper
seal between the seat and the stainless steel ball and between the seat and the
stainless steel valve body.
Post-fire analysis of the partially consumed valve revealed that there was,
in fact, no seal ring at the OPEN position port; and instead of a side disc, there
was a side ring at the CLOSE position, as shown in Fig. 6(b).
FIG. 6—Proper (a) and improper (b) configuration of PTFE valve seat and
support discs.
Known Valve Leakage—Two additional illustrations of the improperly con-

figured valve are shown in Fig. 5. The valve was known to have a leak which
flowed from the OPEN position port to the CLOSE position. In this configura-
tion, the valve is closed, but the gas leaking from the inlet side of the valve,
pressurized to approximately 2200 psi (15 MPa) maximum, would flow
between the unsupported PTFE seat and the stainless steel ball.
Cellulose Particulate Contamination—A third important observation from

the failure investigation was that cellulose particles were present in the high-
pressure oxygen-enriched portion of the system. Paper (cellulose) filters were
used upstream of the compressors in the product gas system and in the high-
pressure portion of the system, and fiber/particle remnants of a paper (cellu-
lose) towel were found (Fig. 7). The fiber and particle diameters ranged from
30 to 60 lm.
It is postulated that cellulose particulate and fibers migrated from these
sources in the high-pressure system to the damaged three-way vent valve and
became trapped in the lubricated space between the ball and the seat, as indi-
cated in Fig. 5.
Possible Ignition Scenario

With most unintended oxygen-enriched fires it is extremely difficult, if not
impossible, to determine the actual root cause of the fire. The main reason is
that most, if not all, of the evidence at the point of ignition is typically con-
sumed in the fire, leaving very little physical evidence from which to investi-
gate or reconstruct the scenario before the fire started. As a result, it is
necessary to identify the known possible ignition mechanisms in oxygen-
enriched environments and eliminate as many as possible based upon the
system operational parameters and procedures, witness accounts, and physical
evidence. Those remaining sources of ignition, or some unknown source of
heat, must then be investigated as the probable source of ignition for the fire
FIG. 7—View of cellulose fibers and particles observed in high-pressure por-

tion of the system. The fiber and particle diameters range from 30 to 60 lm.
event. ASTM Manual 36 [1] lists several ignition mechanisms known to have
caused fires in oxygen-enriched environments. That list is shown in Table 1.
Ignition Mechanisms Ruled Out—Using the evidence provided to the

WSTF failure investigation team, a number of different hypothetical ignition
sources were considered and ruled out as possible causes of the fire, because
the characteristic elements or necessary conditions for those ignition mecha-
nisms to be active were not present in this scenario.
Particle Impact—If the three-way valve had been opened while pressur-
ized, sonic velocity oxygen-enriched product gas would have impinged the op-
posite side of the stainless-steel ball at up to a 90 deg impact angle at a high
velocity. However, this valve was not opened during or just prior to the fire
event, and as such, product gas was not flowing through the valve at the time
TABLE 1—List of possible ignition mechanisms in oxygen-enriched environments [1].
Ignition Mechanisms Description

Particle impact Heat generated when particles strike a material with sufficient
velocity to ignite the particles and or the material.
Heat of compression Heat generated when a gas is rapidly compressed from a low
pressure to a high pressure.
Flow friction Heat generated when oxygen flows across or impinges upon a
nonmetal (usually a polymer) and produces erosion, friction,
and/or vibration.
Mechanical impact Heat generated as a result of single or repeated impacts on a
material.
Mechanical friction Heat generated as a result of friction and galling at the rubbing
interface as two or more parts are rubbed together.
Fresh metal exposure Heat generated by oxidation when metal is exposed to oxygen.
Static discharge Heat generated as a result of static discharge that is sufficient
to ignite proximate materials.
Electrical arc Heat generated by an electrical arc that is sufficient to ignite
proximate materials.
Chemical reaction Heat generated between a combination of chemicals sufficient
to ignite surrounding materials.
Thermal runaway Some materials, notably certain accumulations of fine par-
ticles, porous materials, or liquids, may undergo reactions that
generate heat.
Resonance Heat generated by acoustic oscillations within resonant
cavities.
External Heat Heat generated by any external heat sources such as lightning,
explosive charges, personnel smoking, open flames, shock
waves from tank rupture, fragments from bursting vessels,
welding, and exhaust from internal combustion engines.
of the fire event. Therefore, particle impact was not considered a possible cause
of this event.
Heat of Compression—Based upon witness statements and knowledge of

the system configuration prior to the fire event, it was determined that the max-
imum temperature that could have possibly occurred at the valve inlet due to
rapid pressurization was 81 C (177 F). The autoignition temperature of the sil-
icone grease in the ball/seat assembly is 216 C (421 F). Clearly, the fire event
was not caused by the heat from rapid pressurization alone.
Flow Friction—This ignition mechanism will be discussed in conjunction

with the possible spontaneous heating and ignition of the cellulose particles
entrapped in the valve ball/seat assembly.
Mechanical Impact—Was not present for obvious reasons. Mechanical

friction was eliminated because no motion of the three-way valve had occurred
immediately prior to the fire event. Fresh metal exposure was deemed not pos-
sible because there was no evidence of such exposure in the ball or valve body
at the location of the initiation of the fire at the upstream side of the burned
PTFE valve seat. Static discharge was considered not possible because the
metal body of the valve was grounded to the inlet and outlet fittings and lines,
and the PTFE seat was surrounded by metal parts that were grounded to one
another. Electrical arc was ruled out because the three-way vent valve was not
powered electrically. Chemical reaction was eliminated from consideration
because no other chemical process was ongoing in the oxygen-enriched prod-
uct flow stream. Thermal runaway was considered not possible because there
were no external heaters or heat-generating processes in the vicinity of the
three-way valve. Resonance heating was eliminated because the burned three-
way valve had no configurations known to generate acoustic oscillations.
Finally, there were no sources of external heat, eliminating that as a source of
ignition.
Most Probable Ignition Scenario

After careful and extensive examination of the evidence (which included the
burned hardware, photographs, witness statements, and lab analyses), the fol-
lowing was selected by the team as the most likely scenario to have caused the
fire in the three-way valve.
• Cellulose particles were deposited at the entrance point to the leak path
in the valve as indicated in Fig. 5.

• The temperature of the particle deposit, the PTFE seal, or the silicone
grease was increased to as high as 81 C (177 F) by the heat of rapid

pressurization, as indicated previously.
FIG. 8—Illustration of the damaged valve stem/PTFE seat assembly.
• The warm cellulose particle/silicone lubricant deposit began to self-heat

via the oxidation process. Fresh oxidizer was supplied and the products
of oxidation were moved away by the leaking oxygen-enriched product
gas.
Note: At this point in the process, the only macroscopic observation that
one could have made was that the valve was leaking. Since a fire event ulti-
mately occurred, it is reasonable to attribute this fire event to the valve leaking
gas past a polymer seal at the time of the fire (i.e., flow friction was occurring).
However, while that leak was necessary for the ignition scenario described
herein, it was not sufficient to cause the fire event on its own.
• The cellulose particles/silicone lubricant accumulation self-heated to its
autoignition temperature and ignited (spontaneous ignition).

• The burning cellulose particle and silicone lubricant ignited the PTFE
seat packing surrounding the leak path, producing the damage illustrated
in Fig. 8 and pictured in Fig. 9. The burning progressed to the side ring
improperly located at the CLOSE position of V-9.
• The burning cellulose particles, PTFE seat, and silicone ignited the
sharp edge of the stainless steel ball at location A in Fig. 8 and Fig. 9,
sending a jet of burning stainless steel/particles/PTFE/silicone across
the hole in the ball to far side (location B in Fig. 8 and Fig. 9). As the
fire consumed more of the edge of the stainless steel ball, the hole
FIG. 9—Close-up view of the damaged valve stem/PTFE seat assembly.
increased in size with an accompanying increase in the flow of oxygen

and a subsequent change in direction of the flame jet. The flame jet
ignited, burned, and eroded the inside of the ball (see location C in
Figs. 8 and 9) and the fire flowed out of the ball, consuming the inside
diameter of the valve body at the outlet port.
Discussion
The proposed ignition mechanism is a combination of the heat from pressuriza-
tion of the product gas, spontaneous heating of very small cellulose fibers and
particles, and the effects of the very small flow associated with the product gas
leak across the valve seat. Admittedly, this proposal is derived from an unveri-
fiable set of circumstances; but in its defense, this theory matches with all the
observed and verifiable evidence. All other ignition mechanisms considered;
on the other hand, can be dismissed based upon the known facts related to the
configuration of the burned three-way valve, the system operational configura-
tion, and the witness statements. As one sleuth asked his faithful assistant,
“How often have I said to you that when you have eliminated the impossible,
whatever remains, however improbable, must be the truth?” (Sherlock Holmes
in The Sign of Four by A.C. Doyle). Perhaps that sentiment applies here, as
well.
Estimation of Critical Radius of Small Cellulose Particles in Flowing Oxygen

In order for the proposed ignition scenario to be plausible, it is necessary to
demonstrate that small cellulose fibers and particles can undergo spontaneous
heating in the configuration at the inlet of the leak path between the stainless
steel ball and the unsupported PTFE seat annotated in Fig. 5. To do so, the crit-
ical radius of cellulose particles in oxygen-enriched, high pressure, flowing gas
would need to be in the range of 0.1 to 0.2 cm. This amounts to approximately
a two-order-of-magnitude decrease in the critical radius for white pine sawdust
as shown in Fig. 1. The authors propose that three additive factors will produce
this decrease in critical radius: (1) higher oxygen concentration (purity and
pressure) compared to ambient pressure air, (2) forced convection (from a
leak) as compared to natural convection, and (3) small cellulose particles com-
pared to white pine sawdust.
Effect of Oxygen Purity and Pressure—Considering an accumulation or

pile of cellulose particles, as the temperature of the pile increases, the rate of
heat generation can increase exponentially according to the Arrhenius equation
(Eq 1).
E
q_ ¼ QqAe RT (1)
where:
q_ is the heat generation rate per unit volume (J/sm3),
Q is the heat of reaction (kJ/g),
q is the bulk density of the material (kg/m3),
A is the pre-exponential factor (s1),
E is the apparent activation energy (J/mol),
R is the universal gas constant (¼ 8.314 J/molK), and
T is the temperature (K) of the particles within the pile.
The Frank-Kamanetskii model for spontaneous ignition describes a Dam-
kohler number (d) as an estimate of the heat generation within an accumulation
of combustible material [10] (Eq 2).
qQA Er 2 E
d¼ e RT (2)
k RTo
where:
k is the thermal conductivity of the pile of particles (W/(mK)),
r is the characteristic length of an accumulated pile (radius for a sphere)
(m), and
To is the flowing gas temperature (K).
Another calculated value, known as delta critical, is derived from heat loss
mechanisms. If the Damkohler number is less than delta critical, the rate of
heat loss is greater than the rate of heat generation and spontaneous ignition
cannot occur. If heat generation is greater than heat loss (Damkohler > delta
critical), the temperature can rise exponentially to ignition.
Rearranging Eq 2 produces Eq 3:
E Er 2
d ¼ QqAe RT (3)
kRTo
Substituting the first group with q_ yields (Eq 4):
Er 2
d ¼ q_ (4)
kRTo
Setting the radius of the pile of accumulated particles to the critical radius
makes d equal to the Damkohler number (Eq 5).
Erc2
dc ¼ q_ (5)
kRTo
Carras and Young [11] found that in the early stages of carbonaceous material
oxidation, the reaction proceeds as follows:
Carbonacous material þ O2 ! CO2 þ heat
The rate of the reaction can be expressed as (Eq 6):

E
r0 ¼ ½O2 Ae RT (6)
where:
r0 is the rate of change in the concentrations of the reactants and products
(kmol/(m3 s) and
[O2] is the oxygen concentration (kmol/m3).
Therefore, the rate of reaction is proportional to the rate of heat production
and can be substituted into Eq 5 to yield (Eq 7):
E Erc2
dc / ½O2 Ae RT (7)
kRTo
Since dc, A, E, k, R, T, and To are constants, they can be ignored in the propor-
tionality and Eq 7 becomes (Eq 8):
1
rc / pffiffiffiffiffiffiffiffi (8)
½O2
Equation 8 shows that for a collection of combustible particles that is on the

tipping point of spontaneous heating and ignition, if the rate of heat production
is quadrupled by increased oxygen concentration, the critical radius will only
decrease by a factor of two.
The critical radii data shown in Fig. 1 are postulated assuming air at ambi-
ent conditions. The ratio of oxygen partial pressure at 2200 psi (15.2 MPa) and
95 % purity divided by the oxygen partial pressure at ambient conditions
results in a 677 fold increase in oxygen concentration. Applying this increase
to Eq 8 predicts a 26 fold decrease in critical radius due to this increase in oxy-
gen concentration.
Effect of Forced Convection—Figure 10 [12] shows the temperature sensi-

tivity of a packed bed of fine (<45 lm) carbonaceous material to air flow. The
vertical axis describes a runaway temperature which is defined as the tempera-
ture increase through the packed bed above the inlet temperature of 493 K. The
plateau at approximately 1 L/min translates into a linear velocity across the
bed particles of approximately 27 in/s (69 cm/s).
The ratio of outlet temperature to inlet temperature in air is approximately
1.28. Assuming this ratio applies to enriched oxygen and applying it to cellu-
lose at 353 K (177 F) yields an estimated flowing oxygen exit temperature of
451 K. Substituting the rate of reaction for the oxygen concentration in Eq 8
yields (Eq 9):
FIG. 10—Runaway temperature as a function of airflow [12].

1
rc / pffiffiffiffi (9)
r0
Using Eq 1 with an apparent activation energy of 87 900 J/mol for cellulose

[10] to calculate a ratio of reaction rates at the two different temperatures, and
applying the results to Eq 9, yields an estimated decrease in critical radius of
26 fold.
Effect of Characteristic Particle Dimension—Figure 11 [12] shows the

temperature sensitivity of a packed bed of carbonaceous material to various
particle sizes. The vertical axis describes a runaway temperature which is
defined as the temperature increase through the packed bed above the inlet tem-
perature of 493 K. The assumed characteristic dimension or particle size of
white pine sawdust is in the range of 200 to 800 lm, and the measured fiber
and particle diameters of the cellulose in the fire event oxygen system is 30 to
60 lm.
The ratio of outlet temperature to inlet temperature is approximately 1.54.
Applying this ratio to cellulose at 353 K (177 F) yields an estimated flowing
oxygen exit temperature of 543 K.
FIG. 11—Runaway temperature as a function of particle size [12].

Using Eq 1 with an apparent activation energy of 87 900 J/mol for cellulose

[10] to calculate a ratio of reaction rates at the two different temperatures, and
applying the results to Eq 9, yields an estimated decrease in critical radius of
188 fold.
Combined Effects of Factors—The combined effect of oxygen concentra-

tion (26 fold), forced convection (26 fold), and particle dimension (188 fold) is
a 240-fold decrease in the critical radius. Therefore, the critical radius for white
pine sawdust (which is cellulose and thus comparable to the cellulose fibers in
the damaged oxygen system) at 177 F (81 C) in Fig. 1 will be reduced from
15 in. (38 cm) to 0.062 in. (0.16 cm), as shown in Fig. 12. This dimension fits
well within the dimensions of the inlet of the valve as indicated in Fig. 5. It is
reasonable to conclude that the ignition scenario presented herein is plausible.
FIG. 12—Projected surface temperatures versus critical radius in spontaneous

ignition of very small cellulose particles in elevated oxygen concentration with
forced flow.
Conclusion
The “flow friction” ignition mechanism is a postulated ignition mechanism that
is unsubstantiated by direct test data. Even so, significant circumstantial evi-
dence that ignition events have occurred in oxygen-enriched atmospheres can
be attributed to oxygen-enriched gas leaking past polymer seats and seals. The
spontaneous ignition of accumulations of small particulate and fibrous debris
in a leak path has been postulated as a contributing ignition source in a
stainless-steel-body hand valve with a PTFE seal and silicone lubricant known
to be contaminated with small cellulose particles. It was surmised that the leak-
ing oxygen-enriched gas provided the physical transport mechanism to remove
oxidation products and supply fresh oxidizing gas to the accumulated deposit
of cellulose particles. It is further surmised that this ignition mechanism would
exist for other material combinations than stainless steel and PTFE and for
other contaminant particles than cellulose. This adds an important aspect to
the discussion of the “flow friction” ignition mechanism in oxygen-enriched
atmospheres.
Recommendation for Future Tests

It is recommended that the values of critical radii as a function of surface tem-
perature in Fig. 12 be validated by testing.
References
[1] Beeson, H. D., Smith, S. R., and Stewart, W. F. “Manual 36,” Safe Use of
Oxygen and Oxygen Systems: Handbook for Design, Operation, and
Maintenance, 2nd ed., ASTM International, West Conshohocken, PA,
Oct 2007.
[2] Gallus, T. D. and J. M. Stoltzfus. “Flow Friction Fire History and
Research,” Flammability and Sensitivity of Materials in Oxygen-Enriched
Atmospheres: 11th edition, ASTM STP 1479, D. B. Hirsch, R. Zawieru-
cha, T. A. Steinberg, and H. Barthlelemy, Eds., ASTM International,
West Conshohocken, PA, 2006, pp. 151–162.
[3] Hooser, J. D., Wei, M., Newton, B. E., and Chiffoleau, G. J. A., “An
Approach to Understanding Flow Friction Ignition: A Computational
Fluid Dynamics (CFD) Study on Temperature Development of High-
Pressure Oxygen Flow Inside Micron-Scale Seal Cracks,” Flammability
and Sensitivity of Materials in Oxygen-Enriched Atmospheres: 12th edi-
tion, ASTM STP 1522, H. Barthelemy, T. A. Steinberg, C. Binder, and
S. Smith, Eds., ASTM International, West Conshohocken, PA, 2009, pp.
429–449.
[4] Stoltzfus, J. M., “Fires in P-3 Aircraft Oxygen Systems,” Flammability

and Sensitivity of Materials in Oxygen-Enriched Atmospheres, 11th edi-
tion, ASTM STP 1479, D. B. Hirsch, R. Zawierucha, T. A. Steinberg, and
H. Barthlelemy, Eds., ASTM International, West Conshohocken, PA,
2006, pp 176–188.
[5] Whitaker, A. F., Thompson, R. L., Brunell, R. C., Parr, R. A., Stoltzfus,
J. M., and Leonard, R. Y., Final Report of the Board of Investigation for
Mishap that Occurred on Jan 15, 1997 in Building 4623 at Marshal Space
Flight Center. Failure Analysis, Marshall Space Flight Center: George C.
Marshall Space Flight Center, AL, 1997.
[6] Forsyth, E. T., Eaton, D. J., and Newton, B. E., “Oxygen Fire Cause and Ori-
gin Analysis of the CUMA V2 Underwater Breathing Apparatus,” Flamma-
bility and Sensitivity of Materials in Oxygen-Enriched Atmospheres: Tenth
Volume, ASTM STP 1452, T. A. Steinberg, H. D. Beeson, and B. E. Newton,
Eds., ASTM International, West Conshohocken, PA, 2003.
[7] Primer on Spontaneous Heating and Pyrophoricity, DOE-HDBK-1081-
94, Dept. of Energy, Washington, DC, 1994.
[8] Hossfeld, R. J. and Hatt, R. PRB Coal Degradation – Causes and Cures,
http://www.prbcoals.com/pdf/paper_archives/56538.pdf (last accessed
2/8/12).
[9] Tsuchiya, Y., and Sumi, K., “Spontaneous Ignition,” CBD-189, Canadian
Building Digests, NRC-IRC Publications, 1977.
[10] Tamburello, S. M., “On Determining Spontaneous Ignition in Porous
Materials,” Final Thesis/Master of Science in Fire Protection Engineering,
Univ. of Maryland, College Park, MD. 2011. http://www.google.com/
search?q=Tamburello+spontaneous+ignition+&rls=com.microsoft%3A*%
3AIE-SearchBox&oe=UTF-8. &sourceid=ie7&rlz=1I7GGIE_en&oq=Tamburello
+spontaneous+ignition+&aq=f&aqi=&aql=&gs_sm=e&gs_upl=8359l18389l
0l18483l30l30l0l18l18l0l407l2596l2-6.2.1l9l0 (last accessed 2/8/12).
[11] Carras, J. N. and Young, B. C., “Spontaneous Heating of Coal and
Related Materials: Models, Application and Test Methods,” Progr.
Energy and Combustion Sci. Vol. 20, 1994, pp. 1–15.
[12] Malhotra, V. M. and Crelling, J. C., “Effect of Particle Size And Air
Flow Rate on the Runaway Temperature of Bituminous Coal at 290 K
< T < 700 K,” Dept. of Physics/Dept. of Geology, Southern Illinois Univ.,
Carbondale, IL, http://www.anl.gov/PCS/acsfuel/preprint%20archive/Files/32_
4_NEW%20ORLEANS_08-87_0070.pdf (last accessed 2/8/12).
STP 1561, 2012
DOI:10.1520/STP20120002
Matthias Meilinger1
Ignition Testing by Fracture of Oil- and

Particle-Contaminated Aluminum Pipes
in High Pressure Oxygen at Ambient
and Elevated Temperature
REFERENCE: Meilinger, Matthias, “Ignition Testing by Fracture of Oil- and

Particle-Contaminated Aluminum Pipes in High Pressure Oxygen at Ambient
and Elevated Temperature,” Flammability and Sensitivity of Materials in Oxy-
gen-Enriched Atmospheres on September 19–21, 2012 in Montreal, QC;
STP 1561, Samuel Edgar Davis and Theodore A. Steinberg, Editors, pp. 81–
104, doi:10.1520/STP20120002, ASTM International, West Conshohocken,
PA 2012.
ABSTRACT: Aluminum and its alloys are the preferred material in air separa-
tion units (ASUs) and are submitted to strict cleanliness requirements to
avoid combustion reactions in oxygen. Linde Engineering Division has inves-
tigated the role of particle contamination for the ignition of aluminum pipes in
high pressure oxygen. Tests were carried out with oil coated aluminum pipes
in combination with a particle contamination. Accumulations of following par-
ticle types were tested: molecular sieve 13X, filter deposits (rust), perlite,
quartz sand, and aluminum swarf. The contaminated pipes were pressurized
with oxygen up to 150 bar and fractured in a hydraulic apparatus. The condi-
tions and probability of an ignition reaction compared with oiled pipes without
particles were of interest. Additionally, similar high pressure tests with a mix-
ture of liquid and gaseous oxygen were performed. A combination of oil and
particles enhanced the probability of an ignition at elevated temperatures.
The results indicated that temperature is the dominating factor for an oil igni-
tion, not the oxygen pressure. At room temperature, an oil/aluminum ignition
could never be initiated. The results give a good understanding regarding the
mechanism of an aluminum ignition and help to clarify causes of damage in
the ASU industry.
Manuscript received December 9, 2011; accepted for publication April 27, 2012; published
1
R&D scientist, Research and Development Division, The Linde Group, Linde Engineering
Division, Dr.-Carl-von-Linde-Str. 6-14, 82049 Pullach, Germany.
Copyright V
19428-2959.
81
KEYWORDS: oxygen, aluminum, ignition, oil, particles, contamination,

elevated temperature, particle impact, high pressure, pipe fracture, molecular
sieve, perlite, swarf, rust, quartz sand
Introduction
Is there any pressure limitation for safe application of aluminum as the pre-
ferred material in air separation units (ASUs)? This is a crucial question for the
Linde Engineering Division due to the increased demand of high pressure oxy-
gen above 100 bar, especially thin structures of aluminum. For example, thin
fins used in brazed fin heat exchangers could ignite more easily than massive
piping material. Such heat exchangers submitted to temperature stress tests are
exposed to fatigue at the welding seams. Is a fracture at this point in combina-
tion with the exposure of fresh metal surface and the subsequent release of
high pressure oxygen capable of igniting the adjacent aluminum? If no reaction
during pipe fracture could be initiated, then conditions were searched for which
enhanced the probability of an ignition. These issues were the motivation for
ignition tests with aluminum pipes pressurized with 150 bar oxygen. To be
clear: The goal was not to demonstrate the rapid consumption of aluminum in
high pressure oxygen—the propagation of combustion has been reported by
many authors in the past—but the preconditions for an occurring ignition.
Other requirements comprise the extent and type of contamination with foreign
material, sample temperature, and oxygen pressure. Special attention was paid
to the type of particulate matter and its distinct property to initiate an oil/alumi-
num combustion.
Beside tests in gaseous oxygen (GOX) tests with liquid oxygen (LOX)
were also carried out.
Maximum Pressure Recommendation

Safety recommendations of the German Statutory Accident Insurance Institu-
tion for Chemical Industry (Berufsgenossenschaft Rohstoffe und Chemie):
“Aluminum and aluminum alloys can be used as material for piping with gase-
ous oxygen (O2 > 70 %) up to a working temperature of 80 C and up to a
working pressure of 63 bar overpressure in the area of cryogenic plants.” These
recommendations are based upon an old survey of the oxygen producing indus-
try in 1963 and are still valid [1]. This statement does not rely on experimental
test data but on the experience of good practice at this point in time. No other
standards designate oxygen compatibility based pressure limits for the applica-
tion of aluminum.
Background
The accumulation of hydrocarbons represents a serious hazard for air sepa-
ration units due to their potential capability to promote the ignition of
MEILINGER, doi:10.1520/STP20120002 83
internals made of aluminum. Typical hydrocarbons present in ASUs origi-

nate from undesired residual lubricants from manufacturing, but also unin-
tentionally left wood, papers, plastics, and textiles during erection. In this
context, Ref [2] reports the ignition probability of thin oil films after impact
of high pressure oxygen. Additionally highly volatile hydrocarbons like
methane, ethane, ethylene, and propane along with plugging components
like nitrous oxide and carbon dioxide enter the cryogenic plant as impurities
via process air and can accumulate during improper operation of an air pre-
purification unit and oxygen reboiler to a dangerous extent [3]. Also air-
borne combustible aerosols can pass through the air prepurification unit. In
the range of 1 lm and smaller, the flammable aerosol particles can deposit
within the ASU [4].
During inspection of an ASU prior to commissioning one can find debris
like welding slag, metal fibers, cloth material, and dust. Not all of them can be
removed by wiping and bursting procedures. Later on during operation of
the plant fine particles of outside dust, molecular sieve attrition, or perlite
ingress through pipe leakage can contaminate the oxygen sections. Typically,
particles are considered as plugging components or in high velocity systems
(velocity > 50 m/s) as ignition source by a particle impact.
Aluminum is considered an appropriate material for oxygen producing
plants due to its remarkable ignition resistance that is attributed to the high
thermal conductivity of aluminum and the commonly present protective oxide
layer on the surface. A good overview about oxygen compatibility data of alu-
minum is given in Ref [5]. Reference [6] describes the mitigation effect of
thicker oxide layers on the combustion of aluminum rods. The authors of
Ref [7] performed promoted ignition tests with tubular aluminum samples to
investigate their behavior at elevated oxygen pressures. Approximately 2000 C
are required to initiate an ignition, which is hard to reach due to the high ther-
mal conductivity of aluminum. But once ignited, aluminum combusts violently
at rather low oxygen pressures. Therefore, contamination with a flammable
promoting material with a low auto ignition temperature has to be avoided. A
likely ignition mechanism for such organic contaminants is “adiabatic
compression.” The resulting real temperatures were investigated by several
authors [8–10].
Experimental
Ignition Test Procedure

Aluminum pipes of different diameters with or without an oil/particle-contami-
nation were slowly pressurized with oxygen to 150 bar and subsequently frac-
tured. Ignition tests with either heated pipes (150 C) or non-heated pipes
(GOX-tests: 30 C–40 C þ LOX-tests) were carried out.
TABLE 1—Dimensions of specimen for the GOX-tests.
Pipe Outer Diameter Wall Thickness Length Weight Volume

(mm) (mm) (mm) (g) (cm3)
34 2 600 400 424
12 2 600 100.5 30.2
Specimen
Aluminum pipes (Dimensions see Tables 1 and 2)
Alloys:
34 2 mm pipe: AlMG4.5 MN (en aw-5083)
12 2 mm pipe: AlMG3 (EN AW-5754)
Oxygen
Oxygen cylinders of the purity 99.95 vol. % were used throughout. The purity
of each cylinder was analyzed by gas chromatography ranging from
99.94–99.99 vol. % O2 with approximately equal concentrations of nitrogen
and argon. The assumed oxygen concentration during ignition tests was
approximately 99.9 % due to contamination with ambient air. To ensure a suffi-
cient supply of oxygen during fracture a buffer vessel of 1 liter volume was
inserted into the oxygen supply line.
Pressure
Oxygen pressure was measured with a calibrated pressure sensor (0–250 bar)
and transmitter (Wika Tronic Line, type 909.40.500). The pressure trend was
recorded. General procedure: After attaining the desired temperature the pipe
was pressurized manually to 150 bar within 0.5 min by means of a pressure
regulator.
Temperature
For tests at elevated temperatures the pipes were heated with two electric heat-
ing tapes (2 m long, Tmax.: 450 C, power: 500 W, supplier: Horst) which were
TABLE 2—Dimensions of specimen for the LOX-tests.
Dimensions (mm)
Pipe Outer
Diameter (mm) Wall Thickness (mm) A B C D Weight (g) Volume (cm3)
34 2 100 75 120 460 697 700

12 2 0 70 130 499 135 40.5
FIG. 1—Shape of the specimen for the LOX-tests.
wrapped around the specimen left and right of the placed scratch where the
fracture should occur. The temperature was controlled with a NiCr-Ni—
thermocouple (1 mm diameter, supplier: THERMOCOAX) 2 cm adjacent to
the scratch. The end of the thermocouple wire was attached at the pipe with an
aluminum adhesive tape and finally fixed by a surrounding copper wire. For
every test the temperature trend was recorded. Ignition test sequences at two
temperature levels were performed: ambient temperature (30 C–40 C after
pressurization) and 150 C. Additionally LOX tests were carried out with an
initial temperature of about 0 C at the middle not LOX immersed section of
the pipe. (LOX-Specimen see Fig. 1.)
FIG. 2—Different particle types, scale in cm.

Oil Contamination
Most pipes for GOX tests were contaminated internally with mineral oil prior
to test. Also the LOX test specimen were oil contaminated except their
U-section.
Type: Kaeser Sigma Fluid MOH screw compressor oil used:
boiling range determined by simulated distillation: 319–608 C,
50 % by weight boil at 498 C,
viscosity: 325 Pa s (20 C), 1.78 Pa s (200 C).
Particle Contamination
Particles were selected for ignition tests which can typically contaminate
piping systems and internals of an ASU plant (see Table 3 and Fig. 2).
Subsequently, scanning electron microscopy pictures of the applied
particles: (see Figs. 3–14).
Experimental Procedures
At the beginning every specimen was internally purged by flushing the pipes
with acetone (oil-, particle-removal). After solvent removal and drying the
TABLE 3—Selected particle contamination.
Bulk density
Type of Particle Appearance Color (g/cm3) Composition
Molecular powder, 100–200 lm, high bright beige 0.80 aluminum
sieve 13Xa amount of fines: 10 lm silicate
Perlite cracked fragile surface white 0.06 SiO2, Al2O3,
bubbles Na2O
Quartz sand fine powder, 50–250 lm, beige 1.5 SiO2
smooth sharp-edged rocks
Aluminum serrated edges, 0.1–1 mm grey, silver 0.91 alloy
swarf long, 50–100 lm thick AlMg4.5Mn
Aluminum very fine, 325 mesh grey 1.1 alloy Al 99.5
powder (44 lm)
Filter deposit 100–500 lm, single heterogeneous, 1.4 Fe3O4, SiO2, Al,
blackb welding beads single metal Cr, Ni, Mo
fibers
Filter deposit fine powder: 100–200 lm, red-brown 1.7 Fe2O3, pure rust
redc fines: 10–20 lm
Insects, dry flies, spiders, moths grey-brown 0.03 organic
a
Molecular sieve beads crushed in a mortar.
b
Filter deposit black: stainless steel slag mixed with aluminum fibers and dust, received from Air Liquide
Germany.
c
Filter deposit red from carbon steel piping, received from Air Liquide Germany.
FIG. 3—Molecular sieve 13X.
pipes were completely filled with oil. Subsequently, the oil was drained and the
contaminated specimen weighed back. Dependent on draining time (2–12 h)
and temperature (15–22 C) a quite defined oil contamination could be gener-
ated in the range of 10–35 g/m2. Particles were then added through the GOX
inlet by a funnel under simultaneous rotating of the pipe to attain a uniform
particle distribution. Nevertheless, in some cases an accumulation of the par-
ticles at the opposite pipe end could not be avoided.
FIG. 4—Molecular sieve 13X, enlarged.

FIG. 5—Perlite.
Afterwards the contaminated pipe was mounted in a hydraulic bending ap-

paratus, connected with the GOX supplying line and wrapped with the heating
tapes. (See Fig. 15.) In the uncovered middle section a scratch was placed
(1–2 cm long) where the fracture should occur later on. This was done by a
small saw. After attaching the thermocouple the entire piping system was set
under vacuum and then subjected to several pressurizing/depressurizing cycles
for purging (maximum oxygen cycle pressure: 9 bar absolute). Subsequently,
the specimen was heated to 150 C within 20 min at an oxygen pressure of 4 bar
FIG. 6—Perlite, enlarged.

FIG. 7—Quartz sand.
absolute. After reaching the desired temperature the oxygen pressure was
increased to 150 bar within 0.5 min and the pipe fractured hydraulically.
In contrast to the above depicted GOX tests for the LOX tests 200 cm3 liq-
uid oxygen was filled into the precooled U-section of the pipe after purging
with oxygen. The scratch was placed at the straight, not LOX immersed pipe
section. (See Fig. 16.) After pressurizing with GOX to 150 bar from both sides
the pipe was broken hydraulically. During fracture a strong LOX/GOX jet was
ejected from the cut.
FIG. 8—Quartz sand, enlarged.

FIG. 9—Filter deposit black.
Results
Ignition of Oil Contaminated Aluminum Pipes with and without Particles at

150 C
The probability for an ignition of oil and particle contaminated aluminum pipes
during pressurization with 150 bar oxygen exceeded 50 %, almost independent
of the particle type. In contrast, in absence of particles the ignition probability
was reduced to 4 %. This result indicates a catalytic effect of particles for an
oil ignition at elevated temperatures, especially for these particles with a rough
surface.
FIG. 10—Filter deposit black, enlarged.

FIG. 11—Filter deposit red.
Particle Type Determines the Ignition Mechanism

The observed oil ignitions could occur during pressurization with oxygen or
thereafter during the fracture of the pipe. A contamination with fragile, rough
particles like molecular sieve, perlite, and rust resulted mostly in an oil ignition
during pressurization. These oil ignitions were not only initiated at high pres-
sures but in the entire range of 4–150 bar oxygen. A similar catalytic behavior
was observed with aluminum swarf.
FIG. 12—Filter deposit red, enlarged.

FIG. 13—Aluminum swarf.
In contrast to the quartz sand crystals, oil wetted sand caused a predomi-
nantly oil/aluminum ignition during fracture. This property could be explained
with a particle impact mechanism when compact crystals hit the fresh surface
at the fracture resulting in a massive burnout around the crack. Simultaneously,
molten aluminum was ejected and the room was filled with a fine, white
Al2O3-smoke which started to precipitate not until 10 min.
FIG. 14—Aluminum swarf, enlarged.

FIG. 15—Test apparatus for the GOX tests.
Oil Ignition: Temperature as the Dominant Factor

Oil ignitions occurred in the entire pressure range of 4–150 bar but never at
temperatures below 95 C. Also in regard of the LOX tests, no ignition reaction
could be observed. Thus, an oil combustion can be prevented effectively with
low ambient temperatures. However, only a very small amount of heat in the
lJ-range on a tiny spot is required to initiate a hydrocarbon combustion.
Results in Detail
Ignition Tests with Oil and Particle Contamination at 30–40 C

Twenty one tests with oil coated aluminum pipes additionally contaminated
with either molecular sieve 13X, perlite or Al-fibers were carried out at room
FIG. 16—Test apparatus for the LOX tests.
temperature (see Table 4, columns 1 þ 2). Commonly, after pressurization with

oxygen to 150 bar a wall temperature of 30–40 C was reached, depending on
the pipe diameter. All pipes were fractured subsequently without any indication
for a temperature rise or an ignition. Obviously, there is a minimum tempera-
ture required for an oil ignition.
Ignition Tests with Oil in Absence of Particles at Elevated Temperature

A total of 24 tests with oily aluminum pipes in a temperature range from
130 C–190 C were performed (see Table 4, column 3). Only in one case a
34 mm pipe burst during pressurization. This occurred at 147 C and 30 bar
oxygen pressure. Apparently, the oil’s ignition resistance in high pressure oxy-
gen at the chosen conditions is quite high. In 23 tests a fine oil mist was ejected
during fracturing of the pipe in 150 bar oxygen.
Ignition Tests with Oil and Molecular Sieve 13X Contamination at Elevated
Temperatures
In eight of 14 tests with crushed molecular sieve 13X an ignition during pres-
surization was observed (see Table 4, column 4). During three tests the pipes
did not fracture hydraulically because the resulting ignitions destroyed the
pipes prior to that. During one test molten aluminum was ejected. With 95 C
and 10 bar oxygen pressure the mildest conditions for an oil ignition were
received during one test. The wall temperature jumped from 95 C to 112 C
indicating an incomplete oil combustion.
Ignition Tests with Oil and Perlite Contamination at Elevated Temperatures

In four of eight tests with perlite an ignition during pressurization was observed
(see Table 4, column 4). The occurrence of a couple of independent single
ignitions was remarkable. Apparently, within the 34 mm pipes there were some
distinct perlite/oil accumulations which ignited separately generating tempera-
ture spikes of 30 C–80 C for every single event. In spite of the small surface to
volume ratio the perlite supported an oil combustion like a catalyst. 124 C and
4 bar oxygen pressure were the mildest conditions for an ignition.
Ignition Tests with and without Oil and Quartz Sand Contamination at
Elevated Temperatures
In only one of eight tests with quartz sand an ignition occurred during pressur-
ization (see Table 4, column 5). Combustion reactions occurred during hydrau-
lic fracture of the pipe in four of eight tests resulting in an erosion of the
crevice and the ejection of molten aluminum. Simultaneously white smoke of
aluminum oxide was generated. Comparative tests without oil contamination
did not result in any aluminum combustion (see Table 5, column 3). Obviously,
during facture of the pipe the oil attached on the sand crystals ignited first
followed by an aluminum combustion in the region of fracture. A scanning
electron microscopy investigation revealed the smooth surface and compact
nature of the quartz sand crystals. Thus, the smooth shape is the reason for a
reduced catalytic activity of quartz sand regarding oil ignition. On the other
hand, the compact sand “rocks” transfer their kinetic energy effectively to the
impingement site and make a particle impact ignition mechanism very likely.
Especially the ejected sand crystals which were very abrasive and enhanced
the exposure of fresh metal close to the fracture crevice. (See Fig. 17.)
Ignition Tests with Oil and Aluminum Swarf Contamination at Elevated

Temperatures
In four of seven tests with aluminum swarf an ignition occurred during pressur-
ization (see Table 5, column 2). Simultaneously molten aluminum was ejected
TABLE 4—Ignition test results with aluminum pipes with different contamination at ambient and elevated temperatures.
Contamination With Contamination With Contamination With

Oil and MS13X Oil and Mixture of Contamination With Oil and MS13X Contamination With
or Perlitea,b Al-powder and Al-swarfc,d oil Without particlese or Perlitef Oil and Quartz Sandg,h
30–40 C 30–40 C 150 C 150 C 150 C
Tests total 11 10 24 22 8
No ignition, no DT 11 10 23 9 3
Oil ignites during 0 0 1 6 0

pressurization, DT, no
burnthrough of the pipe
Oil/Al ignite during 0 0 0 6 1

pressurization, DT,
burnthrough of the pipe
Oil/Al ignite during pipe 0 0 0 1 4

fracture
Occurrence of molten 0 0 0 3 4
aluminum, T > 660 C
a
12 mm pipes: oil: 0.3 g (20 g/m2) þ particles: 0.3 g Perlite.
b
34 mm pipes: oil: 1.0–1.9 g (18–34 g/m2) þ particles: 1.5 g Perlite or 3 g MS 13X.
c
12 mm pipes: oil: 0.1–0.2 g (6–13 g/m2) þ particles: aluminum powder þ swarf, each 0.2 g, one test with 1.2 g swarf þ 1 g powder.
d
34 mm pipes: oil: 1.1–3 g (19–53 g/m2 þ particles: 1.3 g aluminum powder þ 1.5 g aluminum swarf.
e
12 mm pipes: oil: 0.11–0.54 g (7–36 g/m2), 34 mm pipes: oil: 1.6–1.9 g (28–34 g/m2).
f
12 mm pipes: oil: 0.2–0.5 g (13–33 g/m2) þ particles: 0.3 g Perlite or 0.5 g MS 13X, 34 mm pipes: oil: 0.7–2.1 g (13–37 g/m2), 3 samples with 80 g/m2 oil þ particles: 1.5 g
Perlite or 3 g MS 13X.
g
12 mm pipes: oil: 0.27 g (18 g/m2) þ particles: 2 g quartz sand.
h
34 mm pipes: oil: 1.5–2 g (27–35 g/m2) þ particles: 10 g quartz sand.
TABLE 5—Ignition test results with aluminum pipes with different contamination at elevated temperatures.
Contamination With Contamination With Contamination With Contamination With

Oil and Filter Oil and Aluminum Quartz Sand Without Insects Without
Deposita Swarfb,c Oild Oile
150 C 150 C 170 C 150 C
Tests total 8 7 4 2
No ignition, no DT 3 3 4 0
Oil ignites during pressurization, DT, no burnthrough of the pipe 1 0 – 2

Insects ignite
Oil/Al ignite during pressurization, DT, burnthrough of the pipe 3 4 0 0
Oil/Al ignite during pipe fracture 1 0 0 0
Occurrence of molten aluminum, T > 660 C 2 4 0 0

a 2 2
12 mm pipes: oil: 0.31–0.36 g (21–24 g/m ) þ particles: 1.5 g filter deposit red or 0.8 g filter deposit black, 34 mm pipes: oil: 1.7 g (30 g/m ) þ particles: 3.5 g filter deposit
red or 1.2 g filter deposit black.
b
12 mm pipes: oil: 0.30 g þ particles: 0.8 g aluminum swarf.
c
34 mm pipes: oil: 1.7 g þ particles: 1–1.5 g aluminum swarf.
d
12 mm pipes: particles: 2 g quartz sand.
e
12 mm pipes: contamination with 0.17 g dry insects.
MEILINGER, doi:10.1520/STP20120002
97
FIG. 17—Typical ignition test result with oil and quartz sand contamination at
elevated temperatures.
and a white smoke of Al2O3 arose. The mildest conditions for an oil ignition
were 142 C and 120 bar oxygen pressure. Aluminum swarf tests in the past with-
out oil at 200 C did not result in any aluminum ignition. Obviously this event
requires a second component with a low auto ignition energy (AIT) like a hydro-
carbon. In the presence of oil and oxygen the aluminum fibers easily triggered
an oil/aluminum ignition. In contrast to all other particles tested aluminum swarf
is itself combustible. Aluminum swarf also has a low surface to volume ratio.
Ignition Tests with Oil and Filter Deposit Contamination at Elevated
Temperatures
Two different types of filter deposits found in oxygen pipes were subjected to
ignition tests (Table 5, column 1). Especially, the red-brown rust particles were
very fine with a large amount in the 10–20 lm size range. In five of eight tests
an oil combustion was observed whereas in two tests molten aluminum was
generated. The mildest ignition conditions were 120 C and 10 bar oxygen pres-
sure. Obviously, also filter deposit is catalytically active regarding oil ignition.
Ignition Tests with Insects without Oil Contamination at Elevated

Temperatures
Twelve mm aluminum pipes were filled with a mixture of dry insects and subse-
quently pressurized (Table 5, column 4). In both tests a couple of separated sin-
gle ignitions were noticed at temperatures above 120 C generating temperature
spikes of 5 C–20 C each. No burnthrough of the 2 mm wall was received. The
hazard of organic debris in the outlet nozzle of oxygen cylinders is well known.
TABLE 6—LOX ignition test results with aluminum pipes with different contamination.
Contamination With a Mixture Contamination With Oil and

of Al-powder and Al-swarf a Mixture of Al-powder and
Without Oila,b Al-swarfc,d
0 C 0 C
Tests total 9 11
No ignition, no DT 9 11
Oil ignites during pressurization, – 0
DT, no burnthrough of the pipe
Oil/Al ignite during pressurization, 0 0
DT, burnthrough of the pipe
Oil/Al ignite during pipe fracture 0 0
Occurrence of molten aluminum, 0 0
T > 660 C
a
12 mm pipes: particles: 0.3–0.8 g aluminum powder þ 0.3–0.8 g aluminum swarf.
b
34 mm pipes: particles: 1.3–2 g aluminum powder þ 1.5–2 g aluminum swarf.
c
12 mm pipes: oil: 0.1 g (8 g/m2) þ particles: 0.3 g aluminum powder and aluminum swarf, each.
d
34 mm pipes: oil: 0.8 g (18 g/m2) þ particles: 2 g aluminum powder and aluminum swarf, each.
Ignition Tests with GOX/LOX Mixtures

Totally 20 test were performed, nine without and 11 with oil contamination
(Table 6). A mixture of aluminum powder and swarf was applied throughout.
FIG. 18—LOX ignition tests with oil and aluminum powder/swarf

contamination.
TABLE 7—Scanning electron microscopy analysis of the black combustion deposit.
34 2 mm aluminum pipe: 34 2 mm aluminum pipe: 1.5 g

Specimen 1.5 g perlite þ 2.1 g oil ¼ 37 g/m2 aluminum swarf þ 1.6 g oil ¼ 28 g/m2
Ignitions occurred at, 1. 125!212 C/16 bar O2 142 C/120 bar O2
temperature spikes 2. 214!295 C/26 bar O2
Result Pipe bursts at 236 C/130 bar O2 Pipe bursts at 142 C/120 bar O2
Residual oil (%) 0.5 0.6
Appearance of extract Yellow solution with soot particles Bright yellow solution
Composition (wt. %) Black spot Clean surface nearby Black spot Clean surface nearby
C 60.2 8.8 33.3 7.3
Al 30.0 79.9 57.1 82.0
O 7.1 5.5 5.5 5.9
Mg 2.2 5.0 3.6 4.8
All pipes were pressurized and hydraulically fractured without any indication
of an ignition. Prior to fracture, a temperature of approximately 0 C was
recorded in the pipe section not immersed in LOX. (See Fig. 18.)
Black Deposit Analysis

After violent combustions often a black deposit was discovered on the inner
walls of the aluminum pipes. For clarification of the chemical composition, a
scanning electron microscopy investigation was carried out. Additionally, the
pipes were washed with acetone to determine the residual oil content. The anal-
ysis of the black spots and, in parallel, of the clean surface nearby proved that
the black deposit consisted of soot resulting from an incomplete oil combus-
tion. (See Table 7 and Fig. 19.)
FIG. 19—Ignition result GOX test: 34 2 mm pipe, contaminated with: 1.5 g

aluminum swarf þ 1.6 g oil, investigation of the black deposit.
FIG. 20—Conditions for the oil ignition depending on particle type, oxygen
pressure, and pipe temperature, first ignition if there are several.
FIG. 21—Probability of an oil/aluminum ignition at approximately 150 C

depending on particle type.
Summary and Conclusions

– The ignition of an aluminum pipe is promoted by the simultaneous pres-
ence of a particle and oil contamination at elevated temperatures
(>95 C). If one factor is missing aluminum ignition becomes very
unlikely.
– Only one of 24 tests oil combusted during pressurization and fracturing
at 150 C demonstrating a relatively high ignition resistance provided
particles are missing. (See Fig. 21.)
– The presence of a particle contamination enhances the probability of an
oil ignition dramatically caused by a suspected catalytic property of the
particles. Molecular sieve, perlite, aluminum fibers, and rust turned out
to be catalytically active.
– The ignition of the oil occurred often during pressurization of the pipes.
However oxygen pressure is a less dominating factor than temperature.
Oil ignitions were obtained above 4 bar gauge oxygen pressure (applied
minimum pressure), but not below 95 C. (See Fig. 20.) Especially an
“adiabatic compression” mechanism is capable of raising the tempera-
ture remarkably; for example, after repeated starts of a compressor. In
this context, the thick wall of a pipe can remain cold in contrast to the
gas phase and loosely attached flammable debris with much lower heat
capacities.
– Oil contaminated, abrasive quartz sand caused a heavy burnout and ero-
sion at the fracture site due to a particle impact ignition mechanism.
Simultaneously, liquid aluminum was ejected and white fine Al2O3
smoke was generated.
– No oil/aluminum ignition was observed when the specimen was not
heated. Thus, during the LOX tests the oil and particle contaminated pipes
could be hydraulically fractured at 150 bar without any further reaction.
– Precondition of an aluminum ignition is the presence of an organic con-
taminant with a low auto ignition temperature. Such a promoter is
required to reach approximately 2000 C for a subsequent aluminum
ignition. This temperature could not be attained by a simple fracture or a
particle impact in absence of a burning hydrocarbon.
– The assumed kindling chain is: (1) Catalytic ignition of the oil; (2) burn-
ing oil ignites/heats the aluminum; (3) pipe bursts under ejection of mol-
ten aluminum.
– Also dry insects/organic debris can easily ignite in oxygen at elevated
temperatures (>120 C) and should be regarded as flammable contami-
nants. However, an ignition of massive aluminum caused by burning sin-
gle insects is unlikely due to their low weight and low heat of
combustion. The situation is different for organic debris adjacent to
polymers like gaskets in cylinder filling industry.
– The larger the pipe diameter the higher were the attained wall tempera-
tures after pressurization. Result: DT ¼ 30 C 40 C for 34 2 mm pipe
after reaching 150 bar.
In conclusion no oxygen pressure safety limit for the application of aluminum
in air separation industry can be stated. The actual temperature was a more crit-
ical factor, provided an oil and particle contamination was present.
Oil migration followed by an accumulation of oil pose a severe hazard. In
general, flammable organics in combination with particulate contaminants
enhance the risk of oxygen systems made of aluminum substantially and have
to be minimized as far as possible.
References
[1] German Statutory Accident Insurance for Chemical Industry, Oxygen, ed. 6,
Booklet: M034e, Jedermann Verlag, Heidelberg, Germany, 2010, p. 61.
[2] Pedley, M. D., Pao, J.-H., Bamford, L., Williams, R. E., and Plante, B.,
“Ignition of Contaminants by Impact of High-Pressure Oxygen,” Flam-
mability and Sensitivity of Materials in Oxygen-Enriched Atmospheres:
Vol. 3, ASTM STP 986, D. W. Schroll, ed., ASTM International, West
Conshohocken, PA, 1987, pp. 305–317.
[3] IGC Document 145/08, globally harmonized, “Safe Use of Brazed Alu-
minium Heat Exchangers for Producing Pressurized Oxygen,” Booklet:
Air Separation, CGA G-4.9, European Industrial Gases Association, Brus-
sels, Belgium, Chapter 11: Design and Operational Considerations.
[4] This is a Air Separation Standard in Form of a Booklet With 10–20
authors. IGC Document 147/08, globally harmonized, “Safe Practice
Guide for Cryogenic Air Separation Plants,” based upon CGA P-8, Euro-
pean Industrial Gases Association, Brussels, Belgium, Chapter 6: Intake
Air Quality.
[5] Werley, B. L., Barthelemy, H. M., Gates, R., Slusser, J. W., Wilson, K.
B., and Zawierucha, R., “A Critical Review of Flammability Data for
Aluminum,” Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres: Vol. 6, ASTM STP 1197, Joel M. Stoltzfus and
Dwight Janoff, eds., ASTM International, West Conshohocken, PA, 1993,
pp. 300–345.
[6] Binder, Chr., Kasch, T., Brock, T., Beck, U., Weise, M., and Sahre, M.,
“Promoted Combustion Testing of Pure and Ceramic-Coated Metals in
High Pressure Oxygen by the Federal Institute of Materials Research and
Testing (BAM),” J. ASTM Int., Vol. 3, No. 4, Paper ID JAI13533.
[7] Millon, J. F., Zawierucha, R., and Samant, A. V., “Promoted Ignition-
Combustion Behavior of Tubular Aluminum Samples in Liquid and Gase-
ous Oxygen Environments,” Flammability and Sensitivity of Materials in
Oxygen-Enriched Atmospheres: Vol. 10, ASTM STP 1454, T. A. Stein-

berg, H. D. Beeson, and B. E. Newton, eds., ASTM International, West
[8] Barragan, M., Wilson, D. B., and Stoltzfus, J. M., “Adiabatic Compres-
sion of Oxygen: Real Fluid Temperatures,” Flammability and Sensitivity
of Materials in Oxygen-Enriched Atmospheres: Vol. 9, ASTM STP 1395,
T. A. Steinberg, B. E. Newton, and H. D. Beeson, eds., ASTM Interna-
tional, West Conshohocken, PA, 2000, pp. 256–265.
[9] Newton, B. E. and Steinberg, T. A., “Adiabatic Compression Testing-Part
I: Historical Development and Evaluation of Fluid Dynamic Processes
Including Shock-Wave Considerations,” J. ASTM Int., Vol. 6, No. 8, doi:
10.1520/JAI102304.
[10] Newton, B. E., Chiffoleau, G. J. A., Steinberg, T. A., and Binder, C.,
“Adiabatic Compression Testing-Part II: Background and Approach to
Estimating Severity of Test Methodology,” J. ASTM Int., Vol. 6, No. 8,
doi: 10.1520/JAI102297.
STP 1561, 2012
DOI:10.1520/STP20120004
T. Tillack,1 C. Binder,1 T. Brock,1 N. Treisch,1

and P. Dielforder1
Fire Without Cause of an Oxygen Component

With an Aluminum Seal
REFERENCE: Tillack, T., Binder, C., Brock, T., Treisch, N., and Dielforder, P.,
“Fire Without Cause of an Oxygen Component With an Aluminum Seal,”
on September 19–21, 2012 in Montreal, QC; STP 1561, Samuel Edgar Davis
and Theodore A. Steinberg, Editors, pp. 105–116, doi:10.1520/STP20120004,
ASTM International, West Conshohocken, PA 2012.
ABSTRACT: It is common, even for high pressure service, for most of the
components in oxygen systems to contain nonmetallic materials. The risk of
an oxygen fire is low if the design of the component and its metallic and non-
metallic materials are properly selected. From a safety point of view, it is
obvious that a component made completely of metal with metallic seals
should have a lower risk of fire. However, during a test series performed at
the outdoor test site of the German Federal Institute for Materials Research
and Testing (BAM), “BAM Test Site Technical Safety,” an oxygen component
burned out that contained metallic seals only. In the test, a bursting metallic
diaphragm produces a single pressure shock at a specified oxygen pressure.
In the incident, an aluminum diaphragm with a thickness of 1.5 mm was used
at a burst pressure of 171 bar. At first, investigations at BAM focused on the
question of whether a bursting aluminum diaphragm is able to ignite a bulk
metal. Stress rupture tests were performed on this particular aluminum. A
high-speed infrared camera helped reveal information on the temperature
rise in the rupture zone. Furthermore, other possible ignition sources such as
contamination or flow friction were considered, also. This paper discusses
the results of the investigation, and it tries to answer the question of whether
aluminum sheet materials increase the fire risk in oxygen components.
KEYWORDS: aluminum, high-pressure oxygen, bursting diaphragm, metallic
material, fire risk, ignition source, accident research, safety engineering
November 2012.
1
Specialists of Working Group “Safe Handling of Oxygen,” Division 2.1, “Gases, Gas Plants,”
Federal Institute for Materials Research and Testing (BAM), Berlin, Germany 12205.
Copyright V
19428-2959.
105
Introduction
To ensure the burn-out safety of components in oxygen systems, only certain
nonmetallic and metallic materials that are suitable for the operating conditions
should be used. The design of the component has a strong influence on the
burn-out safety, too.
From a safety point of view, it should be obvious that a component com-
pletely made of and sealed with metallic materials should have a lower risk of
causing a fire or a burn-out than a component with nonmetallic sealing materials.
However, metallic materials also can be ranked regarding their compatibility
with oxygen.
In particular, it is well known that the flammability of aluminum in oxygen
is much higher than that of copper or Monel [1,2]. The unique burning behav-
ior of aluminum was often part of investigations in the past [3–11]. It was
found that a different physical model is needed to describe a burn-out with alu-
minum and oxygen than what can be used for other metallic materials such as
stainless steel [12–16].
In general, fuel, an ignition source, and oxygen have to be present in order
for a fire to occur. Each of these main components is affected by additional fac-
tors, which Slusser et al. [8] implicated in a so-called “extended fire triangle,”
shown in Fig. 1.
During a test series performed at the outdoor test site of the German Fed-
eral Institute for Materials Research and Testing (BAM), an oxygen component
burned out that was made of and sealed with metallic materials only. This com-
ponent contained a bursting metallic diaphragm that produced a single oxygen
pressure shock.
Two of the three sides of the fire triangle are easily detectable in this
incident:
• high-pressure oxygen as an oxidizer, and
• metallic materials as potential fuel.
FIG. 1—Fire triangle.

TILLACK ET AL., doi:10.1520/STP20120004 107
However, at first glance, there is no obvious ignition source. Therefore,

investigations focused on finding out what kind of ignition source caused the
incident.
Test Set-up and Description of the Incident

The BAM Technical Safety Test Site (BAM TTS) is a unique open-air test fa-
cility in Europe. In many cases, laboratory test results or numerical modeling is
not enough to describe the possible risks associated with materials, composi-
tions, or technical components. Often, it is necessary to compare laboratory
test results with data from real-world and large-scale tests. Therefore, special
test facilities were built at BAM TTS for testing and evaluating materials and
components for oxygen service. Figure 2 presents an overview of the oxygen
supply system and the test building. A piping system 70 m in length with pipe
diameters between 12.7 mm (1/2 in.) and 50.8 mm (2 in.) allows tests under
flowing oxygen conditions. This is shown in Fig. 3.
In a research program, to test the burn-out safety of large-scale compo-
nents, an oxygen component is used that produces a single oxygen pressure
shock by means of a bursting diaphragm. In a first step, bursting diaphragms of
different metallic materials and different thicknesses were tested in order to
determine their burst pressure. In this step, only the bursting diaphragms were
tested, without any additional oxygen components. For statistical purposes,
each of these experiments was performed ten times. The assembly drawing of
the diaphragm holder is shown in Fig. 4.
During the test series with an aluminum diaphragm, an incident happened:
a big bang, accompanied by a fireball, indicated a severe reaction with oxygen.
As a consequence, the main part of the diaphragm holder showed typical signs
of metal combustion. The aluminum diaphragm was completely burnt. The
protective wall clearly indicated the pattern of the burning metal. The counter-
part of the diaphragm holder was found nearly 70 m away, stopped only by the
safety fence. As a consequence of flying pieces of hot and glowing metallic
FIG. 2—Oxygen test site (overview).

FIG. 3—Piping system.
FIG. 4—Assembly drawing of the diaphragm holder; all parts but diaphragm
made of stainless steel.
material, there were several small grass fires. Figures 5–9 show the extent of
the incident. The test parameters are given in Table 1.
Investigations
In general, possible ignition sources were contamination, flow friction, heat,
external fires, igniters, etc. A time-consuming investigation was started, but
there was no indication that one of the mentioned possibilities had caused the
incident.
Assembly of the degreased aluminum diaphragm had always been per-
formed with the operator wearing clean white gloves. Neither particles nor rust
or similar particulates were found in the tube or on the original diaphragms.
Both the oxygen and the holder with the diaphragm were at ambient tempera-
ture and were not heated.
Another possible cause of the fire could have been fragments of the ruptured
aluminum disk. If pieces had impacted the stainless steel test holder, subsequent
FIG. 5—Protective wall.
ignition of the holder assembly might have occurred. However, this is very
unlikely, as the holder has a length of only 0.05 m. Nevertheless, the aluminum
disk very often produced light flashes at the moment of bursting. Figure 10
shows this.
In this test, all four pre-notched segments of the diaphragm just opened but
did not detach, as Fig. 11 depicts. However, aluminum diaphragms also rup-
tured in such a way that the pre-notched segments partially or completely
detached (see Fig. 12).
In an effort to avoid possible ignition via aluminum particle impact, the
stainless steel holder was shortened for future tests. Therefore, further investi-
gation focused on the question of whether the bursting aluminum diaphragm
FIG. 6—Diaphragm holder.
FIG. 7—Diaphragm holder.

FIG. 8—Diaphragm holder, counterpart.
itself was able to produce sufficient energy to ignite a bulk metal. For stress
rupture tests, a standardized tension profile was manufactured on this particular
aluminum.
Commonly, the stress rupture test is a standardized destructive material
test. It helps measure material parameters such as the yield stress, tensile
strength, (breaking) elongation, and coefficient of elasticity. In this investiga-
tion, the stress rupture test was performed to specify the temperature rise in the
rupture zone. A high-speed infrared (IR) camera was used for contactless mea-
surement of the temperature.
Figure 13 shows the material behavior over three time steps. The elonga-
tion results in an area of heating located at the middle of the profile. Shortly
before rupture, the temperature is located in a smaller boundary layer. After
rupture, the high temperature is found at the sharp edges of both profile parts.
Several tests were performed in which the frame rate, the resolution, and
the temperature range of the camera were varied. Even if a high-speed IR cam-
era is used, the pictures deliver mean temperature values over the recording
time and over the pixel size only. Thus it is not surprising that the measured
FIG. 9—Flight range.
TABLE 1—Test parameters in the incident.
Material Aluminum
Tube diameter, mm 76.1 12.5 (DN50)
Diaphragm diameter, mm 80.0
Diaphragm thickness, mm 1.5
Oxygen content, vol. % 99.5
Burst pressure, bar 171
FIG. 10—Bursting of an aluminum diaphragm with light flash; pictures taken

from a video sequence.
FIG. 11—Ruptured diaphragm.
FIG. 12—Examples of diaphragms with partially and completely detached

pre-notched segments.
FIG. 13—Temperature rise in the rupture zone.

maximum temperature at the fracture zone is only 130 C. However, according

to the knowledge of specialists in the stress rupture test, this mean temperature
value is followed by peak temperatures of several hundred degrees Celsius at
the sharp edges.

The technical community knows that aluminum has a limited field of applica-
tion in high-pressure oxygen service. In addition, the incident described herein
shows that even for burst disks and metallic diaphragms, the use of aluminum
as a material has to be reconsidered.
In the past, incidents happened, e.g., in air separation plants or in oxygen
vaporizers, in which the root causes were not clear. These incidents and the
one presented in this paper have in common the use of aluminum.
The investigation of the incident described is still in progress. It is intended
to measure the local maximum temperature at the sharp edges of the bursting
diaphragm more accurately. Several tests that have not led to the same severe
reaction as in this incident show distinct light flashes at the moment when the
aluminum diaphragm bursts. This might be the source of ignition the authors of
this paper are looking for. In that case, all conditions of the three sides of the
fire triangle would be fulfilled, and a fire would be possible.
The intention of this paper is to reveal that with use in high-pressure oxy-
gen, even components or equipment made entirely of metal could pose a fire
risk. This has been the case in testing at BAM’s outdoor test site with a burst-
ing aluminum diaphragm. However, further investigations are necessary to ver-
ify that the maximum temperature/energy at the sharp edges of a bursting
diaphragm is high enough to ignite bulk material.
References
[1] Zabrenski, J. S., Werley, B. L., and Slusser, J. W., “Pressurized Flammabil-
ity Limits of Metals,” Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres, Vol. 4, ASTM STP 1040, J. M. Stolzfus, F. J.
Benz, and J. S. Stradling, Eds., ASTM International, West Conshohocken,
PA, 1989.
[2] Stoltzfus, J. M., Homa, J. M., Williams, R. E., and Benz, J. B., “ASTM
Committee G-4 Metals Flammability Test Program: Data and Dis-
cussion,” Flammability and Sensitivity of Materials in Oxygen-Enriched
Atmospheres, Vol. 4, ASTM STP 986, D. W. Schroll, Ed., ASTM Interna-
tional, West Conshohocken, PA, 1988.
[3] Chiffoleau, G., Newton, B., Holroyd, N. J. H., and Havercroft, S.,
“Surface Ignition of Aluminum in Oxygen,” Flammability and Sensitivity
of Materials in Oxygen-Enriched Atmospheres, Vol. 11, ASTM STP

1479, D. Hirsch, R. Zawierucha, T. Steinberg, and H. Barthélémy, Eds.,
ASTM International, West Conshohocken, PA, 2006.
[4] Tillack, T., Binder, C., Brock, T., Beck, U., Weise, M., and Sahre, M.,
“Promoted Combustion Testing of Pure and Ceramic-Coated Metals in
High Pressure Oxygen by the Federal Institute for Materials Research and
Testing (BAM),” Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres, Vol. 11, ASTM STP 1479, D. Hirsch, R. Zawieru-
cha, T. Steinberg, and H. Barthélémy, Eds., ASTM International, West
Conshohocken, PA, 2006.
[5] Tillack, T., Binder, C., Brock, T., Beck, U., Weise, M., and Sahre, M.,
“Combustion Tests under High Pressure Oxygen: Promoted Ignition
Combustion Test versus Metallic Disk Ignition Test,” Flammability and
Sensitivity of Materials in Oxygen-Enriched Atmospheres, Vol. 11,
ASTM STP 1479, D. Hirsch, R. Zawierucha, T. Steinberg, and H. Barthé-
lémy, Eds., ASTM International, West Conshohocken, PA, 2006.
[6] Barthélémy, H., “Oxygen Fires in Aluminum Alloy Cylinders,” Flamma-
bility and Sensitivity of Materials in Oxygen-Enriched Atmospheres, Vol.
11, ASTM STP 1479, D. Hirsch, R. Zawierucha, T. Steinberg, and H.
Barthélémy, Eds., ASTM International, West Conshohocken, PA, 2006.
[7] Fano, E., Faupin, A., and Barthélémy, H., “Selection of Metal for Oxygen
Valves,” Flammability and Sensitivity of Materials in Oxygen-Enriched
Atmospheres, Vol. 9, ASTM STP 1395, T. Steinberg, B. Newton, and H.
Beeson, Eds., ASTM International, West Conshohocken, PA, 2000.
[8] Slusser, J. W. and Miller, K. A., “Selection of Metals for Gaseous Oxygen
Service,” Flammability and Sensitivity of Materials in Oxygen-Enriched
Atmospheres, Vol. 1, ASTM STP 812, B. L. Werley, Ed., ASTM Interna-
[9] Werley, B. L., Barthélémy, H., Gates, R., Slusser, J. W., Wilson, K. B.,
and Zawierucha, R., “A Critical Review of Flammability Data for Alumi-
num,” Flammability and Sensitivity of Materials in Oxygen-Enriched
Atmospheres, Vol. 6, ASTM STP 1197, D. D. Janoff and J. M. Stolzfus,
[10] Benning, M. A., Zabrenski, J. S., and Le, N. B., “The Flammability of
Aluminum Alloys and Aluminum Bronzes as Measured by Pressurized
Oxygen Index,” Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres, Vol. 4, ASTM STP 986, D. W. Schroll, Ed.,
[11] McIlroy, K., Zawierucha, R., and Drnevich, R. F., “Promoted Ignition
Behavior of Engineering Alloys in High-Pressure Oxygen,” Flammability
and Sensitivity of Materials in Oxygen-Enriched Atmospheres, Vol. 4,
ASTM STP 986, D. W. Schroll, Ed., ASTM International, West Consho-
hocken, PA, 1988.
[12] Wilson, D. B. and Stoltzfus, J. M., “Modeling of Promoted-Ignition Burn-

ing: Aluminum,” Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres, Vol. 8, ASTM STP 1319, W. T. Royals, T. C.
Chou, and T. Steinberg, Eds., ASTM International, West Conshohocken,
PA, 1997.
[13] Wit, J. R., De, W., Steinberg, T., and Haas, J. P., “ASTM G124 Test Data
For Selected Al-Si Alloys, Al-Composites, Binary Alloys and Stainless
Steel,” Flammability and Sensitivity of Materials in Oxygen-Enriched
Atmospheres, Vol. 9, ASTM STP 1395, T. Steinberg, B. Newton, and H.
Beeson, Eds., ASTM International, West Conshohocken, PA, 2000.
[14] Lowrie, R., “Oxygen Compatibility of Metals and Alloys,” Flammability
and Sensitivity of Materials in Oxygen-Enriched Atmospheres, Vol. 6,
ASTM STP 1319, D. D. Janoff and J. M. Stolzfus, Eds., ASTM Interna-
[15] Sato, J. and Hirano, T., “Behavior of Fire Spreading Along High-
Temperature Mild Steel and Aluminum Cylinders in Oxygen,” Flamma-
bility and Sensitivity of Materials in Oxygen-Enriched Atmospheres, Vol.
2, ASTM STP 986, M. A. Benning, Ed., ASTM International, West Con-
shohocken, PA, 1986.
[16] Benz, F. J., Shaw, R. C., and Homa, J. M., “Burn Propagation Rates of
Metals and Alloys in Gaseous Oxygen,” Flammability and Sensitivity of
Materials in Oxygen-Enriched Atmospheres, Vol. 2, ASTM STP 986, M.
A. Benning, Ed., ASTM International, West Conshohocken, PA, 1986.
STP 1561, 2012
DOI:10.1520/STP20120006
Frédéric Crayssac,1 Jean Christophe Rostaing,1

Etienne Werlen,2 Lianming Sun,3 Patrick Houghton,4
William Kleinberg,4 and Frédéric Coste5
Pressure Dependence of Aluminum Ignition

in Gaseous Oxygen and Possible Ignition
Mechanisms in Brazed Aluminum Heat
Exchangers
REFERENCE: Crayssac, Frédéric, Rostaing, Jean Christophe, Werlen, Eti-

enne, Sun, Lianming, Houghton, Patrick, Kleinberg, William, and Coste,
Frédéric, “Pressure Dependence of Aluminum Ignition in Gaseous Oxygen
and Possible Ignition Mechanisms in Brazed Aluminum Heat Exchangers,”
ABSTRACT: The ignition of aluminum foils in gaseous oxygen was experi-
mentally tested using a diode laser as the energy source, which provided a
well-controlled, accurate, and reproducible method of ignition. The tests were
conducted under different conditions in terms of oxygen pressure, oxygen pu-
rity, aluminum thickness, and gas velocity. The aluminum foils tested were
between 0.2 mm and 0.45 mm thick, a range typical of fins contained in
brazed aluminum heat exchangers (BAHXs) used in air separation units
(ASUs). The experimental apparatus was composed of a pressure vessel in
which a single aluminum test sample was placed. The vessel contained an
optical window that allowed a short laser pulse of known power to be applied
to the aluminum sample. The energy dose was systematically varied in order
to identify the threshold ignition energy, defined as the point at which the
Manuscript received January 31, 2012; accepted for publication June 21, 2012; published online
November 2012.
1
CRCD, Air Liquide R&D, Les-Loges-en-Josas 78350, France.
2
Air Liquide E&C Cryogenics Standard Plants, Vitry-sur-Seine 94781, France.
3
Air Liquide Engineering, Champigny-sur-Marne 94503, France.
4
Air Products and Chemicals, Inc., Allentown, PA 18195-1501.
5
PIMM Laboratory CNRS, Arts & Métiers ParisTech, Paris 75013, France.
Copyright V
19428-2959.
117
probability of aluminum combustion with propagation beyond the laser spot

was 50 %. The experimental results show that O2 pressure has no significant
effect on the ignition energy of aluminum over the pressure range tested (10
bar to 120 bar). This conclusion holds for both standard commercial grade
purity O2 (99.8 %) and high purity O2 (99.99 %), as well as for gas velocities
higher than typically encountered in ASU BAHXs. Heat conduction calcula-
tions indicate that aluminum ignition occurs when the laser spot temperature
reaches the melting point of the passivating oxide layer (about 2200 K to
2300 K). The heat conduction model accurately explains the dependence of
the ignition energy on the aluminum sample thickness. These test results
have been used in assessing the risk of ignition of BAHXs used in high pres-
sure oxygen service in ASUs.
KEYWORDS: oxygen, ignition, flammability, aluminum, air separation unit
Introduction
Brazed aluminum heat exchangers (BAHXs) have been used to boil liquid oxy-
gen and to warm gaseous oxygen (GOX) from cryogenic temperatures to ambi-
ent temperature for the past 50 years. Prior to the late 1980s, the processing of
oxygen was done in two separate heat exchangers: one to boil the oxygen, and
one to warm the oxygen to ambient temperature. Typically, this was carried
out at a pressure of less than 1.5 bar.
There have been no industrial accidents reported in BAHXs used to warm
GOX from cryogenic temperatures to ambient temperature. A number of inci-
dents that have been reported involved BAHX reboilers, as shown by a survey
of industrial gas companies in 1998 [1]. In some of these incidents there was
evidence of a small amount of aluminum combustion, and there were three
major incidents that involved a breach of the cold box [1,2]. Investigations
showed that the most common cause was the widespread accumulation of
hydrocarbons, and that combustion of these hydrocarbons could in turn lead to
aluminum combustion. Design and operating practices have been defined so as
to avoid recurrences of these incidences [1].
Since the late 1980s, BAHXs have been used such that oxygen is boiled
and warmed in a single BAHX. In this case, the oxygen is typically pumped to
the required pressure so that oxygen compression can be eliminated. The air
separation industry has built up extensive experience (>4000 plant years) in
operating heat exchangers in this manner. A significant portion of this experi-
ence has been obtained with pressures in the range of 30 bar to 60 bar. Operat-
ing pressures for new plants are now exceeding 100 bar.
To date, there has been one known incident involving this type of
exchanger. This incident occurred in March of 2011 and involved an exchanger
designed to warm O2 at approximately 80 bar. Ignition occurred in the cold
end of the exchanger at the point where liquid oxygen (LOX) enters the
BAHX. The resulting energy release breached the cold box, but damage exter-
nal to the cold box was minimal. There were no injuries and there was no
measurable pressure wave created by the release.
CRAYSSAC ET AL., doi:10.1520/STP20120006 119
A significant amount of work on the testing of aluminum flammability can

be found in the literature, in most cases related to studies of combustion and
propagation [3,4]. It has been demonstrated that combustion and propagation
of aluminum fins can occur at atmospheric pressure. An aluminum–oxygen
reaction in a BAHX operating at an elevated oxygen pressure will likely lead
to a very rapid and violent reaction. The reaction rate in GOX increases with
increasing O2 pressure [5].
The safe operation of a BAHX used to boil and/or warm oxygen at elevated
pressure relies on the absence of ignition mechanisms in the process. This has
led to recognition that a better understanding of the ignition of aluminum at
elevated pressure is necessary if one wishes to avoid serious incidents with
BAHXs operating at elevated oxygen pressures. Specifically, it is necessary to
understand whether the energy required for aluminum ignition changes at ele-
vated oxygen pressure and whether the energy released from an ignition mech-
anism changes at higher operating pressure.
There are no published results on the variation of the ignition energy of
aluminum with pressure. The main purpose of this study was to determine the
energy required for the ignition of aluminum in oxygen at various pressures,
O2 purity levels, O2 velocities, and aluminum thicknesses, using a diode laser
as an energy source.
In this paper, the most likely ignition mechanisms in a BAHX used to boil
and/or warm oxygen are discussed, along with the dependence of each mecha-
nism on the oxygen pressure.
Experimental Testing
Description of Testing Bench

The testing apparatus was designed to create a transient hot spot of small,
nearly constant diameter by focusing a high power laser beam on an aluminum
foil placed in a pressurized oxygen atmosphere. This highly localized, con-
trolled (by adjusting power or duration) heating is supposed to be representa-
tive (for the current purposes) of a real life ignition cause, such as particles
impinging on an aluminum fin with a very high initial velocity or already burn-
ing in the gas phase. With this apparatus, it is possible to determine the mini-
mum amount of thermal energy that needs to be deposited on the hot spot in
order to ensure the ignition of different aluminum samples as a function of
pressure (up to 120 bar), accounting for the effects of other parameters such as
oxygen purity and local flow velocity in the vicinity of the sample.
The test setup is composed of a pressure vessel containing optical view-
ports for laser beam passage and light emission sampling (glass windows),
inside which the aluminum sample is placed on a removable support
(see Fig. 1). The aluminum sheet, in the form of a square coupon, is 10
mm 10 mm in size with a range of thicknesses representative of fins in a real
BAHX (0.20 mm, 0.35 mm, and 0.45 mm). Test samples were cut from real
aluminum sheets intended for BAHX manufacturing that were always stored in
a dry and clean atmosphere and which were thoroughly degreased. The sample
surface condition was checked for a few samples taken from different original
lots using x-ray photoelectron spectroscopy and two nuclear profiling techni-
ques: Rutherford backscattering and elastic recoil detection analysis. As
expected, the oxide superficial layers corresponded to the spontaneously
formed clean, anhydrous oxide. The average thickness was estimated as 80 nm.
An oxygen feedstock and manifold panel allow one to pressurize the vessel
and optionally sustain a localized tangential gas flow. This is accomplished by
impinging a calibrated gas jet on the sample edge through a 4 mm diameter
channel within the sample holder arm. In such a blowing regime, the flow is
regulated by a mass flow controller, and the pressure is controlled and main-
tained using a back-pressure regulator that discharges the excess gas to a vent-
ing outlet. The aluminum sample is held in place on its support by means of
screwed clamps.
The mechanical contact of sample edges with the high-thermal-conductiv-
ity copper support provides a kind of heat sink boundary condition that to some
extent approaches the thermal behavior of a real aluminum fin (extended struc-
ture) in a BAHX.
The laser diode power source module emits two main spectral lines cen-
tered on 931 nm and 972 nm. At such wavelengths, optical absorption by GOX
under pressure is negligible.
Photometric and Optical Diagnostics for Laser Beam and Sample Burning
The testing vessel is equipped with three photodiodes and a high speed camera
with light sampling in the line of sight of the laser beam entrance window or
FIG. 1—Schematic of test bench.

the viewports (see Fig. 2). Photodiodes are appropriately positioned with spe-
cific optical filters to detect the thermal optical emission in the visible spectrum
that is the signature of the aluminum sample’s combustion, and the presence of
the incident laser beam enables one to know exactly the time origin of the sam-
ple’s irradiation. The high speed camera focused on the back (un-irradiated)
side of the sample provides visible evidence of the burning event and can re-
cord up to 4000 frames per second.
Calibration Method and Laser Beam Characterization

Important parameters to characterize in order to obtain accurate control of the
laser energy input are the laser spot diameter, the laser incident power, and the
power absorption by the aluminum sample.
The axial position of the focal spot (i.e., of minimum size) was determined,
and its dimension was quantified as a function of pressure. Thus, it was possi-
ble to control the distance from the beam output of the laser optical head to the
target so as to maintain reproducible beam impact conditions regardless of
small gas refractive index variations with density.
The spot diameter was in all cases close to 3 mm. It is important to note
that no attempt at stronger focusing was made, in order to avoid the potential
occurrence of the direct ablation of matter due to the laser radiation pressure,
as indicated by laser physics theory. Indeed, this physical effect of non-thermal
origin is specific to the laser (very high pressure of the photon gas), and no evi-
dent parallel can be made to real-life ignition mechanisms considered for a
BAHX. The absence of ablation is proven, as is shown later, by the fact that all
results can be interpreted within a purely thermal model, with no singularity
that would have been indicative of a non-thermal contribution to breaching of
the passivating oxide layer.
In order to retain precise control of the amount of available incident laser
power impinging on tested samples, the actual delivered laser power was meas-
ured by a calorimeter. The laser beam enters an almost integral absorber in the
form of a hollow cone-shaped matte-black-painted structure. The generated
heat is in turn transferred to a water circuit, and the differential temperature
FIG. 2—Photometric measurements.

measurement readily yields the absolute incident power value. The correspon-
dence between the setpoint of the laser power and measured values was
rechecked periodically. Power values were measured with the laser beam en-
trance window both in place and removed, allowing the corresponding beam
attenuation to be determined.
The optical absorption coefficient of the sample (in other terms, the ther-
mal power input used to create the hot spot) was determined differentially by
measuring the reflected laser power. Because reflection can be partially non-
normal (because of the geometry) or diffuse (in general, the sample does not
have a specular surface), an integrating sphere was used. A first laser shot for
reference on a mirror (reflection coefficient greater than 99 %) yielded a first
value of reflected energy R1. Then a second laser shot on the sample to be
tested under a forced flow of oxygen resulted in a second value, R2. The
absorption coefficient was calculated as Abs ¼ (1 R2/R1). Figure 3 shows the
configuration used for the absorption measurements. Note that it was not possi-
ble to determine the sample absorption coefficient once the aluminum sample
started burning because thermal optical emissions at the laser wavelength were
superposed with the reflected beam signal. Practically, this was not a serious li-
mitation, because generally there was little overlap of the irradiation and com-
bustion regimes.
A value of 0.20 was obtained for the absorption coefficient, with an uncer-
tainty of about 65 %. One possible explanation for this somewhat large uncer-
tainty is that the absorption coefficient of aluminum is sensitive to the surface
conditioning of the sample.
FIG. 3—Absorption measurements.

FIG. 4—Different types of effects on samples caused by laser irradiation.
Determination of Threshold Ignition Energies

In order to determine the threshold energies for self-sustained combustion
under different operating conditions (pressure, oxygen purity, oxygen flow,
sample thickness), it was necessary to obtain positive test results (meaning at
least partial combustion of sample) and negative ones (without observable
combustion of sample) as a function of the amount of thermal energy deposited
on the sample.
A positive ignition of a sample corresponds to a self-sustained combustion
that propagates beyond the zone affected thermally by laser light absorption
(roughly beyond the laser beam spot boundary), with either total or partial
destruction (see Fig. 4).
A sufficient number of tests under the same operating conditions were con-
ducted in order to determine the significant statistical fraction of positive
results for each energy level. At least five positive results and five negative
results were required in order for the threshold energies to be defined.
In practice, to ensure reasonably fast convergence to a threshold energy
value with a probability for ignition of 50 % (Ethreshold), the test sequence was
conducted following an “up and down” method with a power variation step of
100 W. The principle is that the next testing energy value is determined by the
result of the current test. For example, if the result of the test just completed is
positive, the next test is performed with a negative power increment of 100 W.
Conversely, if the current result is negative, the next test is performed with a
positive increment of 100 W.
The experimental points (the percentage of positive results as a function of
deposited energy) were first fit to a logistical function, and the threshold values
were then estimated at 50 % positive tests (see Fig. 6).
For each set of tests, the pulse duration and beam diameter were fixed, and
the deposited energy was varied by changing the incident laser power. This
method ensured that energy losses from the laser spot due to conduction
remained consistent as the energy dose was varied; this is discussed in more
detail later.
Peak Initial Time and Duration Extracted from Photodiode Measurements

Recorded photodiode signals allowed the assessment of the occurrence and du-
ration of combustion, as the strong thermal optical emission peak featured very
abrupt rising and decaying fronts. The beginning and ending of combustion
were defined as the points at which the respective fronts attained 20 % of the
maximum peak intensity. The duration of combustion was extracted from the
difference between these two points, and the absorbed energy required for igni-
tion could be estimated from the time between the start of the laser pulse and
the onset of combustion (see Fig. 5). An estimate of the combustion propaga-
tion “velocity” could be obtained as well.
Experimental Test Results
Overview of Experimental Data

All the results for the threshold energies and the corresponding conditions of
experiments are summarized in Table 1. These tests covered a pressure range
of 10 bar to 120 bar, thicknesses typically used for fins in BAHXs (0.2 mm,
0.35 mm, and 0.45 mm), velocities similar to or higher than those encountered
in normal BAHX operations, and two different oxygen purities (standard grade
with 99.8 % oxygen and a very high purity with 99.99 % oxygen).
In addition to static tests (with no oxygen flow), experiments were done
with a flow velocity chosen to be inversely proportional to the square root of
the pressure. This was in accordance with the design rules recommended in
Ref 2. The goal is to maintain as constant the qV2 term, that is, the maximum
dynamic pressure. Note that in one case the velocity was increased by a factor
of 5 over the usual value in order to determine whether any effect on the igni-
tion energy could be detected at extremely high velocity.
Tests were performed for both standard grade oxygen and high purity oxy-
gen because it is known that argon, the main residual impurity in commercial
oxygen, can reduce the ignition probability and propagation rate of aluminum,
presumably by accumulating into a barrier layer impeding oxygen transport to-
ward freshly molten or vaporized metal.
The pulse durations used were fixed at the minimal values that produced
positive ignition events in every set of considered operational conditions. In
most cases, this corresponded to a selected value of 60 ms. Only in the case of
FIG. 5—Example of photodiode signals for a typical experiment.
CRAYSSAC ET AL., doi:10.1520/STP20120006
125
TABLE 1—Ignition tests on aluminum samples in gaseous oxygen.
Pressure, Pulse Duration, Thickness, Flow Rate, Purity, EThreshold, Number of

bar ms mm m=s % W Tests Performed
50.5 60 0.2 1.41 99.99 19.8 17
25 60 0.35 Static 99.8 30.8 8
30 60 0.35 Static 99.8 30.0 31
50 60 0.35 Static 99.8 30.5 27
120 60 0.35 Static 99.8 31.8 26
9.5 60 0.35 3.24 99.8 27.8 30
119.5 60 0.35 0.91 99.8 30.3 32
10.5 60 0.35 3.09 99.99 27.0 33
120.5 60 0.35 0.91 99.99 29.5 36
121 60 0.35 0.91 99.99 31.3 15a
50 60 0.35 7.07 99.99 30.0 14
49.5 90 0.35 Static 99.8 38.3 14
50 90 0.45 1.41 99.8 46.5 17
50 120 0.45 1.41 99.8 58.0 30
120 120 0.45 0.91 99.8 63.0 24
a
As an exception, this sample has dimensions of 10 mm 5 mm. The consequences are discussed in the text.
thicker samples was it necessary to increase the pulse duration to 90 ms and then
to 120 ms. The reason is simply that laser heating conditions for such samples
were less adiabatic, with increased thermal losses due to internal heat conduction.
Figure 6 shows several examples of the Ethreshold determination method at
various pressures under static and flowing conditions in the case of standard
grade O2 purity. The number next to each data point indicates the total number
of trials at that particular energy dosage, and the dashed lines correspond to the
fitting curves with a logistical function. Despite the fact that the number of data
points had to be reasonably limited, one can see that the trends in the evolution
of ignition probability are quite consistent.
Effects of Pressure
Figure 7 shows that the effects of pressure on the ignition threshold energy are
very small, and the conclusion remains the same for different flow velocities
and high oxygen purities.
Effects of Sample Thickness

From Fig. 8, it can be seen that the ignition threshold energy is almost doubled
when the sample thickness is increased from 0.35 mm to 0.45 mm. This is the
result of increased thermal losses due to heat conduction for the case of the
thicker (less adiabatic) sample. Despite the higher ignition energy, the impor-
tant fact is that the variation of the ignition energy with pressure remains small.
FIG. 6—Examples of determination of threshold energies at 50 % positive tests.

127
FIG. 7—Effects of pressure on threshold ignition energies for different velocities and O2 purities.
FIG. 8—Threshold energies for different sample thicknesses.

129
TABLE 2—Percentage of full destruction tests relative to the total number of positive ignition tests.
Test Conditions Pressure, bar
Flow Rates O2 Purity, % Thickness, mm 10 25 30 50 120

Static 99.8 0.35 0% 71% 80% 100% 100%
10=P0.5 99.8 0.35 67% 100%
10=P0.5 99.99 0.35 100% 100%
10=P0.5 99.8 0.45 100% 100%
Note that the testing of thicker samples had to be done at an increased pulse du-
ration; otherwise, ignition would not have been possible for a 0.45 mm thick
sample at the maximum available laser power.
Combustion Propagation
At this stage we have shown that the ignitability of thin aluminum sheets does
not depend on pressure. However, the combustion propagation rate is more
strongly influenced by the pressure, and complete sample destruction is more
and more likely as the pressure is increased, as shown in Table 2. Indeed, 100
% sample metal consumption is achieved at 50 bar and higher pressures, and
partial destruction only can be obtained at lower pressures and lower
velocities.
In Fig. 9, we see a very substantial increase of the propagation rate for the
small aluminum samples when the pressure increases from 10 bar to 50 bar.
However, the variation is much less pronounced between 50 bar and 120 bar;
namely, the increase in the propagation rate is only by a factor of 1.5. This is in
accordance with data presented in the literature [5–7].
Figure 9 also shows that both the flow velocity and the oxygen purity
increase the propagation rate. This can be explained by the fact that an increase
of the oxygen flow velocity enhances the supply of oxygen to the molten or
vaporized aluminum and also scavenges the accumulated argon boundary layer
in the case of 99.8 % oxygen. When argon is absent (99.99 %), combustion
propagation becomes, of course, even faster.
Heat Conduction Modeling

Previous laser ignition experiments [8] have indicated that aluminum ignites at
temperatures in the 2150–2250 K range, based on direct temperature measure-
ments obtained over a fairly wide pressure range (up to 68 bar). These ignition
temperatures are just below the melting point of the thin aluminum oxide layer
covering the aluminum surface (2318 K). The melting point of aluminum is
933.5 K, which means that at the point of ignition, molten aluminum exists
FIG. 9—Propagation rate (inversely proportional to the peak duration) as a function of pressure, flow rate, and O2 purity.
131
below the surface oxide layer. The physical picture of the ignition process is that
the aluminum oxidation rate rapidly accelerates when the oxide layer is breached
through either melting or mechanical means once the layer is close to the melting
point. At that point, the aluminum temperature is high enough that the heat
released during oxidation exceeds the rate of heat loss from the reaction zone,
and the reaction process speeds up rapidly as a result of this feedback. In the
analysis that follows, 2230 K is used as the temperature at which ignition occurs.
In our experiments, the aluminum is heated by the laser, and there are sev-
eral mechanisms of heat loss, including conduction, convection, and radiation.
We have found that the energy required for ignition using a short laser pulse
can be calculated with reasonable accuracy by considering only heat conduc-
tion away from the laser spot and ignoring losses due to heat convection and
radiation. To estimate the ignition energy quantitatively, a transient heat con-
duction problem was solved to determine the length of time required in order
for the aluminum surface under the laser spot to reach a centerline temperature
of 2230 K. The initial condition is uniform temperature with a constant heat
flux qo applied within a circular zone on the surface of the aluminum sample at
time zero. The problem’s solution was obtained numerically using a three-
dimensional (3-D) finite difference approach that is described in this section.
The sample geometry is defined in Fig. 10. The diameter of the laser beam
at the surface of the aluminum sample was maintained as relatively constant in
our experiments at a value of 3 mm. This was confirmed by measurements
taken as a function of laser power and pulse duration, as explained previously.
The coordinate system for the mathematical problem defines the X ¼ 0 plane as
the “back” side, facing away from the laser, and the laser impinges on the sam-
ple in the plane defined by X ¼ d, where d is the sample thickness. The problem
was solved in only one quadrant [Y > 0 and Z > 0, the upper right-hand quad-
rant in Fig. 10(a)] because of symmetry.
The fraction of the laser energy reflected by the aluminum sample was
measured and determined to be about 80 % (with an absorption coefficient of
0.20). In addition, the fraction of the laser energy that was either reflected or
absorbed by the chamber window was measured and determined to be 5.5 %.
The remainder of the laser energy is absorbed by the sample and provides the
energy for ignition.
The mathematical problem in dimensionless form is given by

@H @ qx 2 @ qy @ qz
¼ þe þ (1)
@t @x @y @z
where:
H ¼ enthalpy,
q ¼ heat flux, and
e ¼ ratio of the aluminum sample thickness to the sample width, d/b.
FIG. 10—Sample geometry: (a) front view (left) and (b) side view.
133
An enthalpy formulation is used to make it easier to incorporate melting of

the aluminum sample. The numerical procedure is to evaluate the x, y, and z
heat fluxes for each grid point using a second order accurate central difference
formula. Then an explicit time step is taken in enthalpy using Eq 1 to get an
updated enthalpy field, starting from a known initial temperature and enthalpy
field. The temperature field at the new time is then evaluated from the enthalpy.
The process is repeated until the desired final time or temperature is reached.
The melting process is assumed to happen over an interval defined by an input
parameter DTmp, which was chosen as 0.5 C. Solid properties are used for
T < Tmp DTmp, and liquid properties are used for T > Tmp þ DTmp. Linear
interpolation is used for temperatures in between T DTmp and T þ DTmp. Sim-
ilar functionality is assumed for all of the physical properties (k, Cp, and q).
The dimensionless boundary conditions are listed in Eq 2, and the physical
properties used in the calculations are given in Table 3. As an approximate
way to account for the copper supports, the temperature at the surface of the
TABLE 3—Physical properties and parameters used in calculations.
Variable Description Value Units

a Width in z-direction 0.005 m
b Height in y-direction 0.005 m
Cp,L Specific heat of liquid aluminum 1086 J=kg=K
Cp,s Specific heat of solid aluminum 901 J=kg=K
dlase Diameter of laser spot 0.003 m
Hfus Heat of fusion of aluminum 399 600 J=kg
kL Thermal conductivity of liquid aluminum 100 W=m=K
ks Thermal conductivity of solid aluminum 237 W=m=K
Plaser Laser power 2700 (typical) W
Tmp Aluminum melting point 933.5 K
To Initial temperature of slab 300 K
Zholder Coordinates of point at which sample is 0.003 m
held between copper supports
aL Thermal diffusivity of liquid aluminum 0.00003874 m2=s
as Thermal diffusivity of solid aluminum 0.00009742 m2=s
d Slab thickness in x-direction 0.002, 0.0035, or 0.0045 m
qL Density of liquid aluminum 2377 kg=m3
qs Density of solid aluminum 2700 kg=m3
reflection Fraction of laser energy reflected 80%
by aluminum sample
window Fraction of laser energy reflected or 5.5%
absorbed by the chamber window
DTmp Temperature range for change in properties 0.5 K
from solid to liquid
sample was assumed to stay constant at the initial temperature in the zone
where the sample was held by the supports. Heat losses from the sample due to
convection and radiation were ignored.
T ¼ 1 @ t ¼ 0
@T
¼ 0 @ y ¼ 0 and þ b
@y
@T (2)
¼ 0 @ z ¼ 0 and þ a
@z
@T
¼ 0 @ x ¼ 0 and z < Zholder
@x
T ¼ 1 @ x ¼ 0 and z Zholder
@T q0 d pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ @ x ¼ 1 for 0 y2 þ z2 Rbeam
@x k To
@T pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ 0 @ x ¼ 1 for z < Zholder and Rbeam < y 2 þ z2
@x
T ¼ 1 @ x ¼ 1 and z Zholder

Plaser 1 treflection ð1 twindow Þ
qo ¼ p 2
d
4 laser
The numerical method used to solve the equations was checked for accu-
racy in several ways. First, simple heat conduction problems with known ana-
lytical solutions were calculated. Second, results for the melting of water
reported in the literature were reproduced [9]. Last, the number of finite differ-
ence grid points was varied to confirm that the change in the calculated temper-
ature profiles was small.
An example of the calculations is shown in Fig. 11 for a case with a nomi-
nal laser power of 2700 W. The temperature at the centerline of the laser beam
reaches 2230 K in about 63 ms. This calculation is in very good agreement
with the experimental results, which indicate that for a 60 ms pulse and a 0.35
mm thick sample, 2700 W is close to the 50 % ignition probability threshold.
According to these calculations, melting of the aluminum beneath the surface
oxide layer begins to occur about 10 ms after the start of the laser pulse. For
approximately the next 10 ms, the rate of temperature rise at the laser spot cen-
terline slows because of the heat required for melting. After about 20 ms, the
rate of temperature rise at the laser spot centerline increases because the melt-
ing boundary has moved far enough from the surface that the temperature rise
FIG. 11—Surface temperature at laser spot centerline, 2700 W laser power,

0.35 mm sample.
at the surface is dictated by the thermal conductivity (and diffusivity) of the

liquid, which is lower than that of the solid.
Figure 12 shows that the 3-D heat conduction model with aluminum melt-
ing and a constant ignition temperature (2230 K) captures the way the ignition
energy changes as the laser pulse is changed and as the sample thickness is
changed. Predictions of the ignition energy for a given sample thickness are
created by running the model at different laser powers and determining the
amount of time necessary in order for the laser spot centerline surface tempera-
ture to reach 2230 K. The energy for this condition is calculated from the prod-
uct of the power delivered to the sample and the time to reach the ignition
temperature.
FIG. 12—Effect of changing the laser pulse time and sample thickness.
The qualitative trends in Fig. 12 are easy to explain. At high laser power,
the sample heats up quickly and the temperature gradients are high. This
implies that less of the sample heats up, so less energy is required for the cen-
terline temperature to reach the ignition point, and the laser pulse time neces-
sary for ignition is short. To put it another way, the energy delivered to the
sample remains localized near the laser spot. At low laser power, the trend is
the opposite. The energy delivered to the sample has time to spread out because
of heat conduction, and the temperature gradients are low. It takes a longer
laser pulse and more delivered energy to raise the centerline surface tempera-
ture to the ignition point. Neglecting heat losses also leads to less accuracy in
this case.
Discussions of Ignition Mechanisms

The ignition mechanisms in a BAHX used to boil and/or warm oxygen are dis-
cussed in Ref 2. These include the following:
• Particle impact
• Fractures that expose fresh aluminum
• Promoted ignition due to:
• hydrocarbons from the air that accumulate in the heat exchanger
• lubricant used in manufacture and assembly
• foreign material inadvertently left in the BAHX before and after
installation
The existence of sealed cavities can also be a source of aluminum/oxygen
ignition, as noted in Ref 2, although the fundamentals of the ignition process
are not well understood. A sealed cavity is a volume completely enclosed by
aluminum parts that are joined together. Such sealed cavities can be formed
when support plates are welded to the exterior of a BAHX, and they can also
be formed during the fabrication process used to isolate passages and attach
headers to the brazed matrix. A sealed cavity becomes a potential hazard if it is
submerged in LOX and there is a defect in the seal. A hazard develops if LOX
slowly fills the cavity over a relatively long time scale, and if the BAHX is sub-
sequently warmed relatively quickly (10 C to 20 C is sufficient), causing the
pressure within the cavity to rapidly increase. The rapid pressure rise can cause
the cavity to rupture, leading to mechanical failure in the BAHX, and under the
right circumstances this can lead to aluminum ignition through some combina-
tion of the three mechanisms listed above.
Ignition due to particle impact and fractures that produce fresh aluminum,
as well as their dependence on the oxygen pressure, are discussed in this paper.
Promoted ignition is not discussed, as the associated mechanisms are either in-
dependent of the oxygen pressure or even less probable at higher pressures as a
result of the higher solubilities of hydrocarbons. A description of factors
affecting hydrocarbon accumulation in BAHXs used for producing pressurized

oxygen, as well as methods for mitigating the hazard of specific hydrocarbons,
is given in Ref 2.
Ignition by Particle Impact

Residual particles present in oxygen in BAHXs are generally of relatively
small size. The filtration level recommended in Ref 2 for particles in an HP
GOX BAHX limits the maximum particle size to 600 lm, and particles this
size can acquire a velocity inside a BAHX comparable to that of the oxygen
flow. The oxygen flow velocity in a BAHX is relatively low (a few meters per
second) during normal operating conditions but can be very high (a few hun-
dred meters per second) in the event of sudden mechanical rupture.
Impingement of particles could possibly ignite aluminum fins via energy
inputs from the conversion of incident particle kinetic energies and/or the com-
bustion of incident combustible particles ignited either by conversion of their
own kinetic energies or by conversion of the kinetic energies of other, larger
incident particles.
Aluminum ignition, whatever the mechanism, requires thermal energy
build-up that leads to the melting of metal. Then fast, exothermic aluminum
oxidation can develop following breaching, removal, or surface displacement
of the initially solid passivating aluminum oxide layer.
An accurate assessment of energy inputs required for aluminum ignition by
means of particle impact is not an easy task. Fast interactions between the par-
ticles and aluminum fins subjected to particle impacts involve the conversion
of incident kinetic energy to elastic and plastic deformation, heat exchanges
between the particle and the fin, and a loss of integrity of the passivating oxide
layer and its kinetic competition with oxide reformation. The occurrence of
these multiple, complex phenomena on a small geometrical scale are not easy
to simulate. Moreover, no suitable experimental means of real-time in situ
physical diagnostics of the particle–fin interfacial zone seem to exist.
Herein, only the orders of magnitude of the involved energies are estimated
for a comparative assessment of ignitability in different situations considering
the case of aluminum particles.
The following energies are defined for comparisons of orders of
magnitude:
• Eignit represents the maximum energy necessary to ignite an aluminum
particle, assuming that combustion will start only after all the aluminum
composing the particle is heated above the melting temperature of alu-
minum oxide.
• Ecombust is the maximum aluminum oxidation enthalpy released by com-
plete combustion of a particle.

• Ekin is the kinetic energy of incident particles.
TABLE 4—Examples of kinetic and combustion energies of aluminum particles.
Al Particle Size, mm Ekin, J Eignit, J Ecombust, J scombust, s [10]

1
1600 [2] 5 10 (420 m=s) 11 176 >5
600 6 105 (20 m=s) 0.6 9 %1
6 103 (200 m=s)
30 6 105 103 <103
Orders of magnitude for energies for different particle sizes and velocities,
as well as the combustion time scombust, are summarized in Table 4.
From Table 4, it can be seen that a particle with a 600 lm diameter imping-
ing at 20 m/s or 200 m/s has a kinetic energy Ekin ¼ 6 105 J or 6 103 J.
This is lower by a factor of 106 or 104 than the typical laser input energy neces-
sary to ignite a fin (around 30 J). Therefore, fin ignition following the conver-
sion of only particle kinetic energies would be extremely improbable.
Consider now the possibility of fin ignition via combustion of an impacting
particle ignited by its own impact with the fin. In the worst case—a 600 lm
particle having a velocity of 200 m/s—the kinetic energy, Ekin ¼ 6 103 J, is
smaller by a factor of 100 than the corresponding maximum ignition energy
(Eignit ¼ 0.6 J). Actually, the ignition of the particle is likely to require less than
0.6 J because of premature degradation of the passivating oxide layer integrity
under impact.
The experimental works performed at the National Aeronautics and Space
Administration White Sands Test Facility [10] showed that the ignition of an
aluminum target is possible only for very large aluminum particles (1600 lm)
impinging at a supersonic velocity of at least 350 m/s to 450 m/s. The corre-
sponding kinetic energies were around 0.5 J, about 1/20th of the ignition
energy (Eignit ¼ 11 J), meaning that only a limited volume of material close to
the particle–target interface was thermally and/or mechanically affected
enough at impact to allow for ignition. The energy present at this stage comes
exclusively from conversion of the incident particle kinetic energy.
These results suggest that the combustion of a 600 lm aluminum particle
impinging at a velocity of 200 m/s is quite unlikely but cannot be ruled out.
However, considering that only a fraction of the combustion energy (9 J) can
be converted, and thus that the ignition energy available is well below the fin
ignition energy (30 J), it can be concluded that direct fin ignition caused by the
mechanical impact of a single aluminum particle would not be possible in the
considered conditions.
The possibility of fin ignition via particle impacts has to be considered in
the form of kindling chains between several particles. Particles of small enough
diameter, because of their very high effective surface area, are extremely reac-
tive in oxygen and can be ignited by means of impact or friction (with larger
particles, for example). Then, through contact, these burning particles might be
able to ignite other particles of substantially larger size following conversion

of the combustion energies. The elemental kindling mechanism can be repeated
several times involving particles of increasing diameter, leading finally to burn-
ing particles having sufficient combustion energies to ignite a fin through con-
tact. These particles have time to reach the fin wall far before they are fully
consumed by combustion in the free space of the passage, as the combustion
time will be on the order of 1 s for a 600 lm particle.
The probability of the occurrence of such a compound event is very low in
the case of a GOX BAHX during normal operations in which all appropriate
rules and good practices have been observed for cleanliness. However, in an
incident involving mechanical failure, large amounts of particles with a broad
size distribution might be generated, making this mechanism more likely. In
the limit, a dense particle cloud could form, which is a medium of well-known
high reactivity and adiabaticity and accelerated burning kinetics, resulting in
an intense confined fire inside a passage.
It can be noted that the risk of fin ignition via the above-described kindling
mechanism of particle impacts is not fundamentally affected by a high operat-
ing pressure. This is because the presence or generation of particles in the
BAHX is not expected to increase at higher pressure (assuming similar safety
margins for the mechanical integrity of the BAHX) and the velocity of
entrained particles does not increase with increasing pressure, as it is a com-
mon practice to design BAHXs at constant dynamic pressure (i.e., at constant
PV2 or qV2).
Aluminum Rupture Mechanism

The exposure of a fresh surface that is not protected by the thin oxide film nor-
mally present on the surface of aluminum (typically 10 to 30 Å thick) has long
been proposed as a potential mechanism for aluminum ignition [4]. The
kinetics of aluminum oxidation has been well studied for temperatures up to
about 600 C, which is just below the aluminum melting point [11]. At all tem-
peratures there is an initial rapid rate of oxidation followed by a period of rela-
tively constant oxide film growth at a rate that is temperature dependent. After
this initial formative stage, the oxidation rate decreases to a negligible value,
and this behavior occurs for the entire temperature range studied. It is apparent
that at least for temperatures less than about 600 C, the rate of heat release
from the surface oxidation reaction is more than balanced by heat transfer
away from the surface and by the increasing resistance to O2 transfer as the ox-
ide film forms, so that no “runaway” condition exists that would lead to igni-
tion and/or oxidation of the entire sample. If aluminum rupture is to cause
ignition, there must also be a mechanism that increases the temperature to the
point at which the heat generated by the oxidation reaction overcomes the heat
transfer away from the surface and the retarding effect of the oxide layer itself
[12].
A fresh surface will be exposed if a crack propagates through any of the
aluminum components of the BAHX. For example, a fresh surface might be
formed in a BAHX by means of the cracking or tearing of the thin aluminum
fins; the separation of parting sheets or side bars; or the failure of headers, pip-
ing, or bars welded to the cores for mechanical support.
If a crack initiates and starts to spread, stresses are produced ahead of the
crack tip. When the local stresses exceed a threshold value, a plastic deforma-
tion zone forms in the local area of the crack tip and convects along with the
crack tip as it propagates. The plastic zone near the tip is a site of heat genera-
tion, as nearly all (90 %) of the plastic work produced in the plastic zone dis-
sipates into the surrounding metal as heat. The temperature rise resulting from
this energy dissipation has been analyzed using fracture mechanics and is com-
pared to experimental measurements in a number of publications [13–16]. The
primary focus of these investigations was to determine whether the fracture
process was affected by the change in temperature. For example, the increase
in temperature might be expected to change the stress and strain distribution
near the crack tip or change the local fracture toughness. Temperature increases
near a rapidly propagating crack tip have been measured in the 10 C–200 C
range for various materials (mostly steel alloys), and even higher temperature
rises have been calculated under certain conditions.
The work of Rice and Levy [15] was used to estimate the maximum tem-
perature rise for aluminum in a BAHX (Table 5). Stress and strain distributions
in nonhardening materials were used in their analysis, and the plastic zones
around crack tips were regarded as distributed sources of heat proportional to
the plastic work rate. The work rate is proportional to the crack propagation ve-
locity, so estimates of the propagation velocity are necessary to determine the
temperature rise. The increase in temperature was determined by employing
the fundamental solution to the heat transfer problem of a point source moving
at constant speed equal to the crack velocity Vcrack and assuming that the me-
chanical and thermal properties remain constant with temperature. The plastic
TABLE 5—Nomenclature and physical properties used for temperature rise estimation.
Variable Description Value Units

E Young’s modulus ¼ 2G (1 þ ) 71 526 MPa
G Shear modulus for 3003-H18 aluminum 26 890 MPa
KI Stress intensity factor; dependent on crack formation details Less than 21 MPa*m0.5
Vcrack Crack propagation velocity; estimated from Fig. 13 1200 max m=s
ro Yield stress for 3003-H18 aluminum 190 MPa
Poisson’s ratio for 3003-H18 aluminum 0.33
FIG. 13—Aluminum crack speed estimates [17].
zone is assumed to remain fixed in size as the crack advances, and the expres-
sion for the temperature rise for a running crack is given in Eq 7.
pffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi ð1 2 Þ KI ro Vcrack
DT ¼ 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (7)
E qs Cp;s ks
The limiting crack speed for steel and aluminum sheets has been calculated
in terms of the applied load and material properties [17], and the crack velocity
was found to be only weakly dependent on the load. For aluminum, the limiting
crack speed was approximately 0.3(G/q)0.5, where (G/q)0.5 is the shear wave
velocity; the results for crack speed as a function of the applied load are repro-
duced in Fig. 13.
The temperature rise calculated from Eq 7 is given in Fig. 14 as a function
of the stress intensity factor, assuming a crack velocity of 1200 m/s. The stress
intensity factor will vary somewhat depending on the details of where and how
the crack is formed. One approach to estimating KI is to utilize the relationship
between the plastic zone size x and the stress intensity factor determined by
Rice and Levy [15], x ¼ (p/8)*(KI/ro)2, and then recognize that the plastic
zone size is limited by the fin height. For a typical fin height of about 5 mm,
this would imply KI 21 MPa*m0.5 and about a 100 K temperature rise based
on Fig. 13. The thin fins in the BAHX that are perhaps most likely to fail have
a thickness of only 0.2 mm, and stress intensity factors would be even lower
based on this length scale.
The temperature rise of about 100 K estimated above is of an order of mag-
nitude similar to that seen in the measurements made for steel sheets [14–16]
and is far below the temperature needed for aluminum ignition, whether this
FIG. 14—Crack tip temperature rise for two aluminum alloys.
might be the melting point of about 933 K or the 2230 K inferred in the “Heat
Conduction Modeling” section of this paper. Our conclusion is that the fin rup-
ture mechanism, by itself, is not the most likely mechanism for BAHX ignition.
This conclusion is consistent with the fact that ignition via fracture in GOX has
never been demonstrated, despite the many attempts that are summarized in
the review by Werley et al. [4].
It should also be noted that this mechanism would not be affected by
increasing operating pressure. Assuming that similar design factors are used in
the heat exchanger design at both high and low pressure, the probability of
ignition via the fin rupture mechanism is expected to be similar whether the
BAHX is operating at low pressure, for which there is a large industrial experi-
ence base, or high pressure, with which there is less experience. It follows that
assessing whether the design factors used at high pressure are in fact similar to
those used at low pressure is key in assessing ignition risks.
Conclusions for Air Separation Unit Safety

The safe operation of plants using a BAHX to vaporize LOX pumped to high
pressure (up to about 30 bar) is supported by an extensive operating experience
(>4000 plant years) without any ignition incidents. Experience with intermedi-
ate pressures (between 30 bar and 60 bar) is less extensive than that at lower
pressure, but nonetheless it is relatively significant. Experience with operation
above 60 bar is still rather limited, and the one ignition incident reported in a
BAHX used to vaporize LOX pumped to high pressure occurred in an
exchanger designed to operate at about 80 bar. It is therefore important to eval-
uate whether there is any fundamental reason for the probability of aluminum
ignition to be different at high pressure.
The aluminum ignition mechanisms generally considered relevant for

BAHXs vaporizing pressurized LOX are the following:
• Fin rupture (fresh aluminum exposure)
• Hydrocarbon accumulation
• Initial contaminant
• Particle impact/combustion
Each of these mechanisms has been considered individually

(i) By comparing the temperature necessary for ignition to the tempera-
ture rise achievable through fin rupture (using fracture mechanics
models), the fin rupture mechanism can be eliminated, regardless of
the pressure.
(ii) At higher pressure, the gas/liquid phase equilibrium becomes more
favorable, and the solubility of hydrocarbons and plugging agents,
such as CO2 and N2O, increases. The tendency to accumulate hydro-
carbons from the process air will therefore decrease with increasing
pressure. This mechanism is therefore a lower concern at high pres-
sure than at lower pressure.
(iii) The initial contaminant mechanism is mitigated by means of ensur-
ing the appropriate cleanliness. Thin oil films (<100 mg/m2) are
insufficient to ignite aluminum at any pressure. Trapped solid con-
taminants with enough chemical energy to ignite aluminum are not
present because of upstream filtration.
(iv) The particle impact/combustion mechanism cannot be eliminated as
an ignition source from a theoretical point of view. Our test results
show that the level of energy necessary and sufficient to initiate alu-
minum combustion/propagation does not decrease with pressure. It is
assumed that the level of energy supplied by a particle is equal to the
energy from combustion of the particle, and thus the quantity of
energy is independent of pressure. The amount of energy transferred
to the aluminum fin is also independent of pressure. The frequency of
particle impact events is not expected to change significantly in a
BAHX designed for high pressure operation. The probability of igni-
tion via particle impact is therefore not a function of pressure, pro-
vided that the fin thickness, particle velocity, and warm O2 piping
design are consistent with standard design practices.
The analysis in this paper shows that in isolation, these events are not
likely to lead to aluminum ignition, but it is possible that ignition could occur
as a result of a combination of these events. Mechanical failure might lead to
scenarios involving more than one of the mechanisms addressed above acting
in combination. The only known example of aluminum ignition in a pumped
LOX BAHX provides an illustration of this point. The incident likely involved
the failure of the internal transverse bars of the BAHX caused by a sealed cav-
ity type event in the volume between two transverse bars separating the high
pressure LOX from an inactive region below the bars. The failure of the trans-
verse bars to contain the high pressure oxygen is thought to have led to a com-
bination of events, including the exposure of fresh aluminum and particle
impingement both on the thin aluminum fins and on other particles.
Further work is necessary to better understand the risks associated with me-
chanical failure and how they relate to oxygen pressure and to the mechanical
design of the BAHX. The safe operation of BAHXs at high pressure requires
that users avoid ignition mechanisms and properly address the risks associated
with the mechanical failure of end bars or parting sheets.
References
[1] “Safe Operation of Reboilers/Condensers in Air Separation Units,” IGC

Document 65/06/E.
[2] “Safe Use of Brazed Aluminum Heat Exchangers for Producing Pressur-
ized Oxygen,” IGC Document 145/10/E.
[3] Werlen, E., Crayssac, F., Longuet, O., and Willot, F., “Ignition of Conta-
minated Aluminum by Impact in Liquid Oxygen, Influence of Oxygen
Purity,” J. ASTM Int., Vol. 6, No. 10, 2009, pp. 49–67.
[4] Werley, B. L., Barthelemy, H., Gates, R., Slusser, J. W., Wilson, K. B.,
and Zawierucha, R., “A Critical Review of Flammability Data for Alumi-
num,” Flammability and Sensitivity of Materials in Oxygen-enriched
Atmospheres, ASTM STP 1197, Vol. 6, ASTM International, Philadel-
phia, PA, 1993, pp. 300–345.
[5] Benz, F. J., Williams, R. E., and Armstrong, D., “Ignition of Metals and
Alloys by High Velocity Particles,” Flammability and Sensitivity of Mate-
rials in Oxygen-enriched Atmospheres, ASTM STP 910, Vol. 2, M. A.
Benning, Ed., ASTM International, Philadelphia, PA, 1986, pp. 16–37.
[6] Mench, M., Haas, J., and Kuo, K., “Combustion and Flame Spreading of
Aluminium Tubing in High Pressure Oxygen,” Fifth International Sympo-
sium on Special Topics in Chemical Propulsion: Combustion of Energetic
Materials, Stresa (Italy), 2000, pp. 438–452.
[7] Mench, M., Kuo, K., Sturges, J., Hansel, J., and Houghton, P., “Flame
Spreading and Violent Energy Release (VER) Processes of Aluminium
Tubing in Liquid and Gaseous Oxygen Environments,” Flammability
and Sensitivity of Materials in Oxygen-enriched Atmospheres, ASTM
STP 1395, Vol. 9, ASTM International, Philadelphia, PA, 2000, pp.
402–424.
[8] Bransford, J. W., “Ignition and Combustion Temperatures Determined by
Laser Heating,” Flammability and Sensitivity of Materials in Oxygen-
enriched Atmospheres, ASTM STP 910, Vol. 2, M. A. Benning, Ed.,
ASTM International, Philadelphia, PA, 1986, pp. 78–97.
[9] Comini, G., Del Guidice, S., Lewis, R. W., and Zienkiewicz, O. C.,
“Finite Element Solution of Non-Linear Heat Conduction Problems with
Special Reference to Phase Change,” Int. J. Numer. Methods Eng., Vol. 8,
1974, pp. 613–624.
[10] Wu, J., Fang, H., Kim, H. J., and Lee, C., “High Speed Impact Behavior
of Al Alloy Particle onto Mild Steel Substrate during Kinetic Deposition,”
Mater. Sci. Eng., A, Vol. 417, 2006, pp. 114–119.
[11] Smeltzer, W. W., “Oxidation of Aluminum in the Temperature Range
400 –600 C,” J. Electrochem. Soc., Vol. 103, 1956, pp. 209–214.
[12] Yuen, W. W., “A Model of Metal Ignition Including the Effect of Oxide
Generation,” Flammability and Sensitivity of Materials in Oxygen-
enriched Atmospheres, ASTM STP 910, Vol. 2, ASTM International,
Philadelphia, PA, 1986, pp. 59–77.
[13] Tzou, D. Y., “The Thermal Shock Phenomena Induced by a Rapidly
Propagating Crack Tip: Experimental Evidence,” Int. J. Heat Mass Trans-
fer, Vol. 35, 1996, pp. 2347–2356.
[14] Li, W., Deng, X., and Rosakis, A. J., “Determination of Temperature
Field around a Rapidly Moving Crack-tip in an Elastic-Plastic Solid,” Int.
J. Heat Mass Transfer, Vol. 39, 1996, pp. 677–690.
[15] Rice, J. R. and Levy, N., “Local Heating by Plastic Deformation at a
Crack Tip,” Physics of Strength and Plasticity, A. S. Argon, Ed., MIT
Press, Cambridge, MA, 1969, pp. 277–292.
[16] Guduru, P. R., Zehnder, A. T., Rosakis, A. J., and Ravichandran, G.,
“Dynamic Full Field Measurements of Crack Tip Temperatures,” Eng.
Fract. Mech., Vol. 68, 2001, pp. 1535–1556.
[17] Kanninen, M. F., “An Estimate of the Limiting Speed of a Propagating
Ductile Crack,” J. Mech. Phys. Solids, Vol. 16, 1968, pp. 215–228.
STP 1561, 2012
DOI:10.1520/STP20120008
Vladimir V. Leschevich,1 Oleg G. Penyazkov,1

and Jean-Christophe Rostaing2
Auto-Ignition and Combustion Properties

of Iron/Steel Micro-Particles in Oxygen
Atmospheres Heated by Rapid Compression
REFERENCE: Leschevich, Vladimir V., Penyazkov, Oleg G., and Rostaing,

Jean-Christophe, “Auto-Ignition and Combustion Properties of Iron/Steel
Micro-Particles in Oxygen Atmospheres Heated by Rapid Compression,”
ABSTRACT: Accumulated deposits of metal particles contaminating high-
pressure oxygen piping systems represent a strongly reactive and highly
adiabatic medium. They are prone to ignition following an initial thermal
perturbation or heat confinement and then will apply an extremely high tem-
perature pulse to adjacent surfaces, thus becoming a factor enhancing the
severity of accidental fires. In order to investigate metal powder layers’ ignit-
ability and burning, a rapid compression machine was used for generating, at
a given time-origin, heated oxygen atmospheres at programmed pressures
of 0.5 to 28 MPa and temperatures of 550 to 1100 K. Critical conditions for
auto-ignition were determined as a function of particle size for two commer-
cial iron powders with particle sizes ranging from 1 to 5 lm and three real
residual powders from industrial installations. A first major finding is that iron/
steel particle deposits can be easily ignited in rapidly heated oxygen atmos-
pheres at temperatures considerably lower than the iron melting point. More-
over, this critical temperature diminishes significantly with increasing oxygen
pressure. The temperature of burning particles was measured with a time
resolution of 4 ls by means of a novel photoemission technique. It was seen
that the temperature can quickly rise up to 3100 K and then fall to 1850 6 50 K.
An extensive database is available for the development and validation of
models describing fires in various piping elements with specified particulate
November 2012.
1
A.V. Luikov Heat and Mass Transfer Institute, Minsk, Belarus 220072.
2
Air Liquide Corporate R&D, les Loges-en-Josas, France F78354.
Copyright V
19428-2959.
147
contamination, to be used in order to identify circumstances for especially

severe accidents.
KEYWORDS: iron micro-particle powders, oxygen, rapid compression
machine, auto-ignition
Introduction
All gases under high pressure and temperature conditions present potential
safety hazards. This is especially true for oxygen. Although it is not flammable,
is not shock sensitive, and will not decompose, oxygen is a strong oxidizer and
vigorously supports combustion. Moreover, its reactivity increases with
increasing pressure, temperature, and concentration. As a result, with sufficient
pressure and ignition energy, almost all materials are combustible in pure oxy-
gen, including substances that are not usually regarded as flammable, e.g., met-
als. Moreover, all flammable substances will burn more actively at higher
temperatures in a high-pressure oxygen atmosphere because of the enhance-
ment in the oxygen transport and supply to the reacting surface. Therefore,
special knowledge and understanding of material properties in different condi-
tions are required in order for one to design safe high-pressure oxygen piping
systems.
Potential ignition mechanisms and ignition sources are in many cases con-
nected with contaminants such as oil, grease, solvents, and materials such as
dust, lint, and metal chips that can be found in pipeline systems. Organic sub-
stances and fine metal particles are ignited extremely easily and can act as a
kindling source for other, more difficult-to-ignite materials. There are different
scenarios of ignition in closed piping systems. The main ones are particle
impact and pneumatic impact (adiabatic compression).
Oxygen flowing at high speed through valves and piping systems can pro-
pel contaminants with such force that friction or impact between particles or on
solid metal walls can raise their temperature to the ignition point. The transfer
of heat from these burning particles might be enough to produce the ignition of
plain metal wall sections and cause a serious accident. This ignition mechanism
was extensively studied with particles of various materials, velocities, and sizes
[1,2].
Another important case for practice is when the ignition energy in an oxy-
gen system comes from adiabatic compression. When the gas is quickly com-
pressed inside a closed system, such as a container or piping (from a valve
opening too suddenly, for instance), local temperatures inside the system can
rise sharply and might be high enough to cause the ignition of fine metal par-
ticles or other kindling intermediates such as polymer seals, valve seats, etc.,
which in turn could play a role in the initiation and propagation of fire acci-
dents. This is why adiabatic compression test (pneumatic impact) methods
[3,4] have been widely used to qualify high-pressure equipment in oxygen
service. Usually this test involves examining, analyzing, and evaluating
LESCHEVICH ET AL., doi:10.1520/STP20120008 149
different oxygen system components (regulators, valves, hoses, etc.) under con-
ditions they are expected to withstand in real installations. Standards put forth
by several organizations (International Organization for Standardization,
Centre d’Etudes Nucleaire, ASTM) describe how to perform the adiabatic
compression tests for each component. The test setup and protocol are designed
in such a way as to obtain the maximum intensity of the compression effect in
order for the test conclusions to be more conservative. There is no intended
access to information that would be required in order for one to fully under-
stand the phenomena of the fire initiation involving kindling intermediates, in
particular metal particles that ignite spontaneously through adiabatic oxygen
heating. The validity of the test is difficult to assess when metal particles, in
general in arbitrary amounts, are purposely introduced into the component to
be evaluated (“pill test”). In parallel, the collective burning of close assemblies
or residual metal particles in large industry piping networks can give rise to
intense pulses of confined heat. In the likely case of a subsequent burn-through
event, the propagation of the fire will be considerably enhanced and will con-
sume a very large mass of metal. It is thus highly suspected that the uncon-
trolled presence of particulate pollution has been the origin of the most severe
accidents in the history of large oxygen pipeline operation. Here also, the risk
is not well quantified. Therefore, additional data on the auto-ignition tempera-
ture and its dependence on the oxygen pressure, particle size, and material are
needed.
A number of experimental methods have been used to determine the igni-
tion temperature of solid metals and alloys in oxygen gas, and a wide range of
ignition temperatures have been reported for the same or similar materials [5].
Authors explain the variations in measured ignition temperatures with refer-
ence to the concept that the heat balance of a system can greatly influence the
measured temperature. Therefore, it is highly desirable to monitor and study
the desirable phenomena under highly controlled and reproducible tempera-
ture/pressure conditions, applying accurate and high-time-resolution measure-
ment methods. It is clear that complicated phenomena such as the auto-ignition
and combustion of metal powders consisting of different kinds of particles
(material and size) in rapidly heated oxygen cannot be studied via the stand-
ard adiabatic compression test. For this purpose, the advantages of a rapid
compression machine (RCM) were exploited in this work. In the RCM, the
use of a compression piston allows one to isolate the desirable phenomena
and generate specified thermodynamic conditions in the test volume. More-
over, the RCM test chamber offers good access for different measurements
and observations.
The first extensive experimental particle auto-ignition and combustion
studies using the RCM have been conducted on synthetic iron powders from an
industrial materials supplier [6]. The critical conditions that can provoke the
combustion of a particle layer placed in a rapidly heated oxygen atmosphere
were determined for three powders (20–45 lm, 56–63 lm, and 80–125 lm),
sorted by sieves, at pressures of 1.5 to 28 MPa and temperatures of 660 to 1100
K. The smallest particles showed a strong ability to auto-ignite at temperatures
of less than 700 K and oxygen pressures of more than 22 MPa. Moreover, for
iron powder particles 20–45 lm in size, the dependence of ignition delay times
on the oxygen temperature was established at pressures ranging from 20 to 30
MPa. This paper extends the previous study. New data were obtained for finer
synthetic iron particle powders and compared to previous results. Additionally,
several powders consisting of real residual particles sampled by Air Liquide at
some critical locations in one of its large industrial oxygen distribution net-
works were examined.
Experimental Setup and Materials

The RCM facility used in this study is described in detail elsewhere [6]. There-
fore, only key features and modernizations of the experimental setup are pre-
sented here. The high compression ratio (80:1) of the RCM generates a heated
oxygen atmosphere with pressures of up to 30 MPa and temperatures of up to
1500 K in the test chamber for a duration of 500 ms. The experimental time is
limited only by heat loss through the chamber’s walls. A Kistler 6031U18
high-temperature quartz pressure sensor with a special diaphragm protected
from temperature shock was used for precise pressure measurements. The
transducer was installed, using a recess mount, into the cylindrical wall of the
combustion chamber. Such installation protects the sensor from the effects of
high flash temperatures and burning particle impingement. Also, a special
inlet/outlet pneumatic valve was installed in order to minimize dead volume,
as it can affect combustion behavior and disturb the flow field. For optical
measurements and observations, a quartz window was mounted into the cylin-
drical wall of the test volume. Several photometric devices (photomultipliers
and a pyrometer) were connected to the mandrel of the window by a bifurcated
optical fiber light guide. Neutral filters were installed in front of the photoca-
thode of the photomultipliers. This enabled the detection of both weak lumines-
cence from local auto-ignitions of individual particles and strong luminescence
from the combustion of the whole particle layer.
The first experimental series were performed for two metallurgical grades
of ultra-fine iron powders, with the main particle fractions being 1–3 lm and
1–5 lm, respectively. The size of the particles contained in the powder is one
of the most important parameters determining the combustion properties of
particle layers. Therefore, the powders and samples studied previously [6]
were examined using a certified “Mini-Magiscan” picture scanner. The results
of accurate granulometric analysis are presented in Table 1. It includes the size
TABLE 1—Results of granulometric analysis of studied powders.
Main Fraction Size Average Size
by Quantity by Quantity
Size Range of Particles of Particles
(Minimum– (Horizontal/ (Horizontal/Vertical
Maximum), Vertical Projection), by Mass, Projection), by Mass,
Powder lm lm lm lm lm
Metallurgical 0.2–9.2 1–3/1–3 2–4 2.59/2.16 3.4
ultra-fine iron powders 0.2–16.4 1–5/1–3 4–5 4.01/3.08 4.7
Powders separated
by sieves, lm
45
56 and 63 9.0–98.1 20–40/30–40 30–60 42.92/30.56 43.8
80 and 125 6.7–180.3 60–90/70–90 70–90 82.76/59.01 80.6
32.8–268.1 110–140/100–110 130–160 155.48/114.8 140.2
range of particles contained in an entire particle sample, the size range of the
main fraction by quantity and mass, and the average size by quantity and mass.
It was found that the main fraction size range of powders separated by
sieves is slightly different from the sieve mesh sizes. The contributions of a
main fraction to the mass of a powder and the average by mass sizes are calcu-
lated values. The accuracy of these data is lower than the accuracy of the
parameters as determined from pictures of the powders. Therefore, the main
fraction sizes (the size range of particles generally presented by quantity in a
powder) are used herein to identify each powder. It is also seen (Table 1) that the
size distributions determined by horizontal and vertical projections are approxi-
mately the same, especially for small particle sizes. This is because small par-
ticles take the form of balls, whereas among the big particles there are a lot of
oblong ones. This was also seen in photos of powder taken using a microscope.
Moreover, big particles have complex surfaces with a lot of cavities and bumps.
The combustion characteristics of particle layers can also strongly depend
on the particle material. In order to eliminate the effects connected with the
powder composition, a comprehensive elemental analysis was performed using
a roentgen fluorescent spectrometer and the standard approaches for individual
elements. The results of this analysis are shown in Table 2. The amount of oxy-
gen was determined using the method of reduction of oxygen in a hydrogen
flow. It was found that in all powders, the main material is iron (98.131 % to
99.587 %) and the amount of additives does not exceed 2 %, with oxygen rep-
resenting the highest content among them (0.2 % to 0.48 %). The heavy par-
ticles separated by sieves contain more additives than the ultra-fine ones. In
this series of experiments, the particle samples were placed in a small ceramic
cup mounted in the test chamber of the RCM. The sample mass was 0.05 g in
all tests.
TABLE 2—Elemental composition of studied iron powders.
Metallurgical Ultra-fine Iron Powders Separated by Sieves, lm
Parameter Iron Powders 45 56 and 63 80 and 125

Main fraction size in 1–3 1–5 20–40 60–90 110–140
horizontal projection, lm
Fe, % 99.587 99.206 98.131 98.379 98.198
O, % 0.20 0.42 0.48 0.40 0.37
C, % 0.112 0.073 0.090 0.066 0.053
S, % 0.001 0.001 0.009 0.015 0.019
Si, % 0.1 0.1 0.2 0.2 0.3
Cr, % 0.1 0.1 0.1
Cu, % 0.2 0.3 0.1 0.1
Mn, % 0.1 0.1 0.2
Ni, % 0.1 0.1 0.1
Mo, % 0.01
The second part of this work was conducted using three powder grades
supplied by Air Liquide. These powders are the actual residual particles
sampled by Air Liquide at some locations in an oxygen distribution pipeline
system. The chemistry of each powder type was analyzed with a focused beam
electron microscope (CZEISS SUPRA 55). Results are presented in Table 3. It
was found that Fe, O, and C constituted approximately 96 % of all powder
types; the remainder in all powders generally consisted of Tb and Si. More-
over, in the first powder was found a small amount of Mn, Al, and Mo; in the
second Cu; and in the third Mn, Al, and Cu. In addition, the analysis of par-
ticles’ sizes and surface structures was performed using a focused beam elec-
tron microscope. The photos were taken with different magnifications (from
323 to 89 700). It was found that all powders consisted of a wide size range
of particles. Powder 1 contained particles ranging from 7 to 260 lm, powder 2
TABLE 3—The mean atomic composition of particle powders supplied by Air Liquide.
Element Powder 1 Powder 2 Powder 3

O 60.8 60.6 44.9
Fe 23.4 24.7 33.3
C 13.6 12.9 17.9
Tb 1.1 1.4 1.4
Mn 0.6 0.7
Al 0.4 1.1
Si 0.6 0.5 0.7
Mo 0.6
Cu 0.2 0.4
ranged from 4 to 240 lm, and powder 3 ranged from 15 to 550 lm. Almost all
the particles were ellipsoidal with a fairly smooth, undeveloped surface. The
high concentration of oxygen atoms in the test powders and the existence of
almost ideal spherical particles indicate that incomplete oxidation of the pow-
ders had already occurred.
It is a well-established fact among practitioners that metal powders and
scraps found in large-scale oxygen pipeline networks supplying large industries
such as steelmaking and petrochemistry are oxidized to a variable degree,
sometimes almost completely. In fact, it is not possible to maintain an
extremely clean and pure ambient atmosphere inside large-diameter piping sec-
tions that can extend over hundreds of kilometers. Consequently, RCM ignition
experiments on “real” powder samples should yield quite different results than
the ones on fairly pure commercial metallurgical powders. However, besides
the necessity of a scientific baseline, there is a practical justification for studies
on non-oxidized iron and steel particles. Namely, at the level of factory use,
oxygen distribution systems are of a comparatively small scale, made of higher
quality engineering alloys and well conditioned internally to ensure controlled
specifications of the delivered oxygen purity. This means that accidental dense
particulate pollution is uncommon, although not excluded, and micro-particles
will not generally be notably oxidized.
Oxygen Temperature Calculation Procedure

The direct measurement of gas temperatures in an RCM is impossible because
of the insufficiently fast response of the instrumentation. In fact, the tempera-
ture is determined indirectly from an experimental pressure record. It is usually
assumed that the effect of wall heat loss during and after the compression
stroke is limited to a thin boundary layer along the reaction chamber wall and
that the majority of the reaction chamber volume remains unaffected by heat
losses. The unaffected region of the chamber is called a “core region,” in which
a uniform temperature distribution is assumed to exist. Therefore, although the
compression event is not completely adiabatic, the core region of the chamber
can be assumed to be compressed adiabatically. This assumption enables one
to estimate the temperature of low-density and high-temperature gas, because
it can be assumed to be ideal for these conditions. Gas compressed to 25 MPa
oxygen at temperatures lower than 600 K cannot be considered as ideal. For
these conditions, the thermodynamic parameters were defined by using the
equation of state for real gases [7].
Auto-ignition Conditions of Synthetic Iron Powders

The results of experimental runs conducted in this study, together with the pre-
vious results for the larger scale particles, are plotted in Fig. 1 in terms of the
FIG. 1—Oxygen pressures and temperatures at which auto-ignitions of pow-

ders were observed (solid symbols) and were not observed (open symbols). The
main fractions of powders consisted of particles 1–3 lm (powder 1), 1–5 lm
(powder 2), 20–40 lm (powder 3), 60–90 lm (powder 4), and 110–140 lm
(powder 5) in diameter.
oxygen pressures and temperatures generated by the RCM. Conditions at which

auto-ignition and combustion were observed are marked by solid symbols. The
critical parameters of oxygen needed for auto-ignition are connected by lines
in Fig. 1. It is seen that the critical temperature strongly depends on the oxygen
pressure, and this dependence intensifies with decreasing particle size. It was
found that iron micro-particles can be easily ignited in a rapidly heated oxygen
atmosphere at temperatures much lower than the iron melting temperature.
Moreover, this critical temperature decreases significantly with increasing oxy-
gen pressure. The behavior of this dependence changes greatly with decreasing
particle size. Thus, the fraction of fine particles 1 to 3 lm in size was ignited at
an oxygen temperature of 620 K and a pressure of 10 MPa. At pressures lower
than 5 MPa, the critical temperature starts to rapidly increase and reaches 700
K at 1 MPa (Fig. 1). For coarser fractions, this tendency is not so pronounced,
and the demarcation line between “burning” and “not burning” conditions
moves toward higher temperatures.
Temperature of Burning Iron Powder

A photoemission pyrometer was used to measure the temperature of the burn-
ing particles. This method, based on analysis of the photoelectron energy distri-
bution, is described in Ref 8. Before temperature measurements were made,
the applicability of this method was proved via observation of the emission
spectrum during the ignition and combustion of particle samples. The spectrum
was recorded for a duration of 20 ms. It was found that the energy distribution
of light emitted during the auto-ignition and combustion of particle samples
was continuous for the registered wavelengths of 300–600 nm and similar to
the emission spectrum of solids. The color temperature, calculated from the
slope of the curve f(k,T) ¼ (C2/T) k, is equal to 3041 K. This value agrees
well with the maximal temperature of 3100 K determined from the pyrometer
signal via the photoemission method. The advantage of the used method is the
possibility of obtaining temperature measurements with a temporal resolution
of up to 4 ls. It allows one not only to detect the maximal temperature value
but also to see temperature variations during auto-ignition and combustion.
The measured temperatures are plotted in Fig. 2, together with the pressure and
photomultiplier and pyrometer signals. The 1–3 lm diameter iron powder was
tested in these experiments.
The detected temperatures near the onset of particle auto-ignition vary
from 2450 K to 3000 K (Fig. 2(a)) or 3100 K (Fig. 2(b)) and then fall to 1850
K, which is close to the melting temperature of iron (1812 K) and iron oxides.
The maximal temperature at auto-ignition of 1–3 lm diameter iron powder did
not exceed 3100 K, which is close to the boiling temperature of iron. Thus, the
iron boiling probably limits the maximal temperature as a result of its high
latent heat of vaporization.
Auto-ignition Tests With Real Powders

The experimental series involving particle powders supplied by Air Liquide
(Table 3) was conducted at a compression ratio of 65:1. The corresponding
post-compression oxygen pressure/temperature conditions were 1.2–1.4 MPa/
1100–1160 K. The particle samples were placed on a concave stainless steel
shelf fixed at mid-height of the combustion chamber (Fig. 3). This is a better
simulation of real situations that can occur in pipeline systems than experi-
ments with ceramic cups. The initial mass of the particle samples was 0.1 g
in experiments using powders 1 and 2 and 0.05 g in all experiments using
powder 3.
It was observed that local auto-ignition events occured in all powders
before the end of the compression stroke. This corresponds to a temperature of
about 990 K and a pressure of about 0.7 MPa for powder 1, 980 to 1020 K and
0.7 to 0.9 MPa for powder 2, and 1020 to 1144 K and 0.8 to 1.1 MPa for pow-
der 3. The complete combustion of the entire powder sample, like that which
occurred with ultra-fine synthetic powders, was not observed in these tests.
First of all, this result is connected to the presence of large-scale particles and
contaminants in these powders. Moreover, as is seen from the presented
FIG. 2—Pressure (1), luminescence (2), pyrometer signals (3), and temperature
(4) registered at auto-ignition and combustion of powder particles 1–3 lm in
diameter in oxygen with parameters of (a) 11.3 MPa, 630 K and (b) 4.7 MPa,
630 K.
photographs, particles were scattered during the tests by the vibration of the
RCM. Of course, this scattering deeply affects the complete combustion of the
whole particle assembly.
In order to reduce the scattering of particles, the powders were placed
directly on the bottom of the combustion chamber. The initial mass of the parti-
cle samples was increased from 0.1 g to 0.4 g in order to obtain a thicker parti-
cle layer, thus reinforcing the collective effects. The local ignition initiated the
combustion of the whole particle layer only in experiments with powder 3
(Fig. 4). This can be explained on the basis of the results of chemical analysis,
because a low content of oxygen atoms was found in this sample relative to the
FIG. 3—Photos (a) before and (b) after experiments with powder 2.
FIG. 4—Photos (a) before and (b) after experiments with powder 3.
others. This means that powder 3 contains more fuel material than the other
powders. Despite the fact that real powder samples are consistently more diffi-
cult to ignite than pure iron ones, ignition was achieved in one of these three
samples; all of them had been collected at locations chosen at random.
Conclusions
The rapid compression machine (RCM) appears to be a very powerful tool for
detailed investigation of the auto-ignition and combustion of closely packed
collections of small metallic particles in strictly controlled and time-resolved
temperature and pressure conditions. In this work, the auto-ignition and com-
bustion data for layered synthetic ultra-fine iron powders in rapidly heated oxy-
gen atmosphere have been collected for oxygen pressures varying from 0.5 to
28 MPa and temperatures varying from 550 to 1100 K. The critical conditions
that can provoke ignition and complete combustion were determined. It was
found that iron micro-powders can be easily ignited in rapidly heated oxygen
atmosphere at temperatures much lower than the iron and iron oxide melting
temperatures. Moreover, this critical temperature decreases significantly with
increasing oxygen pressure and decreasing particle size. This is quite surprising
and represents an extremely important finding for oxygen system safety. The
influence of oxygen pressure and temperature, as well as of particle size, on the
ignition delay time of an entire powder test sample was established. The varia-
tion of the temperature during the combustion of 1 to 3 lm diameter iron pow-
ders was measured with a temporal resolution of 4 ls using a photoemission
pyrometer. It was found that the temperature quickly rises to 3100 K just after
auto-ignition and then slowly falls to the value of 1850 K, which is close to the
melting temperature of iron (1812 K) and iron oxides. The collected data are
important for the development of new technological processes during which
iron micro-particles and a heated oxygen atmosphere can appear together.
Moreover, the obtained database is useful for the development and validation
of mathematical models describing auto-ignition and combustion phenomena
of iron particles in a heated oxygen atmosphere. One of the first applications is
the adiabatic compression test, for which an upgraded version is proposed with
intentional particulate contamination; the standard version remains at present
to a great extent a “black box” from the standpoint of physics. The second area
of interest is the anticipation of the strong aggravating character of residual
particles in causing accidents in large industry oxygen piping systems. The col-
lective combustion of such particles could result in the extreme confinement of
heat, the initiation of strong thermal perturbations, or a massive delocalized
transfer of heat to a pipe wall, resulting in extended surface burning modes.
However, the spectacular conclusions from experiments on pure (non-
oxidized) iron particles are not fully valid for metal powders accumulating in
large industry supply pipeline networks, which in general are partially oxidized
and dirty. Here, however, for one such real powder sample containing more
iron than the others, local ignition effectively initiated the ignition and combus-
tion of the entire particle layer. The two other samples showed resistance to
complete combustion initiated by local auto-ignition as a result of the presence
of large-scale particles and contaminants.
More reliable conclusions about the specific fire-enhancement risk in large
industry pipeline systems could be drawn if RCM ignitability experiments
were performed on a number of other real powder samples or on iron powder
purposely pre-oxidized in the laboratory.
Acknowledgments
This work was sponsored by Air Liquide (France).
References
[1] Benz, F. J., Williams, R. E., and Armstrong, D., “Ignition of Metals and
Alloys by High-Velocity Particles,” Flammability and Sensitivity of Mate-
rials in Oxygen-Enriched Atmospheres, ASTM STP 910, M. A. Benning,
Ed., ASTM International, West Conshohocken, PA, 1986, pp. 16–37.
[2] Williams, R. E., Benz, F. J., and McIlroy, K., “Ignition of Steel Alloys by
Impact of Low-Velocity Iron/Inert Particles in Gaseous Oxygen,” Flam-
mability and Sensitivity of Materials in Oxygen-Enriched Atmospheres,
ASTM STP 986, D. W. Schroll, Ed., ASTM International, West Consho-
hocken, PA, 1988, pp. 72–84.
[3] Newton, B. and Steinberg, T., “Adiabatic Compression Testing—Part I:
Historical Development and Evaluation of Fluid Dynamic Processes
Including Shock-wave Considerations,” J. ASTM Int., Vol. 6, No. 8, 2009,
pp. 337–361.
[4] Newton, B., Chiffoleau, G., Steinberg, T., and Binder, C., “Adiabatic
Compression Testing II—Background and Approach to Estimating
Severity of Test Methodology,” J. ASTM Int., Vol. 6, No. 8, 2009, pp.
362–384.
[5] White, E. L. and Ward, J. J., “Ignition of Metals in Oxygen,” Defense
Metals Information Center Report No. 224, Battelle Memorial Institute,
Columbus, OH, 1966.
[6] Leschevich, V. V., Penyazkov, O. G., Fedorov, A. V., Shulgin, A. V., and
Rostaing, J.-C., “Conditions and Ignition Delay Times of Iron Micro-
particles in Oxygen,” J. Eng. Phys. Thermophys., Vol. 85, No. 1, 2012,
pp. 139–144.
[7] Frolov, S. M., Kuznecov, N. M., and Kryuger, S., “Properties of Real
Gases—n-Alkanes, O2, N2, H2O, CO, CO2 and H2 at Diesel Engine
Conditions,” Supercritical Fluids: Theory and Practice, Vol. 4, No. 3,

2009, pp. 56–105 (in Russian).
[8] Baranyshyn, Y. A., Penyazkov, O. G., Kasparov, K. N., and Belaziorava,
L. I., “Photoemission Measurements of Soot Particles Temperature at Py-
rolysis of Ethylene,” Nonequilibrium Phenomena: Plasma, Combustion,
Atmosphere, Torus Press Ltd., Moscow, 2009, pp. 87–93.
STP 1561, 2012
DOI:10.1520/STP20120007
Walter Ciscato,1 Christian Binder,2 Olaf Hesse,2

Siegfried Lehné,2 and Thomas Tillack2
Another Dangerous Practice in the

SCUBA Diving Community
REFERENCE: Ciscato, Walter, Binder, Christian, Hesse, Olaf, Lehné, Sieg-

fried, and Tillack, Thomas, “Another Dangerous Practice in the SCUBA
Diving Community,” Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres on September 19–21, 2012 in Montreal, QC; STP
1561, Samuel Edgar Davis and Theodore A. Steinberg, Editors, pp. 162–
PA 2012.
ABSTRACT: The widely held incorrect belief in the self-contained underwater
breathing apparatus (SCUBA) diving community, that oxygen-enriched gas
mixtures, so-called NITROX, with an oxygen content of up to 40 vol. % can
be considered as regular air, has already been refuted in many papers. Now,
another dangerous practice has drawn the attention. In a market study, more
than 60 representative SCUBA cylinder valves were examined. Sales offices
claim that all of those valves can be used for oxygen-enriched gas mixtures.
The fact is that some of these cylinder valves are for air use only. By
exchanging the nonmetallic materials and applying additional cleaning proce-
dures, these air valves become so-called “oxygen clean.” Then, the valves
are on sale for oxygen-enriched service. This procedure is dangerous
because the labeling pretends a pseudo-safety. All of the 60 SCUBA cylinder
valves were tested applying the standardized oxygen pressure surge tester
at BAM. As suspected, many of the cylinder valves are not burn-out safe. In
addition, different test results were received for actual new valves, for so-
called new but temporarily stored valves, and for used valves. This paper
reveals another dangerous practice and wants to alert the SCUBA diving
community.
KEYWORDS: SCUBA diving, pneumatic impact, burn-out safety, oxygen,
NITROX, modified air valves, aging of diving valves, oxygen compatibility
November 2012.
1
Expert, Gentoo Divers Intl., Duebendorf 8600, Switzerland.
2
Specialist, Working Group “Safe Handling of Oxygen,” Division 2.1, “Gases, Gas Plants,”
BAM Federal Institute for Materials Research and Testing, Berlin 12205, Germany.
Copyright V
19428-2959.
162
CISCATO ET AL., doi:10.1520/STP20120007 163
Introduction
In the SCUBA (self-contained underwater breathing apparatus) field, gas mix-
tures with more than 21 vol. % oxygen became more and more popular in
recent years. The physical advantage for the human body is the reduction of
nitrogen during and after the dive. On the one hand, those benefits reduce the
health risks of decompression sickness after diving. On the other hand, this is
connected with additional costs for the extra gases and for the oxygen compati-
ble equipment.
There is still the widely held incorrect belief in the SCUBA diving commu-
nity that oxygen-enriched gas mixtures, so-called NITROX, with an oxygen
content of more than 21 vol. % can be considered as regular air. The historical
reason for this is that training manuals of diving organizations refer and cite
the U.S. Department of Labor Occupational Safety and Health Administration
Regulations 1910.430(i)(1) [1]. Therein, the common statement is made that
gas mixtures up to 40 vol. % oxygen can be considered as regular air. This has
been already refuted in many papers, e.g., Binder et al. [2].
In the last 10 years, the quantity of so-called “technical divers” increased.
In general, these divers use gas mixtures with more than 40 vol. % oxygen for
following reasons:
- as decompression gases during the surfacing phase of the diving event,
- to reduce the dive time, and
- to reduce the nitrogen and/or helium decompression stress, especially the
risk of decompression sickness in the human body after diving.
From the practical point of view, most of those “technical divers” use the
same valves for oxygen-enriched and for pure-oxygen applications. The reason
for this is the compatibility of the connection threads and, of course, having
only one equipment type to buy. In case of an emergency under water, SCUBA
breathing regulators have to be switched from one to another cylinder. There-
fore, the diving community prefers to use only one connection thread type for
all gas mixtures, as described in technical reports, such as in Ref 3. In contrast,
it is common practice for the gas industry and for the medical sector to use dif-
ferent connection threads.
In the past, the thread types for diving valves have been standardized accord-
ing to ISO 12209-1 [4] and ISO 12209-2 [5]. The G 5/800 and the so-called “INT”
threads are the most common connections for diving valves. However, at the end
of the 1990s, in the EEU, the corresponding CEN standardization committee
decided to adopt the common practice for the gas industry and for medical valves.
In addition, the EEU 97/23/EC, “Pressure Equipment Directive” [6], and the
“ADR” (European Agreement Concerning the International Carriage of Danger-
ous Goods by Road) [7] also apply to SCUBA practice. Therefore, equipment for
service with oxygen concentrations higher than 23 vol. % have to be treated as
being used for pure oxygen and have to pass the oxygen compatibility tests.
In 2003, the new CEN standard EN 144-3 [8] was created for oxygen-
enriched and for pure-oxygen diving valves. Therein, on the one hand, the new
connection thread M 26 2, and, on the other hand, the corresponding oxygen
test method according to EN 13949 [9] were combined.
This test method is nearly equal to the known oxygen test according to ISO
10297 [10] used for all gas industry and medical valves. In practice, however, the div-
ing community has not accepted those new standards up to now. In this way, the basic
concept of those new standards to enhance safety by testing the valves for burn-out
safety and by using the new thread-type are avoided by the diving community too.
In general, it should be well known, that not only the high pressure itself
but also the oxygen may cause problems. Oxygen is a nonflammable gas, but it
does enable and promote combustion. Even a slight enrichment beyond the
21 vol. % in atmospheric air may cause a more violent combustion and
increase the combustion velocity. Additionally, oxygen is able to cause sponta-
neous combustion. Also, in SCUBA practice, a lot of incidents happen every
year. Two examples are shown in Figs. 1 and 2.
Now, another dangerous practice has to be considered critical. In a market
study carried out by BAM Federal Institute for Materials Research and Testing,
and Gentoo Divers International, more than 60 representative cylinder valves
for SCUBA diving were examined. Sales offices claim that all of those valves
are suitable for oxygen-enriched gas mixtures.
The fact is that some of these cylinder valves are designed for air use only.
Replacing the nonmetallic materials and applying additional cleaning proce-
dures, these original air valves become so-called “oxygen clean.” Thereafter,
the valves are placed on the market as suitable for oxygen-enriched service.
This procedure is dangerous because the labeling pretends a pseudo-safety.
Experimental
The intention of the standardized oxygen impact test is to simulate compressive
heating or pressure shocks that may occur during service of oxygen systems
FIG. 1—Incident with 50 vol. % oxygen after opening the SCUBA valve.
FIG. 2—Incident with oxygen after opening the SCUBA valve.
and components. The test method is described for different gas cylinder valve
types in ISO 10297 annex D and in EN 13949 clause 5.2 for diving valves.
However, there is a discrepancy between both standards. The total number
of consecutive oxygen pressure shocks that have to be performed during testing
is, for diving valves, 50, and, for cylinder valves, only 20.
The tester consists of a connecting tube with an internal diameter of 5 mm
and with a length of 1000 mm between the quick opening valve and the sample
holder. In general, pure oxygen is used for the pressure shocks. Within this
examination, valves were tested with pure oxygen, as well as with nitrogen/
oxygen mixtures with an oxygen content of 40 vol. %.
The intention was to investigate differences in the materials’ reactions of the
same valve type. The oxygen temperature was 60 C. By opening of the quick-
opening valve, the samples were exposed to gaseous oxygen impacts from atmos-
pheric pressure to the final test pressure. The pressure rise time was 17.5 ms 6 2.5 ms.
For this market study, 61 representative SCUBA valves were examined; 16
were oxygen valves with threads, according to EN 144-3. The other 45 valves
were modified air valves with the other thread type. They were sold as so-
called “oxygen clean” and “oxygen compatible” valves for all diving applica-
tions with oxygen.

According to the standards, there are two possible test results: A burn-out
occurs and the valve fails or the valve passes the test. Significant traces of igni-
tion detected by visual inspection on the nonmetallic materials are the criterion
for a burn-out.
Because of the following test results, the authors of this paper, however,
think that an additional criterion called “visible reaction” might be justified.
This reaction describes a detectable change on the nonmetallic material that is
different to charring or ignition. This “visible reaction” may be a change in

color, distinct mechanical marks on the nonmetallic material (partial), melting
of the nonmetallic material, or other distinct material loss, as well.
Figures 3–5 are examples for typical burn-outs, whereas Figs. 6 and 7
show valve seats that did not burn or char, but show distinct changes on the
nonmetallic material. Although the criterion for a burn-out is not fulfilled, simi-
lar pressure shocks in real-life situations could lead to a malfunction of the
valve, e.g., a leakage.
Including the criterion “visible reaction,” Table 1 gives an overview on the
oxygen pressure shock test results for all valves. Table 2 gives an overview on
the oxygen pressure shock test results for the oxygen valves only.
There is a distinct large number of burn-outs. This may be caused by the dif-
ferent ages of so-called new valves. In some cases, they had been stored in a shelf
for up to 5 years. One can assume that aging of the nonmetallic materials, espe-
cially of the lubricants, may have had a strong influence on the burn-out safety.
On the other hand, 45 modified air valves for diving applications were
tested too. Table 3 gives an overview on the oxygen pressure shock test results
for those modified air valves.
At a first look, the test results seem to be quite good and acceptable. Only
five clear burn-outs did occur. This is a distinct lower percentage rate in
FIG. 3—Burn-out of a valve.

FIG. 4—Burn-out of another valve.
comparison to the test results of the oxygen valves. On closer examination, this
finding is explainable.
On the one hand, most of the modified air valves were tested right after their
manufacturing. The influence of aging on the nonmetallic seat could be detected
for a particular valve type that had been tested twice: new and older from the
manufacturing line. The new version passed the test, the old version failed.
On the other hand, the brand new valves revealed another interesting obser-
vation after disassembling. Although only a few technical documents were
available, distinct differences between old and new versions of the valves were
obvious. Apparently, technical improvements were done for the same valve
type. For example, a lubricant was used in the new valves that was more com-
patible with oxygen than in the older ones.
Some changes in the nonmetallic seat materials or in the inner geometry of
the valve were also detected. Although these modifications effect a lower quan-
tity of burn-outs, this practice is dangerous for the user, as he does not know
what kind of version he has bought and what sort of modifications have been
done to his valve.
In addition to the above-mentioned tests, five different valve types were
tested with NITROX-40 that is equal to an oxygen content of 40 vol. %. These
valve types had failed the oxygen impact test before and the intention now was
to find out if differences in the materials’ reactions do occur or not. Table 4
FIG. 5—Burn-out of a third valve.
FIG. 6—Traces of reaction on a valve seat.

FIG. 7—Traces of reaction on a valve seat.
TABLE 1—Overview on the oxygen pressure shock test results for all valves.
Total Amount of Tested Valves 61

Burn-out 13
Visible reaction 12
Passed 36
TABLE 2—Overview on the oxygen pressure shock test results for oxygen valves.
Total Amount of Tested Oxygen Valves 16

Burn-out 8
Visible reaction 2
Passed 6
TABLE 3—Overview on the oxygen pressure shock test results for modified air valves.
Total Amount of Tested Modified Air Valves 45

Burn-out 5
Visible reaction 11
Passed 29
TABLE 4—Overview on the oxygen pressure shock test results for valves with an oxygen content of
40 vol. %.
Total Amount of Tested Valves with a Content of 40 Vol. % 11

Burn-out 1
Visible reaction 4
Passed 6
gives an overview on the pressure shock test results with an oxygen content of
40 vol. %.
Only one valve failed the test indicated by a distinct burn-out. In four
valves, reactions on the seat materials were detected. However, six valves,
which had failed the pressure shock test with pure oxygen, now passed the
same test but with an oxygen content of 40 vol. %.

Many results of this examination confirm findings and comments of previous
publications. NITROX mixtures with an oxygen content of more than 21 vol.
% should be handled carefully like pure oxygen. In general, the regulations and
standards for pure oxygen should be also adopted for NITROX mixtures.
However, this paper reveals another dangerous practice in the SCUBA div-
ing community. From the safety point of view, the term “oxygen clean valves”
is misleading and dangerous. Divers assume the valves are both oxygen com-
patible and safe for all oxygen-enriched mixtures up to pure oxygen. Therefore,
there should be no labeling, such as “oxygen clean” or “oxygen compatible,”
on modified air valves.
In addition, aging of nonmetallic materials, especially of lubricants in div-
ing valves, may have a strong influence on the burn-out safety. Therefore, it
makes sense to check diving valves on a regular basis, e.g., perform mainte-
nance once a year.
Manufacturer and supplier of diving valves should inform users about
changes of nonmetallic materials or about a new version of the diving valve
type if changes are carried out for safety reasons. It is not tolerable that similar
looking valve versions have a different level of safety.
Diving valves and components should be tested for burn-out safety accord-
ing to national and international standards, as do valves for the gas industry or
medical valves.
But for the diver, it does not matter whether a distinct burn-out is detecta-
ble performing the oxygen pressure shock test or whether it is because of other
errors that the diving valve does not function properly. The fact is that errors,
which occur under water during diving, could have immediate consequences,
and even cause death to the diver. Therefore, it is hard to understand the com-
mon approaches of the SCUBA community.
References
[1] OSHA 1910.430, 1996, “Equipment Used With Oxygen or Mixtures Con-
taining Over Forty Percent (40%) by Volume Oxygen Shall Be Designed
for Oxygen Service,” U.S. Dept. of Labor Occupational Safety and Health
Administration, Washington, D.C.
[2] Binder, C., Brock, T., Hesse, O., Lehné, S., and Tillack, T., “Ignition Sen-
sitivity of Nonmetallic Materials in Oxygen-Enriched Air (NITROX): A
Never Ending Story in SCUBA Diving?” Flammability and Sensitivity
of Materials in Oxygen-Enriched Atmospheres: Vol. 12, ASTM STP 1522,
H. Barthélémy, T. A. Steinberg, and S. Smith, Eds., ASTM International,
West Conshohocken, PA, 2009.
[3] Ciscato, W., Szypkowski, M., Aris, S., and Nadeschdin, C.,
“Nitroxnormen—Was Gilt Nun?” Divesinde (German), Vol. 8, 2008, pp.
51–55.
[4] ISO 12209-1, 2000, “Gas Cylinders—Outlet Connections for Gas Cylin-
der Valves for Compressed Breathable Air—Part 1: Yoke Type Con-
nections,” International Standards Organization, Geneva, Switzerland.
[5] ISO 12209-2, 2000, “Gas Cylinders—Outlet Connections for Gas Cylin-
der Valves for Compressed Breathable Air—Part 2: Threaded Con-
nections,” International Standards Organization, Geneva, Switzerland.
[6] European Parliament, 1997, Pressure Equipment Directive (97/23/EC),
http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CONSLEG:
1997L0023:20031120:en:PDF (Last accessed December 7, 2011).
[7] United Nations, 1957, ADR, European Agreement Concerning the Inter-
national Carriage of Dangerous Goods by Road, http://www.unece.org/
trans/danger/publi/adr/adr_e.html (Last accessed December 7, 2011).
[8] EN 144-3, 2003, “Respiratory Protective Devices—Gas Cylinder
Valves—Part 3: Outlet Connections for Diving Gases Nitrox and Oxy-
gen,” European Committee for Standardization, Brussels, Belgium.
[9] EN 13949, 2003, “Respiratory Equipment—Open-Circuit Self-Contained
Diving Apparatus for Use with Compressed Nitrox and Oxygen—
Requirements, Testing, Marking,” European Committee for Standardiza-
tion, Brussels, Belgium.
[10] ISO 10297, 2006, “Transportable Gas Cylinders—Cylinder Valves—
Specification and Type Testing,” International Standards Organization,
Geneva, Switzerland.
STP 1561, 2012
DOI:10.1520/STP20120005
T. Tillack,1 C. Binder,1 O. Hesse,1 S. Lehné,1

and W. Ciscato2
Safe for Oxygen but Not for Oxygen-Enriched

Gas Mixtures in SCUBA Diving
REFERENCE: Tillack, T., Binder, C., Hesse, O., Lehné, S., and Ciscato, W.,
“Safe for Oxygen but Not for Oxygen-Enriched Gas Mixtures in SCUBA Div-
ing,” Flammability and Sensitivity of Materials in Oxygen-Enriched Atmos-
pheres on September 19–21, 2012 in Montreal, QC; STP 1561, Samuel
Edgar Davis and Theodore A. Steinberg, Editors, pp. 172–181, doi:10.1520/
STP20120005, ASTM International, West Conshohocken, PA 2012.
ABSTRACT: In a market study, more than 60 representative cylinder valves
that are used in SCUBA diving were tested for burn-out safety according to the
current standard ISO 10297 [2006, “Transportable Gas Cylinders—Cylinder
valves—Specification and Type Testing,” International Standards Organiza-
tion, Geneva, Switzerland]. This standard requires that oxygen cylinder valves
have to be pressure-surge tested to prove their ignition resistance. After test-
ing, the samples are disassembled and the nonmetallic materials shall not
show any traces of ignition for the valve to pass the test. This is usually done
by visual inspection. The existing pass–fail criteria, however, cause problems
sometimes when the test results have to be interpreted or are not sufficient to
ensure a safe use of the valve. This market study reveals what changes a non-
metallic material may show after pressure-surge testing. It also shows the
influence of the different design geometries in the seat area on the nonmetallic
materials. After testing, some nonmetallic materials do not show any distinct
traces of ignition or combustion. Only a change in color, notable surface deteri-
oration, and loss of material is visible. Occasionally, the valve is no longer gas-
tight, although no traces of ignition can be detected. However, according to the
standard, the valve has passed the test. This paper provides various reasons
why the pass–fail criteria of the oxygen pressure surge need to be more tight-
ened. This applies not only to the current version of the standard ISO 10297,
November 2012.
1
Specialist of Working Group “Safe Handling of Oxygen,” Division 2.1 “Gases, Gas Plants,”
BAM Federal Institute for Materials Research and Testing, Berlin 12205, Germany.
2
Expert of Gentoo Divers Intl., Duebendorf 8600, Switzerland.
Copyright V
19428-2959.
172
but for any standard that requires performance of the pressure-surge test for
oxygen components.
KEYWORDS: valve seat, SCUBA diving, pneumatic impact, burn-out safety,
oxygen, NITROX, modified air valves, aging of diving valves, oxygen
compatibility
Introduction
At BAM Federal Institute for Materials Research and Testing, the working
group “Safe Handling of Oxygen” tests nonmetallic, as well as metallic, mate-
rials for reactivity with oxygen. Furthermore, the burn-out safety of oxygen
components, such as valves, hoses, fittings, and other components for oxygen
service, are tested also.
According to numerous national and international standards, pressure-
surge testing is the only way to check the burn-out safety of oxygen compo-
nents for high-pressure service. Generally after testing, the samples are
disassembled and visually inspected. The nonmetallic materials shall not show
any traces of ignition for the oxygen component to pass the test. This pass–fail
criteria sometimes causes problems if the test results are equivocal. Then it is
difficult to decide whether the oxygen component can be considered burn-out
safe or not.
Using a pneumatic impact tester that allows oxygen test pressures up to
450 bar [1], BAM performed a market study. Sixty-one representative SCUBA
valves were tested for burn-out safety. A closer look at this pool of data and
pictures of the disassembled valves caused the authors to question the criterion
for a burn-out according to the existing standards. In contrast to the standards,
which consider various traces of ignition, BAM applied additional means and
indications to determine a reaction. These include the following: illumination
and optical magnification, a change in color, distinct mechanical marks or ther-
mal impacts on the nonmetallic material, melting of the nonmetallic material,
and other distinct material losses.
Figure 1 presents a nonmetallic seat material that shows no traces of igni-
tion. However, distinct baring or channeling of the material is detectable. In
Fig. 2, the seat surface is destroyed without traces of ignition.
According to the criterion of the existing standards, those valves would
have passed the test even if the shown damages could result in a limited or in a
complete malfunction of the SCUBA valve.
Especially in the diving sector, this might pose an enormous risk. If a
SCUBA valve does not function properly, the end result could be the death of
the diver. Even if in other applications, the consequences are not as clear as in
the diving sector, there are enough reasons why the pass–fail criteria of the
oxygen pressure shock test should be more stringent.
FIG. 1—Valve seat with baring or channeling of the material.
Experimental
In general, the intention of the oxygen impact test is to simulate compressive
heating or pressure shocks that may occur during filling processes or in prac-
tice. The test method itself is described in several standards, e.g., in the ISO
10297 [2].
FIG. 2—Valve seat with destroyed surface.

A connecting tube, with a length of 1000 mm and with an inner diameter

of 5 mm, was used between the test sample and the quick-opening valve. By
opening of the hydraulically operated quick-opening valve, the test samples
were exposed to oxygen pressure shocks. Usually, the oxygen temperature was
60 C. The pressure rise time between ambient pressure and the final test pres-
sure was 17.5 ms 6 2.5 ms.
According to the standard, a minimum of three identical test samples of the
same valve type have to be tested. It is not only required that samples are iden-
tical in design features, but also of the same production period and production
bench.
After the test, the valves were disassembled. First, the standardized pass–-
fail criterion was checked. Thereafter, the nonmetallic materials were checked
for all kinds of damages in more detail. Damages and observations were noted
or documented photographically. For this purpose, BAM uses a magnifying
glass with a magnification of 2.0 to 2.5. This is sufficient to help put details of
the surface closer to the human eye. By doing so, the number of artifacts in the
surface is only slightly increased. A higher magnification would lead to more
details and allow more room for interpretation.
At BAM, an additional common practice is to check the technical drawings
of the valves and to compare them with the real-life conditions. If differences
are found, they are also noted.

Sixty-one valves were examined. Significant traces of ignition or, in other
words, a burn-out according to ISO 10297 [2], were detected in 13 valves.
After disassembling the valves, not only any trace of ignition but also any
detectable change on the nonmetallic materials was qualified as “visible reac-
tion.” Generally, up to now, the latter valves would not have been classified as
failed but would have passed the test according to the criterion for a burn-out.
However, using “visible reaction” as an additional criterion for the test
results, 12 of 61 valves of the above-mentioned market study are affected. On
the other hand, 36 SCUBA valves showed neither traces of ignition nor any
other “visible reactions.” Therefore, a closer look at those with “visible reac-
tions” would make sense for safety reasons.
A beginning reaction of a nonmetallic material with oxygen may be indi-
cated by a change in color. Figure 3 shows an original white seat of a SCUBA
valve. In contrast, Fig. 4 presents the same seat after the oxygen impact test. In
this case, the blackening is considered a “visible reaction,” even if no charring
or ignition of the material occurred.
Another kind of “visible reaction” is indicated by distinct mechanical
marks or thermal influences on the nonmetallic material. The valve seat in
FIG. 3—Original fresh valve seat.
Fig. 5 is heavily damaged without showing traces of ignition on the nonmetal-

lic material. In this particular case, the valve would have no longer been gas-
tight. However, according to the existing pass–fail criterion of the standard,
this SCUBA valve would have passed the test, because no safety-related reac-
tion with oxygen is detectable by visual inspection.
FIG. 4—Valve seat after oxygen impact test.

FIG. 5—Mechanically damaged valve seat.
Figure 6 shows the influence of thermal heating on the valve seat. As

shown in Fig. 2, the surface of the nonmetallic material is affected in such a
way that the valve seat is no longer able to work properly.
FIG. 6—Valve seat after thermal heating.

FIG. 7—Melting of a valve seat.
Figures 7 and 8 serve as an example of melting of the nonmetallic seat ma-

terial and another example of distinct material loss. In both cases, this may be
linked to some sort of reaction of the nonmetallic material under the test condi-
tions. If the number of test samples is increased, simultaneously increasing the
level of confidence, this result could also be a burn-out.
FIG. 8—Material loss after testing.

FIG. 9—Evident differences between so-called identical test samples (white

circles).
Checking technical drawings is an important tool to determine if and how

the quality assurance system of the manufacturer works. As an example, Fig. 9
presents distinct differences between so-called “identical” SCUBA valves.
Obviously, the bore holes have different angles. On one hand, this shows
an insufficient quality assurance during manufacturing. On the other hand, the
results of the oxygen pressure shock test shifted from pass to fail for the same
valve type. A possible explanation is that the different bore angles produce dif-
ferent flow conditions inside the valve and therefore pose a different risk of a
burn-out for the seat material.
The examples show that the requirements of the standards need more spec-
ification in detail. From a safety point of view, at least, the above-mentioned
“visible reactions” are easily observed and verified and should be used as
“failed” criteria because of the possible dangerous consequences for the user,
especially for divers.
Figure 9 shows that the quality of manufacturing also seems to be an issue.
The importance of quality assurance and of batch testing on nonmetallic mate-
rials for oxygen service was previously presented by the authors [3]. In addi-
tion, it shall be clear that differences found by checking the drawings have to
be considered as a “failed” criterion.
The certification of a valve by an accredited independent third party might
be a solution to the problem. In this context, it includes design type testing of
the valve, assessment, and monitoring of the manufacturer’s quality assurance
system and production.

Components for use in high-pressure oxygen service have to be tested for burn-
out safety. For example, the current standard ISO 10297 [2] describes the test
procedure for cylinder valves. After testing, the samples are disassembled and
the nonmetallic materials shall not show any traces of ignition for the valve to
pass the test. This is usually being done by visual inspection. However, the
existing pass–fail criterion seems not to be sufficient as experience shows,

especially in the diving community.
The revised version ISO=CD 10297 [4] of the above-mentioned standard
solves this problem by specifying the pass–fail criterion in more detail. In the
revised version a close visual examination of the nonmetallic materials is
required. However, there is no explicit description on what the term “close vis-
ual examination” means in practice. Examples for traces of ignition are given
as notable surface deterioration, change in surface texture, change in color,
and=or material loss. In addition, the test sample will also fail if parts of the
component are displaced, non-functional, or missing.
In a market study, more than 60 representative cylinder valves that are
used in SCUBA diving were tested. The results were used to check the existing
and the revised pass–fail criterion.
The revised version of the standard is a good step forward to enhance the
safety of oxygen cylinder valves. However, the authors of this paper recom-
mend the phrase “close visual examination” be more specified. Additional
means, such as optical magnification, should be permitted and mentioned in
the standard. A magnification factor of about 2.0 to 2.5 seems to be most prac-
ticable for examination. In addition, photos may help to implement the correct
use of the pass–fail criterion.
But this is a first step only. In this paper, additional criteria such as “visible
reaction” and “check of technical drawing” are presented and evaluated. Imple-
menting these criteria into the standard and using the possibility of certification
should be a further step to enhance the safety of oxygen components.
The valve type presented in Fig. 9 with the different bore holes geometries
should fail the test for previously mentioned reasons. It is not possible to per-
form the test with three identical samples. On the other hand, the discrepancies
between the technical drawings and the manufactured valve are an indication
of inadequate quality assurance.
The intention of this paper is to reveal the problem that components like
cylinder valves are being considered burn-out safe for oxygen service but are
not suitable for use in diving. This is shown by consequently applying the
actual pass–fail criterion according to the standard.
References
[1] Binder, C., Kieper, G., and Herrmann, P., “A 500 Bar Gaseous Oxygen
Impact Test Apparatus for Burn-Out Testing of Oxygen Equipment,”
Flammability and Sensitivity of Materials in Oxygen-Enriched Atmos-
pheres: 7th Volume, ASTM STP 1276, D. D. Janoff, W. T. Royals, and
M. V. Gunaji, Eds., ASTM International, West Conshohocken, PA, 1995.
[2] ISO 10297, 2006, “Transportable Gas Cylinders—Cylinder Valves—

Specification and Type Testing,” International Standards Organization,
Geneva, Switzerland.
[3] Binder, C., Arlt, K., Brock, T., Hartwig, P., Hesse, O., and Tillack, T.,
“The Importance of Quality Assurance and Batch Testing on Nonmetallic
Materials Used for Oxygen Service,” Flammability and Sensitivity of
Materials in Oxygen-Enriched Atmospheres: 12th Volume, ASTM STP
1522, H. Barthélémy, T. A. Steinberg, and S. Smith, Eds., ASTM Interna-
[4] ISO=CD 10297, 2011, “Transportable Gas Cylinders—Cylinder Valves—
Specification and Type Testing,” Committee Draft Document N881 of
ISO=TC 58=SC 2=WG 6, International Standards Organization, Geneva,
Switzerland.
STP 1561, 2012
DOI:10.1520/STP20120017
Joel M. Stoltzfus,1 Nathan Jeffers,2

and Timothy D. Gallus2
Burning of CP Titanium (Grade 2)

in Oxygen-Enriched Atmospheres
REFERENCE: Stoltzfus, Joel M., Jeffers, Nathan, and Gallus, Timothy D.,
“Burning of CP Titanium (Grade 2) in Oxygen-Enriched Atmospheres,”
ABSTRACT: The flammability in oxygen-enriched atmospheres of commer-
cially pure (CP) titanium rods as a function of diameter and test gas pressure
was determined. Test samples of varying diameters were ignited at the bot-
tom and burned upward in 70 % O2/balance N2 and in 99.5þ % O2 at various
pressures. The burning rate of each ignited sample was determined by
observing the apparent regression rate of the melting interface (RRMI) of the
burning samples. The burning rate or RRMI increased with decreasing test
sample diameter and with increasing test gas pressure and oxygen concen-
tration. The oxygen concentration had a much more marked affect on the
burning rate than oxygen pressure.
KEYWORDS: flammability, oxygen, titanium, RRMI
Background
Even though titanium and its alloys are known to be easily ignited [1] and,
once ignited, burn vigorously in oxygen-enriched atmospheres [2–4], parts
Manuscript received February 14, 2012; accepted for publication July 20, 2012; published online
November 2012.
1
Oxygen Group Project Leader, Materials and Components Laboratories Office, NASA Johnson
Space Center White Sands Test Facility, Las Cruces, NM 88012.
2
Oxygen Group Engineer, NASA Test and Evaluation Contract, NASA Johnson Space Center
White Sands Test Facility, Las Cruces, NM 88012.
Copyright V
19428-2959.
182
made from these materials are sometimes exposed to that environment. For
example, titanium alloy blades are used in turbine engines which, by virtue of
their high pressures, could be considered oxygen enriched. The conditions
were chosen to aid in a failure analysis (the details of which are proprietary),
as described in the next paragraph. The turbine engine reference is provided
only as an example of commercially pure (CP) titanium in an O2-enriched
environment.
To assist in a failure analysis, the NASA Johnson Space Center White
Sands Test Facility (WSTF) Oxygen Group was asked to determine the flam-
mability of CP titanium as a function of test sample diameter, test gas pressure,
and oxygen concentration.
Experimental
Test Samples
The test material was CP titanium (Grade 2), UNS R50400, with a composition
(weight %) as follows: C 0.1 max, Fe 0.3 max, H 0.015 max, N 0.03 max,
O 0.25 max, and Ti 99.2.
Test samples were configured as 6.0-in.- (15.2-cm-) long rods having diam-
eters of 1/32 in. (0.8 mm), 1/16 in. (1.6 mm), 1/8 in. (3.2 mm), 3/16 in.
(4.8 mm), and 1/4 in. (6.4 mm). With the exception of the 1/32-in.- and 1/16-
in.- (0.8-mm- and 1.6-mm-) diameter wires, the samples were prepared for test
by machining a 0.030-in.- (0.762-mm-) wide slot through the bottom end
through which the igniter wire could be passed to ensure ignition (see Fig. 1).
Ignition Promoters
CP titanium (Grade 2) wire with a 0.010-in. (0.254-mm) or 1/32-in. (0.8-mm)
diameter was used as the ignition promoter. This wire was wrapped around the
bottom end of the test sample and resistively heated to induce melting and sub-
sequent ignition of the igniter wire. Within a fraction of a second of applying
power to the ignition promoter, it ignited. The burning igniter wire heated,
melted, and ignited the sample to which it was attached.
For the 1/32-in. (0.8-mm) sample, 16 wraps of 0.010-in.- (0.254-mm-) di-
ameter wire were wound around the bottom end of the test sample (Fig. 2). For
the 1/16-in. (1.6-mm) and 1/8-in. (3.2-mm) samples, four wraps of 1/32 in.
(0.8-mm) diameter were used (Fig. 2). For the 3/16-in. (4.8-mm) diameter sam-
ple, a total of eight wraps of 1/32-in. (0.8-mm) diameter were used: four wraps
were wound through the slot and four wraps were wound on top of the wraps in
the slot (Fig. 3). For the 1/4-in.- (6.4-mm-) diameter sample, a total of eight
wraps of 1/32-in. (0.8-mm) diameter were used: five wraps were wound
through the slot and three wraps were wound on top of the wraps in the slot
FIG. 1—Photo of CP titanium samples prior to test.
(Fig. 3). The purpose of the igniter was to initiate burning, without introducing
a foreign material. The authors had no interest in the ignition energy; hence, no
estimates of igniter energy were made.
Test Apparatus
The tests were conducted in the WSTF High Pressure Promoted Combustion
Test Facility. It comprised a 757-in. [3] (12-L) test chamber (Fig. 4), a 6000-
psi (41-MPa) oxygen supply system, a 3000-psi (21-MPa) mixed-gas supply
system, pressure recording equipment, and 200-frame-per-second high-speed
digital camera. The digital camera was used to record test sample burning by
observing the regression rate of the melting interface (RRMI) as it progressed
up the test sample after ignition. The video images were recorded through a
FIG. 2—1/32-, 1/16-, and 1/8-in.- (0.8-, 1.6-, and 3.2-mm-) diameter samples
with ignition promoters installed.
2-in.- (5-cm-) diameter sapphire sight glass located adjacent to the bottom end
of the test sample.
The test samples were suspended vertically and ignited at the bottom using
multiple wraps of resistively heated Grade 2 titanium wire. If the test sample
did not propagate, a post-test inspection of each sample verified that it had
been melted by the igniter wire to ensure the test was valid.
Test Procedure
The samples were washed in soapy, deionized water for 5 min, rinsed ten times
with deionized water, and blown dry with filtered nitrogen. The samples were
individually heat sealed in clean polyethylene bags.
FIG. 3—3/16- and 1=4-in.- (4.8- and 6.4-mm-) diameter samples with ignition
promoters installed.
FIG. 4—Schematic of WSTF 757-in.3 (12-L) promoted combustion test

chamber.
Just prior to testing, each sample was removed from its bag and the igniter
wire was installed. The sample was then mounted vertically in the chamber,
being secured at the top (Fig. 5). The igniter leads were connected to the
electrical power supply feedthroughs (Fig. 5), and the chamber was sealed.
FIG. 5—Test sample installed in test chamber.

The chamber was purged three times to 500 psig (3.5 MPa) using the test gas
and vented to ambient pressure. After purging, the test chamber was pressur-
ized to the test pressure, the data-acquisition equipment was turned on, and the
igniter was initiated by applying power. The test results were observed and
recorded. The test chamber was vented, purged, and opened, allowing the sam-
ple’s post-test condition to be observed. The burn rates (RRMI) were calcu-
lated using measurements taken from the digital video recordings.
Test Conditions
Tests were conducted on samples of each test diameter at 25 and 200 psia (0.2
and 1.4 MPa) in 70% O2/balance N2. In addition, tests were conducted using
1/8-in.- (3.2-mm-) diameter samples in 99.5þ % O2 at 12.4, 25, 50, 100, and
200 psia (0.09, 0.2, 0.4, 0.7, and 1.4 MPa). At least one test, and in some cases
two tests, was conducted at each condition. In all tests, the initial test sample,
igniter wire, and test chamber temperatures were ambient.
Results
Uncertainty Analysis
The reported RRMI for each test was determined by analysis of the digital
video recordings. Data sets of elapsed time after the camera trigger point and
pixel location were obtained at the first moment of drop separations. Calcula-
tions from drop-to-drop differences of melting interface locations divided by
the corresponding difference of elapsed time produced the reported RRMIs.
This method contains four sources of measurement error, which give rise to
RRMI uncertainty.
Uncertainty in the RRMI was introduced by the error associated with jitter
in the camera clock; pixel bleed because of brightness of the event; melting
interface position determination by operator judgment; and tilt of the melting
interface caused by precessional motion of the molten burning mass attached
to the sample rod [5]. Of these error sources, the last two are by far the domi-
nant sources of error.
The digital camera used to record video images has a highly accurate
clock; and because the images were recorded at 200 frames per second of a
slow event, the timing error adds very little to the uncertainty of the RRMI
determinations. The bleed of brightness from the molten droplet over the melt-
ing interface area of the image could have introduced the dominant source of
position determination error. However, by adjusting the brightness and contrast
of each video recording before the melting interface location was obtained,
that error source was reduced to an insignificant level.
FIG. 6—Burning droplet cycle for test number 7402 with melting interface tilt indicated.
The two remaining sources of error were the major contributors to uncer-
tainty in the five rod diameter cases, and both are related to distance measure-
ments. At the zoom level used for video recordings, an operator could only
reliably determine position of any point to within 61.5 pixels. This translated
into a worst case error of 62.7% for distance and, therefore, RRMI for all
recordings.
The natural random tilt of the melting interface because of the precession
of the molten, burning droplet around the bottom end of the test sample was
another dominant source of error. Figure 6 shows how the melting interface tilt
varied throughout test #7402. An operator could estimate the average position
of an interface that was tilted on an axis parallel with the camera view. How-
ever, in every frame it is uncertain whether or not the melting interface is tilted
around an axis perpendicular to the camera view. The maximum tilt of the
melting interface appeared constant through all video recordings for all sample
sizes at 615 degrees. With the maximum tilt constant, larger rod diameters
allowed for a larger total vertical offset uncertainty of the melting interface.
At the smallest two rod diameters, the uncertainty caused by operator
melting interface determination dominated. At the larger three sample rod
diameters, uncertainty caused by melting interface tilt dominated. The esti-
mated total uncertainties are listed in Table 1.
Tests in 70% O2/balance N2 at 25 psia

Test samples with diameters ranging from 1/32 to 3/16 in. (0.8 to 4.8 mm)
supported burning in 70% O2/balance N2 at 25 psia (0.2 MPa) (Table 2). The
RRMI decreased with increasing test sample diameter. The 1/4-in.- (6.4-mm-)
diameter sample ignited but did not support self-sustained burning in all nine
tests conducted. In all other cases, the samples burned completely once ignited.
A plot of the RRMI as a function of test sample cross-sectional area is
shown in Fig. 7. A least-squares curve fit through the non-zero burning rate
data yields a power curve, which describes the burning rate as follows (Eq 1):
RRMI ¼ 0:0025 A0:705 (1)
TABLE 1—RRMI uncertainty as a function of sample rod diameter.
Rod Diameter
in. mm RRMI Uncertainty (6%)

1/32 0.8 2.7
1/16 1.6 2.7
1/8 3.2 5.1
3/16 4.8 7.6
1/4 6.4 10
TABLE 2—Test results in 70% O2/balance N2 at 25 psia (0.2 MPa).
Sample Diameter Cross-Sectional Area RRMI
Test No. in. mm in.2 cm2 in./s cm/s

7390 1/32 0.0313 0.7950 0.0008 0.0052 0.38 6 0.010 0.97 6 0.025
7391 1/32 0.0313 0.7950 0.0008 0.0052 0.45 6 0.012 1.14 6 0.031
7392 1/16 0.0625 1.5875 0.0031 0.0199 0.14 6 0.004 0.36 6 0.010
7393 1/16 0.0625 1.5875 0.0031 0.0199 0.13 6 0.004 0.33 6 0.010
7394 1/8 0.1250 3.1750 0.0123 0.0794 0.07 6 0.004 0.18 6 0.010
7395 1/8 0.1250 3.1750 0.0123 0.0794 0.06 6 0.003 0.15 6 0.008
7377 3/16 0.1875 4.7625 0.0276 0.1781 0.03 6 0.002 0.08 6 0.005
7378 3/16 0.1875 4.7625 0.0276 0.1781 0.03 6 0.002 0.08 6 0.005
7359–7367 1/4 0.2500 6.3500 0.0491 0.3168 0a 0a
a
Samples ignited but did not propagate. A total of nine tests were conducted at this test condition.
where:
RRMI ¼ regression rate of the melting interface (burning rate) [in./s] and
A ¼ test sample cross-sectional area [in.2].
When the cross-sectional area of the test sample was less than 0.0123 in.2
(0.0794 cm2), or the diameter of the test sample was less than 1/8 in. (3.2 mm),
the burning rate increased dramatically.
FIG. 7—Plot of burning rate (RRMI) in 70% O2/balance N2 at 25 psia

(0.2 MPa) as a function of test sample cross-sectional area. (The uncertainty in
each data point is smaller than the height of the symbol and is, therefore, not
indicated on the plot.)
TABLE 3—Test results in 70% O2/balance N2 at 200 psia (1.4 MPa).
Sample Diameter Cross-Sectional Area RRMI
Test No. in. mm in.2 cm2 in./s cm/s

7385 1/32 0.0313 0.7950 0.0008 0.0052 1.81 6 0.049 4.60 6 0.125
7383 1/16 0.0625 1.5875 0.0031 0.0199 0.48 6 0.013 1.22 6 0.033
7384 1/16 0.0625 1.5875 0.0031 0.0199 0.50 6 0.014 1.27 6 0.036
7381 1/8 0.1250 3.1750 0.0123 0.0794 0.22 6 0.011 0.56 6 0.028
7382 1/8 0.1250 3.1750 0.0123 0.0794 0.21 6 0.011 0.53 6 0.028
7380 3/16 0.1875 4.7625 0.0276 0.1781 0.15 6 0.011 0.38 6 0.028
7379 1/4 0.2500 6.3500 0.0491 0.3168 0.10 6 0.010 0.25 6 0.025
Tests in 70% O2/balance N2 at 200 psia

All the test samples supported burning and burned completely in 70% O2/bal-
ance N2 at 200 psia (1.4 MPa) (Table 3). The RRMI decreased with increasing
test sample diameter.
A plot of the RRMI as a function of test sample cross-sectional area is
shown in Fig. 8. A least-squares curve fit through the non-zero burning rate
data yields a power curve, which describes the burning rate as follows (Eq 2):
RRMI ¼ 0:0116 A0:705 (2)
FIG. 8—Plot of burning rate (RRMI) in 70% O2/balance N2 at 200 psia

(1.4 MPa) as a function of test sample cross-sectional area. (The uncertainty in
TABLE 4—Test results of 1/8-in.- (3.2-mm-) diameter Ti rods in 70% O2/balance N2.
Test Pressure RRMI
Test No. psia MPa in./s cm/s

7394 25.0 0.2 0.065 6 0.003 0.165 6 0.008
7401 50.0 0.4 0.10 6 0.005 0.254 6 0.013
7402 100.0 0.7 0.16 6 0.008 0.406 6 0.020
7381 200.0 1.4 0.215 6 0.011 0.546 6 0.028
7382 200.0 1.4 0.21 6 0.011 0.533 6 0.028
where:
Similar to the 25 psia data, when the cross-sectional area of the test sample
was less than 0.0123 in.2 (0.0794 cm2), or the diameter of the test sample was
less than 1/8 in. (3.2 mm), the burning rate increased dramatically.
The test results of 1/8-in.- (3.2-mm-) diameter CP titanium (Grade 2) rods
in 70% O2/balance N2 as a function of test gas pressure are shown in Table 4.
As expected, the burning rate (RRMI) increased with increasing pressure and
all the samples burned completely, once ignited.
A plot of the RRMI as a function of test gas pressure is shown in Fig. 9. A
least-squares curve fit through the burning rate data yields a logarithmic func-
tion, which describes the burning rate as follows (Eq 3):
RRMI ¼ 0:0732 lnð AÞ0:1769 (3)
FIG. 9—Plot of burning rate (RRMI) of 1/8-in. (3.2-mm) diameter CP Ti rods

in 70% O2/balance N2 as a function of test gas pressure. (The uncertainty in
TABLE 5—Test results of 1/8-in. diameter Ti rods in 95.5þ % O2 as a function of test pressure.
Sample Diameter Test Gas Pressure RRMI
Test No. in. mm psia MPa in./s cm/s

7396 1/8 0.1250 3.2 12.4 0.09 0.31 6 0.016 0.787 6 0.041
7397 1/8 0.1250 3.2 25 0.2 0.49 6 0.025 1.245 6 0.064
7398 1/8 0.1250 3.2 50 0.4 1.26 6 0.064 3.20 6 0.163
7399 1/8 0.1250 3.2 100 0.7 2.41 6 0.123 6.121 6 0.312
7400 1/8 0.1250 3.2 200 1.4 3.22 6 0.164 8.179 6 0.417
where:
Tests in 99.5þ % Oxygen

The test results of 1/8-in.- (3.2-mm-) diameter CP titanium (Grade 2) rods in
99.5þ % O2 as a function of test gas pressure are shown in Table 5. Every sam-
ple burned completely once ignited and, as expected, the burning rate (RRMI)
increased with increasing pressure.
A plot of the RRMI as a function of test gas pressure is shown in Fig. 10.
A least-squares curve fit through the burning rate data yields a logarithmic
curve, which describes the burning rate as follows (Eq 4):
FIG. 10—Plot of burning rate (RRMI) of 1/8-in. (3.2-mm) diameter CP Ti rods

in 100% O2 as a function of test pressure.
RRMI ¼ 1:1135 lnð AÞ2:8163 (4)
where:
Discussion
It is clear that CP titanium will undergo complete, self-sustained burning if
ignited in 70% O2/balance N2. In addition, it is evident that the burning rate, as
indicated by the RRMI, increases as the cross-sectional area decreases. In
test gas pressures of 25 and 200 psia (0.2 and 1.4 MPa), the burning rate
increased dramatically when the cross-sectional area was less than 0.0123 in.2
(0.0794 cm2), which corresponds to a test sample diameter of 1/8 in. (3.2 mm).
Effect of Test Gas Pressure (70% O2/balance N2)

When test gas pressure was increased, the burning rate increased for all diame-
ters of CP titanium tested. The magnitude of the increase ranged from 4.8 times
for the 1/8-in.- (3.2-mm-) diameter samples to 3.3 times for the 3/16-in.- (4.8-
mm-) diameter samples. A comparison of the burning rates in 70% O2/balance
N2 is shown in Fig. 11.
FIG. 11—Comparison of the burning rates of CP Ti rods in 70% O2/balance

N2 as a function of test gas pressure. (The uncertainty in each data point is
smaller than the height of the symbol and is, therefore, not indicated on the
plot.)
FIG. 12—Comparison of burning rates of 1/8-in. (3.2-mm) diameter CP Ti in

pure oxygen versus mixed gas. (The uncertainty in each 70% O2/balance N2
data point is smaller than the height of the symbol and is, therefore, not indi-
cated on the plot.)
Effect of Oxygen Concentration

The effect of oxygen concentration on the burning rate of 1/8-in.- (3.2-mm-)
diameter rods of CP titanium is shown in Fig. 12. At 25 psia (0.2 MPa), the
rods tested in 99.5þ % oxygen burned approximately 5 times the rate of sam-
ples tested in 70% O2/balance N2. At 50 psia (0.4 MPa), the rods tested in
99.5þ % oxygen burned approximately 12.5 times the rate of samples tested
in 70% O2/balance N2. At 100 and 200 psia (0.7 and 1.4 MPa), the rods
tested in 99.5þ % oxygen burned approximately 15 times the rate of samples
tested in 70% O2/balance N2.
The burning rate in 200 psia (1.4 MPa) of 99.5þ % oxygen was 6.6 times
greater than the burning rate in 25 psia (0.2 MPa) of 99.5þ % oxygen. In con-
trast, the burning rate in 200 psia (1.4 MPa) of 70% O2/balance N2 was only
3.3 times greater than the burning rate in 25 psia (0.2 MPa) of 70% O2/balance
N2. It is probable that this was because of the presence of the 30% N2 in the
mixed gas test atmosphere. It is surmised that the N2 significantly increased the
heat transfer from the burning test sample at elevated pressure, reducing the
effect of the greater quantity of oxygen available at that pressure.
Direct comparison to previous flammability studies [6,7] of titanium and ti-
tanium alloys in O2-enriched atmospheres is not possible, because in those
studies burning rates were not measured. They presented threshold pressures
(the minimum pressure required to support self-sustained burning) but did not
record RRMI. It was observed in the McIlroy paper [7] that 1/8-in.- (3.2-mm-)
diameter CP titanium (ERTi-2) welding filler rods will support burning at
oxygen concentrations lower than 70% O2/balance N2. They report that in 60%
O2/balance N2, CP titanium (ERTi-2) will not support burning at 24.7 psia
(0.17 MPa) but will burn at 39.7 psia (0.27 MPa) once ignited. In 50% O2/bal-
ance N2 it will not burn at 39.7 psia (0.27 MPa) but will at 64.7 psia
(0.45 MPa). In 40% O2/balance N2, they report no self-sustained burning at
114.7 psia (0.79 MPa) but self-sustained burning at 214.7 psia (1.48 MPa). And
finally, they observed no self-sustained burning at up to 214.7 psia (1.48 MPa)
in 30% O2/balance N2.
Conclusion
CP titanium (Grade 2) is flammable in 70% O2/balance N2 pressures as low as
25 psia (0.2 MPa) up to 3/16-in. (5.8-mm) diameter. At 200 psia (1.4 MPa), test
samples with diameters up to 1/4 in. (6.4 mm) supported self-sustained burning.
The burning rate increased as cross-sectional area decreased.
For all diameters of CP titanium and oxygen concentrations tested,
the burning rate increased as the test gas pressure was increased. In 70%
O2/balance N2, the magnitude of the increase ranged from 3.3 to 4.8 times for
the various diameter samples. In 95.5þ % O2, the burning rate of test samples
with 1/8-in. (3.2-mm) diameters increased over an order of magnitude when
the test gas pressure was increased from 12.4 to 200 psia (0.09 to 1.4 MPa).
The oxygen concentration had a much more marked affect on the burning
rate of CP titanium wire and rods than did oxygen pressure.
References
[1] Key, C. F. and Reihl, W. A., “Compatibility of Materials with Liquid

Oxygen,” NASA Technical Memorandum (TM) X-985, NASA Marshall
Space Flight Center, Huntsville, AL, 1964.
[2] Laurendeau, N. M., “The Ignition Characteristics of Metals in Oxygen-
Enriched Atmospheres,” Test Report TR-851, Princeton University,
Princeton, NJ, 1968.
[3] McKinley, C., “Experimental Ignition and Combustion of Metals,” Oxy-
gen Compressors and Pumps Symposium, Compressed Gas Association,
Chantilly, VA, 1971, pp. 27–34.
[4] Pelouch, J. J., “Characteristics of Metals that Influence System Safety,”
ASRDI Oxygen Technology Survey, Volume VII, NASA Special Publica-
tion SP-3077, Aerospace Safety Research and Data Institute, National
Aeronautics and Space Administration, Washington, D.C., 1974.
[5] Edwards, A. P. R., Chiffoleau, G. J. A., Maes, M. J., and Steinberg, T. A.,
“Instantaneous RRMI of Iron Rods in Reduced Gravity,” Symposium on

pheres: Tenth Volume, ASTM STP 1454, T. A. Steinberg, B. E. Newton,
and H. D. Beeson, Eds., ASTM International, West Conshohocken, PA,
2003, pp. 91–101.
[6] Gunaji, M. V., Sircar, S., and Beeson, H. D., “Ignition and Combustion of
Titanium and Titanium Alloys,” Symposium on Flammability and Sensi-
tivity of Materials in Oxygen-Enriched Atmospheres: 7th Volume, ASTM
STP 1267, D. D. Janoff, W. T. Royals, and M. V. Gunaji, Eds., ASTM
International, West Conshohocken, PA, 1995, pp. 81–85.
[7] Zawierucha, R., McIlroy, K., and Million, J. F., “Promoted Ignition-
Combustion Behavior of Light Metals in Oxygen Enriched Atmospheres,”
Symposium on Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres: 7th Volume, ASTM STP 1267, D. D. Janoff, W. T.
Royals, and M. V. Gunaji, Eds., ASTM International, West Consho-
hocken, PA, 1995, pp. 69–80.
STP 1561, 2012
DOI:10.1520/STP20120009
Martina Ridlova,1 Jean-Christophe Rostaing,1

Alain Colson,2 and Olivier Longuet3
Flammability Evaluation of Metals in Piping

Configurations Under Flowing Conditions:
Hollow-Vessel Promoted Ignition Test
REFERENCE: Ridlova, Martina, Rostaing, Jean-Christophe, Colson, Alain,

and Longuet, Olivier, “Flammability Evaluation of Metals in Piping Configura-
tions Under Flowing Conditions: Hollow-Vessel Promoted Ignition Test,”
ABSTRACT: The industry experienced burn-through and failure of large oxygen
piping systems corresponding to geometry and flow pattern conditions very
different from the ASTM G124 Promoted Ignition Test standard. In Air
Liquide’s hollow-vessel promoted ignition test, puncturing ignition is effected
by a pyrotechnic promoter inserted into the wall of a hollow structure confin-
ing high-pressure gaseous oxygen. This is in strong contrast to a rod burning
vertically in static atmosphere. The critical flow of oxygen through the breach
both supplies efficiently oxygen to the peripheral burning front and blows out
the formed oxides, so that combustion propagates with the breach diameter
expanding. The facility can supply approximately 9000 Nm3/h oxygen, main-
taining the internal pressure at the breach, similar to the case of a real-life
piping accident. The final extension of the breach is studied as a function of
pressure from 2.8 to 40 barg (40 to 580 psi). Even at the lowest pressure,
self-sustained radial propagation is unambiguously evidenced from initial di-
ameter close to promoter size. However, the breach extension indicative of
the severity of damages, is, on average, scaled to pressure, and, moreover,
substantially larger for carbon steel. Also at highest pressure (30–40 barg),
local flow patterns at the breach-like turbulence, recycling, and swirling of
Manuscript received January 31, 2012; accepted for publication July 20, 2012; published online
November 2012.
1
Air Liquide Corporate R&D, CRCD, Les-Loges-en-Josas, France.
2
Air Liquide Engineering, Champigny-sur-Marne, France.
3
Air Liquide Center for Technology and Expertise, le Blanc-Mesnil, France.
Copyright V
19428-2959.
198
RIDLOVA ET AL., doi:10.1520/STP20120009 199
strongly heated oxygen tend to enhance metal combustion. It is further

observed that increasing the burning metal wall thickness (e.g., from 5.5 mm
to 12.5 mm) enhances the burning intensity and aggravates damages.
Results give a different vision and understanding regarding the flammability
of stainless steel compared to ASTM G124 standard where larger rod section
is found to have a mitigating effect.
KEYWORDS: oxygen compatibility, promoted ignition, heat affected, fire
propagation, pipe wall burn-through, stainless steel, carbon steel
Introduction
It is currently general practice to use the ASTM G124 [1] promoted ignition
test (PIT) to assess the compatibility of metallic materials with oxygen service,
for instance, high-pressure, high-flow oxygen piping networks for supplying of
large industry plants. The principle of PIT [2] is to attempt to burn an elongated
cylindrical rod of the tested material suspended vertically in a pressurized oxy-
gen ambient (Fig. 1) without intentionally established flow. Ignition is effected
at the bottom extremity by means of a specific calibrated pyrotechnic promoter,
letting then combustion free to propagate upward. A “burn” versus “no-burn”
result corresponds to the consumption of a minimal length of the rod, generally
30 mm. Historically, the term “exemption pressure” (EP) has been derived
from the well-known “S-shaped” transition curve [3,4] of burned length versus
pressure (Fig. 2) for a statistical series of PIT experiments. For any use at a
pressure inferior to the exemption value, the material is supposed not to sustain
combustion.
The application of PIT following such principles (or possibly slightly less
conservative ones) has a very long history (about 20 years) with really consid-
erable accumulation of statistical data. The PIT is simple in its principles and
FIG. 1—Schematic view of ASTM G124 promoted ignition test (PIT) setup.
FIG. 2—The “S-shaped” transition curve being the basis of the practical
CGA/EIGA exemption pressure definition from ASTM G124 PIT.
its interpretation is straightforward for practitioners. However, there remain a

number of concerns:
– The burn length has been for most of the published results the only
recorded information. Metal regression rates during propagation of com-
bustion are not available in general.
– There have been extremely few attempts of detailed real-time diagnostic
and modeling [5–7] compared to the enormous amount of generated sta-
tistical “burn–no burn” data-points. So there is no clear scientific inter-
pretation of PIT regarding the concrete safety of a real-life installation.
This leaves room for subjective and arbitrary considerations.
– The real definition of the exemption pressure is the pressure below which
there is no flow velocity limitation [8], in other terms particle impact is
considered unable to induce ignition. Strangely, a criterion pertaining to
ignition in flowing conditions is used in place of a propagation-related
one in static conditions. For justification it is advocated that ranking of
engineering alloys by particle impact test and by PIT follow the same
order with very few exceptions [9]. Anyway the practical implementa-
tion of the original definition based on particle impact tests would have
been impossible: particle impact tests are very complex and fastidious
and far more costly than PIT. Moreover there seems to exist very few
facilities worldwide (if not only the NASA WSTF) which prevents any
round-robin implementation. So it is considered that relative ranking of
materials by PIT is acceptable.
– Nevertheless, it remains disturbing to apply the same criterion to very
different practical situations, in particular without consideration of flow.
It is reasonably presumable that flowing conditions may affect propaga-
tion in the PIT test.
– It is also a false opinion to consider that an installation is totally safe if

operated below the exemption pressure as per EIGA/CGA. Accident his-
tory confirms that hazards can be present while using metals even at rela-
tively low pressure if the appropriate specific ignition conditions are
met.
The understanding of exemption pressure can be improved if more informa-
tion is taken into account to decide its value than the simple and sole considera-
tion of the limit of the “upper shelf” of the S curve in the static PIT. Indeed,
industry experience should be considered in addition to a complete re-evaluation
of the PIT data.
However, there is another complementary approach that would consist of
performing tests in a different geometry that is more specifically representative
of the real situation of an accidental fire affecting a piping element. This is in
contrast to the rod PIT, which can be viewed as illustrative of a burning valve
stem. A piping fire starts generally via primary local outward burn-through of
the pipe wall induced by an ignition cause inside. Puncturing leaves an initial
orifice with melted metal and oxide at its periphery, whereas high-pressure ox-
ygen starts to flow out massively. A burning front can then progress radially
thanks to convective forcing of the oxygen supply and continued displacement
of the produced liquid oxides by the dynamic pressure. The breach can expand
to a final limit depending on operating parameters. This is the principle of the
hollow-vessel promoted ignition test (HV-PIT), where puncturing ignition will
be effected, instead of a real-life cause like high-velocity particle impingement,
by means of a specific pyrotechnic promoter derived from that of the standard
PIT.
The main objective of developing such a test is to assess how the burning
events in this geometry may be classified in function of conditions, and in par-
ticular if it is possible do define a parametric domain within which expansion
of the initial breach diameter by self-sustained combustion cannot take place at
all. After preliminary trials as early as 1986 [10], Air Liquide has built a new
complete facility and the current test program has been conducted on the
2006–2011 period. For the present work, 218 data points were compiled.
Experimental Installation
Structure and Environment of the Burning Samples

The basic element of the HV-PIT setup is a metal enclosure connected to pres-
surized oxygen supply means. Ignition and puncturing are obtained by means
of a pyrotechnical promoter derived from that of PIT adapted at a given loca-
tion on the vessel wall. Initially, oxygen is confined inside the feeding line and
vessel, with the minor exception of a calibrated oxygen leak ensuring a small
constant purging flow. The static pressure initially established corresponds to

the set-point value of the test. After the promoter firing, once the wall burn-
through has taken place, the oxygen supply system will regulate the feeding
flow to compensate for the loss of oxygen to the outside and sustain the pres-
sure close to the set-point inside the vessel and at the breach. This is intended
for reproducing to the best extent the conditions that would prevail for a real
punctured pipeline, with virtually unlimited oxygen supply capabilities (con-
nected production plant and long distance lines acting as a buffer of extremely
large capacity).
Primarily, tests were conducted on 1.00 -m-long DN80 tubing sections
with the burn-through achieved at the lateral pipe wall, as shown in Fig. 3. The
choice of this geometry seemed implicit to reproduce accurately a section of a
real oxygen distribution installation. However, using a complete tube each time
represented a waste of material and in addition repeated handling of heavy
pieces of metal (45 kg) was very inconvenient for operators.
For this reason, an alternative configuration has been devised where the
sample is only a limited element of thin wall, precisely a DN125 disc tightly
clamped by a flange closing the DN80 oxygen supply line at its downstream
extremity. In fact the line forms a 90 elbow upstream of the flange and the
disc is normally installed horizontally in bottom position. It is, however, possi-
ble to tilt the disc for some experiments to show specific effects.
FIG. 3—Test sample structure in, respectively, tube and disc configuration.
The question, of course, arises of the equivalence of the tube and disc con-
figurations. Comparative series of results will be shown further in the text. It is
considered that there is no major artifact introduced for the kind of information
we are interested in, whereas it proved considerably easier and more productive
to turn definitively to the disc configuration.
Oxygen Supply System and Utilities

The oxygen supply facility (Fig. 4) has been designed to yield for the short du-
ration of a burn-through test a high-pressure flow rate of an order of magnitude
consistent with operation of a real-life oxygen pipeline delivery system,
namely, about 9000 Nm3/h. The system has two branches in parallel upstream
of the dome-loaded pressure regulator and test vessel feeding line. Figure 5
gives an idea of the impressive dimensions of the components.
Each branch is composed first of a 9 B50 cylinder bundle loaded nominally
at 200 bar, that constitutes the main feedstock, with the regulatory safety output
flow restrictor of 4 mm diameter. However, in order not to suffer from this flow
restriction at the level of delivery, there is provided downstream a buffer in the
form of a single B50 cylinder, with an outlet diameter expanded to 12 mm.
During a burn-through test, as the buffer is discharging massively into the test
feeding line through the pressure regulator, it is being simultaneously replen-
ished from the bundle. So the pressure upstream of the regulator remains fairly
constant during the short experiment and, therefore, the flow supply capacity
cannot be exhausted for the same time. It is sufficient to maintain the set-point
pressure inside the punctured vessel for a notable breach diameter even at
30–40 bar. The sample can be allowed to burn in all cases to the ultimate possi-
ble value of the breach diameter (absolute maximum about 100 mm) because
the test unit is enclosed within a reinforced concrete casemate with a triple
steel belt hardening structure, proofed against any violent energy release event.
Instrumentation, Diagnostics and Measurements

Besides the equivalent final diameter of the breach on each tested sample,
which is the measure of the extension of the fire propagation, some other pa-
rameters are acquired during the burning event. Three fast-response pressure
sensors are installed: one on the primary branches (bundle/buffer pressure),
one downstream of the pressure regulator, and one within the test vessel. Such
pressure measurements allow checking to what extent the oxygen supply facil-
ity is capable of maintaining the set-point pressure at the breach during a test.
In addition a thermocouple probes the sample wall temperature just at the
external location opposite to the promoter implantation inside the vessel. In the
first generation of the testing setup, another thermocouple was provided inter-
nally in the immediate vicinity of the promoter to detect the sharp temperature
rise indicative of the time origin of the promoter firing. More recently, a
FIG. 4—Schematic of the ultra-high flow (9000 m3/h) gaseous pressurized oxygen supply.
FIG. 5—(a) Picture of the giant dome-loaded pressure regulator and safety
shutoff valves, and (b) schematic of gaseous oxygen feedstock and metering,
feeding line, and utilization.
convenient alternative has been the direct acquisition by the system controller
of the promoter firing electric signal. Finally, the burning event is also imaged
externally to the hollow vessel by a web camera (30 frames/s), and also in a
few cases by a high-speed camera (10|000 frames/s).
Promoter Design
The design and sizing of the promoter for the HV-PIT follows the same general
philosophy as that of the ASTM G124 PIT, however, here the functionalities
are somewhat different and more constraints are present. Ideally, the HV-PIT
promoter should ensure outward burn-through with minimal initial diameter,
leaving just after puncture the periphery of the orifice with a positive net rate
of thermal energy input (heating by promoter, plus enthalpy creation because
of the onset of metal oxidation, minus thermal losses). This corresponds to the
“ignition” criterion.
In practice, however, it is not possible to design a promoter capable to
induce burn-through of an initially plain wall of tube or disc. This is because it
is not sufficient to inject enough thermal energy into said metal wall, but there
must be some mechanism ensuring the removal of the formed metal oxides;
otherwise combustion cannot progress steadily in the direction normal to the
metal wall. There is no possibility of such oxide removal for a promoter that
would simply form a close mechanical contact with the plain wall. This has
been amply confirmed experimentally. Therefore, practically, the promoter
must be of the traversing type. It is postulated that provided the diameter of the
traversing element can be kept small enough, this does not introduce a notable
artifact with respect to real-life ignition mechanisms involving of course no
pre-existing modification of the metal wall, The following optimized design
that was definitively adopted after evaluating a number of configurations is
depicted schematically in Fig. 6 and a photo appears in Fig. 7.
If we consider for instance the 5.5-mm-wall thickness version, the travers-
ing burning element is a 6-mm-diameter aluminum screw with its head pro-
truding internally from the disc wall surrounded by an ignition structure
consisting of two concentric aluminum and magnesium rings. The magnesium
ring is ignited by a wound Pyrofuze wire fired electrically. The total energy of
the promoter is 47 kJ, notably higher than the standard PIT. This can be
explained by enhanced thermal losses by conduction in the metal wall, and
also by convective cooling because of the steady oxygen purge flows estab-
lished prior to promoter firing.
Nevertheless, only a small fraction of this total promoter energy is actually
yielded to the disc wall and contributes to ignition. The screw diameter of
6 mm is the minimum for complete puncturing to be affordable for all consid-
ered conditions. For the 12.5-mm-disc thickness version, it was decided to
keep the same format and energy of promoter so that the screw is not fully tra-
versing and some free space is left inside the channel on the outer side.
Heat-Affected Zone and Discrimination of the Regime of Self-Sustained

Combustion
The heat-affected zone (HAZ) is a key notion for determining whether self-
sustained combustion has occurred or not. Overall, the HAZ definition will not
change for HV-PIT with respect to standard PIT. HAZ is the spatial region
inside which metal oxidation has been notably influenced by the initial heat
release from promoter operation. Beyond the boundary surface of the HAZ, the
only driving force for combustion progress is the liberated enthalpy of oxida-
tion that induces gradual melting of new fresh metal (pure propagation regime).
In the case of standard PIT, theoretical and experimental (axially distributed
thermocouples) analyses of the thermal wave resulting from promoter burning
[11] have led to the determination of a HAZ extension representing the first
30 mm of the rod. This in passing gives an objective justification of the “burn”
criteria expressed as minimal consumption of 30 mm of the rod length. In the
FIG. 6—Cross-sectional view of ignition promoter design for 5.5 - and 12.5-mm-thick steel walls.
RIDLOVA ET AL., doi:10.1520/STP20120009
207
FIG. 7—Photograph of ignition promoter mounted on disc sample for firing.
case of the HV-PIT, the geometry (3D versus 2D) and sequences of ignition
are different and more complex.
There is no possible equivalent of thermocouples setting-up used in PIT.
Only thermal imaging would be affordable, but it has not yet been imple-
mented. Here our determination of the HAZ extension will be mainly theoreti-
cal. For numerical simulation of ignition by promoter, we formulate the
following hypotheses:
– The electrically-fired Pyrofuze ignites the promoter instantaneously.
– The upper part of promoter outside the drilled channel (Mg/Al rings þ Al
screw head) burns in situ either in liquid phase, vapor phase or a mix
(depending on oxygen pressure and pre-flow rate).
– Heat is transferred to the traversing section of the Al screw but also
partly to the encompassed zone of the inner side of the disc.
– Oxygen confinement is lost (puncture) when either Al screw threads
become fragile under heating and fail mechanically, or screw starts melt-
ing, or screw begins burning inside the channel.
– Ejected hot aluminum/magnesium oxides travelling over the channel
walls deposit enough heat to start sample metal oxidation in surface.
– Radial propagation of combustion starts because of continuous blowing
out of liquid oxides as they are formed (dynamic pressure of outflowing
oxygen against liquid oxide surface tension and viscosity).
For the numerical calculations, we presuppose that the promoter burns in
liquid phase. This represents the potential “worst case” in terms of HAZ radial
extension because the mass of melted Mg/Al oxides will undergo notable vol-
ume expansion (whereas in the case of vapor phase burning combustion will
occur almost fully in situ).
We display in Fig. 8 the graphical results of simulation at the stage that the
500 C isotherm intercepts the bottom end of the screw. This corresponds to
the earliest possibility of puncturing, with the screw still solid and thermal-
mechanical failure of threads.
FIG. 8—Crosscut drawing of promoter zone showing computed isotherms after

promoter firing, just before puncturing.
To check the simulation results against experiments, the visual aspects of

breach on both disc sides after one test are shown in Fig. 9. This test (stainless
steel disc, 12.5 mm thickness) was done at the minimal pressure of use of the
promoter (i.e., allowing burn-through), of 2.8 bar.
FIG. 9—Photo showing aspect of the breach on both sides for the minimum
tested pressure 2.8 bar, stainless steel thickness is 12.5 mm.
Let us first consider the inner side (location of promoter):

– We observe some signs of surface melting and burning in the area adja-
cent to the breach, initially beneath the promoter head. This means that
in actuality puncture occurred later than considered in the simulation.
– However, this surface burning cannot have occurred before puncture,
because at that time there was no possibility to remove oxides.
– Moderate heating (iridescence fringes) visible over slightly larger diame-
ter than breach of about 20 mm indicates that the hypothesis of liquid
phase burning of Mg/Al is quite valid.
Now about the breach appearance on the outer side:
– The channel longitudinal section initially free from Al metal (non tra-
versing screw, blind hole) was of course not pre-heated. The self-heated
radial zone, implicitly from pure propagation, is very narrow.
– Despite that the diameter at the outer side is not smaller (even slightly
larger) than at the inner side.
From this analysis, we conclude that a diameter of ca 14 mm represent the
very upper limit for the radial extension of the heat-affected zone.
Test Protocols and Realization
Specifications of Materials and Gases

Three engineering alloys were investigated: stainless steel 316 L, carbon steel
A106-GradeB and aluminum bronze at 11% Al CuAl10Ni5Fe4. Detailed com-
positions are given in Table 1.
Oxygen is of standard grade with 99.5% purity.
Practice of Tests
The oxygen feeding lines and test vessel are classically purged (5 cycles at
10 bar) before establishing the set-point static pressure prior to a test. However,
TABLE 1—Nomenclature designation and composition of engineering alloys intended for first series
of HV-PIT trials.
Stainless steel 316L (in wt. %)

C Mn Si P S Cr Ni Mo N
0.026 0.87 0.61 0.038 0.006 16.81 11.50 2.11 –
Carbon steel A106-B (in wt. %)

C Mn Si P S Cr Ni Mo Cu V; Ti Nb/Cb
0.16 0.78 0.22 0.019 0.006 0.12 0.08 0.03 0.16 0.0 0.0
Aluminium CuAl 10 Ni 5 Fe4 bronze (in wt. %)

Al Ni Fe Mn Cu Pb Sn Si Zn
10.68 4.55 4.44 0.12 80.09 0.010 0.02 0.009 0.029
the additional small continuous purge of about 25-slm flow maintained before
and during promoter firing has been evidenced to be absolutely mandatory.
Otherwise, in any case, the promoter will not burn completely and trial will
fail. It is likely that some desorbed residues and moisture have to be effectively
scavenged to avoid perturbing the fast oxygen mass transfer needed to the
burning promoter. For the same reason, it must be also ensured that a minute
leak flow escapes via the threads of the promoter screw. If necessary, the screw
has to be slightly trimmed with a file.
All tests were conducted at ambient outdoors temperature inside casemate
between 0 C and 25 C.
Results
Nature of Gathered Information

For each test event, the equivalent averaged diameter of breach (which is not,
in general, of very regular shape) is determined. Previously with the tube sam-
ples graphical projections of the inner and outer breach sections were used. Af-
ter the introduction of the disc samples, it has become possible to use a
differential weighing procedure, which is much more convenient. In compila-
tion and interpretation of results, the equivalent diameter is used preferentially
to the metal mass lost, because it is more intelligibly comparable to standard
PIT results expressed as minimum burn length.
Another important indication is provided by the pressure drop within the
vessel one second after puncture, representing most often the typical maximum
duration of hollow-vessel combustion events. This latter information is indica-
tive of how effective the oxygen supply system was in compensating for the
oxygen outflow through the breach; hence, how close the trial was to a real-life
accident situation.
Last, the combustion duration is estimated from video records. It is consid-
ered that the combustion lasts as long as massive projections of blown-off,
high-temperature-melted meal/oxide droplets are observed (Figs. 10 and 11).
Then the breach lips return to the very slow solid phase oxidation regime, with
the possible exception of a few remaining localized hotter spots giving rise to
tiny residual incandescent effusions. However, this transition is generally very
short so that combustion duration values determined by processing videos are
highly inaccurate, except in the quite rare cases that a high-speed camera was
used.
General Features or Fire Events and Characteristics of Burnt Samples

Figures 12(a) and 12(b) and 13(a) and 13(b) illustrate the internal and external
side aspect of, respectively, carbon steel and stainless steel disc samples after
FIG. 10—View of the violently effusing jet of hot oxygen and melted/burning
metal/oxide droplets for the tube configuration.
testing. Usually, breaches resulting from burn-through of carbon steel walls ex-
hibit quite a regular, quasi-circular shape. Conversely, stainless steel samples
bear orifices or much more irregular geometry, although roughly circular on
average in general.
Figures 12(a) and 12(b) correspond to a nominal testing pressure of 10 bar,
having lead to almost total consumption of the metal disc. In fact, the oxygen
flow had to be shut off by operator to avoid further extension of the fire that
would have resulted in severe damages to the test bench itself.
Figures 13(a) and 13(b) pertain to a trial at 40 bar. As usual, stainless steel
appears substantially less prone than carbon steel to combustion, as the average
FIG. 11—View of the violently effusing jet of hot oxygen and melted/burning
metal/oxide droplets for the disc configuration.
FIG. 12—(a) Photo of internal side of carbon steel disc sample (12.5 mm thick)
tested at 10 bar. Oxygen supply had to be prematurely shut off by the operator
because unlimited combustion was about to extend to the experimental installa-
tion itself. Breach equivalent diameter was already 85 mm. (b) Photo of external
side of carbon steel disc sample (12.5 mm thick) tested at 10 bar. Oxygen sup-
ply had to be prematurely shut off by the operator because unlimited combus-
tion was about to extend to the experimental installation itself. Breach
equivalent diameter was already 85 mm.
equivalent breach diameter here is only 53.9 mm (test E209). The breach lateral
walls are remarkable by their tormented, sometimes locally reentrant shapes.
Moreover, the external face of the disc has also a very characteristic appear-
ance of deep irregular erosion. This is interpreted as surface combustion being
caused by forced turbulent convection, recirculation, and swirling of the
extremely hot jet just at the outlet of the breach. This mechanism supplies
the disc’s external surface with very large amounts of heat and oxygen, and at
the same time establishes the dynamic overpressure ensuring the removal of
melted oxides.
This specific effect manifests itself for the supersonic oxygen outflowing
regime taking place from 30–40 bara pressure value inside the testing vessel.
Moreover, it is to be noted that when the bolted flange holding the disc in place
is also made of stainless steel, at such high pressure it may start to burn as
well, yielding a massive amount of heat contributing to exacerbate the radial
propagation of combustion at the breach lips. This is the reason why some trials
where made with an Inconel 600 flange, which is much less prone to burning,
thus suppressing the aforementioned artifact. However, the erosion of the
disc’s external face still remains.
FIG. 13—(a) Photo of external side of stainless steel disc sample (12.5 mm
thick) tested at 40 bar, showing the strong front erosion assumed to be caused
by hot jet turbulence, swirling, and recirculation in supersonic outflowing con-
ditions. Equivalent breach diameter is 53.9 mm. (b) Photo of internal side of
stainless steel disc sample (12.5 mm thick) tested at 40 bar.
Burn-Through Damage Extension With Oxygen Pressure for Stainless Steel

Figure 14 shows the correspondence between the pressure established inside
the test vessel just before puncturing, and the final equivalent diameter of the
breach for trials on stainless steel samples. These include tubes of 5.5-mm-wall
thickness as well as disks of 5.5 mm and 12.5 mm. A first major observation is
FIG. 14—Breach equivalent diameter as a function of pressure for stainless

steel: tube and disc configurations for 5.5-mm thickness and disc configuration
for 12.5-mm thickness.
that, even at the lowest tested pressure value of 2.8 bara, there always exist
data points for which the final breach diameter widely exceeds the estimated
extension of the heat-affected zone, namely a maximum of 14 mm. In other
terms, in such cases, there has been a very significant contribution to the forma-
tion of the final breach, of the radial self-sustained combustion propagation
regime.
To make the parallel with the standard ASTM G124 PIT (rod) test, one
does not find any threshold pressure value below which one would observe
only “no-burn” events. Puncturing fire initiation can always lead to subsequent
expansion of the orifice diameter. However, although raw data appear consider-
ably scattered, on average there is a trend toward regular if not proportional
growing of the final equivalent breach diameter value with the operating pres-
sure. One also notes a few especially singular points with “abnormally” large
or conversely small breach extension with respect to the corresponding
pressures.
In parallel, it is observed that the pressure drop one second after puncture
does not always scale very consistently with the final breach diameter. We esti-
mate this may be ascribed to two possible causes for the corresponding trials.
First, the remaining oxygen pressure in the bundle may have been low because
of normal consumption along a testing campaign. Second, in some cases, the
breach may have opened more rapidly, with the demand in flow delivery to the
regulator reaching prematurely high values. In this case, an increased released
oxygen inventory may have started to exhaust the buffers. It is to be recalled
that at this time there is no mean to measure in real time the variation of the
flow or breach diameter during the combustion. This information cannot be
inferred solely from the combustion time and final breach diameter.
As to the consistency of the disc sample data versus the tube sample ones,
their respective distributions fit within the general dispersion for the complete
set of results. Moreover, the increase in burning-through wall thickness is not
found at all to be a factor of attenuation of the extension of damages. Even, for
the range of pressure beyond 30 bar, almost systematically the breach diame-
ters are very significantly larger for 12.5-mm-thick disc samples than for the
5.5 mm ones. This of course will have to be interpreted differently from the
standard PIT results, for which an increase of the rod section diameter appears
to have a mitigating effect.
Finally, the small number of tests performed on aluminum bronze show a
very similar behavior to stainless steel, as seen in Fig. 15.
Figure 16 depicts for all the same experiments the duration of the strong re-
gime of combustion as defined in the preceding section. Although the combus-
tion duration is effectively in almost all cases inferior to one second, one can
see that the distribution of values is extremely broad within this interval and
there is no apparent correlation for instance to the final breach diameter. It has
been identified some infrequent singular burning events that proceed partly by
FIG. 15—Comparison of breach equivalent diameter versus pressure between

stainless steel and aluminum bronze (disc samples, 12.5 mm thick).
surface modes and do not follow the general kinetics. However, such cases
remain exceptional and the uncertainty in data points can be most often
ascribed to the fact that the extinction time fits between two consecutive frames
for standard videos. So it was extremely difficult and uncertain to identify the
true end of combustion, except in the rare cases high-speed video was
available.
In addition to standard experiments of downward burning-through of a
sample positioned horizontally, a few trials were achieved with a 12.5-mm-
thick disc tilted by 45 . It has been observed in some real-life accident situa-
tions that melted metal/oxide transport under effect of gravity may have an
FIG. 16—Duration of the combustion regime (i.e., strong incandescent projec-

tions) as a function of breach diameter for stainless steel and aluminum
bronze.
FIG. 17—Comparison between breach equivalent diameter, respectively, for

standard horizontal disc configuration, and for 45 tilted configuration.
influence on the form and extension of damages. In the case of the HV-PIT
test, it is worth noticing that initial puncturing ignition by promoter is not pos-
sible when either the metal disc sample is vertical, or it is situated in top posi-
tion with the inserted promoter intended to burn upward. In fact, in these
situations, the internal protruding part of the promoter (aluminum screw head
with the aluminum and magnesium surrounding rings) invariably detaches and
falls down once it has started burning. Thus, the traversing screw insert is pre-
vented from being further heated and will weaken thermomechanically, melt,
and/or burn. Therefore, the puncturing ignition process aborts.
Conversely, with the disc metal wall at 45 attempting to burn through
downward by the promoter, there was an extremely spectacular enhancement
of damage extension (Fig. 17), in addition to a strong asymmetry of the breach.
One notes that the final dimensions of the breach show widely dispersed val-
ues; however, the effect is so strong, in general, that the flange is severely dam-
aged and, in one case, the hollow-vessel body itself was partially destroyed,
necessitating important repairs.
Burn-Through Damage Extension With Oxygen Pressure for Carbon Steel

In the case of carbon steel tests, one major difference compared to stainless
steel is the much reduced scattering of data and relatively well marked trend to
linear scaling of the breach diameter with pressure. This is at least true for the
5.5-mm-thick samples, either the tube ones or the disc ones [18,19]. For the
12.5-mm-thick disc samples, a very strong radial combustion propagation
enhancement effect is observed. This is up to a point that even at pressure of
10 and 20 bar, as we mentioned earlier (Fig. 12), the oxygen supply had to be
interrupted prematurely to avoid further extension of propagation and potential
FIG. 18—Diagram of equivalent breach diameter versus pressure for carbon

steel 5.5-mm-thick samples (tube and disc) and carbon steel 12.5-mm-thick
samples (disc). In the latter case, oxygen supply had to be prematurely shut off
by the operator as fire propagation had no foreseeable limit.
severe damages to the whole testing installation. So the limits of the damage,
as far as they would exist, could not be explored in this case, which is anyway
very far from the recommended pressure limits for use of carbon steel.
For the combustion duration in the case of carbon steel it is noted as well
that values are dispersed within the interval of 1 s, but also that there are three
very singular points with a considerable increase to 2–3 s. Again, this cannot
be correlated to other data at present.
Discussion and Interpretations
Understanding of HV-PIT in Terms of Potential Accident Severity

The standard ASTM G124 PIT test is solely interpreted on the basis of definite,
abrupt discrimination between “burn” and “no burn” events. Following the
EIGA-CGA definition, below the exemption pressure an installation can be
operated with a null or acceptably infinitesimal risk of fire event with substan-
tial consumption of metal.
In the case of HV-PIT test, as aforementioned no minimum pressure value
is found for the possible occurrence of a self-sustained combustion propagation
regime, either for stainless steel or carbon steel. This means, appreciable dam-
ages are always possible, with opening of a breach and loss of confinement, as
well as violent effusion of hot gas and melted/burning metal/oxide droplets sus-
ceptible to at least seriously harm personnel or deteriorate nearby infrastruc-
tures. However, on average metal loss and damage extension; hence, severity
of accident consequences, are scaled quite regularly to operating pressure. This
FIG. 19—Estimated combustion duration versus final breach diameter for carbon
steel.
would mean that for any operating pressure value, the level of mitigation and
protection measures can be optimized accordingly by designers. In the case of
carbon steel, this assumption seems fairly valid with little dispersion and no
singular points. Confidence seems less justified for stainless steel because of
much more scattered data, in addition to some singular points remaining preoc-
cupant and unexplained.
Consider for instance on Fig. 14 the two trials on 5.5–mm-thick tube at
20 bara set-point that have resulted in effective breach diameters of 38 and
39 mm whereas eight other trials at the same 20 bara are grouped between 20
and 30 mm. What is even more intriguing with these two singular points is that
in fact, because of control system failure during each test the dynamic flow had
been off during most of the combustion and so the pressure drop after puncture
had been only partly compensated. However, inherently, such situations could
not be reproduced later. At this time a limit cannot be put on the possible
extension of damages whatever the circumstances at this pressure, or even at
higher ones. This concern comes in addition to the effect of local geometry
such as exemplified by the 45 tilting experiments or wall thickness.
Qualitative Influence of Physical Parameters

The expansion of a breach proceeds via radial propagation of a combustion
front in the liquid phase at the breach lips. The basic mechanism, similar to the
case of the standard PIT test burning rod, is the oxidation of liquid metal by ox-
ygen diffusing through the external melted oxide layer. Heat generated as oxi-
dation enthalpy serves to continuously melt adjacent new fresh metal to sustain
the combustion progression process.
It is well known that if the produced liquid oxide is not permanently
removed by some mechanism, combustion will rapidly stop by inhibition of
oxygen transport to the reaction region of metal with oxygen. In rod PIT oxide
removal is effected by periodic detachment of the oxide droplet under competi-
tion between gravity and viscosity/surface tension forces. In HV-PIT oxide is
displaced and blown off by the dynamic pressure shearing forces of the out-
flowing oxygen, countering the oxide/metal melt viscosity and surface tension
forces. The qualitative effect of the involved elemental parameters is evident
but their quantitative hierarchy has never been modeled in this geometry.
If the test nominal set-point pressure; hence, the static pressure at the
breach becomes higher, the oxygen concentration in the metal-oxidation region
also increases, the oxidation kinetics is accelerated and so is the generation of
heat to further melt fresh metal. In parallel, the dynamic pressure also increases
and the melted oxide/metal is displaced and removed more efficiently. There-
fore, the oxide layer is thinner and oxygen is more easily transported toward
fresh melted metal to feed the oxidation reaction. However on the opposite,
faster melt removal could be also detrimental to the combustion propagation
kinetics because more heat would be carried away and no longer available to
sustain further melting, or even liquid metal could be withdrawn prematurely
from the system before having time to oxidize and generate heat.
The quantitative resultant of such competitive mechanisms cannot be read-
ily assessed and modeling appears challenging, among others because the
involved physical coefficients both for metals and oxides (viscosity and surface
tension, heat conduction, oxygen diffusivity, oxidation kinetic constants, etc.)
vary in a widely unknown way in the concerned range of very high temperature
(3000–4000 K) characterizing the self-sustained combustion regime of a fer-
rous alloy. However, the hierarchy of the competitive factors is obviously not
the same for carbon steel and stainless steel.
There seems to be no upper limit to the carbon steel combustion enhancing
effect of increasing oxygen “blowing strength.” Conversely a different behavior
prevails for stainless steel. Supersonic outflowing at sufficiently high pressure
has been evidenced to have a tendency to quench combustion, which is never
observed for carbon steel. Also in some real accidents it has been observed that
metal combustion under turbulent convection, meaning moderate local dynamic
pressure fluctuations, may have a much more destructive effect than impact of a
very strong jet representing a much higher local dynamic pressure.
One may, therefore, wonder if in some marginal fortuitous situations, the
physical system parameters ruling the radial progression of the burning front
may follow a specific evolution sequence during the breach diameter expansion
leading to uncommonly large final dimensions. This could explain the afore-
mentioned singular points at 20 bara. So as long as the complete burn-through
process is not understood quantitatively, it may not be ruled out that exception-
ally the breach extension can reach much higher limits than usual average.
At least however, the effect of metal wall thickness will always be in the
same way. The mechanism responsible for the spectacular combustion
propagation enhancement when the thickness is notably increased resides in a

reinforced radiating cavity effect inside the channel of closer geometry. This
effect does not persist for long as propagation proceeds however, what mainly
matters is the much higher initial temperature of the channel walls and the
memory of this temperature is kept over a much longer distance by the propa-
gation regime.
Potential Influence of Promoter Design

Until then, we have implicitly supposed that promoter operation had no notable
influence on the subsequent propagation regime and so on test results. In fact,
there exist factors that could potentially cause the puncturing ignition to depart
from the near-ideal case. Coming back to Figs. 8 and 9 and the corresponding
text section, it is very likely that before puncturing some melted oxides remain
confined in the channel just above the top of the remaining traversing part of
the aluminum screw. In this liquid oxides pool, the temperature distribution
has no reason to be highly symmetrical and furthermore heat and mass trans-
fers dominated by free convection will not be very efficient in restoring homo-
geneity. So, just as puncture occurs the initial temperature on the channel walls
can be somewhat irreproducible and inhomogeneous. One may reasonably
assume that the subsequent radial combustion propagation regime will keep the
memory of this initial temperature distribution over a notable distance from the
initial puncturing location.
To further support this argument let us come back to the results of tests on
12.5-mm-thick stainless steel discs tilted by 45 . It does not seem realistic to
ascribe the strong anisotropic enhancement of combustion to liquid oxide trans-
port and removal becoming strongly dissymmetrical during the propagation
because of tilting of discs in the gravity field. In effect, as soon as puncturing
has occurred and the propagation regime is established, the oxide removal pro-
cess by very strong supersonic blowing-off at 40 bara will totally dominate the
gravity effect. However, before puncturing the strong anisotropy because of
gravity is present in the static oxide melt, and translates to a considerable rup-
ture of symmetry of the initial temperature distribution on the channel walls
just at the time of puncture. It is particularly impressive that, in this case, the
memory of the effect be kept by the propagation at a distance comparable to
the whole disc radius. On this basis, it can be reasonably accepted that most
generally, the promoter operation in the current design may have an influence
on the results.
Conclusion
Over the 2006–2011 period, the new generation HV-PIT test platform allowed
to accumulate a considerable amount of results covering a wide variation range
of operational parameters. From the analysis of the current 218 data-points

considered in this work, statistical trends can be assessed with good confidence.
The optimization of the ignition promoter design received great attention in a
way whenever possible consistent with the approach taken for the standard PIT
test. The notion of heat affected zone (HAZ) was also transposed very carefully
from the 2D problem of standard PIT to the 3D problem of HV-PIT to unam-
biguously discriminate the occurrence or not of a self-sustained combustion
propagation regime, here radial instead of linear.
A first major finding is that HV-PIT does not show any kind of behavior
that would allow for defining a quantity equivalent to the exemption pressure
as per CGA/EIGA based on the standard PIT test. Self-sustained combustion
propagation always contributes to the breach radial expansion, even at the low-
est tested pressure of 2.8 bar, and the propagation distance can be substantial
compared to the HAZ radius for 10 bar and less. This is of course even more
valid for carbon steel that always shows very notable consumption of metal
beyond the HAZ boundary.
However, a second very important specificity is that, most generally, the
final breach diameter as indicative of the damage extension scales quite line-
arly on statistical average with the operating pressure value. This gives an indi-
cation of the potential consequences of an accident for given conditions of use
and helps in deciding the appropriate level of protection measures. From this
standpoint, results of HV-PIT may be viewed as input information complemen-
tary to industry experience and accident history to be taken into account in
revising the mode of interpretation of the standard ASTM G124 PIT.
At this time, the HV-PIT allows to compare engineering alloys differently
than standard PIT by estimating the concrete level of risk from their use in a
real industrial installation. It is for instance very valuable information to find
that aluminum bronze is not more “dangerous” than stainless steel.
Besides the extension of the range of tested metallic materials, further
work should have the objective of improving the confidence in HV-PIT results
by reducing scattering of data and explaining the few highly singular points. In
this respect, completing a numerical model of the radial combustion propaga-
tion regime should allow to assess the quantitative hierarchy of the elemental
phenomena in the self-sustained radial combustion propagation regime and
give definite prediction of the absolute maximum breach extension. However,
it is correlatively indispensable to improve the promoted ignition process
reproducibility, at first by further refining the promoter design based on simula-
tions. The target is to reduce the perturbation introduced by the accumulation
of melted oxides in the traversing channel prior to puncturing, an artifact that
has no equivalent in the standard PIT.
More generally, because there is no equivalent of an exemption pressure in
HV-PIT, it must be interpreted not only in terms of potential damage extension,
but also of ignitability. From this standpoint, the mechanisms of initial
localized energy deposition followed by puncturing ignition and transition to

the radial propagation regime must be understood more in depth. For this, an
alternative energy source for ignition experiments is necessary, with high con-
trollability and reproducibility, convenient handling, and necessitating no pre-
traversing arrangement. Obviously, a focused power laser beam could be the
solution of choice, together with in situ real-time physical diagnostics, like fast
imaging and pyrometer.
Acknowledgments
Special thanks are due to Dr. V. Martynenko and Dr. S. Shabunya from A.V.
Luikov Heat and Mass Transfer Institute, Minsk, Belarus, for computing the
heat transfer upon promoter burning.
References
[1] ASTM G124-95, 2003, “Standard Test Method for Determining the
Combustion Behavior of Metallic Materials in Oxygen-Enriched Atmos-
pheres,” Annual Book of ASTM Standards, Vol. 14.04, ASTM Interna-
tional, West Conshohocken, PA.
[2] McIIroy, K., Zawierucha, R., and Drnevich, R. F., “Promoted Ignition
Behavior of Engineering Alloys in High Pressure Oxygen,” Flammability
and Sensitivity of Materials in Oxygen-Enriched Atmospheres: Third
Volume, ASTM STP 986, D. W. Schroll, Ed., ASTM International, West
[3] Zawierucha, R., McIlroy, K., and Mazzarella, R. B., “Promoted Ignition-
Combustion Behavior of Selected Hastelloys in Oxygen Gas Mixtures,”
Flammability and Sensitivity of Materials in Oxygen-Enriched
Atmospheres: Fifth Volume, ASTM STP 1111, J. M. Stoltzfus and K.
McIlroy, Eds., ASTM International, West Conshohocken, PA, 1991, pp.
270–287.
[4] Zawierucha, R., Samant, A. V., and Million, J. F., “Promoted Ignition
Behavior of Cast and Wrought Engineering Alloys in Oxygen-Enriched
Atmospheres,” Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres: Tenth Volume, ASTM STP 1454, T. A. Steinberg,
H. D. Beeson, and B. E. Newton, Eds., ASTM International, West Con-
shohocken, PA, 2003, pp. 164–176.
[5] Hirano, T. and Sato, J., “Fire Spread along Structural Metal Pieces,” J.
Loss Prevent. Proc. Ind., Vol. 6, No. 3, 1993, pp. 151–157.
[6] Chiffoleau, G. J. A., Steinberg, T. A., and Veidt, M., “Ultrasonic Mea-
surement Technique of Burning Metals in Normal and Reduced Gravity,”
pheres: Tenth Volume, ASTM STP 1454, T. A. Steinberg, H. D. Beeson,
and B. E. Newton, Eds., ASTM International, West Conshohocken, PA,

2003, pp. 75–90.
[7] Steinberg, T. A., Mulholland, G. P., Wilson, B. D., and Benz, F. J., “The
Combustion of Iron in High-Pressure Oxygen,” Combust. Flame, Vol. 89,
1992, pp. 221–228.
[8] EIGA/CGA, 2011, “Oxygen Pipeline and Piping Systems,” Globally
Harmonized Document IGC Doc 13/2011/E, European Industrial Gases
Association AISBL, Brussels, Belgium.
[9] ASTM G94-92, 1998, “Standard Guide for Evaluating Metals for Oxygen
Service,” Annual Book of ASTM Standards, Vol. 14.04, ASTM Interna-
[10] Diéguez, J. M., Bothorel, L., de Lorenzo, A., and Faufin, A., “Ignition
Testing of Hollow Vessels Pressurized With Gaseous Oxygen,” Flamma-
bility and Sensitivity of Materials in Oxygen-Enriched Atmospheres:
Third Volume, ASTM STP 986, D. W. Schroll, Ed., ASTM International,
West Conshohocken, PA, 1988, pp. 368–388.
[11] Sparks, K. M., Stoltzfus, J. M., Steinberg, T. A., and Lynn, D.,
“Determination of Burn Criterion for Promoted Combustion Testing,” J.
ASTM Int., Vol. 6, No. 10, doi: 10;1520/JAI102351.
STP 1561, 2012
DOI:10.1520/STP20120012
Chris Dobinson,1 Peter Nelson,1 and Ted Steinberg2
Situational Non-Flammable Control Valves

for Oxygen Service
REFERENCE: Dobinson, Chris, Nelson, Peter, and Steinberg, Ted,

“Situational Non-Flammable Control Valves for Oxygen Service,” Flammabil-
ity and Sensitivity of Materials in Oxygen-Enriched Atmospheres on Septem-
ber 19–21, 2012 in Montreal, QC; STP 1561, Samuel Edgar Davis and
Theodore A. Steinberg, Editors, pp. 225–232, doi:10.1520/STP20120012,
ABSTRACT: Past work has clearly demonstrated that numerous commonly
used metallic materials will support burning in oxygen, especially at higher
pressures. An approach to rectify this significant safety problem has been
successfully developed and implemented by applying the concept of situa-
tional non-flammability. This approach essentially removes or breaks one leg
of the conceptual fire triangle, a tool commonly used to define the three
things that are required to support burning: a fuel, an ignition source, and an
oxidiser. Because an oxidiser is always present in an oxygen system, as are
ignition sources, the concept of situational non-flammability essentially
removes the fuel leg of the fire triangle by only utilising materials that will not
burn at the maximum pressure, for example, that the control valve is to be
used in. The utilisation of this approach has lead to the development of a
range of oxygen components that are practically unable to burn while in serv-
ice at their design pressure, thus providing an unparalleled level of first safety
while not compromising on the performance or endurance required in the
function of these components. This paper describes the concept of situa-
tional non-flammability, how it was used to theoretically evaluate designs of
components for oxygen service and the outcomes of the actual development,
fabrication, and, finally, utilisation of these components in real oxygen sys-
tems in a range of flow-control devices.
Manuscript received February 6, 2012; accepted for publication March 30, 2012; published
1
OXYCHECK Pty. Ltd., Brisbane, Queensland, Australia.
2
School of Chemistry, Physics and Mechanical Engineering, Queensland Univ. of Technology,
Brisbane, Australia.
Copyright V
19428-2959.
225
Background
Numerous fires in oxygen systems and components over the years highlighted
the significantly increased potential for burning for almost all materials in
oxygen-enriched atmospheres when ignition sources are present. Indeed, all
non-metallic materials and many of the important engineering metallic materi-
als will support burning if ignited in oxygen-enriched atmospheres (especially
at higher pressure and=or temperature and in small geometries). These fires are
often much larger, progress much faster, consume greater amounts of nearby
fuel, and produce much greater damage than fires when enriched oxygen is not
present. This has lead, over the years, to the development of many international
standards and test methods that assist in defining when a particular material
will burn when specific (and often different) ignition mechanisms are present.
Though some industrial standards, (for example, CGA G-4.4, or EIGA 33-02)
focus on selecting materials below an exemption pressure (that is, where they
cannot burn), other standards focus on the minimization of ignition mecha-
nisms as the most important technique to mitigate the risk of a fire occurring
with the materials used in an oxygen-enriched environment. Even with the
guidance from the numerous applicable standards, fires still occur because of
the fact that the materials being used to fabricate these systems will often sup-
port burning, if ignited, under the conditions these systems typically operate at.
Fire Triangle and the Concept of Situational Non-Flammability (SNF)

The fire triangle has been used for many years by numerous researchers to
describe the requirements for a fire to occur with each leg of the triangle repre-
senting a critical element. A typical fire triangle is shown in Fig. 1, where the
three legs shown represent (1) the fuel, (2) the oxidiser, and (3) the ignition
event. It was recognised in early work that the fire triangle for air and oxygen
were critically different because the oxidiser leg, in air, is 21 % oxygen and
79 % nitrogen and, in oxygen, was pure (100 %) oxygen. This difference
means many materials that would not burn in air vigorously support burning in
pure oxygen. Indeed, in hazard mitigation approaches that recognised this dif-
ference (that is, that the fuel and oxidiser are always present in an oxygen sys-
tem), the emphasis is normally placed on having an understanding of, and
mitigating the active ignition mechanisms. The reason these approaches
emphasize the reduction and mitigation of the ignition sources is that then the
fuel and oxidiser (that are always present) can be used without a fire occurring.
Ignition sources, however, will always be present in an oxygen system
(particularly during flow when the oxygen is being used). The oxygen pressure,
flow rate, flow friction, movement of particles, compression processes and
random “unexpected” accidents (electrical discharge, human error, etc.)
always create some ignition risk in the oxygen system that can lead to fires.
DOBINSON ET AL., doi:10.1520/STP20120012 227
FIG. 1—Fire triangle for an oxygen-enriched atmosphere. Each leg represents

a requirement for a fire to occur.
A more-comprehensive listing of ignition mechanisms, or factors within an

oxygen system that can cause fires, is given in Section 7 of ASTM G88.
Because ignition mechanisms are inherent in some types of oxygen compo-
nents, such as control valves, another approach to mitigating the risk of a fire
occurring was developed. In this approach, the concept of situational non-
flammability (SNF) is used to “break” the fire triangle and, therefore, prevent
the possibility of a fire occurring. Utilisation of the concept of SNF recognises
that ignition mechanisms will always be present at certain locations and with
certain types of components in oxygen and, instead, focuses on breaking the
fire triangle by only using materials that will not support burning at these loca-
tions, at actual use conditions, where ignition is more probable. This essentially
removes the fuel leg from the fire triangle and, without fuel, the occurrence of
a fire is precluded.
Application of SNF to Oxygen Valve Development

Applying the concept of SNF to the development of a useful oxygen compo-
nent requires the use of materials that will not support burning if aggressive
ignition mechanisms are present. Because all currently available non-metallic
materials and lubricants will support burning in oxygen, especially when
exposed to oxygen at pressures above ambient, they need to be removed from
the oxygen-wetted regions if the entire component is to be designed using the
concept of SNF. Additionally, only metallic materials that have been shown to
resist self-sustained burning (SSB) at the component’s design pressure, as evi-
denced and supported by specific test data, would be permitted (especially at
locations where ignition mechanisms are active). When these design con-
straints are applied, then there is no fuel present to support burning and the risk
of a fire is mitigated completely through design practices utilizing the principal
of SNF. It is important to note that an associated disadvantage to the use of
SNF in the selection of a component’s metallic materials is often the cost
increase associated with the use of these more burn-resistant metallic

materials.
The principal of SNF and how it eliminates the risk of a fire are straightfor-
ward concepts, however, it was not clear if adherence to the engineering con-
straints imposed would lead to the development of useful (and useable) oxygen
componentry. This is because SNF requires materials that will not support
burning when ignition sources are present and these are typically metallic
materials (in reasonable thicknesses) and are often lacking in their ability to
form robust sealing surfaces in parts that move over the life of the component.
That is, their performance relative to currently available componentry (that
would support burning, if ignited) is often traded for increased burn resistance.
Therefore, to evaluate the application of SNF to the development of flow-
control componentry for gaseous oxygen, a program was started to produce a
check valve that could be utilised for several different applications (system fill
valve, manifold valve insert, standard check valve, etc.).
An initial design=use pressure of 34.5 MPa (5000 psia) was selected along
with a use environment that was 100 % oxygen. Using SNF requires no non-
metallic materials or lubricants to be used in the specific component because
they are all flammable at this pressure in pure oxygen. Additionally, all metal-
lic materials being considered must be shown to be unable to support burning.
To evaluate the suitability of a metallic material, promoted ignition test data
obtained in accordance with ASTM G124 was used. ASTM G124 methodology
is consistent with the similar ISO (Part 4 of ISO 14624-4:2003) and NASA
(NASA-STD-(I)-6001A Test 17) test methodology for evaluating a metallic
materials propensity to support burning in pressurised oxygen. In these tests,
an upright (3.2-mm-diameter) cylindrical rod is ignited at its bottom end. Posi-
tive ignition is indicated by the melting away of the bottom end of the sample
rod in the pressurised oxygen. If the test sample is able to support burning at
the pressure being evaluated, the test sample will be consumed as the burning
progresses up the rod. Conversely, if the test sample is not able to support burn-
ing, the test sample will extinguish and the sample rod will remain. This test-
ing, in accordance with these test methods, has been performed for many years
and the results for numerous metallic materials are publicly available [1–3]. A
sample of the results for several metals, tested as 0.32-cm-diameter cylindrical
metal rods and in smaller (mesh) configurations is presented in Table 1. Pre-
sented in Table 1 is the metallic material and its specific threshold pressure.
The threshold pressure, or lowest burn pressure, is the lowest pressure the me-
tallic materials will support burning along the test sample for a burn length
greater than 30 mm (for a given test pressure).
A check valve, in its most general description, is a valve that checks the
flow of a fluid, that is, it allows the flow to proceed in one direction only. The
checking of the flow can be accomplished by several methods. However, for
the current application, a spring was selected to provide the bias required.
TABLE 1—Lowest burn (threshold) pressure and highest no-burn pressure for several metallic mate-
rials [4].
Material Lowest burn Highest no-burn pressure

(0.32-cm-diameter rod) (threshold) pressure MPa (psia) MPa (psia)
Carbon steel 0.7 (100) None
316 Stainless steel 1.4 (200) 0.8 (111)
Inconel 600 17.2 (2500) 13.8 (2000)
Monel K-400 None >68.9 (10 000)
Nickel 200 None >68.9 (10 000)
Brass
Material Threshold Highest no-burn pressure

(sintered or wire mesh) (lowest burn) pressure MPa (psia) MPa (psia)
316 Stainless steel 0.082 (12.4) None

Monel K-400 0.69 (100) 0.082 (12.4)
Nickel 200 None >68.9 (10 000)
Bronze None >68.9 (10 000)
The spring was critical because it required both certain essential material prop-
erties and had to be burn resistant up to (or above) 34.5 MPa (5000 psia) in its
actual use configuration when there could be credible ignition sources present.
For a typical check valve, there is typically present the potential for ignition by
particle impact, by adiabatic compression on any soft goods or contaminants
that could be present from inadequately filtration or perhaps contaminant pro-
motion from the presence of non-volatile residue (there is, typically, a small
pressure differential across a normal check valve-based on its cracking
pressure-making high velocities unlikely except perhaps during start-up were
initial pressure transients can lead to high velocities and the mobilisation of
particulate). Because of the presence of these credible ignition mechanisms
and using the design principal of situational non-flammability required that the
body materials as well as the small diameter spring material not be able to sup-
port burning at 34.5 MPa (5000 psia). These constraints, and the data presented
in Table 1, led to the development and use of a modified Nickel 200 alloy that
was specially processed to provide the required results. As shown in Table 1,
Nickel 200 is unable to support burning in small configurations similar to the
spring configuration utilised.
The first application for the checking device was to be a filling port for an
oxygen system. Figure 2 shows the associated schematic of the fill port along
with the materials selected for each part (and the associated threshold pressure
for that material as specified in Table 1). As can be seen, only metallic materi-
als are used and the materials selected have been shown to not support self
sustained burning in pure oxygen at 34.5 MPa (5000 psia) as per the values
provided in Table 1.
FIG. 2—Schematic of check valve in fill-port application. Shown are the mate-
rials of construction and their associated highest pressure where no burning
occurred, as indicated by testing in accordance with ASTM G124 and specified
in Table 1.
Figure 2 provides a general schematic for a fill port utilising a check valve
designed by applying the concept of SNF, where the materials are selected to
resist self-sustained burning (according to ASTM G124) in a pure oxygen envi-
ronment when credible ignition sources are present up to its intended use pres-
sure of 34.5 MPa (5000 psia)—indeed, as indicated by the metallic materials’
burning parameters as given in Table 1, perhaps even up to 70 MPa (10,000
psia) when produced in accordance with standard mechanical strength require-
ments (as shown by both standard theoretical calculations and proven during
normal hydrostatic demonstration). It should be noted that the data presented in
Table 1 is applicable because all metallic material thicknesses in the developed
component are greater than or equal to the test sample diameter (0.32 cm) of
the cylindrical test sample used in ASTM G124 except the spring and the filter.
As already noted, the spring was made of Nickel 200 (1 mm diameter) and
the filter made of sintered bronze so that both these subparts also have thresh-
old pressures (as given in Table 1) that indicate the they will not support burn-
ing up to (or greater than) the design pressure of 34.5 MPa (5000 psia), that is,
through applying the design principal of situational non-flammability, a check
valve that cannot burn can be produced.
Engineering Information
Shown in Fig. 3 is an actual fill valve and several other check valves=check-
valve applications that were produced from the schematic and materials indi-
cated in Fig. 2 (or, as in the case of the in-line check valve, similarly acceptable
metallic materials, here the brass housing, were replaced with a Monel K-400
housing). All the componentry shown in Fig. 3 has the same functional check
FIG. 3—Photo of various prototyped componentry produced from the materi-

als and functional schematic shown in Fig. 2.
element within them and bore diameter (meaning the flow performance of the
components would be the same). The componentry shown would be expected
to exhibit burn resistance up to the pressures indicated, but still required valida-
tion of the performance as indicated by flow rate through the valve as a func-
tion of differential pressure, reverse leak rate as a function of back pressure,
extended life cycle testing, and mechanical limit testing. The performance of
the checking mechanism developed was then rigorously evaluated in laboratory
trials under normal (ambient) and extreme (cold, heated, excessive flow, and
pressure) conditions and shown to be equal to or better than the performance of
similar componentry that contain materials that can support self-sustained
burning if ignition sources are present. The developed valve has since been
utilised in many applications as fill ports, check valves, compressors, and
manifold inserts, and shown to perform well in service.
Summary
All non-metallic materials and many metallic materials will burn if ignited in a
pressurized oxygen-enriched atmosphere making the fire safety of related com-
ponents and systems constructed from these materials of critical importance. A
design approach to ensure the fire safety of oxygen components was described
and applied to the design, development, and demonstration of a new flow-
check device. The design approach used, situational non-flammability, essen-
tially breaks the fire triangle on the “fuel” leg by using only materials that will
not support self-sustained burning at the use conditions, even in the presence of
aggressive ignition mechanisms. A check insert that was developed in this
way, by definition, is practically unable to burn, and shown to be able to

achieve equivalent performance characteristics (e.g., pressure rating, cycle
characteristics, and endurance=life, all with very low (or zero) reverse leak
rates) to similar components that use materials known to support burning if
ignited in oxygen. The developed configuration has been successfully used in
industry in a number of applications and also on military aircraft as fill valves.
References
[1] Steinberg, T. A., Rucker, M. A., and Beeson, H. D., “Promoted Combus-
tion of Nine Structural Metals in High-Pressure Gaseous Oxygen; A Com-
parison of Ranking Methods,” Flammability and Sensitivity of Materials
in Oxygen-Enriched Atmospheres: Vol. 4, ASTM STP 1040, J. M. Stoltz-
fus, F. J. Benz, and J. S. Stradling, Eds., ASTM International, West
[2] McIlroy, K., Zawierucha, R., and Drnevich, R. F., “Promoted Ignition
Behaviour of Engineering Alloys in High-Pressure Oxygen,” Flammabil-
ity and Sensitivity of Materials in Oxygen-Enriched Atmospheres: Third
Volume, ASTM STP 986, D. W. Schroll, Ed., ASTM International, West
[3] Zawierucha, R. and McIlroy, K., “Promoted Ignition-Combustion Behav-
iour of Selected Engineering Alloys in Oxygen Gas Mixtures,” Flamma-
bility and Sensitivity of Materials in Oxygen-Enriched Atmospheres: Vol.
4, ASTM STP 1040, J. M. Stoltzfus, F. J. Benz, and J. S. Stradling, Eds.,
[4] ASTM Manual 36, Safe Use of Oxygen and Oxygen Systems: Handbook
for Design, Operation, and Maintenance, 2nd ed., H. D. Beeson, S. R.
Smith, and W. F. Stewart, Eds., ASTM International, West Consho-
hocken, PA, pp. 18–21.
STP 1561, 2012
DOI:10.1520/STP20120011
Owen Plagens,1 David Lynn,1 Martin Castillo,1

Todd Paulos,2 and Theodore Steinberg1
Combustion Products of Bulk Aluminum

Rods Burning in High-Pressure Oxygen
REFERENCE: Plagens, Owen, Lynn, David, Castillo, Martin, Paulos, Todd,

and Steinberg, Theodore, “Combustion Products of Bulk Aluminum Rods
Burning in High-Pressure Oxygen,” Flammability and Sensitivity of Materials
QC; STP 1561, Samuel Edgar Davis and Theodore A. Steinberg, Editors,
pp. 233–244, doi:10.1520/STP20120011, ASTM International, West Con-
shohocken, PA 2012.
ABSTRACT: Characterization of the combustion products released during
the burning of commonly used engineering metallic materials may aid in
material selection and risk assessment for the design of oxygen systems.
The characterization of combustion products in regard to size distribution
and morphology gives useful information for systems addressing fire detec-
tion. Aluminum rods (3.2-mm-diameter cylinders) were vertically mounted
inside a combustion chamber and ignited in pressurized oxygen by resistively
heating an aluminum/palladium igniter wire attached to the bottom of the test
sample. This paper describes the experimental work conducted to establish
the particle-size distribution and morphology of the resultant combustion
products collected after the burning was completed and subsequently
analyzed. In general, the combustion products consisted of a re-solidified oxi-
dized slag and many small hollow spheres of size ranging from about 500 nm
to 1000 lm in diameter, surfaced with quenched dendritic and grain-like
structures. The combustion products were characterized using optical and
scanning electron microscopy.
KEYWORDS: combustion products, bulk aluminum, particle size, morphol-
ogy, aluminum, aluminum oxide
Manuscript received February 5, 2012; accepted for publication April 18, 2012; published online
xx, xxxx.
1
School of Chemistry, Physics and Mechanical Engineering, Queensland Univ. of Technology,
Brisbane, Queensland, Australia.
2
Alejo Engineering Inc., 15562 Graham St, Huntington Beach, CA 92649.
Copyright V
19428-2959.
233
Introduction
Numerous studies on the combustion of aluminum have focused on the com-
bustion behavior of powders and small single particles in oxygen or oxygen/
nitrogen mixture at atmospheric pressure [1–9]. The combustion behavior of
aluminum under these conditions and of this configuration is difficult to apply
to bulk aluminum configurations; however, many of the same phenomena,
such as phase changes and smoke production, translate well. The combustion
behavior of aluminum particles has been shown in several publications to be a
multiple-stage process characterized by several distinct stages. Dreizin explains
that stage 1 starts with rapid heating to the boiling point of aluminum at
2730 K (above the melting point of the Al2O3 passivation layer at 2323 K) and
the solution consists of dissolved oxygen and molten aluminum combining into
an oxygen-lean solution with gaseous aluminum present [1]. Stage 2 consists
of gaseous aluminum, oxygen-lean solution (<1 at. %) and an oxygen-rich
solution (10 at. %) at a temperature of approximately 2510 K. The combus-
tion behavior during stage 2 exhibits typically strong oscillations in observed
radiation, and the flame and smoke cloud are no longer symmetrical or
spherical. Stage 3 consists of a noticeable drop in radiation intensity, where the
temperature drops to about 2300 K, below the melting point of Al2O3. The pre-
cipitation of the final product, Al2O3 occurs at this stage [1].
Wilson et al. [7] investigated the combustion products of bulk zinc, vana-
dium, molybdenum, tungsten, and silicon at pressures ranging from 3.44 MPa
(500 psi) to 68.9 MPa (10 000 psi) and found that each metal tested produced
a powder of varying particle shape and size distribution. Additionally, the
effect of pressure was investigated, which showed that increased pressure
generally yielded larger particle sizes. Flagan and Lunden [8] investigated
the sizes of nanoparticles condensing out of a vapor and found that the key
factors dictating the particle sizes are the initial partial pressure of the vapor
and cooling rate. In contrast to the results of Wilson et al., Flagan and Lun-
den found that increased pressure decreases particle size. Additionally,
increased cooling rate decreases particle size and increased residence time
increases particle size.
Experiment
Cylindrical aluminum rods, of 60-mm length and 3.2-mm diameter were verti-
cally mounted inside a combustion chamber, and ignited in pressurized oxygen
by resistively heating an aluminum/palladium igniter wire attached to the bot-
tom of the test sample. Pure oxygen (99.9 %) was used at 2.76 MPa (400 psi) to
ensure that the samples would burn. The combustion products were collected
from the chamber—the slag was collected in a copper cup and the particles
were washed off the inner surfaces of the chamber using acetone, which was
PLAGENS ET AL., doi:10.1520/STP20120011 235
subsequently allowed to evaporate. The dry samples were then stored in glass
sample vials.
To investigate the morphology of individual particles, samples were pre-
pared for scanning electron microscopy (SEM) analysis by placing the dry
samples on adhesive carbon tape mounted on SEM sample holders. In the case
of investigating the inner surfaces of hollow particles, a hollow particle was
carefully placed on adhesive carbon tape mounted on an SEM sample holder
and was then crushed with a metal spatula to reveal the inner surface.
To investigate the size distribution, samples for optical microscopy were
prepared by spreading an entire sample on several adhesive carbon circular
strips mounted on glass microscope slides. Images of the samples were taken
and analyzed.
FIG. 1—SEM images of the surfaces of aluminum-oxide particles, where (a)

exhibits typical particles smaller than 500 lm, (b) shows the same particle at
higher magnification, (c) displays a typical particle larger than 500 lm, and
(d) exhibits the same particle at higher magnification.
Results
Demonstration of Morphologies
SEM analysis was used to image the exterior and interior surfaces of the
hollow particles. Figure 1 shows the exterior surfaces of typical particles
produced during the experiment. From visual inspection of SEM images,
particles smaller than 500 lm generally exhibited rapidly quenched
dendritic surfaces, as shown in Figs. 1(a) and 1(b). Particles larger than
500 lm generally exhibited non-uniform grain morphologies, as shown in
Figs. 1(c) and 1(d).
Figure 2 displays the interior surfaces of a particle 500 lm in diameter.
The interior surfaces of the particles contained areas of ordered rapidly
FIG. 2—SEM images of the interior surfaces of aluminum-oxide particles

showing the (a) interior surface of the particle, (b) physically ordered den-
drites, (c) physically disordered dendrites, and (d) the cross section of the shell
of the particle.
quenched dendritic structures, as shown in Fig. 2(b), where the directions of

growth are similar. Figure 2(c) displays rapidly quenched dendrite structures
formed in a physically less ordered fashion than that in 2(b). Figure 2(d)
depicts the fracture surface of a hollow spherical formed particle, where the
bottom side is the interior of the particle.
Figure 3 displays several particle morphologies obtained using an opti-
cal microscope. Unlike electrons in SEM, the visible light is able to pene-
trate the surface of the particles revealing that the particles are hollow and
translucent. Additionally, bubbles trapped in the shell and the degree of
transparency, relating to the thickness of the particle shell, are able to be
observed.
Size Distribution
Figure 4 displays graphs of the size distributions measured. The first graph was
created using a series of images taken by an optical microscope at 40 zoom
of all collected particles (14 000 individual particles) from a single test. The
second graph was created from an SEM image of a cluster of small particles
FIG. 3—Optical photographs of the rapidly quenched hollow spherical

aluminum-oxide particles formed.
FIG. 4—Graphs displaying the particle-size distribution: (a) graph obtained

by automated particle counting of whole sample (14 000 individual particles)
overlaid with a fitted lognormal curve (l ¼ 4.8339, r ¼ 0.7464), and (b) graph
obtained by manual particle counting of a sample of very small particles on the
side of a large particle (900 individual particles) overlaid with a fitted log-
normal curve (l ¼ 0.3207, r ¼ 0.6029).
(900 individual particles) on the side of a large particle (>500 -lm diameter)
similar to the micrograph shown in Fig. 6. Overlaid on both these graphs are
lognormal distributions produced from calculations of the mean and standard
deviation of the data.
Discussion
Morphology
The aluminum-oxide particles were generally hollow white spheres; however,
some were shades of orange possibly because of traces of other elements, such
as copper, as the reaction chamber contains copper components in the vicinity
of the combustion [although energy dispersive spectroscopy (EDS) analysis
found no difference between white and orange particles]. On closer examina-
tion under an optical microscope, they appeared glassy and translucent
because of ridges on the exterior surfaces and dendritic structures on the inte-
rior surfaces. These ridges and boundaries were mostly because of the granular
or dendritic morphology of the outer and inner surfaces of the particles. Addi-
tionally, in the larger particles, gas bubbles were present in the particle shell.
The thickness of the particle shell varied from particle to particle judging from
the range of transparency of the particles. Additionally, the thickness of the
particle shell was non-uniform for each particle. The particle shown in Fig. 2
had a diameter of 500 lm and a thickness varying from 5 lm to 60 lm.
The outer surfaces of the particles were glassy in appearance but rough on
the micrometer scale. The larger particles, those with a diameter greater than
500 lm, tended to have a granular surface, as shown in Fig. 1(c), however those
smaller than 500 lm tended to have a dendritic surface as shown in Fig. 1(a).
The inner surfaces of the hollow particles were clearly dendritic. Some areas of
the inner surfaces had very ordered arrays of dendrites with direction of growth
towards the center of the hollow particle as shown in Fig. 2(b). However,
within the same particle, another area on the inner surface may have more dis-
ordered dendrites as shown in Fig. 2(c).
The observations of the rapidly quenched dendrites suggest that there was
sufficient time and temperature for the material to begin to transition into an
ordered dendritic structure in the rapid combustion process. This verifies that
the material formed retains enough energy to begin the heterogeneous nuclea-
tion of the dendritic fingers. Although partial cooling could have taken place
from contact with the reaction chamber, there was no observation of flattened
sides or non-perfect spherical geometry of the particles.
Most of the aluminum-oxide particles viewed under the optical microscope
appeared to be hollow. An important question was brought up during this
investigation: Why are the particles hollow? For a particle with an average
thickness of 30 lm and a diameter of 500 lm, about 32 % of the volume is
solid aluminum oxide (possibly porous). Possible factors include oxygen solu-
bility of liquid aluminum and liquid aluminum oxide, and other forms of excess
oxygen, which release gaseous oxygen during cooling. The variation of the sol-
ubility of oxygen in liquid aluminum caused by temperature suggests contrac-
tion rather than expansion during cooling because liquid aluminum is able to
dissolve more oxygen as temperature decreases [1]. Also, only a very small
amount of oxygen can dissolve in liquid aluminum when compared to other
metals with estimates ranging from 0.1 at. % [1] to 0.003 wt. % or less [10].
Corresponding values for the solubility of oxygen in liquid Al2O3 is not avail-
able. Additionally, there is evidence that several sub-oxide species, namely
AlO, Al2O, and Al2O2, are formed during combustion [1]. These compounds
have a lower oxygen-to-aluminum ratio than Al2O3, thus not contributing to
excess oxygen when reacting to form a stable oxide, however, they may have
differing oxygen dissolution properties. Other ways oxygen can be held are
adsorption and the formation of high-oxide ions. Iron, for example, has been
shown to produce a range of high-oxide ions during combustion, namely
FeO21, Fe2O54, FeO33, and Fe2O78 [11]. These high-oxide ions were
shown to be a mechanism for oxygen transport to the reaction zone, which in
the case of iron is the solid/liquid interface. The combustion behavior of alumi-
num differs from that of iron as a significant amount of gas-phase oxidation
occurs during the burning of aluminum. However, it is still possible that
oxygen is transported to the freshly melted aluminum in this manner. It is
important to note that most of the aluminum combustion studies are done at
1 atm of pressure where some phenomena may be repressed. This study was
done at about 27.2 atm (400 psi) of pressure.
Size Distribution
The size distribution of particles collected ranged over three orders of magni-
tude, from 1 lm to 1000 lm. Many of the larger particles were most likely
produced because of ejecta from the molten mass during combustion and tail-
ing of the molten mass, where the molten mass separates from the end of the
rod as it becomes too heavy, and as it detaches, a set of smaller liquid drops are
created. Additionally, it is possible that some of the smaller particles may
have been produced by the ignition of the aluminum/palladium igniter wire as
small particles of aluminum oxide and palladium in intimate contact were not
uncommon in previous experiments by the author, as shown in Fig. 5.
This micrograph was taken from another experiment using the same alumi-
num/palladium igniter wire; however, the sample did not contain aluminum.
Optical microscopy was used to determine a size distribution by analyz-
ing an entire sample. Only one level of zoom was used to compromise reso-
lution, amount of data and depth of field, and as a result, the analysis
produced sound results only for particles larger than 100 lm for images taken
by this method. The counting procedure consisted of some simple image
processing on each micrograph image and determining the Feret diameter
(longest line from edge to edge) for each particle detected. For accuracy, this
was compared to a smaller sample of about 2500 particles which were man-
ually measured and counted, which produced a similar distribution but with a
FIG. 5—SEM micrograph of an aluminum-oxide particle (lower) and a palla-

dium particle (upper) in contact.
slightly lower mean and standard deviation. A histogram produced from the
particle count, shown in Fig. 4(a), reveals a peak around 100 lm where,
because of the level of zoom, the method begins to become inaccurate, sug-
gesting that the actual peak should be much more toward the left. This is
also evidenced by SEM micrographs similar to Fig. 6 of numerous small
FIG. 6—SEM micrograph of small particles on the surface of a large particle.

aluminum-oxide particles of 1–10 lm on the surface of larger particles. This

is consistent with observations of Bucher et al., where small spheres of
Al2O3 with diameters of 100–200 nm were found within the high-density
smoke region around a small burning aluminum particle [5]. Particles of this
size could be present in the sample, but because of the much longer time-
frame and higher pressure of this experiment, they most likely grew or coa-
lesced to form larger particles. Figure 4(b) shows a histogram of a sample of
small particles from an SEM image similar to Fig. 6, which shows a peak
around 1 lm.
The distribution of particle sizes fit well into lognormal distributions as
shown in Fig. 4. It is unsure, however, if there is one encompassing lognormal
distribution for all particles collected or if, because of several mechanisms,
there exists two or more overlapping distributions. It is expected that the
particle-size distribution would be lognormal because many studies concerning
particle-size distributions of finely divided systems across a multitude of disci-
plines result in finding a lognormal distribution [9].
Implication on Fire Detection

In this experiment, particles formed under no-flow conditions. This allowed
nucleation and coalescence of combustion products to progress with little hin-
drance to create a wide distribution of particles. If the experiment was repeated
with the sample subjected to a flow of oxygen to simulate a realistic scenario,
it is expected that coalescence of combustion products would be reduced,
leading to a smaller average particle size and a narrower distribution. Some
particles formed could be very small and very light because of high porosity
(possibly a thin membrane filled with oxygen), allowing them to remain
airborne for extended periods of time. These particles could be detected by
properly tuned detectors (using information about the size distribution and
morphology of combustion products) and appropriate suppression measures
can be taken. Another hazard posed by these small particles is that Al2O3 can
become an abrasive hazard to surrounding systems and the detection of these
particles can prompt investigation into whether other systems have been
affected.
Conclusion
The combustion of bulk aluminum rods in high-pressure oxygen typically pro-
duces white hollow spheres of Al2O3. This experiment demonstrated that the
size distribution produced ranged from about 500 nm to 1000 lm and followed
a lognormal distribution with a hypothesized peak in the order of 1–10 lm. The
exterior surfaces of particles were glassy and translucent in appearance and
were typically granular for particles larger than 500 lm and dendritic for
particles smaller than 500 lm. The interior surfaces were lined with physically
ordered and non- ordered dendritic structures. These dendritic structures on the
interior surfaces of the particles, in particular, have not been previously investi-
gated in the literature in any way.
References
[1] Dreizen, E. L., “Phase Changes in Metal Combustion,” Prog. Energ.

Combust. Sci. Vol. 26, 2000, pp. 57–78.
[2] Badiola, C., Gill, R. J., and Dreizin, E. L., “Combustion Characteristics of
Micron-Sized Aluminum Particles in Oxygenated Environments,” Com-
bust. Flame, Vol. 158, 2011, pp. 2064–2070.
[3] Bucher, P., Yetter, R. A., Dryer, F. L., Parr, T. P., Hanson-Parr, D. M.,
and Vicenzi, E. P., “Flame Structure Measurement of Single,
Isolated Aluminum Particles Burning in Air,” Twenty- Sixth Symposium
(International) on Combustion, The Combustion Institute, 1996, pp.
1899–1908.
[4] Bucher, P., Yetter, R. A., Dryer, F. L., Parr, T. P., and Hanson-Parr,
D. M., “PLIF Species and Ratiometric Temperature Measurements of
Aluminum Particle Combustion in O2, CO2 and N2O Oxidizers, and
Comparison With Model Calculations,” Twenty-Seventh Symposium
(International) on Combustion, The Combustion Institute, 1998, pp.
2421–2429.
[5] Bucher, P., Yetter, R. A., Dryer, F. L., Vicenzi, E. P., Parr, T. P., and
Hanson-Parr, D. M., “Condensed-Phase Species Distributions about Al
Particles Reacting in Various Oxidizers,” Combust. Flame, Vol. 117,
1999, pp. 351–361.
[6] Dreizin, E. L., “Experimental Study of Stages in Aluminum
Particle Combustion in Air,” Combust. Flame, Vol. 105, 1996, pp.
541–556.
[7] Wilson, D. B., Sircar, S., Hornung, S., and Stoltzfus, J. M., “Analysis of
Metals Combustion through Powder Production,” Flammability and Sen-
sitivity of Materials in Oxygen-Enriched Atmospheres: Eighth Volume,
ASTM STP 1319, W. T. Royals, T. C. Chou, and T. A. Steinberg, Eds.,
American Society for Testing and Materials, 1997.
[8] Flagan, R. C. and Lunden, M. M., “Particle Structure Control in Nanopar-
ticle Synthesis from the Vapor Phase,” Mater. Sci. Eng.: A, Vol. 204,
1995, pp. 113–124.
[9] Kiss, L. B., Söderlund, J., Niklasson, G. A., and Granqvist, C. G.,
“The Real Origin of Lognormal Size Distributions of Nanoparticles in
Vapor Growth Processes,” Nanostruct. Mater., Vol. 12, 1999, pp.
327–332.
[10] Wilson, D. B. and Stoltzfus, J. M., “Modelling of Promoted-Ignition

Burning: Aluminum,” Flammability and Sensitivity of Materials in
Oxygen-Enriched Atmospheres: Eighth Volume, ASTM STP 1319, W. T.
Royals, T. C. Chou, and T. A. Steinberg, Eds., American Society for Test-
ing and Materials, 1997, pp. 258–271.
[11] Steinberg, T. A., Kurtz, J., and Wilson, D. B., “The solubility of Oxygen
in Liquid Iron Oxide during the Combustion of Iron Rods in High-
Pressure Oxygen,” Combust. Flame, Vol. 113, 1998, pp. 27–37.
STP 1561, 2012
DOI:10.1520/STP20120014
Bradley S. Forsyth,1 Gwenael J. A. Chiffoleau,2

Greg A. Odom,1 and Barry E. Newton2
A Comparison of Fire Reaction Effects

of Halogenated Versus Non-Halogenated
Non-Metals in High-Pressure Oxygen
Components
REFERENCE: Forsyth, Bradley S., Chiffoleau, Gwenael J. A., Odom, Greg

A., and Newton, Barry E., “A Comparison of Fire Reaction Effects of Halo-
genated Versus Non-Halogenated Non-Metals in High-Pressure Oxygen
Components,” Flammability and Sensitivity of Materials in Oxygen-Enriched
Atmospheres on September 19–21, 2012 in Montreal, QC; STP 1561,
Samuel Edgar Davis and Theodore A. Steinberg, Editor, pp. 245–261,
doi:10.1520/STP20120014, ASTM International, West Conshohocken, PA
2012.
ABSTRACT: Some high-pressure oxygen-breathing gas components have
been transitioning from predominantly halogenated (fluorinated or chlori-
nated) polymers to non-halogenated materials for seats and seals. The ra-
tionale for the material changes was related to reducing the risk associated
with production of toxic combustion by-products in the event of a component
fire. This paper neither discusses the risks associated with toxicity, nor does
it attempt to determine which risk (ignition or toxicity) is preeminent. Rather,
examples of fires in components from a component test program are exam-
ined to compare the reaction effects for components configured with halo-
genated and non-halogenated materials. A limited comparison of the test
results to actual field fires is also provided. The implications of the ignition
and fire effects are discussed.
KEYWORDS: oxygen fire, reaction effect, halogenated non-metals, non-hal-
ogenated non-metals, polymers
Manuscript received February 9, 2012; accepted for publication July 3, 2012; published online
November 2012.
1
Wendell Hull and Associates, Inc., 5605 Dona Ana Road, Las Cruces, NM 88011.
2
Ph.D., Wendell Hull and Associates, Inc., 5605 Dona Ana Road, Las Cruces, NM 88011.
Copyright V
19428-2959.
245
Introduction
Materials selection for oxygen components has historically been focused on
selecting materials that exhibit the highest degree of compatibility with respect
to ignition resistance and damage potential [1]. For non-metals, this typically
meant selecting materials with the highest auto-ignition temperature (AIT) and
lowest heat of combustion (DHc). Materials with relatively high AIT s (>300 C)
require more energy to ignite, and, therefore, are less likely to ignite in oxygen
service. Materials with a low DHc release less energy when they combust, and,
typically, release the energy at a slower rate, which results in less potential for
propagating combustion to other materials or components if ignited.
Common halogenated polymers and elastomers include materials such as
fluoroelastomers (FKM), perfluoroelastomers (FFKM), fluoropolymers, such as
polytetrafluoroethylene (PTFE), and chlorinated polymers, such as polychloro-
trifluoroethylene (PCTFE). Compared to other non-halogenated materials,
these materials typically exhibit higher AITs, making them, in nearly all cases,
the most ignition-resistant materials available for seal, seat, and similar appli-
cations. Furthermore, these materials often exhibit DHc’s less than other non-
halogenated materials, resulting in less-severe reaction effects if ignited.
Another effect not reflected in the typical material flammability data is burn
rate. Halogenated materials have been observed, if ignited, to burn more
slowly. Therefore, the heat of combustion is delivered over a longer duration,
reducing the peak energy and, therefore, reducing the risk of kindling ignition
to other materials.
When halogenated materials are ignited and combust in an oxygen system,
the fluorine or chlorine in the materials will form some combustion by-
products that are considered to be a toxicity risk if ingested in a sufficient dos-
age [2]. Assessment of the toxicity risk is complex, and involves the mass of
polymer combusted, the amount of potentially toxic by-products produced,
whether the toxic by-products are contained in the flow-path or vented, the
amount ingested by the user over a specified duration, and the likelihood of
severe physiological effects of ingesting that dosage [3].
Much discourse has occurred regarding whether the ignition risk or the tox-
icity risk is preeminent. One perspective is to use the most-ignition-resistant
and lowest-damage-potential materials so that that if ignition is prevented, then
the toxicity risk is, by default, mitigated. Another view is that all non-metals
(halogenated and non-halogenated) are flammable in oxygen, and ignition has
been demonstrated in adiabatic compression materials tests at pressures less
than those used in high-pressure oxygen-breathing gas components (for both
halogenated and non-halogenated materials), and therefore, halogenated mate-
rials provide little reduction in ignition risk [4].
Regardless of which risk might be considered more severe, both halogen-
ated and non-halogenated materials are currently used in high-pressure
FORSYTH ET AL., doi:10.1520/STP20120014 247
oxygen-breathing gas components. Some entities have defined the toxicity risk
as unacceptable, and consequently, established strict requirements or guide-
lines for how halogenated materials can and cannot be utilized in breathing gas
applications. In these applications, non-halogenated materials are often incor-
porated where, if ignition and combustion occur, the by-products would be
contained within the device, or, at least, delivered downstream. Some materials
that have been employed as alternatives to halogenated materials include, but
are not limited to, nylon, ethylene propylene diene monomer (EPDM), silicone
rubber, and polyether ether ketone (PEEK).
A recent test program provided a unique opportunity to study the fire reac-
tion effects in high-pressure oxygen components. As part of the development
of a valve integrated pressure regulator (VIPR) promoted ignition test for the
ASTM G175 standard [5], a round-robin test program was conducted to vali-
date the test method, as performed at various test laboratories. This test method
was intended to create the ignition of the VIPR shut-off valve (SOV) seat, and
evaluate the reaction effect on the component design. Several high-pressure ox-
ygen VIPR manufacturers provided test articles to be used for the round-robin
program. Because the manufacturers and intended markets for the devices var-
ied, some utilized halogenated materials and others did not. For one device, the
manufacturer provided the same device, but in two configurations, one with
halogenated seats and seals, and the other with all non-halogenated seat and
seals. The test program, therefore, provided an excellent opportunity to com-
pare the reaction effects of components configured with halogenated and non-
halogenated materials. VIPRs also represent a severe application regarding
ignition and reaction effects caused by the proximity of the various component
sub-assemblies, which may increase the risk of a fire that propagates through
the device [6]. Because all of the sub-assemblies are contained within a com-
pact device, if one piece ignites, downstream components and materials may
be substantially exposed to the combustion energy, increasing the risk of a
propagating kindling chain.
Test Description
The ASTM G175 promoted ignition test for VIPRs was a test method under
development. The VIPR promoted ignition test objective and approach were
similar to the existing ASTM G175 Phase 2: Regulator Inlet Promoted Ignition
Test. The objective of both methods was to evaluate the “fault tolerance” of
the component being tested. In other words, the test objective was to assess the
ability of the component to contain an ignition event within the device so that
the device neither ruptured nor produced sustained discharge of fire. In each
test, an “ignition pill” was placed upstream of the test article, and the pill
ignited by rapidly pressurizing up to the pill with high-pressure oxygen. For
J_ID: STP DOI: 10.1520/STP20120014 Date: 3-October-12 Stage: Page: 248 Total Pages: 17
TABLE 1—Test parameters.
Test Parameter Specified Value

Test pressure 1.2 times nominal inlet pressure
Test gas Oxygen: 99.5 % (minimum)
Pressurization rate 20 ms (þ0, 5 ms)
Minimum pressure hold period 10 s
Impact tube 1000 mm long 5-mm i.d.
Pressure surge application point VIPR inlet (cylinder port)
Pressure surge cycles 1 cycle
standard pressure regulators in the regulator ASTM G175 Phase 2 test, ignition
of the pill was intended to simulate ignition of a cylinder valve seat upstream
of the regulator. The ignition pill for the existing (regulator) Phase 2 test nomi-
nally delivered 500 cal (2100 J), representing the energy of a typical nylon cyl-
inder valve seat. However, for the VIPR test, the ignition pill nominally
delivered 200 cal (835 J), and was intended only to ignite the VIPR shut-off
valve (SOV) seat, providing little additional energy. In actual VIPR fires inves-
tigated by Wendell Hull and Associates, Inc. (WHA), the origin of most fires
involved ignition of the SOV seat. Therefore, the test evaluated the reaction
effect as initiated by ignition of the SOV seat.
In the round-robin test program each component was tested with the shut-
off valve fully open, and the flow metering device set to the middle of the flow
range. The test pressure was 1.2 times the nominal rated inlet pressure. The test
conditions are summarized in Table 1.
Test Articles
A subset of the test articles tested in the round-robin program is discussed in
this paper to highlight major observations from the test results. The design of
each component included an inlet port, a shut-off valve (SOV), a residual pres-
sure valve (RPV), a filter (as required by ISO 10524-3 [7]), a regulator assem-
bly, and a flow-meter assembly. Figure 1 shows the general configuration of a
typical VIPR as tested by the round-robin program.
Table 2 lists the non-metal materials used for major non-metal parts in
each component. VIPR “A” was configured with all non-halogenated seats
and seals. VIPR “B” was configured with all halogenated non-metal seats
and seals. VIPR “C” was from the same manufacturer and of the same design
as VIPR “B,” but was configured with all non-halogenated seat and seals.
Table 2 also lists typical AIT and DHc values for each type of material for
comparison.
ID: kumarva Time: 17:24 I Path: Z:/3B2/STP#/Vol01561/120160/APPFile/AT-STP#120160

FIG. 1—Valve integrated pressure regulator schematic.
TABLE 2—VIPR test article non-metal materials.
Test Article i.d. Component Material AIT ( C)a DHc (J/g)

A SOV seat Nylon 178 32220
A Filter Sintered bronze n/a
A RPV seat EPDM O-ring 201 47260
A Regulator seat Nylon 178 32220
B SOV seat PCTFE 377 5150

B Filter Sintered bronze n/a
B Relief valve seat FKM O-ring 290 15085
B Regulator seat PTFE 434 4772
C SOV seat Nylon 178 32220
C Filter Sintered bronze n/a
C Relief valve seat EPDM O-ring 201 47260
C Regulator seat Nylon 178 32220
a
All AIT and DHc values are from ASTM Manual 36, “Safe Use of Oxygen and Oxygen Systems,” 2nd Ed.,
Tables 3–12.
TABLE 3—VIPR test results.
VIPR ID SOV Seat Filter RPV Seat Regulator Seat Comments

A Ignited Breached Ignited Ignited RPV piston melted,
and fully and fully and fully regulator O-rings
consumed consumed consumed ignited and consumed,
regulator piston
melted, flow
meter melted
B Ignited Undamaged Undamaged Undamaged Seal in burst disk port
and fully partially consumed
consumed
C Ignited Breached Ignited Ignited Many other soft goods
and fully and fully and fully ignited and partially
consumed consumed consumed consumed within device
Test Results
Table 3 summarizes the test results for the three VIPRs discussed in this report
by characterizing the damage to each sub-assembly. The reaction effect is
described for each sub-assembly of each component. To maintain the confiden-
tiality of the VIPR manufacturers, only limited photographs of similar internal
components are shown in this report. No external photographs are supplied, but
descriptions of reaction effects are provided.
VIPR “A” Test Results

VIPR “A” was configured with non-halogenated materials for all seats and
seals as detailed in Table 2. All oxygen-wetted metal components were con-
structed entirely of brass, except the filter, which was sintered bronze. This
VIPR was tested at a pressure of 3480 psig (240 bar), which represented 1.2
times the rated inlet pressure for this device. It is noteworthy that the nominal
rated pressure of this VIPR, and, therefore, the test pressure, was greater than
the other VIPR tests.
The ignition pill ignited the SOV seat as intended (Fig. 2). The reaction
effect from the seat ignition created a kindling chain that subsequently pro-
duced melt damage or partial consumption of the downstream filter, ignition of
the RPV O-ring seat with partial melting of the RPV poppet (Fig. 3), ignition
and complete combustion of the regulator seat, ignition and complete combus-
tion of the regulator piston large and small O-rings, and substantial melting or
consumption of the regulator piston (Fig. 4). The flow metering device sus-
tained damage as observed by the melting of the brass body at the inlet, and
likely sustained additional internal damage, as was evident by difficulty in turn-
ing the flow selector after the test. Certain locations of the brass VIPR body
were also melted and eroded by the hot combustion by-products. Soot and
FIG. 2—VIPR “A” consumed SOV seat post test.
sparks were discharged from the device, but no flames were observed. Based
on the duration of these observations, this would not be a passing result based
on the test criteria.
At the time of this report, a total of 15 VIPRs of this design and configura-
tion were tested, five each at three different laboratories, all with similar reac-
tion effects as described above. In two of the 15 tests, the regulator spring was
partially ignited and consumed. Ignition of the regulator spring represents a
FIG. 3—VIPR “A” consumed RPV seat and melted piston post test.
FIG. 4—VIPR “A” regulator consumed seat and melted piston post test.
potential worst-case scenario with respect to reaction effects in these types of

devices. Regulator springs are often constructed of carbon steel or stainless steel,
and if ignited, release a relatively large amount of energy due to their mass and
high heat of combustion. In WHA testing of regulators and VIPRs, and observed
in actual field fires, when the regulator spring has ignited, significant damage to
the device has resulted with component breach and discharge of fire.
VIPR “B” Test Results

VIPR “B” was configured with halogenated materials for all seats and seals.
All oxygen-wetted metal components were constructed entirely of brass, except
the filter, which was sintered bronze. This VIPR was tested at a pressure of
2880 psig (200 bar), which represented 1.2 times the rated inlet pressure for
this device.
In most tests, the ignition pill ignited the SOV seat as intended (Fig. 5).
Only one other component in the device sustained any damage. Partial ignition
and combustion of a gasket seal at the burst disk occurred. No ignition or dam-
age occurred to the filter (Fig. 6), regulator seat (Fig. 7) or any other compo-
nent except as noted above. No soot, sparks or flames were observed
discharged from the device. These observations would be a passing result
based on the test criteria.
tion effects as described above. In 27 % of the tests (four of 15) performed
on this VIPR design and configuration, the ignition pill did not ignite the
FIG. 5—VIPR “B” consumed SOV seat post test.
PCTFE SOV seat (Fig. 8). This result was attributed to the ignition resistance
of the SOV seat material, rather than a malfunction of the ignition pill, based
on other observations within the tested VIPR.
VIPR “C” Test Results

VIPR “C” was the same design as VIPR “B”, but was configured with non-
halogenated materials for all seats and seals, as detailed in Table 2. All
FIG. 6—VIPR “B” undamaged filter post test.

FIG. 7—VIPR “B” undamaged low pressure relief device post test.
oxygen-wetted metal components were constructed entirely of brass, except

the filter which was sintered bronze. This VIPR was tested at a pressure of
2880 psig (200 bar) in the same manner as VIPRs “A” and “B.”
The ignition pill ignited the SOV seat as intended. The reaction effect was
described by one test laboratory as explosion-like. The reaction effect from the
seat ignition (Fig. 10) created a kindling chain that subsequently produced melt
damage or partial consumption of the downstream filter (Fig. 11), ignition of
FIG. 8—VIPR “B” undamaged regulator piston post test.

FIG. 9—VIPR “B” undamaged SOV seat post test.
the low pressure relief device spring, melting or consumption of the relief de-
vice poppet and body (Fig. 12), ignition and complete combustion of the regu-
lator seat, ignition and complete combustion of the regulator piston large and
small O-rings, and melting or partial consumption of the regulator piston. The
flow metering device sustained damage, as observed by melting of the brass
body at the inlet, and likely sustained additional internal damage as was evi-
dent by the difficulty turning the flow selector after the test. Certain locations
FIG. 10—VIPR “C” consumed SOV seat post test.

FIG. 11—VIPR “C” consumed filter post test.
of the brass VIPR body were also melted and eroded by the hot combustion by-
products. Soot, sparks, and flames were observed discharged from the device.
These observations would not be a passing result based on the test criteria.
tion effects as described above. In two of the tests of VIPR “C”, the external
plastic shroud of the VIPR was ignited and significantly melted and consumed.
FIG. 12—VIPR “C” low pressure relief device assembly with consumption
and breach post test.
FIG. 13—VIPR “C” consumed regulator seat and melted piston post test.
Comparison to Field Fires

WHA had previously investigated actual fires involving the VIPR “A” design
configured with non-halogenated non-metals consistent with the materials in
the round-robin tests. The kindling chain observed in the field fires and the
extent of damage was considered very similar to the test results described
above. In the field fires, the SOV seat was ignited and consumed (Fig. 14), the
FIG. 14—Consumed SOV seat from field fire involving VIPR “A.”
FIG. 15—Consumed RPV seat and piston melting from field fire involving
VIPR “A.”
filter was melted and partially consumed, the RPV seat O-ring was ignited, the
RPV poppet partially melted (Fig. 15), and the regulator seat ignited and was
totally consumed with substantial melting and consumption of the regulator
piston (Fig. 16). Also, in some actual fires, the plastic shroud of the VIPR was
ignited and substantially melted. Therefore, the results of the VIPR tests are
considered to be representative of the reaction effects observed in the field.
FIG. 16—Consumed regulator seat and melted piston from field fire involving
VIPR “A.”
Summary
The test program to validate the proposed ASTM G175 promoted ignition test
for VIPRs provided an opportunity to study fires in these components. Further-
more, because some test articles were configured with all halogenated polymer
seats and seals, and others configured with all non-halogenated polymer seat
and seal materials, a comparison of the reaction effects based on these materi-
als was available.
In all tests performed, a clear distinction was observed between the reaction
effects of components with halogenated materials versus non-halogenated mate-
rials. In components configured with halogenated non-metals, besides ignition
of the SOV seat (which is the intent of the ignition pill), only partial damage to
one downstream gasket was observed. The regulator seat was not ignited in any
test. Therefore, a kindling chain did not develop within the components config-
ured with halogenated polymers. Also, in some tests, the ignition resistance of
the PCTFE SOV seat prevented ignition of the seat by the ignition pill.
In tests of two different VIPR designs configured with non-halogenated
materials, ignition of every non-metal component downstream of the SOV seat
occurred. Furthermore, melting and consumption was observed for every metal
component downstream of the SOV seat, despite being constructed of highly
burn-resistant brass and bronze. The observed melting of these metal materials,
and transport of the molten slag downstream, significantly contributed to prop-
agation of fire. Metal slag has a high heat capacity and can, therefore, carry
substantial thermal energy to downstream materials or components.
Standard material test data for non-metals indicates that for nearly all
materials, halogenated polymers exhibit superior ignition resistance (high AIT)
and a lower risk of kindling ignition to other materials and components (lower
DHc). Observations from WHA materials testing also indicate that halogenated
polymers burn more slowly, reducing the risk of kindling. The results of the
VIPR promoted ignition tests are consistent with these conclusions. Ignition re-
sistance of halogenated materials, as configured in an actual component (not
just a material test), was demonstrated by the lack of kindling chain develop-
ment in these components, and containment of the ignition within the device.
The ignition resistance was further demonstrated when, in some tests, the halo-
genated SOV seat was not ignited by the ignition pill. In contrast, for the tested
VIPRs configured with non-halogenated materials, a kindling chain developed
that ignited every non-metal component downstream of the SOV. Furthermore,
the kindling chain also involved substantial melting of metal components
which contributed to the propagation of fire. Ignition of the external shroud in
some tests represents propagation of the kindling chain to materials external to
the device. Based on these results, the two VIPRs configured with non-
halogenated materials would not pass the proposed ASTM G175 VIPR pro-
moted ignition test.
At the time of the writing of this paper, WHA had investigated or was
aware of approximately 30 actual fires involving VIPRs, and all of the sub-
ject components were configured with non-halogenated non-metals for in-
ternal seats. Some of these devices used halogenated materials for external
seals. WHA was not aware of any fires of VIPR components with halogen-
ated non-metals for the SOV seat or regulator seat. However, the popula-
tion of VIPR devices in-service may be greater in markets that prefer or
require non-halogenated materials. One of the VIPRs tested (VIPR “A”)
was involved in some of the field fires investigated by WHA. Comparison
of the VIPR “A” test results to the field fires indicated that the reaction
effects observed in the tests were comparable to those observed in the
field.
The ignition history and test data described above are considered relevant
to material selection in assessing the risk of ignition and the consequence of
ignition, whether it is fire propagation, or the production of potentially toxic
combustion by-products.
References
[1] ASTM G63, 2007, “Standard Guide for Evaluating Nonmetallic Materials
for Oxygen Service,” Annual Book of ASTM Standards, Vol. 14.04,
ASTM International, West Conshohocken, PA, pg. 6.
[2] Chiffoleau, G., Forsyth, B. S., and Newton, B. E., “Analysis of Toxic
Combustion Product Exposure in Oxygen Breathing Equipment,” J.
Occup. Environ. Hyg. (to be published).
[3] Chiffoleau, G., Forsyth, B. S., Newton, B. E., and Forsyth, E., “Risk Eval-
uation Approach for Non-Metals in Oxygen Breathing Systems,” Flam-
mability and Sensitivity of Materials in Oxygen-Enriched Atmospheres,
ASTM STP1561, Vol. 13, ASTM International, West Conshohocken, PA
(to be published).
[4] Barthelemy, H., “Combustion Products Toxicity of Non-Metallic Materi-
als Used for Medical Oxygen Equipment,” Flammability and Sensitivity
of Materials in Oxygen-Enriched Atmospheres: ASTM STP 1454, Vol. 10,
T. A. Steinberg, H. D. Beeson, and B. E. Newton, Eds., ASTM Interna-
tional, West Conshohocken, PA, 2003, pg. 17.
[5] ASTM G175-03, 2011, “Standard Test Method for Evaluating the Ignition
Sensitivity and Fault Tolerance of Oxygen Regulators Used for Medical
and Emergency Applications,” Annual Book of ASTM Standards, Vol.
14.04, ASTM International, West Conshohocken, PA.
[6] Forsyth, E. T., Newton, B. E., Chiffoleau, G. J. A., and Forsyth, B. S.,
“Oxygen Fire Hazards in Valve-Integrated Pressure Regulators for
Medical Oxygen,” Paper ID JAI102296, J. ASTM Int., Vol. 6, No. 10,

2009.
[7] ISO 10524-3, 2005, “Pressure Regulators for Use with Medical Gases—
Part 3: Pressure Regulators Integrated with Cylinder Valves: International
Organization for Standardization, Geneva, Switzerland, pg. 9.
STP 1561, 2012
DOI:10.1520/STP20120013
Gwenael J. Chiffoleau,1 Barry E. Newton,1

Bradley S. Forsyth,1 and Elliot T. Forsyth1
Risk Evaluation Approach for Polymers

in Oxygen Breathing Systems
REFERENCE: Chiffoleau, Gwenael J., Newton, Barry E., Forsyth, Bradley

S., and Forsyth, Elliot T., “Risk Evaluation Approach for Polymers in Oxygen
Breathing Systems,” Flammability and Sensitivity of Materials in Oxygen-
Enriched Atmospheres on September 19–21, 2012 in Montreal, QC; STP
1561, Samuel Edgar Davis and Theodore A. Steinberg, Editors, pp. 262–
PA 2012.
ABSTRACT: An approach for fire risk assessment of polymers in oxygen
breathing systems is presented. The approach separates the fire risk of poly-
mers into the probability of ignition and the consequence (severity) of ignition.
Recommended methods for estimating the probability of ignition, the severity of
fire damage, and the severity of any toxic combustion product exposure are pro-
vided. The ignition probability and severity of fire damage are based on ASTM
G63. The severity of toxic combustion product exposure is based on ISO 15001
and a new methodology developed as part of a toxicity risk research program.
Consistent with international risk assessment standards for medical devices
(ISO 14971), the approach then combines the elements of risk according to the
risk evaluation method. To complete the risk assessment, mitigation strategies
are also presented for various results of the risk evaluation.
KEYWORDS: nonmetals, polymers, risk, analysis, oxygen, compatibility,
ignition, probability, consequence, toxicity, fire
Introduction
The selection of non-metallic materials (polymers) in oxygen systems is criti-
cal because they are far more ignitable and flammable design materials than
metals. The oxygen compatibility of polymers is an important factor to con-
sider in the selection process, along with cleanliness, mechanical properties,
Manuscript received February 9, 2012; accepted for publication June 25, 2012; published online
November 2012.
1
Wendell Hull & Associates, Inc., Las Cruces, NM 88007.
Copyright V
19428-2959.
262
CHIFFOLEAU ET AL., doi:10.1520/STP20120013 263
and compatibility with other materials. These factors can all influence the fire
hazard and are influenced by temperature and pressure conditions, wear, me-
chanical loads, and environmental conditions. Oxygen compatibility is defined
as the ability of a material to co-exist with both oxygen and an ignition source.
All polymers (e.g., plastics and elastomers) are considered flammable if ignited
in pressurized oxygen. Oxygen compatibility for polymers is a relative indica-
tor of a material’s ability to resist ignition and burning and to minimize the
combustion energy release. Polymers considered to exhibit the highest oxygen
compatibility might not be suitable in certain applications because of poor me-
chanical properties, which can significantly reduce the oxygen compatibility if
the polymer’s geometry is compromised. Therefore, when selecting polymers
for oxygen service, oxygen compatibility must be considered with other mate-
rial selection factors, such as mechanical properties.
Another material selection factor that has been considered for polymers is
the toxicity of combustion products. This applies to breathing oxygen systems
used in, for example, medical, military, aerospace, and diving applications. If a
polymer ignites and the resulting fire is contained in the oxygen system, the
combustion products might be delivered to the end user (e.g., patient, pilot).
The toxicity of the combustion products will depend on the chemistry and com-
position of the polymer, the quantity of oxygen available, and the reaction tem-
peratures. The toxicological severity of combustion products depends on not
only the types of combustion products but also the quantity, as well as the dura-
tion of exposure. The quantity and exposure duration are defined as the dosage
and are system dependent. Factors such as the mass of polymer, oxygen pres-
sures, the volume of oxygen surrounding the polymer, and flow rates all influ-
ence the potential dosage.
This paper proposes an approach for performing a fire risk assessment of
polymers in oxygen breathing systems and includes a method for incorporating
toxicity data into the risk evaluation. This approach also defines the relation-
ships among ignition probability, fire damage severity, and toxicity of combus-
tion products, and it combines these elements of risk in a manner consistent
with international risk assessment standards for medical devices (ISO 14971)
[1]. To complete the risk assessment, mitigation strategies are also presented
for various results of the risk evaluation.
Background
Risk is defined (ISO 14971, Section 2.16) as the combination of (1) the proba-
bility of the occurrence of harm and (2) the severity of that harm (i.e., conse-
quence) [1]. This does not mean that one arrives at a risk value by multiplying
the two factors. Each component of risk—i.e., probability and consequence—
should be analyzed separately. When analyzing the risks of polymers for use in
high-pressure medical oxygen equipment, the primary hazard is ignition. With-

out ignition, there can be no combustion products or fire damage. An ignition
must occur before combustion products can cause harm (i.e., internal physio-
logical damage to people) and before fire can cause harm (i.e., burn injury or
damage to equipment). Fire and toxicity are the consequences of ignition.
Therefore, the probability of the occurrence of harm is based on the ignition
probability, and the severity of that harm is based on the severity of the com-
bustion products inhaled (exposure) and the severity of the fire damage. Risk
estimation is defined by ISO 14971 as the process used to assign values to the
probability of the occurrence of harm and the severity of that harm [1].
Risk evaluation is defined by ISO 14971 as “the process of comparing the
estimated risk against a given risk criteria to determine the acceptability of the
risk” [1]. This standard does not specify what constitutes acceptable risk, but it
presents the use of risk matrices to describe the probabilities and severities of
the risk associated with a given hazardous situation. The acceptable risk criteria
are organization dependent and generally are defined by organizations. Com-
pletion of the risk evaluation is followed by the risk control step if the eval-
uated risks are found to be unacceptable. Risk control measures can reduce the
severity of the harm or the probability of occurrence of the harm, or preferably
both. Ideally, a polymer material will exhibit a low probability of ignition and
a low severity of ignition (both fire and toxicity severity). Depending on the
organizational risk matrix, a polymer material might still be considered accept-
able if (1) the probability of ignition is low but the severity is high or (2) the
probability of ignition is high but the severity is low. There might be excep-
tions when the severity is significantly high such that no level of ignition prob-
ability is acceptable.
ASTM G63 [2] and ISO 15001 [3] provide guidelines for polymer selec-
tion in oxygen service. However, ASTM G63 provides no guidance, and ISO
15001 provides only limited guidance, on evaluating the risk of toxicity and
methods for incorporating the toxicity risk into the overall fire risk. ISO 15001
provides a test and analysis method to quantify combustion products but pro-
vides limited guidance on incorporation of the toxicity data into a risk evalua-
tion of a medical device.
An ASTM International Committee G4 task group was developed to facili-
tate an industry-sponsored toxicity research program. The program’s goal was
to investigate the toxicity risk of polymer combustion products in oxygen
breathing applications (e.g., medical, scuba, and aerospace devices). A method-
ology for analyzing the severity of toxic combustion product exposure was
developed and consisted of the following three phases:
• Phase 1—Adiabatic compression testing and combustion products
analysis
• Phase 2—Exposure model
• Phase 3—Toxicological risk analysis
Only phase 1 is outlined in ISO 15001, and if phases 2 and 3 are not
included, the severity of combustion products cannot be adequately evaluated
[3]. This methodology was used to study the ignition of a polytetrafluoroethyl-
ene (PTFE) internal seal in a typical medical oxygen breathing application.
The methodology and the results had been submitted for publication at the time
of this writing [4].
Recommended Approach for Fire Risk Assessment of Polymers

The following section proposes a guideline for performing a fire risk assess-
ment for use with polymers in high-pressure oxygen breathing applications.
This guideline is schematically shown in Fig. 1 as a flow diagram. Each of the
steps in the flow diagram is presented and discussed below. This guideline is
consistent with the risk terminology and methods outlined in ISO 14971 [1].
Material Suitability
Objective: evaluate the material’s suitability for the application and ensure that
it can perform the design requirements.
Approach:
a. Evaluate the material’s cleanliness.
i. Contaminants, such as particles and non-volatile residues, can
represent the most ignitable and hazardous materials in an ox-
ygen system. Examples include dislodged flashing, fibers, lint,
dust, machining lubricants, and mold release agents. These
contaminants can remain on the surface of the polymer after
the manufacturing process and can be more ignitable than the
polymer. To avoid contaminant promoted ignition of the poly-
mer, these contaminants must be removed via cleaning.
b. Evaluate mechanical properties.
i. The polymer should exhibit mechanical properties (e.g.,
strength, hardness) that can withstand the design loads and
anticipated wear in the application. Deformation of the poly-
mer in the application could lead to thin cross-sections, leak-
age, or particle generation, all of which increase the risk of
ignition.
c. Evaluate the material compatibility.
i. The polymer should exhibit compatibility with other materials
(not just oxygen) that ensures that the material will not react
chemically or degrade. Polymers can be exposed to other
materials, such as cleaning fluids and lubricants, over their
intended manufacturing and service life in oxygen compo-
nents. Polymers also can be a source of contamination via
FIG. 1—Flow diagram of fire risk assessment of polymers.
physio-absorption processes, whereby additives, such as plas-

ticizers, are extracted from the polymer and re-deposited in
the oxygen system. The diffusion of a plasticizer into an oxy-
gen compatible lubricant can also compromise the oxygen
compatibility of the lubricant.
Likelihood of Ignition
Objective: estimate the probability of ignition.
Approach:
a. Perform an oxygen hazard and fire risk analysis (OHFRA), or
equivalent.
b. Determine, from the OHFRA, which polymers have the highest proba-
bility of ignition based on their material (chemical structure, additives,
fillers [diluent effect], impurities), design configuration, operating con-
ditions (e.g., pressure, temperature, wear, mechanical loads), and expo-
sure to ignition mechanisms.
c. Estimate the likelihood of ignition between different materials using
either of these two methods.
i. Historical data
1. Research historical data that compare the ignition fre-
quency of materials in oxygen equipment.
ii. Testing
1. Perform ignition testing to compare the ignition fre-
quencies of materials under severe conditions. This is
done to minimize the amount of testing required in
order to obtain ignition. Analyze the severity differ-
ence between the test conditions required for ignition
and the normal operating conditions intended in serv-
ice. Use statistical acceleration models that incorpo-
rate this severity difference to predict the ignition
likelihood in actual use conditions [5], e.g.,
a. component testing such as ISO 10524-1 [6]
and ISO 10524-3 [7], or
b. material testing such as ASTM G72 [8] and
ASTM G74 [9].
Severity of Resulting Fire Damage
Objective: estimate the severity of injuries/deaths and damages caused by fire.
Assume ignition of the polymer.
Approach
a. Evaluate the kindling chain progressing from the polymer using an
OHFRA and ASTM G63 [2].
i. Evaluate the polymer’s mass and heat of combustion.
ii. Evaluate the flammability of surrounding materials.
iii. Evaluate the mass and heat of combustion of surrounding
materials.
iv. Determine whether fire can progress through and possibly
breach or exit the oxygen equipment.
b. Evaluate the severity of fire exiting oxygen equipment.
i. Determine whether fire would exit with energetic pressure

release, ejecting materials (hot gases, flames, molten metal,
and fragments).
1. Consult ASTM G63, Table 1, “Reaction Effect
Assessment for Typical Pressure” [2].
2. Consider blast codes for predicting energy release.
ii. Assess the proximity and number of personnel that would be
exposed to this exit point.
1. Estimate the kinds of injuries and damages that would
be sustained.
iii. Determine whether there are other flammable materials that
could be kindled and initiate secondary fires or pressure
breaches (building, common household products, pressurized
vessels, fuel gases/liquids).
iv. Determine the number of people who would be exposed to
these secondary events.
1. Estimate the kinds of injuries and damages that would
be sustained.
Severity of Toxic Combustion Product Exposure

Objective: estimate the severity of injuries/deaths and damages caused by toxic
combustion product exposure [4]. Assume ignition of the polymer.
Approach
a. Determine the conditions created by the fire that could affect toxicity
or dosage.
i. Determine whether the ignition of material(s) identified in
the OHFRA creates a condition that vents the combustion
products external to the device because of a loss of seals due
to fire consumption and/or high pressures from high combus-
tion temperatures.
1. Loss of seal integrity (external seal)
2. Actuation of a relief device
3. Burn-out of the component
ii. Establish whether the ignition of material(s) identified in the
OHFRA is contained within the device.
iii. Evaluate whether all combustion products remain in
the device and can be delivered to and inhaled by the end
user.
b. Consider the use conditions that could affect toxicity or dosage.
i. Determine whether the end user is conscious so that he or
she is able to remove the breathing equipment if irritating
vapors or gases (toxic species) are detected.
ii. Establish whether the device or system has a conserving

mode that provides flow only when the user is breathing.
iii. Determine whether the system continues to flow (and thus
vents byproducts) during user exhalation.
c. Define variables that could affect dosage.
i. Determine the volume at the location where byproducts
would accumulate (such as upstream of a regulator).
ii. Establish the flow rate of gas delivered to the user.
iii. Locate surface areas available for the reaction of toxic spe-
cies prior to their reaching the end user.
d. Define the amount of toxic byproducts that could be generated.
i. Determine the mass of polymer identified in the OHFRA
that could combust.
ii. Determine the mass of toxic compounds produced by the
specific material available for delivery to the user, if data
exist from prior testing and analysis. For example, up to
35% of the fluorine present in PTFE is released in the form
of toxic species when PTFE undergoes complete combus-
tion. The remainder is contained in non-toxic, asphyxiant
compounds (e.g., CF4).
iii. Perform ignition testing and combustion product analysis
[3,9]. In addition to this or as an alternative, use chemical
reaction codes to predict combustion product concentrations,
if no data are available [10].
e. Perform an exposure simulation to evaluate dosage.
i. Employ a mathematical model that calculates the concentra-
tion (dosage) versus time of the gas mixture delivered to the
end user by the system (i.e., dosage).
1. Use the system-specific variables of volume and flow.
2. Utilize the amount of toxic byproducts initially avail-
able for delivery to the user, as identified above.
3. Use the combustion rates of polymers to define the
generation rates of combustion products.
4. Incorporate these parameters into a time-stepping finite
difference calculation that evaluates the gas concentra-
tions in a control volume in which the polymer com-
bustion and combustion gas accumulation would
occur. The exposure model is based on the input of gas
species after ignition occurs (i.e., the generation of
combustion products plus the supply of oxygen) and
the output of gas species (i.e., the mixture of combus-
tion products and oxygen) [4].
f. Perform a toxicological assessment.

ii. Compare the dosage (concentration and exposure duration)
predicted by the exposure simulation to applicable toxicity
laboratory data (e.g., LC50) for each toxic species.
1. Adjust the duration of the dosage to the related time
associated with toxic laboratory LC50 data.
a. Utilize the available LC50 data that represent
experiments performed for durations of 1 to
4 h. Preliminary exposure simulations for
small polymer parts show that dosage would
be within minutes.
2. Assess, for some species, the cumulative effect of
breathing multiple toxic compounds.
Risk Evaluation
Objective: evaluate the overall risk by using risk matrices consistent with ISO
14971 [1]. An example of a risk matrix is provided in Fig. 2.
Approach
a. Risk evaluation is defined by ISO 14971 as “the process of comparing
the estimated risk against a given risk criteria to determine the accept-
ability of the risk.” This standard does not specify what constitutes ac-
ceptable risk, but it presents the use of risk matrices to describe the
probabilities and severities of the risk associated with hazardous situa-
tions. The acceptable risk criteria are organization dependent. The key
in Fig. 2 is usually color coded to represent the organization’s risk
criteria.
b. The likelihood of ignition is captured by the probability levels, often
represented by sequential probability ranges (e.g., 10 4 to 10 5).
FIG. 2—Example of a risk evaluation matrix [1]. The key is usually color
coded to represent the organization’s risk criteria.
c. The severity of ignition is captured by the severity levels, which often

use the severity or number of injuries and deaths and which might also
include the cost of damage. The severity levels for the ignition conse-
quences, resulting fire damage, and toxic combustion product exposure
are obtained from the steps previously outlined.
d. Position the resulting risk in the matrix for each polymer. Each poly-
mer will exhibit two risk values that represent the severity levels for
each ignition consequence. These two risk values will exhibit the same
ignition probability estimate, but the might exhibit different severity
levels based on the estimates from the preceding steps.
e. The resulting positions of the risks (e.g., R1, R2, R3, R4, R5, and R6)
are compared to the acceptability criteria to determine whether the
polymer is acceptable for use.
Risk Mitigation
Objective: develop strategies to reduce the risk to an acceptable level.
Approach
a. If the resulting risk is considered unacceptable according to the risk
matrix, then mitigation strategies should be developed based on the
prior risk analysis performed. Table 1 presents a guide for fire risk mit-
igation based on polymer properties. ASTM G63 provides guidance on
identifying materials based on their ignition resistance and heat of
combustion.
b. Apply the following mitigation strategies (good practices) irrespective
of the scenarios in Table 1:
i. Perform robust qualification testing (e.g., adiabatic compres-
sion testing after endurance testing).
ii. Incorporate design changes that reduce the size/mass of
polymers to such a degree that
1. combustion products produced are of insufficient
quantity to produce a toxicity risk, and
TABLE 1—Risk mitigation strategy for polymers.
Severity
Ignition
Probability Fire Toxicity Toxicity Fire
Low Use materials with low Use materials that produce less toxic
heat of combustion. materials upon combustion.
High Use materials with high Use materials with high ignition resistance
ignition resistance and low that produce less toxic materials upon combustion.
heat of combustion.
2. energy release is insufficient to continue a kindling

chain and cause system burnout.
iii. Ensure all components including polymers are cleaned for
oxygen service and are maintained as clean.
c. Once the risk-mitigating strategies have been implemented, revisit the
risk assessment to demonstrate that the risk is sufficiently reduced to
acceptable levels.
Conclusion
An approach for the fire risk assessment of polymers in oxygen breathing systems
was proposed. The approach included an evaluation of material suitability and sep-
arated the fire risk of polymers into the probability of ignition and the consequence
(severity) of ignition. Recommended methods for estimating the probability of
ignition, the severity of fire damage, and the severity of toxic combustion product
exposure were provided. Consistent with international risk assessment standards
for medical devices (ISO 14971), the approach then combined the elements of risk
according to the risk evaluation method. To complete the risk assessment, mitiga-
tion strategies were also presented for various results of the risk evaluation.
References
[1] ISO 14971, 2007, “Medical Devices—Application of Risk Management

to Medical Devices,” International Organization for Standardization, Ge-
neva, Switzerland.
[2] ASTM G63-99, 2007, “Standard Guide for Evaluating Nonmetallic Mate-
rials for Oxygen Service,” Annual Book of ASTM Standards, Vol. 14.04,
ASTM International, West Conshohocken, PA.
[3] ISO 15001, 2010, “Anesthetic and Respiratory Equipment—Compatibil-
ity With Oxygen,” International Organization for Standardization, Ge-
neva, Switzerland.
[4] Chiffoleau, G., Forsyth, B. S., and Newton, B. E., “Analysis of Toxic
Combustion Product Exposure in Oxygen Breathing Equipment,” J.
Occup. Environ. Hyg. (submitted).
[5] Johnson, R. A., Miller & Freund’s Probability and Statistics for Engi-
neers, 6th ed., Prentice-Hall, Upper Saddle River, NJ, 2000.
[6] ISO 10524-1, 2006, “Pressure Regulators for Use With Medical Gases—
Part 1: Pressure Regulators and Pressure Regulators With Flow-metering
Devices,” International Organization for Standardization, Geneva,
Switzerland.
[7] ISO 10524-3, 2005, “Pressure Regulators for Use With Medical Gases—
Part 3: Pressure Regulators Integrated With Cylinder Valves,” Interna-
tional Organization for Standardization, Geneva, Switzerland.
[8] ASTM G72-09, 2009, “Standard Test Method for Autogenous Ignition
Temperature of Liquids and Solids in a High-Pressure Oxygen-Enriched
Environment,” Annual Book of ASTM Standards, Vol. 14.04, ASTM
[9] ASTM G74-08, 2008, “Standard Test Method for Ignition Sensitivity of
Materials to Gaseous Fluid Impact,” Annual Book of ASTM Standards,
Vol. 14.04, ASTM International, West Conshohocken, PA.
[10] Glenn Research Center, 2010, “What is CEA?” http://www.grc.nasa.gov/
WWW/CEAWeb/ceaWhat.htm (Last accessed February 9, 2012).
STP 1561, 2012
DOI:10.1520/STP20120018
Elliot T. Forsyth,1 Nicholas Linley,2 Gwenael J. A. Chiffoleau,3

Jared Hooser,4 and Barry E. Newton5
Development of Burn Curves to Assist With

Metals Selection in Oxygen
REFERENCE: Forsyth, Elliot T., Linley, Nicholas, Chiffoleau, Gwenael J. A.,

Hooser, Jared, and Newton, Barry E., “Development of Burn Curves to Assist
With Metals Selection in Oxygen,” Flammability and Sensitivity of Materials
274–305, doi:10.1520/STP20120018, ASTM International, West Consho-
hocken, PA 2012.
ABSTRACT: Promoted ignition data from ASTM G124 have been of great
value to the industry, as they provide a fundamental basis for comparing the
relative flammability of metal alloys in oxygen. An industry-sponsored metals
test and analysis program was performed with the goal of applying statistical
analysis methods to ASTM G124 data to create burn curves that plot the esti-
mated average burn length versus pressure for a given alloy. These burn
curves provide a more thorough characterization of the relative flammability of
metal alloys in oxygen and, as such, allow greater understanding of the risk of
self-sustained combustion (i.e., burning) if an alloy is ignited at a given pres-
sure in oxygen. This paper presents the results of the industry-sponsored test
and analysis program, including the burn curves and corresponding statistical
confidence curves that were developed for five alloys: 316 stainless steel, 304
stainless steel, Inconel 625, Inconel 600, and Hastelloy C-276.
KEYWORDS: Metals flammability, oxygen, oxygen-enriched, ASTM G124,
metals selection, burn curve, statistical analysis, average burn length, oxy-
gen compatibility, minimum burn pressure, exemption pressure
Manuscript received February 21, 2012; accepted for publication May 21, 2012; published
1
Technical Consultant, Oxygen Safety Consultants, Inc., Wendell Hull & Associates, Inc., Tulsa,
OK 74104.
2
Electrical Engineer, Wendell Hull & Associates, Inc., 5605 Dona Ana Rd., Las Cruces, NM 88007.
3
Senior Scientist and Test Facility Manager, Wendell Hull & Associates, Inc., 5605 Dona Ana
Rd., Las Cruces, NM 88007.
4
Mechanical Engineer, Wendell Hull & Associates, Inc., 5605 Dona Ana Rd., Las Cruces, NM 88007.
5
VP R&D, Wendell Hull and Associates, Inc., 5605 Dona Ana Rd., Las Cruces, NM 88007.
Copyright V
19428-2959.
274
Introduction
For years ASTM G124 [1] has been used as a foundational test for evaluating
the flammability of metals in oxygen-enriched atmospheres. This test method
determines the lowest pressure that demonstrates self-sustained burning for a
3.2 mm (0.125 in.) diameter rod and generally requires that at least 30 mm
(1.18 in.) of the sample burn in order for it to meet the criteria for self-
sustained burning. In general, metal alloys are considered burn resistant in a
given oxygen system if the application pressure is less than or equal to the low-
est burn pressure and if the metal’s cross-sectional thickness is greater than or
equal to the test sample diameter.
Because many factors can affect the actual flammability of an alloy in a
given oxygen environment and application (for example, cross-sectional thick-
ness, use temperature, flow rate, ignition energy, and others), it is important to
note that ASTM G124 is a configurationally dependent flammability test for
metal alloys and, as such, does not characterize the absolute flammability of a
given alloy in oxygen. It does, however, provide insight into the relative flam-
mability behavior of a given alloy so that comparisons between alloys can be
made and applied, along with other factors, in the selection of metals for oxy-
gen service. For this reason, several leading industry standards that provide
guidance for the selection of metals in oxygen service advocate using ASTM
G124 data to evaluate an alloy’s flammability [2–4]. Further, methods for per-
forming oxygen fire risk analysis and oxygen compatibility analysis also rely
on ASTM G124 data for evaluating metals’ flammability [5,6].
The ASTM G124 test method procedure requires that a minimum of five
tests be performed at a given pressure, unless a single test burns more than 30
mm (1.18 in.). In the event of a “burn,” testing continues at the next lower pres-
sure until a minimum of five tests demonstrate no burn lengths greater than
30 mm (1.18 in.). The data that are typically reported are the lowest burn pres-
sure, formerly termed the threshold pressure, and the highest no-burn pressure.
Although these results give some insight into the relative flammability of the
alloy in the standard test configuration, they do not fully characterize the flam-
mability of the alloy over a pressure range and also are typically not subjected
to statistical analysis. This can often lead an end user to misinterpret these data
as actual flammability limits for the alloy, or at least to assume that these data
are representative of all test data at that pressure.
In order to most effectively understand ASTM G124 flammability data and
use them to select metals for a given oxygen service application, the end user
would benefit from access to the entire population of burn length data over a
pressure range, and further benefit would be obtained if a “burn curve” could
be developed and applied with statistical confidence to the data. A burn curve
could potentially even allow the end user to quantify the degree of burn resist-
ance at a given pressure by considering several factors.
Background
The concept of developing burn curves to show flammability trends for metal
alloys was previously considered by Zawierucha et al. [7]. Zawierucha pro-
posed that metals follow a flammability transition behavior with pressure that
closely resembles a metal’s well-known ductile-to-brittle transition curve, with
three distinct regions: upper shelf, transition regions, and lower shelf. Zawieru-
cha referred to this burn curve as a promoted ignition-combustion transition
curve (see Fig. 1).
Zawierucha suggested that materials on the upper shelf are generally con-
sidered as burn resistant, meaning that even when exposed to a strong ignition
source at the given pressure in oxygen, they would be expected to resist igni-
tion and, further, would be expected to self-extinguish if ignition energy were
sufficiently high to ignite the metal under these conditions. In contrast, materi-
als on the lower shelf are generally considered easier to ignite (though metal
ignition still requires a high-energy source that elevates the alloy temperature
to near its melting temperature) and, if ignited, would be expected to self-
sustain combustion, or continue to burn, in the given conditions. Materials in
the transition zone typically exhibit considerable scatter in burn length meas-
urements, and the burn behavior will appear erratic. Zawierucha suggested that
materials used in this region might or might not continue to burn if ignited at
the given test pressure.
FIG. 1—Zawierucha’s promoted ignition-combustion transition curve.

The Zawierucha work is quite helpful for providing a qualitative character-

ization of metals’ burn behavior in oxygen, but an end user requires a quantita-
tive approach to assist in selecting metals for a given oxygen application. Such
a quantitative approach was proposed by Suvorovs et al. [8], who presented an
analysis of promoted ignition test data based on formal statistical techniques.
Suvorovs applied a logistic regression method to model binary burn–no-burn
test data based upon standard burn criteria from ASTM G124. The logistic
regression model was used to predict the burn probability over a range of pres-
sures. Confidence intervals for the model were also determined, allowing the
uncertainty associated with the predicted burn probabilities to be quantified.
Suvorovs et al.’s work provides a quantitative burn curve using logistic
regression and associated statistical confidence, but it requires that burn criteria
be applied to the ASTM G124 data prior to analysis. A more fundamental and
broader-scoped approach for the end user might be to first provide a compila-
tion of the raw ASTM G124 data for a given alloy that shows burn length ver-
sus pressure for all data tested, and then provide a statistically meaningful
curve fit through the average burn lengths, along with associated statistical con-
fidence curves. This approach would allow an end user to apply whatever burn
criteria are desired (if, for example, the 30 mm burn length were considered
too conservative) and locate the corresponding pressure and confidence limits
associated with the average burn length of the material. An industry-sponsored
metals test and analysis program was conducted in order to develop burn
curves based upon this new approach.
Industry-sponsored Metals Test and Analysis Program

Wendell Hull & Associates, Inc. (WHA) coordinated an industry-
sponsored test program to compile and analyze the promoted ignition com-
bustion test data per ASTM G124 for five alloys: 316 stainless steel, 304
stainless steel, Inconel 625, Inconel 600, and Hastelloy C-276. Program
contributors consisted of U.S. government organizations and private corpo-
rations mostly from the industrial gas sector. Table 1 shows the alloys and
respective test sample thicknesses considered in the program. The study
TABLE 1—Alloys and test sample thicknesses compiled in program.
Alloy Thickness (Diameter)

Stainless steel 304L 3.2 mm (0.125 in.), 6.4 mm (0.25 in.)
Stainless steel 316L 3.2 mm (0.125 in.), 6.4 mm (0.25 in.)
Inconel 625 3.2 mm (0.125 in.)
Inconel 600 3.2 mm (0.125 in.)
Hastelloy C-276 3.2 mm (0.125 in.)
had two main objectives: first, to provide a thorough compilation of all

historical ASTM G124 test data for the selected alloy, and second, to
apply a statistical analysis model to the data compilations in order to de-
velop a burn curve using burn length data versus pressure including statis-
tical confidence intervals.
The study was divided into two phases. Phase 1 consisted of data research
and compilation from both published and unpublished sources. WHA received
cooperation from several test agencies as part of this extensive effort and was
able to compile, review, and sort historical ASTM G124 data according to key
parameters such as sample geometry (rod versus other geometries), pre- and
post-sample length, overall burn length, and test pressure. Though the data were
generated by different test agencies and different test systems, the program
assumed that the test systems all met the standard test requirements and thus pro-
duced comparable data within the discerning capability of the test method. For
most alloys and thicknesses, the historical data compilations still left large pres-
sure gaps where no testing had been performed or where only a few data points
existed. Thus, additional testing was performed in these pressure regions in phase
II of the program to generate a statistically meaningful curve fit. A plot of the
raw burn length data was published as part of the results for each alloy.
Phase II of the program also included the development of burn curves
using statistical analysis methods with confidence intervals. Several different
curve-fitting methods were attempted and refined based upon the results. A
sigma-type curve was identified as a relatively good fit to some historical
data. The approach used analysis techniques appropriate for the manner in
which historic data were collected. The results of both the burn length data
compilations and their developed burn curves are provided further on.
Raw Burn Length Data Compilations

It is noteworthy that the starting test sample lengths varied among certain test
agencies and test configurations. The reason for this is that larger promoted
combustion test chambers allowed for longer test samples. To accommodate
this factor into the analysis, in most cases the test data were truncated to the
shortest test sample used, usually 100 mm.
In general, the data compilations for each of the alloys and their given
thicknesses showed good agreement among test agencies.
Burn Curve Method Development
Approach
This paper describes a procedure for approximating a non-linear regression to
burn length versus pressure data produced by ASTM G124. It is important to
point out that the analysis is retrospective of already collected data. A different
approach might be more appropriate when planning future data collecting
endeavors.
The approximation model used a two-parameter sigma curve fit function
similar to the binomial logistic regression presented in Ref 8. The model fol-
lows a sigma-like curve that has been identified as fitting the data “well” (by
visual observation and by the goodness of fit statistic [9]). Some physical
explanations are presented as a defense for this model, but a more rigorous
application of the science should be undertaken in order to better assess the
usefulness of the model.
The model coefficients are determined using the Levenberg–Marquardt
(LM) algorithm, which provides a solution to the nonlinear least squares mini-
mization through an iterative process [10]. The average burn lengths and the
population mean confidence interval derived from a student’s t-distribution
method are calculated for each pressure interval [11]. These values are then
processed using the LM algorithm to produce the “best fit” coefficients for use
in the selected model. The data processing was achieved using a MathCAD 15
worksheet produced by WHA. This paper highlights some of the MathCAD
code used and provides a general procedure that the MathCAD code follows.
Regression Curve Fit Model

A sigmoidal curve fit was first considered the preferred method for producing
burn curves for these alloys. Though it was observed to fit some data sets very
well, the area of most interest expressed by the users of the data was the no-
burn to transition area, and the method of determining the “best fit” coefficients
to the sigmoidal model resulted in compromises in the no-burn and transition
area as a result of the full burn area of the curve. Thus, an exponential regres-
sion curve fit model was used because it was found to have the best functional-
ity in the no-burn to transition area of the data.
Equation 1 shows the form that the exponential regression model takes.
f ðxÞ ¼ Aeb0 (1)
In order to compile the arrays used to calculate the curve-fitting coefficients

and complete the model in Eq 1, the following terms must be calculated: sam-
ple mean burn length, standard deviation of burn length, standard error of the
sample mean, and population mean 95 % confidence intervals.
Sample Means and Standard Deviations

The sample data were assumed to follow a normal distribution, and therefore
statistical tools that were developed for normal (or Gaussian) curves were
applicable. This assumption is based on the distribution of the data around the
burn length sample means; however, the pressure intervals used were not
selected at random and were determined and gathered with the guidance of
ASTM G124. “ASTM G124 provides guidelines to determine the minimum
test gas pressure (threshold pressure) that will support self-sustained burning of
a 3.2-mm (0.125-in) cylindrical rod” [1]. The sample means with standard
deviation spreads are shown in the results, along with the distinct pressure
intervals that were selected for testing.
The sample mean of each pressure was derived using Eq 2.
P
N
Li
i¼1
xpressure ¼ (2)
N
where:
xpressure ¼ sample mean of the burn lengths at pressure,
L ¼ array of burn length data, and
N ¼ number of tests done at pressure.
The standard deviation of the sample means was computed by assuming a
normal distribution of the data and applying Eq 3 to the data. One standard
deviation contains approximately 68 % of the sample distribution.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u 1 X N
s¼t ðLi xpressure Þ2 (3)
N 1 i¼1
where:
s ¼ sample standard deviation,
L ¼ array of burn length data,
xpressure ¼ sample mean of the burn lengths at pressure, and
N ¼ number of tests done at pressure.
Student’s t-distribution Confidence Interval of the Mean

The standard error and the student’s t-distribution were used to calculate the
95 % confidence interval of the population mean. This is the probable error of
a sample mean; that is to say that if all samples of the material were accounted
for and tested, the 95 % confidence interval derived around the sample means
would most likely contain the mean of the entire population. The application of
this type of confidence interval was used because of the relatively small sample
sizes and the application of the Central Limit Theorem [11]. The Central Limit
Theorem states that in any sampling of a random variable, as the number of
samples increases, the distribution becomes increasingly more normal or Gaus-

sian in shape. That is to say, the frequency of the results tends to be higher
around the middle of the bell curve, and that is where the mean exists.
The standard error is defined as a method of measurement or estimation of
the standard deviation of the sampling distribution associated with the estima-
tion method [11]. The standard error is defined in Eq 4.
s
xpressure ¼ pffiffiffiffi
SE (4)
N
where:
s ¼ sample standard deviation, and
N ¼ number of tests performed at pressure.
The student’s t-distribution is used to derive confidence intervals around the
mean in conjunction with the standard error [11]. The t-distribution has “heavier
tails” (where more probability lies in the tails of the distribution) when the
degrees of freedom (df) is small because of the limited number of samples. It
becomes equivalent to a normal distribution as the df approaches infinity. The t-
distribution is commonly used to compute confidence intervals for populations
that have an unknown population mean with unknown variance. The values of
the student’s t-distribution are generally derived from lookup tables. The user is
typically interested in the t-value, which is used to give the 95 % confidence
interval when multiplied by the standard error. This is generally specified as the
t-critical or t025 value. Excel has a function called TINV() that can be used to
derive the t-critical value. The arguments for the function are in Eq 5.
TINVð0:05; kÞ (5)
where:
0.05 ¼ desired t-value for the two-sided 95 % confidence interval, and
k ¼ df.
The df is specified as the number of samples (N) minus the number of esti-
mated parameters. In this case, the number of estimated parameters is 1 (the
mean), and therefore the df is typically N 1.
MathCAD has a function named qt(p, d) that returns the inverse
cumulative probability distribution of the probability p. This function can also
be used to return the t-critical value for a desired confidence interval as shown
in Eq 6.

1 CI
TCriticalðCI; NÞ:¼ qt CIþ ;N 1 (6)
2
The 95 % confidence intervals of the population mean were calculated and pre-
sented as error bars for calculated sample means in some data compilation plots.
General Procedure
The general procedure for performing the curve fits is as follows:
1. Combine all the burn length data for a material into an array.
2. Truncate the results to account for smaller sample sizes that were used
by some test facilities (i.e., if a sample from one test facility burned
120 mm and samples from another facility were limited to 100 mm,
then the 120 mm burn result is truncated to 100 mm).
3. Calculate the sample means and the 95 % confidence intervals for the
population means at each pressure interval.
4. Compile these calculations into four arrays with corresponding elements:
a. test pressures,
b. sample means,
c. upper bounds of 95 % population mean confidence intervals,
and
d. lower bounds of 95 % population mean confidence intervals.
5. Use an arbitrary 50 mm burn length criteria to determine a subarray
that contains only the necessary data in the no-burn and transition
regions with which to perform the exponential curve fits.
6. Use these four arrays to generate three sets of coefficients to use in
Eq 1 using the LM algorithm.
The model coefficients are determined using the LM algorithm, which pro-
vides a solution to the nonlinear least squares minimization through an iterative
process [10]. Other coefficient solving methods based on maximum likelihood
methods or least squares would also be sufficient for finding the coefficients.
Results
A review of the data revealed strong correlation in burn length variance as the
pressure increased into the transition area, and therefore an approach involving
fitting a curve to the confidence intervals of the individual population means
was used. There were other techniques considered for computing a curve that
represents the confidence intervals; however, these techniques relied on the
assumption of uncorrelated variance with pressure.
The results of the raw burn length data compilations and their associated
burn curves are shown in this section (Figs. 2–22). This paper is intended to
provide the results of this industry-sponsored test and analysis program without
direct application of the results beyond what is discussed above. Through a
simple presentation of the graphical results of the quantified data compilations
for common metal alloys and their corresponding curve fits with confidence
intervals, the data are expected to greatly benefit the current methodologies for
selecting metals in oxygen-enriched atmospheres.
Each alloy’s data compilation shows the number of tests performed at each
pressure interval as a separate graph on top of the burn length plot.
FORSYTH ET AL., doi:10.1520/STP20120018
283
FIG. 2—3.2 mm (0.125 in.) 316 stainless steel, full pressure range.
FIG. 3—3.2 mm (0.125 in.) 316 stainless steel, lower pressure range.
FIG. 4—3.2 mm (0.125 in.) 316 stainless steel, exponential curve fit to lower pressure range.
285
287
289
291
293
295
FIG. 14—3.2 mm (0.125 in.) Inconel 625, full pressure range.

FIG. 15—3.2 mm (0.125 in.) Inconel 625, lower pressure range.

297
FIG. 16—3.2 mm (0.125 in.) Inconel 625, exponential curve fit to lower pressure range.
FIG. 17—3.2 mm (0.125 in.) Inconel 600, full pressure range.

299
FIG. 18—3.2 mm (0.125 in.) Inconel 600, lower pressure range.

FIG. 19—3.2 mm (0.125 in.) Inconel 600, exponential curve fit to lower pressure range.
301
FIG. 20—3.2 mm (0.125 in.) Hastelloy C-276, full pressure range.

FIG. 21—3.2 mm (0.125 in.) Hastelloy C-276, lower pressure range.

303
FIG. 22—3.2 mm (0.125 in.) Hastelloy C-276, exponential curve fit to lower pressure range.
Conclusions
An industry-sponsored metals test and analysis program was performed with
the goal of compiling historical test data from ASTM G124, conducting new
testing where needed, and applying statistical analysis methods to ASTM G124
data in order to create burn curves that plot the average burn length versus pres-
sure for five alloys: 316 stainless steel, 304 stainless steel, Inconel 625, Inconel
600, and Hastelloy C-276. These burn curves provide a more thorough charac-
terization of the relative flammability of metal alloys in oxygen and, as such,
allow greater understanding of the risk of self-sustained combustion (i.e., burn-
ing) if an alloy is ignited at a given pressure in oxygen.
References
[1] ASTM G124-10, 2010, “Standard Test Method for Determining the Com-
bustion Behavior of Metallic Materials in Oxygen-Enriched Atmos-
pheres,” Annual Book of ASTM Standards, Vol. 14.04, ASTM
[2] ASTM G94-06, 2006, “Standard Guide for Evaluating Metals for Oxygen
Service,” Annual Book of ASTM Standards, Vol. 14.04, ASTM Interna-
[3] Compressed Gas Association (CGA), Oxygen Pipeline Systems, CGA G-
4.4 (2003)/EIGA IGC 13-02, CGA, Chantilly, VA/European Industrial
Gas Association, Brussels, Belgium, 2003.
[4] NFPA 53, 2011, “Recommended Practice on Materials, Equipment, and
Systems Used in Oxygen-Enriched Atmospheres,” National Fire Protec-
tion Agency, Quincy, MA.
[5] Forsyth, E. T., Newton, B. E., Rantala, J., and Hirschfeld, T., “Using
ASTM Standard Guide G 88 to Identify and Rank System-Level Hazards
in Large-Scale Oxygen Systems,” Flammability and Sensitivity of Materi-
als in Oxygen-Enriched Atmospheres, ASTM STP 1454, Vol. 10, T. A.
Steinberg, H. D. Beeson, and B. E. Newton, Eds., ASTM International,
West Conshohocken, PA, 2003.
[6] Shoffstall, M. S. and Stoltzfus, J. M., “Oxygen Hazards Analysis of
Space Shuttle External Tank Gaseous Oxygen Pressurization System,”
Atmospheres, ASTM STP 1454, Vol. 10, T. A. Steinberg, H. D. Beeson,
and B. E. Newton, Eds., ASTM International, West Conshohocken, PA,
2003.
[7] Zawierucha, R., Mcllroy, K., and Mazzarella, R. B., “Promoted Ignition-
Combustion Behavior of Selected Hastelloys In Oxygen Gas Mixtures,”
Atmospheres, ASTM STP 1111, Vol. 5, J. M. Stoltzfus and K. Mcllroy,

[8] Suvorovs, T., Ward, N., Steinberg, T., and Wilson, R., “Statistical Evalua-
tion of Promoted Ignition Test Data,” J. ASTM Int., Vol. 4, No. 7, 2007,
101068.
[9] Windmeijer, F. A. G., 1995, “An R-squared Measure of Goodness of Fit
for Some Common Nonlinear Regression Models,” UC Davis Department
of Economics, March 31, http://cameron.econ.ucdavis.edu/research/
je97preprint.pdf (Last accessed February 25, 2011).
[10] “The Levenberg-Marquardt Algorithm,” CiteSeerx Scientific Literature
Digital Library and Search Engine, June 8, 2004, http://citeseerx.ist.psu.
edu/viewdoc/download?doi=10.1.1.135.865&rep=rep1&type=pdf (Last
accessed 25 Feb 2011).
[11] Montgomery, D. C. and Runger, G. C., Applied Statistics and Probability
for Engineers, John Wiley & Sons, New York, 2003.
307
Author Index
B K
Binder, C., 105-116, 162-171, Kleinberg, W., 117-146

172-181
Brock, T., 105-116 L
C Lehné, S., 162-171, 172-181

Leschevich, V. V., 147-161
Castillo, M., 233-244 Lineberry, D. M., 28-46
Chiffoleau, G. J., 245-305 Linley, N., 274-305
Ciscato, W., 162-171, 172-181 Longuet, O., 198-224
Colson, A., 198-224 Lynn, D., 233-244
Coste, F., 117-146
Crayssac, F., 117-146 M
D Meilinger, M., 81-104

Mulkey, H. W., 28-46
Davis, S. E., 1-27
Dielforder, P., 105-116 N
Dobinson, C., 225-232
Nelson, P., 225-232
F Newton, B. E., 245-261, 262-273,
274-305
Forsyth, B. S., 245-261,
262-273 O
Forsyth, E. T., 262-273, 274-305
Odom, G. A., 245-261
G
P
Gallus, T. D., 62-80, 182-197
Paulos, T., 233-244
H Penyazkov, O. G., 147-161
Piña Arpin, C. Y., 47-61
Hesse, O., 162-171, 172-181 Plagens, O., 233-244
Hooser, J., 274-305
Houghton, P., 117-146 R
J Ridlova, M., 198-224

Rostaing, J. C., 117-146, 147-161,
Jeffers, N., 182-197 198-224
308
S T
Sparks, K., 62-80 Tillack, T., 105-116, 162-171, 172-181

Steinberg, T., 225-232, Treisch, N., 105-116
233-244
Stoltzfus, J., 47-80, W
182-197
Sun, L., 117-146 Werlen, E., 117-146
309
Subject Index
A fire risk, 105-116
flammability, 117-146, 182-197
accident research, 105-116 flow friction, 62-80
aging of diving valves, 162-171,
172-181 H
air separation unit, 117-146
aluminum, 81-104, 105-116, 117-146, halogenated non-metals, 245-261
233-244 hardware cleaning procedures, 47-61
aluminum oxide, 233-244 heat affected, 198-224
analysis, 262-273 high pressure, 81-104
ASTM G124, 274-305 high-pressure oxygen, 105-116
auto-ignition, 147-161
average burn length, 274-305
I
B ignition, 81-104, 117-146, 262-273

ignition mechanism, 62-80
bulk aluminum, 233-244 ignition source, 105-116
burn curve, 274-305 iron micro-particle powders, 147-161
burn-out safety, 162-171, 172-181
bursting diaphragm, 105-116
L
C Level 300, 47-61

liquid oxygen, 28-46
carbon steel, 198-224 liquid rocket engine testing, 28-46
combustion products, 233-244
compatibility, 262-273
M
consequence, 262-273
contamination, 81-104 metallic material, 105-116
Metals flammability, 274-305
E metals selection, 274-305
minimum burn pressure, 274-305
elevated temperature, 81-104 modified air valves, 162-171, 172-181
exemption pressure, 274-305 molecular sieve, 81-104
morphology, 233-244
F
N
field shop environment, 47-61
fire, 262-273 NITROX, 162-171, 172-181
fire propagation, 198-224 non-cleanroom, 47-61
310
non-halogenated non-metals, 245-261 Q

nonmetals, 262-273
quartz sand, 81-104
O
R
oil, 81-104
oxygen, 81-104, 117-146, 147-161, rapid compression machine, 147-161
162-171, 172-181, 182-197, 262-273, reaction effect, 245-261
274-305 risk, 262-273
oxygen compatibility, 162-171, RRMI, 182-197
172-181, 198-224, 274-305 rudimentary cleaning, 47-61
oxygen fire, 245-261 rust, 81-104
oxygen hazards, 28-46
oxygen systems, 47-61 S
oxygen-enriched, 274-305
safety engineering, 105-116
SCUBA diving, 162-171, 172-181
P
spontaneous combustion, 62-80
stainless steel, 198-224
particle impact, 81-104
statistical analysis, 274-305
particle size, 233-244
swarf, 81-104
particles, 81-104
perlite, 81-104
T
pipe fracture, 81-104
pipe wall burn-through, 198-224
titanium, 182-197
pneumatic impact, 162-171,
toxicity, 262-273
172-181
polymers, 245-261, 262-273 U
precision cleaning, 47-61
pressurized oxygen-enriched systems, university of alabama huntsville pro-
62-80 pulsion research center, 28-46
probability, 262-273
process, 47-61 V
promoted ignition, 198-224
propulsion testing facility, 28-46 valve seat, 172-181

STP 1561-2012

Uploaded by

Copyright:

Available Formats

You might also like

STP 1561-2012

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

STP 1561-2012

Uploaded by

Copyright:

Available Formats

Selected Technical Papers STP1561

Flammability and Sensitivity

Printed in the U.S.A.

ASTM Stock #: STP1561

Printed in Bay Shore, NY

An Elementary Overview of the Selection of Materials for Service in Oxygen-

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307

Samuel Edgar Davis1

An Elementary Overview of the Selection of

REFERENCE: Davis, Samuel Edgar, “An Elementary Overview of the Selec-

Hazards Inherent in Oxygen-enriched Environments

Oxygen and Materials

FIG. 1—The fire triangle.

Factors Significantly Influencing Combustion Probability and Severity of

Oxygen Concentration Effects—Combustible materials, with very few

TABLE 1—Nonmetal flammability versus oxygen.

Oxygen Burn Length, Burn Rate,

Pressure Effects—Combustible materials, with few exceptions, burn far

Temperature Effects—Temperature effects on combustible materials, with

TABLE 2—Metal flammability versus pressure.

Pressure, Burn Length, Burn Rate,

Generalizations on Effects—The ignition resistance and burn inhibition prop-

TABLE 3—Metal flammability versus temperature.

Pressure, Temperature, Burn Length, Burn Rate,

Conserve Oxygen and Materials—The first step in designing safer oxygen

Maximize Good Materials—Poor material choices can greatly increase the

FIG. 2—Ambient mechanical impact test fixture (from NASA-STD-6001 [1]).

FIG. 3—Ambient mechanical impact tester in operation.

FIG. 4—Ambient mechanical impact tester showing a severe reaction.

FIG. 6—Promoted ignition test fixture (from NASA-STD-6001 [1]).

Tables 4 and 5 provide summaries of data generated at NASA’s MSFC for

Selection of Metals versus Nonmetals

FIG. 7—Promoted ignition tester with sample rod.

TABLE 4—Minimum pressures required in order to ignite or burn metals.

Metal Oxygen Condition Promoted Ignition Test Mechanical Impact Test

TABLE 5—Minimum pressures required in order to ignite nonmetals.

Nonmetal Oxygen Condition Minimum Ignition Pressure

materials are typically incompatible with oxygen. However, laboratory testing

Minimize Ignition Sources and Mechanisms

Contamination—A contaminant, commonly called foreign object debris

Particle Impact—The particle impact ignition mechanism involves heat

Rapid Pressurization—The rapid pressurization ignition mechanism, also

in which c represents the adiabatic index of the gas for O2 ¼ 1.40.

FIG. 9—Rapid pressurization ignition results from the rapid compression of

(2) Rapid pressurization occurs within a fraction of a second.

Mechanical Impact—The mechanical impact ignition mechanism involves

FIG. 10—Mechanical impact ignition results from a wrench falling onto an

The mechanical impact ignition mechanism is most commonly witnessed

Friction—The friction ignition mechanism involves the heat energy gener-

Some components, such as check valves, regulators, and relief valves,

Less Common Ignition Sources and Mechanisms

Ignition Happens but Nobody Knows Why; Formerly Known as “Flow

Further research will be necessary in order to fully elucidate the actual

Importance of Appropriate Cleaning

Trivial Overview of an ASTM Oxygen Compatibility Assessment

Many organizations, including ASTM, have over the years generated a

[1] NASA-STD-6001, “Flammability, Offgassing, and Compatibility

H. W. Mulkey1 and D. M. Lineberry2

Development of a Liquid Oxygen Facility