Refinery & Petrochemical Plant

Maintenance Conference and Exhibition

May 20-23, 2003
Salt Palace Convention Center
Salt Lake City, Utah

MC-03-92 In-Service Robotic AST Cleaning and

Inspection: History and Operational
Experience

Presented By:

Gene Silverman
Chief Technical Officer
InTank, Inc.
Laurel, MD

Frank Furillo
ExxonMobil Research
and Engineering
ExxonMobil Corporation
Fairfax, VA

National Petrochemical & Refiners Association 1899 L Street, NW 202.457.0480 voice

Suite 1000 202.429.7726 fax
Washington, DC www.npra.org
20036.3896
This paper has been reproduced for the author or authors as a courtesy by the National
Petrochemical & Refiners Association. Publication of this paper does not signify that the
contents necessarily reflect the opinions of the NPRA, its officers, directors, members, or staff.
Requests for authorization to quote or use the contents should be addressed directly to the
author(s)
 In-Service Robotic AST Cleaning and Inspection:
History and Operational Experience

Gene Silverman
InTANK Services, Inc.

Frank Furillo
ExxonMobil Oil Corporation

Abstract

Since its first introduction to the petroleum industry in 1996, the use of in-service robots for
determining the remaining life of AST floors through the use of high density ultrasound scanning
has become an important option for a plant’s tank management program. This paper presents the
history of the introduction of this technique into the petroleum industry and summarizes current
applications and case histories. Special emphasis is placed on the use of in-service robots to
address API 653 guidelines and how the results of direct floor thickness measurements can be
integrated into a plant’s risk-based inspection program. A discussion is also provided that
introduces the reader to the use of extreme value statistics for the analysis of corrosion data from
the floor of storage tanks. The economics associated with the use of in-service robotics are
summarized within each case history.
Brief History of Robotics for AST Cleaning and Inspection

The integrity of bulk liquid storage tanks continues to be a major area of concern to the
petrochemical industry. Many tank owners and operators have developed inspection schedules
that respond to their in-house process safety management policies and the inspection
requirements and guidelines established by the Environmental Protection Agency (EPA) and
State governments. Most recently, the EPA published the revised SPCC regulations that require
a formal AST inspection program that must follow industry guidelines (such as API 653).

The schedule of AST inspections depends on a number of factors including the age of the tank,
its proximity to groundwater, the leak record of the tank, the date of the tank’s last integrity test,
the construction material of the tank, the product stored, soil conditions, previous corrosion rate
calculations, etc. One of the critical areas of integrity testing is that of the tank bottom.
Traditional methods of bottom inspection have required the emptying of the tank’s product and a
thorough cleaning and degassing prior to the allowance of personnel entry for a floor inspection.
This method results in significant releases of volatile organic compounds into the atmosphere
and creates a host of other cost and operational problems. There are significant advantages to the
tank owner if proper bottom inspection can be completed while the tank is in-service. One such
in-service method is the use of remotely controlled or robotic devices for scanning the floor of
AST’s while they are full of product and in-service.

MC-03-92
Page 1
The first commercial testing of in-service robotic techniques for determining the integrity of
AST floor bottoms began in 1994. These tests were conducted by three major oil companies.
The results of these tests were used by (1) equipment vendors to improve the performance
capabilities of their equipment, and (2) used by the oil companies to help integrate these new
inspection capabilities into their tank inspection and management programs. A description of the
characteristics of a typical in-service robotic system is provided in Appendix A.

Since the initial tests of in-service tank inspection technology in the early 1990’s, operational
and performance improvements have been made. Today, the technology has emerged as a viable
technique for not only providing an in-service cleaning option to tank owners, but for
determining the integrity and corrosion rate of tank floors without taking a tank out of service.
As long as certain conditions are met, the use of robotic inspection equipment can be used to
satisfy industry guidelines. Furthermore, the use of in-service robotics can be combined with
Risk Based Inspection (RBI) methods for improving the confidence and reliability of RBI
analyses (See discussion below).

Robotics and API 653

Although API 653(1) remains the industry standard relative to tank inspection and maintenance,
the frequency of testing and inspection can also be affected by various state and local
regulations. Robotic technology is now available as a method for acquiring quantitative
information for determining the integrity of aboveground storage tank floors while the tank
remains on-line and in-service. API 653 allows the use of robotics as an alternative method for
assessing the condition of a tank floor as long as certain conditions are met.

API 653 (Third Edition), paragraph 6.4.1.2 states that: “If the internal inspection is required
solely for the purpose of determining the condition and integrity of the tank bottom, the internal
inspection may be accomplished with the tank in-service utilizing various ultrasonic robotic
thickness measurement and other on-stream inspection methods capable of assessing the
thickness of the tank bottom, in combination with methods capable of assessing tank bottom
integrity as described in 4.4.1. Electromagnetic methods may be used to supplement the on-
stream ultrasonic inspection. If an in-service inspection is selected, the data and information
collected shall be sufficient to evaluate the thickness, corrosion rate, and integrity of the tank
bottom and establish the internal inspection interval, based on tank bottom thickness, corrosion
rate, and integrity, utilizing the methods included in this standard. An individual, knowledgeable
and experienced in relevant inspection methodologies, and the authorized inspector who is
responsible for evaluation of a tank, must assure the quality and completeness of the in-service
NDE results.”

MC-03-92
Page 2
In-Service Cleaning and Inspection of Aboveground Storage Tanks – Case Studies

This section provides a summary of the application of robotics to the inspection of five candidate
ASTs, all in different types of service. The inspection objectives are summarized and the results
are presented relative to the technical merits and economic advantages achieved by each
inspection.

Internal Floating Roof – Gasoline Tank Inspection

The primary purpose of this inspection was to conduct an internal floor inspection on a 93-foot
diameter gasoline AST without removing the tank from service. The inspection included the
following: Floor ultrasonic thickness measurements (topside and bottomside), tank floor
elevation measurements and visual inspection of the underside of the fixed roof and top side of
the internal floating roof.
The tank was full of product during the inspection. There was about two to three inches of
sludge and sediment in the tank bottom. Over 362 thousand ultrasonic thickness (UT)
measurements were systematically taken of the entire tank floor. The average measured floor
thickness was .300 inches. The highest measured floor thickness was .322 inches and the
minimum measured floor thickness was .175 inches.
When using the minimum floor thickness calculations, MRT1 from API 653 section 4.4.7, the
life expectancy was determined to be 12.7 years without any repairs. A subsequent analysis of
the data distribution throughout the tank revealed the following:
1. There was a heavier pattern of general corrosion towards the peripheral part of the tank
(closer to the critical zone) than in the tank’s center.
2. The majority of bottomside pitting was near the periphery of the tank.
3. The pattern of general corrosion was shifting from 1 to 2 mils per year based on normal
manufactured plate thickness variations.
4. Some coating failure was found.
Tank floor elevation results demonstrated a maximum measured depth change of 4 inches. This
depth change measurement was found between the center of the tank and the tank shell. A
similar degree of settlement was confirmed by an external shell planar tilt survey. The elevation
readings also revealed a bulge on the floor just north of center.
A component of this inspection included a visual inspection of the underside of the fixed roof
and the floating roof seals. A special-purpose high resolution remotely controlled camera was
used for this inspection. An inspection of the floating roof found that the secondary seal was in
good condition with no noticeable tears or gaps. However, seal penetration failures were found
around a fixed roof center column and an inspection port. These penetration failures were also
confirmed by the robot’s “look-up” camera which recorded “light leaks” from underneath the
product and through the floating roof.

MC-03-92
Page 3
 Following the in-service inspection an out-of-service inspection was conducted in order to
validate the in-service results. The costs associated with a typical out-of-service inspection are
summarized in the chart presented in Appendix A. Using the cost categories as a guideline and
only considering costs that were known, the overall budgetary cost of the out-of-service
inspection was $86,000. The facility cost reduction using the in-service option was more than
50%.

Jet Fuel Tank Inspection

The primary purpose of the inspection of this diesel tank was to conduct an internal floor
inspection to determine corrosion rates without removing the tank from service. The inspection
included floor ultrasonic thickness measurements, tank floor elevation measurements and a
visual inspection of the floor. The tank was on-stream during the inspection. The tank was 120
feet in diameter with approximately 15 feet of Jet Fuel. There was about 8 - 12 inches of sludge
and sediment in the tank bottom.
Over 500 thousand ultrasonic thickness (UT) measurements were systematically taken of the
tank floor. The average nominal floor thickness was .247 inches. The results of this in-service
inspection can be summarized as follows:
1. The minimum floor thickness was determined to be 50% of nominal. Given the
aggressive rate of corrosion the operational life of the tank was determined to be two (2)
years.
2. Tank floor elevation variations were also measured by the robot and the settlement was
determined to be within acceptable limits according to API 653 Appendix B (Evaluation
of tank bottom).
3. Initial stages of corrosion were found to be developing in localized areas of lateral clips
of roof column bases.
4. The internal piping system as well as the lower area of internal ladder appeared to be in
good condition following the review of the video information from the robot’s camera.
5. Localized lower shell areas appeared to be in good condition.
6. The entire bottom floor plates were covered with sediment with the greatest amount
found in localized areas adjacent to column bases and shell walls. The same areas
reflected what appeared to be a lining and/or coating failure (degradation) in different
modes such as lifting, blistering, bubble form, cracking. In some places the coating was
missing.
7. The UT scanning system detected, in localized areas, minimum bottom-side remaining
thickness of between 0.247” and 0.116” considering the original Nominal Plate
Thickness of 0.250”.

Using the cost categories presented in Appendix A as a guideline and only considering costs that were
known, the overall cost of the out-of-service inspection was approximately one-half of the out-of-service
budget.

MC-03-92
Page 4
External Floating Roof – Sweet Crude

A remotely controlled cleaning and inspection robot provided an API 653 internal inspection of
two (2) external floating roof 300-foot diameter tanks containing crude oil (API 35) without
removing the tank from service and without personnel entry. The cleaning and inspection robot
was deployed through a roof manway with the tank full of oil. The inspection consisted of
acquiring over 500 thousand ultrasonic measurements per tank.
The position and orientation of the robot were determined using onboard and external sonar
navigation transducers. The robot also measured tank bottom settlement with onboard pressure
transducers. The analysis provided a thickness map of the floor with identification of both
topside and bottomside corrosion areas. The floor life expectancy was generated from the
readings using the API minimum floor thickness calculations. The inspection also assessed the
general condition of the fiberglass coating.
There was about 6 inches of sludge and sediment in the tank bottom. The minimum recorded
floor thickness was .205 inches. There were 80 separate instances where corrosion was noted
during this in-service tank inspection. When using the minimum floor thickness calculations,
MRT1 from API 653 section 4.4.7, the floor life expectancy was determined to be 9.6 years.
Tank floor elevation readings determined that the maximum measured depth change was
determined to be 10.2 inches. The elevation runs taken did not reveal excessive deviation.
Bottom settlement was within acceptable limits according to API 653 Appendix B (Evaluation of
tank bottom).
Using the cost categories presented in Appendix A as a guideline and only considering costs that
were known, the overall budgetary cost of the out-of-service inspection, not including tank
downtime, discretionary and non-discretionary repair costs, was $52,500 which was more than
50% greater than the in-service alternative.

In-Service Inspection of a Resid Oil Tank at 380°F

The primary purpose of this inspection was to conduct an in-service tank bottom ultrasonic
inspection of a 275-foot diameter X 50-foot high Residual Coker Feed Service tank. The tank
product was maintained at approximately 380°F during the inspection. A special purpose robotic
probe manually deployed through four roof manways was used for the inspection.

The probe unit consisted of a lower unit “car” assembly which carried two 5 MHz ultrasonic
transducers used in the immersion mode. The transducers were rated to operate in temperatures
up to 600°F. The second subsystem was a series of hollow circular steel alloy (4130) tubes
consisting of seven 10’ sections and one 2’ section. A fork assembly connected the car to the
probe tube. The two UT transducers were connected to 100’ high temperature cables that were
routed through the probe tubes.
The third subsystem was a gimbaled seal fitted over the roof manway which also served as a
support point. The gimbaled seal was equipped with a flexible boot and wiper seal that served as
a vapor barrier. The fourth subsystem was a flushing assembly that provided a means for
flushing the area beneath the immersion transducers and the tank floor in order to remove any
residual sludge. The flushing fluid used for this project was vacuum gas oil (VGO) provided by

MC-03-92
Page 5
the plant. The fifth subsystem was a UT pulser-receiver, data acquisition, recording and batch
processing system capable of acquiring a high volume of UT data.
The inspection program involved three phases of operation. The first phase was devoted to the
design and development of an inspection probe that was able to withstand temperatures in excess
of 380°F. The second phase involved the deployment of the equipment in the tank for the
purpose of acquiring limited tank bottom ultrasound readings in the vicinity of each of four roof
manways. The third phase of the project involved the analysis of 13,600 UT readings and the
incorporation of these readings into a Risk Based Inspection analysis program. The UT data
acquisition concentrated only in areas directly beneath each of four manways.

The UT results revealed an average floor thickness of .365 inches. The minimum thickness
recorded was .355 and the maximum thickness recorded was .394 inches. When converting the
UT floor readings into a corrosion rate based on an eleven year old tank, the remaining useful
life (RUL) calculation resulted in excess of 20 years. The RBI results revealed a criticality
rating of “4” which reflects a relatively low rating of failure probability and consequence with a
recommended in-service inspection within 20 years. Criticality ratings range from 1 (where both
the probability and consequence ratings are high) to 5 (where both probability and consequence
ratings are low).
Using the cost categories presented in Appendix A as a guideline and only considering costs that
were known, the savings reported by the plant was in excess of $200,000.

Integrating Risk Based Inspections with Robotic In-Service Inspections

The development of risk based inspection (RBI) methodologies for assessing the integrity of
aboveground storage tanks (AST) is providing tank owners with an alternative to conventional
inspection methods. The publication of API 580(2) and the availability of new, unique RBI tools
are opening up new avenues for assessing the integrity of ASTs. Regardless of the efficacy of
these RBI tools they all have one characteristic in common: their assessment of the integrity of
the tank floor is dependent on the available evidence. That is, accurate corrosion rate calculations
cannot be conducted, the location of potential areas of corrosion or serious floor pitting cannot be
predicted and remaining life calculations cannot be made using actual floor data.
The application and use of RBI analyses provide a straightforward and meaningful avenue of
establishing an initial inspection protocol based on sometimes limited knowledge of tank
condition. Greater confidence and more realistic inspection frequency emerges with the addition
of direct floor ultrasonic measurements. A recent 35-tank study has demonstrated the effects of
the use of direct floor measurements on RBIs(3). Tables 1 and 2, taken from this recent (CY
2002) 35-tank RBI study, summarize the results. Table 1 presents the distribution of RBI results
for all tanks without the use of tank floor ultrasonic data. Table 2 provides the RBI results when
incorporating direct ultrasonic measurements into the RBI.

MC-03-92
Page 6
Table 1. RBI matrix created from a population of 35 tanks where direct floor UT measurements
were not acquired.
CONSEQUENCE

Low Med High Total

High 0 27 6 33
LIKELIHOOD Med 0 2 0 2
Low 0 0 0 0
Total 0 29 6 35

Table 2. RBI matrix created from a population of 35 tanks where direct floor UT measurements
were acquired.
CONSEQUENCE

Low Med High Total

High 0 14 4 18

LIKELIHOOD Med 0 8 0 8
Low 0 7 2 9
Total 0 29 6 35

The results of this study can be summarized as follows:

1. There was a reduction in the average probability of failure.
2. There was an increased level of confidence associated with the risk assessment.
3. Higher averages of remaining useful life were generated.
The more refined distribution presented in Table 2 suggests the use of this approach to update the
inspection program planning horizon (indeed, the primary focus of its use in API 653). Figures 1
and 2 show the graphical results of these distributions.

MC-03-92
Page 7
 30 14

25 12

10
20
8
Frequency 15 Frequency
6
10
4
5 S3 S3
2
S2 S2
0 Likelihood 0 Likelihood
1 S1 S1
1
2 2
3 3
Consequence Consequence

Figure 1. RBI Without Tank Floor UT Data Figure 2. RBI With Tank Floor UT Data

The net effect of moving equipment to lower criticality positions on the matrix shown in Figure 3
is to increase inspection intervals based on criticality and confidence in knowledge of tank
integrity, while at the same time inspecting for the right damage using the right techniques.
Output from the program described in this paper can be readily integrated into existing plant
maintenance management systems.

30
Number of Tanks

25
20
15
10
5
0
1 2 3 4 5
Level of Criticality 1=High; 5=Low

RBI Without Floor UT Data RBI With Floor UT Data

Figure 3. Distribution of Criticality Ratings for Each Tank RBI - With and Without Floor UT
Data

MC-03-92
Page 8
Extreme Value Statistics and the Evaluation of the UT Thickness Distributions

Conventional techniques of analyzing floor UT data requires the inspector to evaluate the total
population of ultrasonic data, determine the values of the deepest pits (or thinnest areas also
known as Tmin), confirm the age of the tank and the condition of the floor since its last
inspection, and calculate the remaining life of the tank by using the calculated corrosion rate.
The inspector may also use the preferred method suggested by API (Third Edition), paragraph
4.4.7. These values, known as Minimum Remaining Thickness (MRT) are used when enough
data has been acquired to characterize areas of topside and bottomside general corrosion and
pitting. One shortcoming with this approach is that unless the inspector is absolutely certain that
they have scanned the entire tank floor so that any pitting has not gone undetected, the corrosion
rate results will not reflect the actual condition of the tank floor. This is especially a problem
where the use of qualitative measurements (such as magnetic flux leakage or eddy current
scanning) may not be as thorough as anticipated – an industry dilemma(4). One approach to
addressing this issue is with the use of extreme value statistics, a technique that is common to
many industries when the assessment of a greater population is necessary with the use of a
limited set of data(4).

The distribution of corrosion on the floor of an AST is in essence a statistical effect governed by
a number of variables. Variations on the surface of an AST floor such as depressions, bulges, the
chemistry of the tank’s product as well as its temperature and a wide range of other factors cause
different forms of corrosion. These different forms of corrosion also create variations in
corrosion rates. When the corrosion is widespread it is usually considered to be general
corrosion. When the corrosion is limited to small areas it usually takes the form of pitting. For
instance, the corrosion environment of the floor of a heavy oil storage tank can vary from the
center of the tank to the periphery where corrosion tends to be heavier(6).

In these peripheral areas or any area where there may be pitting, the assumption that corrosion
rate is uniform is not likely to be accurate. That is, sample thickness measurements are unlikely
to be representative of the distribution of corrosion on the tank floor.

One way to determine the distribution of corrosion throughout a tank floor is to take quantifiable
measurements (i.e. UT) over 100% of the floor’s surface. Under these circumstances, the floor
thickness data will produce a variety of thickness distribution patterns. These patterns may be
normal, exponential, Poisson, lognormal, etc. Data population sets that are 100% represented are
rare and UT scanning 100% of a tank floor is not practical (at least not with most technology
today). As a consequence, statistical techniques can be used with limited data sampling sets for
inferring the greater population of corrosion or pitting behavior of a tank floor. That is, the
statistical distribution of the small sample set may reflect the distribution of its parent population.

One example that is important in the context of inspection is the family of extreme value
distributions similar to those of Gumbel and Weibull(6). In the case of floor thickness readings,
these distributions consist of UT readings that represent the minimum thickness acquired from
the floor and are, therefore, considered to be extreme values. These thin areas or extreme values
may be taken in isolation or from groups of small areas or patches around the tank floor. When
grouped together they form their own unique extreme value distribution.

MC-03-92
Page 9
 Extreme value statistical methods can be used in conjunction with degrees of reliability (and,
therefore can be used to determine risk) since any extreme event (i.e. a measured or predicted UT
thickness reading leading to floor leakage) can be expressed as a degree of probability when
correctly associated with a parent distribution.

Gumbel(7) is credited with developing practical approaches to the use of extreme value statistics
for a wide variety of applications. These applications include the use of extreme values for
gaining inferences into financial markets, determining liabilities associated with insurance risk,
predicting the impact of certain medical treatments on large populations, for instance, and most
recently, for predicting corrosion pitting rates in petroleum piping and AST floors(8).

The use of the extreme value methods has been used by a number of petroleum companies for
extrapolating pitting corrosion from small inspection patches in an above ground storage tank to
the whole tank(9). Sparago(8) has demonstrated how the thickness distributions can be used to
estimate the probability of a wall thickness being below a certain level from ultrasonic thickness
gauge data. Although the use of this method for characterizing the distribution of pitting in
ASTs is common in Japan, it is relatively new to the US petroleum industry. A recent paper,
however, used extreme value statistics for comparing the results of UT sampling by robotic
equipment conducted while a population of tanks remained in-service to the results of out-of-
service surveys. In all cases the extreme value results matched the results scrutinized from the
out-of-service inspections(10).

In general, extreme value statistics used to evaluate AST floor pitting tend to generate thickness
results that are more conservative than if these statistics were not used. That is, the minimum
remaining thickness of a tank floor determined by a limited data set will be thicker than if the
data were subject to an extreme value analysis. This result is illustrated by the data presented in
Table 3 and Figure 4. The data in Table 3 summarize the results of the case study discussed
above (Jet Fuel Tank Inspection). The measured Tmin value was 0.116 (bottomside). Simply
evaluating the isolated pitting does not provide the inspector with a sense of probability of
occurrence. When the data are analyzed using extreme value statistics, however, the probability
of a particular minimum thickness can be determined. This is shown in Figure 4 where there is a
1% probability that there is a Tmin value of less than .100 inches. Percent probability is
determined from the y-axes. Different floor plate thickness can be reported dependent on their
probability of occurrence. The degree of probability is determined by the number of UT
measurements and the area over which the readings were taken (4).

MC-03-92
Page 10
Table 3. Results from the Jet Fuel Case Study. The minimum thickness was determined to be
0.116 inches.

Average Nominal Thickness (inches) 0.247

Max Thickness (inches) 0.260
Minimum Thickness (inches) 0.116 (bottomside)
Number of UT Readings 508,000

Figure 4. The distribution of data from all of the UT readings from the Table 3 data set using an
extreme value analysis technique (Weibull distribution). Three distributions are shown, that is,
topside, bottomside and nominal UT data were determined to fit individual curves rather than
one single curve. The reliability of the curve fit is noted in the r2 column in the legend.

As discussed above, Table 3 summarizes the results of ultrasonic measurements taken from an
AST floor while in service. The population of floor thickness was determined to be normally
distributed given that over 500,000 ultrasonic measurements were acquired on this particular
AST bottom. The results of this distribution of ultrasonic measurements are shown in Figure 5.

MC-03-92
Page 11
Of greatest interest to the inspector are the values associated with the lower left area of the curve
where the thinnest ultrasonic readings were found. The astute inspector would realize that these
data represent a distribution of values with an associated probability. That is, once a distribution
of actual data is created the analyst can now work with the statistics associated with the
probability of certain values (not actually measured) occurring.

Figure 5. Distribution of UT thickness readings typical of the data found in the Jet Fuel Case
Study. Over 500,000 readings were acquired from the tank floor.

For instance, as mentioned above, the inspector will have the greatest interest in the data located
on the left side of the normal distribution in Figure 5. This area is known in statistics as, one or
two sigma depending on the degree of statistical error the inspector is willing to tolerate
(statistical error can also be thought of as degree of acceptable risk). However, it is possible that
the data within a certain segment of this area may not be normally distributed but may fit another
type of distribution. That is, closer scrutiny of the thinnest areas under the curve may show that
they are not normally distributed in a way that would depict the expected patterns of readings
from areas of general corrosion. On the other hand, the data in the two sigma region may fit a
distribution used for evaluating extreme values which may be exactly the population of
ultrasonic readings associated with random pitting. Distributions typically used to evaluate such
pitting are the previously mentioned Gumbel and the Weibull(5).

MC-03-92
Page 12
In order to test this hypothesis, only the readings from the thinnest area of the distribution in
Figure 5 were subjected to an extreme value analysis. If the data fit an extreme value statistic
better than that associated with the normal distribution, then the inspector can not only claim that
deeper pits exist in the tank but the actual value of the pit with a certain degree of certainty or
reliability can be detected. It is beyond the scope of this paper to explain this process, however,
Figure 4 shows the results of extracting only the 20 deepest bottomside pits from the distribution
shown in Figure 5. The fact that the data fit a Gumbel distribution with a reliability coefficient
of .91 suggests that there probably are thinner values present in the tank than the measured Tmin.
As noted by the indications on both the x and y axes, there is about a 1% probability that a Tmin
exists in the tank of .100 inches.

The case studies presented in this paper used extreme value statistical treatment as well as
calculations based on API 653 (MRT1 and MRT2). It is important to evaluate all of the data to
the extent to which the data fit various distributions in order to better characterize the potential
for tank leakage and to determine various corrosion mechanisms that may exist in the tank. For
instance, the ability to make statements regarding the distinction between general corrosion in
the central part of the tank versus pitting in the critical zone or other confined areas can be a
product from this analysis. The use of robotics for high density scanning produces more than
enough data for evaluating such tank bottom corrosion morphology.

MC-03-92
Page 13
 Appendix A
Cost-Benefit Components

Personnel Pump heel from tank with pump truck

Internal VOC monitoring
Facility Operations Squeegee remaining product
Headquarters, Supply, Scheduling, Trading, Water wash
Marine, Exchange Blow area with fan through manway
Procurement Test for vapors continuously
Engineering Sweep, vacuum interior
Tank and Technical Support Sand blast
Environment Health & Safety
Commercial/Distributor Sales Tank Inspection
Dispatch Center Conduct MFE Inspection
External Conduct Contact UT
Pipe, Valve, Fitting Contractor Measure Bottom Settling
Tank Cleaner
Tank Inspector Tank Downtime Costs
Disposal Company Reduce refinery production
Tank Repair Company Marine demurrage, short load freight rates
Additional trucking
Planning Scheduling & Administration Product downgrading
Identify and notify all impacted departments Alternative storage
Review old tank records Alternative product supply
Optimize dates given all internal needs
Obtain insurance Non-Discretionary/ Just-in-Time Tank
Finalize contract administration Repairs
Scope work, write RFP Patch plates over bottom areas with >40% of
Analyze RFP replies corrosion
Qualify contractors Broken side gauges, floating suction
Review State & Federal requirements IFR/EFR repairs impacting current near term
Obtain alternative product supply functionality
Engineer alternative piping/storage
Tank Cleaning Training Discretionary/Premature Tank Repairs
Confined Space Entry Training Patch plates over bottom areas with <40%
corrosion
Tank Piping Modifications, Product
Disposition Return Tank to Service, Misc.
Drain and clean tank for alternative storage Reinstall manways
Construct alternative piping Remove blinds from lines
Tank lock-out tag-out Contract administration
Payables
Tank Cleaning & Disposal Exchange reconciliation
Remove product through fixed piping Hazardous waste tracking
Downgrade product Tank record keeping
Remove manway

MC-03-92
Page 14
 Appendix B

General Description of In-Service Robotic Equipment and Applications

The remotely controlled system that was used for these case studies is composed of a sludge
removal vacuum system, a robotic vehicle with an in-tank navigation system, an eight-transducer
array ultrasonic system, a sonar imaging system and a video system. The vehicle has been
designed to operate in both fixed and floating roof storage tanks, and is designed to fit through a
20” (50 cm) manway. The system operates in water, No. 2 fuel oil, No. 6 bunker C (heated),
lube oil, diesel fuel, jet fuel and similar middle distillates. This equipment was also configured
to operate in medium to high API crude oil. A similar machine for tank floor inspections in
volatile products such as gasoline is currently in operation.

When deployed inside a tank, the robot can simultaneously vacuum sludge and perform an
ultrasonic inspection of the tank floor. The position and orientation of the vehicle are determined
in real-time using onboard and tank external, shell-mounted navigation transducers which enable
the vehicle to be mapped on a display of the tank floor plates. Position resolution is on the order
of +/- 1.0-inch. During the inspection, vehicle location for each bottom plate thickness
measurement is recorded along with the ultrasonic returns from the tank floor. Although some
data evaluation is conducted while the vehicle is in the tank, the majority of the data analysis is
conducted post-test due to the enormous amount of UT data acquired during a typical 12 hour
inspection. In addition to the thickness map of the floor, the UT data analysis software allows
for the identification of bottomside and topside corrosion. The vehicle is also capable of
detecting tank bottom settlement with onboard pressure transducers. The system is deployed with
the tank full of product. Tank blinding is not required.

Typical deployment operations consist of locating the equipment van and associated utilities
adjacent to the tank within the berm area. The deployment process usually requires two crane
lifts to the top of the tank. Equipment located at the top of the tank consists of the submersible
vehicle and umbilical, pumping systems and in-tank deployment gear. Entry of the vehicle is
completed through the roof’s top manway (18 inches in diameter or greater). A 350-foot (107
meters) umbilical is used to support vehicle operation.

While the system is readied for deployment from the top of the tank, the crew locates the tank
navigation transducers at their proper locations around the tank. These locations, as well as the
position of all tank appurtenances, are entered into a CAD system.

During vehicle deployment, a video record is made by the on-board camera in order to ensure the
proper positioning of the vehicle onto the tank bottom. As with most tanks, accurate drawings
are not always available. Consequently, special procedures need to be implemented in order to
determine the location of various objects within the tank. These objects consist of roof supports,
inlet and discharge pipes, sumps, and related internals. Once these objects are annotated into the
CAD drawing and the proper position of the vehicle is determined, then the vehicle is ready for
floor scanning.

MC-03-92
Page 15
A typical UT run consists of capturing data from all eight transducers every 0.16 of an inch while
the vehicle is driven in a straight path for 2 feet. During each individual run, 1,200 ‘A’ scans
(Figure 1) are collected and recorded, unless the product and sludge prevent the two foot run,
then data are collected while the vehicle is stationery or during very short scan runs. The
thickness measurements are then loaded into a spreadsheet where the ‘B’ scan data limits (Figure
2) are checked. Any location with a measured thickness of less than a pre-determined value is
highlighted by proprietary signal processing software and is then manually reviewed. This
review determines the cause of the low thickness measurement. These causes could include
actual component thinning, a gate error, a loss of signal, sludge/sediment, and/or if the vehicle
ran over a weld seam.
For the particular tanks used in these case studies, between 50,000 and 1.5 million UT readings
per tank were taken. These data are taken throughout the tank including the critical zone around
the shell. Clearly, manual review of this volume of data is not practical. Most of the tanks
inspected were found to have sludge at the bottom of the tank that precluded clear video for
identifying some of the floor features. Sufficient video was taken, however, to confirm the
presence of topside corrosion in some cases and the location of particular in-tank obstacles.
Because of the presence of sludge, the material handling system was initiated in certain tanks and
was used throughout most tank bottom scanning. Although the vehicle sludge removal system
provided a clear path for the vehicle, much of the UT data had to be carefully analyzed off-line
in order to acquire proper information regarding the presence of corrosion.

For all of the case studies, post processing of the UT data was completed through the use of a
thickness measurement module software package in conjunction with post processing signals
detection software. B-scan thickness data and thickness histograms were plotted in all cases.
Areas of special interest were manually reviewed with the thickness measurement module and
with 2D plotting software. Minimal and nominal thickness readings, as well as corrosion
information, were recorded in tabular form for later statistical analysis.

The objective of the in-service inspection is to quantitatively establish the parameters that will
set the next internal inspection. Documents such as API 653 typically use the bottom condition,
corrosion rate and thinnest remaining steel, as the governing parameters. The robotic in-service
inspections described in this paper can provide that data. The minimum remaining metal
estimate is provided as a direct result of the inspection and statistical data analysis. The
corrosion rate is computed from that information.

With this information, the tank operator can estimate how much longer the tank can remain in-
service before a bottom leak. Some codes and regulations may specify a non-zero minimum
remaining bottom metal thickness for operations. In either case, the operator can use the in-
service data to make an informed, quantitative decision regarding the appropriate schedule for
the out-of-service inspection and repair.

MC-03-92
Page 16
References

1. American Petroleum Institute Standard 653, Third Edition, December 2001.

2. American Petroleum Institute Standard 580, Risk Based Inspections
3. Silverman, E.B., Merrick, E.A., Tischuk, J.L. Integrating In-Service High Density AST
Bottom Scanning Into Risk Based Inspections, National Institute of Storage Tank
Management Proceedings, May 2002.
4. Guidelines for the use of statistics for analysis of sample inspection of corrosion. The
Welding Institute Ltd., Cambridge U.K., Research Report 016, 2002.
5. Abernathy, R.B. The New Weibull Handbook, 4th Edition, 2000.
6. Kowaka, M. Introduction to life prediction of industrial plant materials – application of
extreme value statistical method of corrosion analysis. Allerton Press Inc., New York,
1994, ISBN 0 89864 073 3.
7. Gumbel E.J. Statistical theory of extreme values and practical applications. US Dept of
Commerce Applied Mathematics series 33 (1954).
8. Sparago, M. Statistical tools for ultrasonic thickness data analysis, Part 2 – remaining
life estimates. Inspectioneering Journal 1999 (March/April) 1-4.
9. Joshi, N.R. Statistical analysis of UT corrosion data from floor plates of a crude oil
aboveground storage tank. Materials Evaluation 1994 (July).
10. Silverman, E.B., Bass, R., Furillo, F., Wolf, A. In-Service oil tank cleaning and
inspection system: Results of eight (8) independent validations, ASNT Proceedings,
October 2000.

MC-03-92
Page 17