Professional Documents
Culture Documents
Fugitive Emissions Guidance Document
Fugitive Emissions Guidance Document
Fugitive Emissions Guidance Document
December 2020
Foreword
Industrial refrigeration systems commonly use anhydrous ammonia as the refrigerant. Determining the
quantity of refrigerant within a given system is essential to determine whether compliance with the
Process Safety Management standard (29 CFR 1910.119) and the Risk Management Plan program (40
CFR 68) is required. The design quantity of refrigerant within a given system also serves as a benchmark
to normalize the refrigerant losses from the system on an annualized basis. Because industrial
refrigeration systems are, principally, custom-engineered and field-erected, the design or intended
refrigerant inventory needs to be determined on a case-by-case basis.
This document provides methods for establishing the design or intended inventory of refrigerant for a
given industrial refrigeration system. It also provides guidance on dynamically tracking the refrigerant
inventory in refrigeration systems as a means of alerting plant personnel to refrigerant losses that may
be occurring. Finally, this document provides approaches that can be used for identifying locations
where fugitive emissions of ammonia from refrigeration systems may be occurring. Based on the field-
evaluation of fugitive emissions at five separate ammonia refrigerated facilities, we believe the two
greatest categories contributing to annual losses of refrigerant are (1) accidental releases and spills
(including relief valves actuating) and (2) discharging ammonia during maintenance activities that
involve pumping out one or more subsystems or components. Improvements in refrigeration system
mechanical integrity will reduce the frequency and likelihood of accidental refrigerant releases.
Recovery and re-use of ammonia in conjunction with maintenance activities that involve opening the
system will reduce refrigerant consumption.
This project was funded by U.S. EPA Region 5 as part of the Pollution Prevention (P2) program under
Grant Number 00E02366. The investigators would like to thank Christine L. Anderson, Region 5
Pollution Prevention Coordinator and Antoinette Hall, Region 5 Project Officer for their project oversight
and support.
Page 1 of 66
Best Practices for Reducing Fugitive Emissions from
Industrial Refrigeration Systems
Contents
Refrigerant Inventory Determination ........................................................................................................... 3
Principles of Inventory Determination...................................................................................................... 4
Properties of Ammonia ......................................................................................................................... 4
Identifying Liquid Refrigerant Locations ............................................................................................... 7
The Inventory Calculation ......................................................................................................................... 8
Vessels................................................................................................................................................... 9
Condensers ......................................................................................................................................... 17
Evaporators ......................................................................................................................................... 19
Piping................................................................................................................................................... 22
Compressors ....................................................................................................................................... 27
Refrigerant Inventory Calculation Considerations .............................................................................. 29
System Operation ............................................................................................................................... 34
Identifying and Tracking Fugitive Emissions ............................................................................................... 36
Benchmarking Refrigerant Losses ........................................................................................................... 37
Estimating Aggregate Refrigerant Losses by Refrigerant Fill Tracking ................................................... 38
Estimating Aggregate Refrigerant Losses by Dynamic Refrigerant Inventory Calculations.................... 40
Quantify Known Losses ........................................................................................................................... 44
Refrigerant losses due to Incidents and accidents ............................................................................. 44
Refrigerant Releases into Liquids ........................................................................................................ 45
Refrigerant losses due to Maintenance activities ............................................................................... 46
Non-condensable gas Purgers ............................................................................................................ 50
Finding Fugitive Emissions ...................................................................................................................... 53
Screening Components ....................................................................................................................... 53
Bagging Components .......................................................................................................................... 58
Fugitive Emissions Field-Survey Results .............................................................................................. 62
Managing Fugitive Emissions .............................................................................................................. 67
Recommendations to reduce Refrigerant Emissions .......................................................................... 68
References .................................................................................................................................................. 70
Page 2 of 66
Refrigerant Inventory Determination
There are various reasons why industrial refrigeration end-users need to know the inventory 1 (charge)
of refrigerant in their refrigeration systems and, in some cases, managing the inventory of refrigerant in
a system is a regulatory requirement. For end-users with refrigeration systems utilizing certain classes
of fluorochemical refrigerants, initial refrigerant inventory determination and ongoing tracking of
refrigerant emissions is required for compliance with regulations enacted by Congress under Title VI of
the Clean Air Act (CAA) Amendments of 1990. Specifically, Section 608 of the CAA defines specific
requirements for those refrigeration systems containing 50 pounds or more of Class I or Class II ozone
depleting substances (see 40 CFR 82 Part F).
A part of the requirements for owners/operators of refrigeration systems using Class I or Class II
fluorochemical refrigerants is to determine system refrigerant inventories and track refrigerant
additions which are, frequently, the consequence of refrigerant leakage 2 from the system. Specifically,
Section 608 of the Clean Air Act requires owners or operators of an industrial process or commercial
refrigeration systems using Class I or Class II refrigerants to repair refrigerant leaks when the cumulative
loss would exceed 30% of the total charge for an industrial system and 20% of the total charge for a
commercial refrigeration system over a 12-month period. To enable estimating refrigerant leak rates on
a percentage basis, the system's normal refrigerant inventory must be established. For large built-up
refrigeration systems, determining refrigerant inventory within a given refrigeration system is best
accomplished by summing the quantity of refrigerant that normally resides within each individual
component or subsystem that, together, comprise the entire refrigeration system. Further information
on the regulations that are applicable to fluorochemical refrigerant-based systems can be found on the
EPA's website at: https://www.epa.gov/section608.
For end-users that operate industrial refrigeration systems using anhydrous ammonia as the refrigerant,
the provisions of Section 608 noted above do not apply since it has no ozone depletion potential.
Rather, quantifying and documenting the normal and maximum intended refrigerant inventory is
necessary to establish whether the system is a covered process and subject to OSHA and EPA regulations
applicable to processes that use toxic (or flammable) substances above a threshold quantity. Both
OSHA’s Process Safety Management of Highly Hazardous Chemicals program (29 CFR 1910.110) and
EPA’s Risk Management plant program (40 CFR 68) have established a threshold quantity of ammonia
for coverage at 10,000 lbm. For industrial ammonia systems that have a refrigerant inventory at or
above this threshold quantity, end-users/operators are required to comply with the provisions of both
PSM and RMP. Even in cases where the total inventory of an ammonia refrigeration system is less than
the threshold quantity, documentation to establish the system’s refrigerant inventory is needed to
demonstrate a given system is not subject to these requirements. Beyond having a documented
refrigerant inventory of an industrial refrigeration system for purposes of being covered by PSM/RMP,
this information is an essential component for hazard management.
A logical extension of developing an initial determination of refrigerant inventory (charge) for a given
system is to update the refrigerant inventory on an ongoing basis during system operation. When
1
The terms “charge” or “quantity” are commonly used interchangeable with “inventory” and represent the
amount of refrigerant contained within a given refrigeration system.
2
Intentional venting of ODS refrigerants is prohibited by Section 608.
Page 3 of 66
properly analyzed, this “dynamic charge calculation” can provide end-users with an indicator that their
refrigeration system is not only losing refrigerant but it can also provide estimates of the annual loss
rate. Although this approach is a “lagging indicator”, it can prompt action by the system’s end-user
operator to investigate source(s) of the refrigerant loss and subsequent triggering repair actions much
quicker than simply waiting until the refrigerant charge within the system has significantly depleted.
Most facilities today periodically “top-off” the refrigerant in their systems only when the system reaches
a low level where continued operation of the system may be disrupted. Such a threshold to trigger
action leads to a substantially larger total loss of refrigerant from the system over time, effectively, the
current approach would be considered less desirable because it is a longer “lagging indicator.”
An alternative approach for inventory determination that is non-invasive uses engineering calculations.
In this approach, first-principles of thermodynamics are applied to establish the refrigerant state within
each given component in a system and established thermodynamic properties (e.g. refrigerant density
as a function of pressure) along with the corresponding physical volume of each given component or
subsystem are used to calculate the mass of the refrigerant residing within that component or
subsystem. Adding the refrigerant inventory (mass) residing in each individual component of a system
enables estimating the entire system’s refrigerant charge. Applying the methods included in this
guidance document for quantitatively estimating the refrigerant inventory can be expected to yield total
system inventory estimates to within single digit percentage accuracy based on our experience in having
detailed inbound refrigerant receipts for initial charging of built-up refrigeration systems.
Properties of Ammonia
Ammonia’s properties allow prioritizing the process of quantifying the refrigerant inventory in a large
industrial refrigeration system comprised numerous components. Density is a property that represents
the mass per unit volume (e.g. lbm/ft3) of a substance. Figure 1 shows the variation in the density of
liquid ammonia as a function of temperature with various levels of subcooling as a parameter. As the
liquid ammonia temperature increases, the refrigerant density decreases. The decrease in density is
rather modest at 8 lbm/ft3 over a 140°F temperature range which represents about an 18% overall
decrease in density or a 0.13% density decrease per °F temperature increase.
Page 4 of 66
Saturation Temperature [°C]
-40 -30 -20 -10 0 10 20 30
44
Anhydrous Ammonia
43 690
20°F Subcooling
Liquid Density [lbm/ft ]
3
42
30°F Subcooling
660
41
Saturated Liquid
645
40
10°F Subcooling 630
39
615
38
600
37
585
36
-40 -20 0 20 40 60 80 100
Saturation Temperature [°F]
Figure 1 – Variation in density for liquid phase anhydrous ammonia.
The variation of ammonia density for a vapor-phase is much more significant than the liquid phase;
however, the actual magnitude of vapor density is substantially smaller compared liquid density
diminishing the importance of accurately accounting for its variability. As the saturation temperature
(pressure) of ammonia vapor increases, the vapor density increases but as the superheat increases at
constant pressure, the density decreases. Figure 2 shows the density variation for ammonia vapor over
the same temperature range as Figure 1. In absolute terms, the change in vapor density is rather small
(about 0.6 lbm/ft3); however, the percentage basis change is quite large (1,200% increase). Another
point to be noted here is that, unlike the liquid density, the vapor density increases with increasing
saturation temperature/pressure.
Page 5 of 66
Saturation Temperature [°C]
-40 -30 -20 -10 0 10 20 30
0.7 11
Anhydrous Ammonia
10
0.6
9
Vapor Density [lbm/ft ]
3
0.5 8
7
0.4
6
10°F Superheat 5
0.3
30°F Superheat 4
Saturated Vapor
0.2 20°F Superheat 3
2
0.1
1
-40 -20 0 20 40 60 80 100
Saturation Temperature [°F]
Figure 2 – Variation in density for vapor-phase anhydrous ammonia.
Figure 3 combines the results shown in the previous two figures and shows the ratio of saturated liquid
density to saturated vapor density over a range of saturation temperatures. The density ratio ranges
from approximately 55 at 95°F to over 1,000 at -40°F! This heightens the importance of identifying
those portions of a refrigeration system containing liquid ammonia and determining the volume of liquid
within those components to accurately estimate the refrigerant inventory compared to those portions
of the system containing vapor-phase ammonia.
Page 6 of 66
Figure 3 – Ratio of saturated liquid to saturated vapor density for anhydrous ammonia.
In summary, accurately quantifying the operating refrigerant charge for a given system requires focusing
on those components and subsystems that hold liquid-phase refrigerant. The rationale for this will
become apparent when comparing mass of refrigerant associated with vapor vs. liquid in the examples
provided in the sections that follow.
Page 7 of 66
Figure 4 – Schematic illustrating a two- temperature level refrigeration system with multiple
evaporator types.
Page 8 of 66
liquid level setpoints, float locations, and actual operating levels of liquid refrigerant. Doing the initial
work on component details will yield dividends in an accurate system inventory estimate.
Vessels
As noted above, vessels are the component category that holds the largest refrigerant charge for most
industrial refrigeration systems. The principle function of vessels in these refrigeration systems is to
separate liquid and vapor phase refrigerant. In addition, vessels provide a buffer to accumulate liquid
refrigerant when transient operation occurs within a system. In some cases, liquid inventory is needed
to create a sufficient head (pressure) to avoid liquid refrigerant pump cavitation. The internal volume of
a typical vessel can be determined by:
2 3
𝜋𝜋 ∙ 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝜋𝜋 ∙ 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣
𝑉𝑉𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 = ∙ �𝑂𝑂𝑂𝑂𝑂𝑂𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 − �+ Equation 1
4 𝑅𝑅𝐷𝐷 6 ∙ 𝑅𝑅𝐷𝐷
where 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 is the vessel diameter, 𝑂𝑂𝑂𝑂𝑂𝑂𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 is the overall length or height of the vessel, and 𝑅𝑅𝐷𝐷 Is the
ratio of the head curvature to the vessel diameter (refer to Figure 12 for an illustration of the vessel
dimensions included in this formula). The ratio of heat curvature to vessel diameter is calculated by the
following:
𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣
𝑅𝑅𝐷𝐷 = Equation 2
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 ℎ𝑒𝑒𝑒𝑒𝑒𝑒 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙ℎ
where “Total head length” is the length of the vessel with the cylindrical portion subtracted off.
Dimensions can be in any consistent unit system, feet, inches, or meters.
Fluctuating refrigeration loads and transient system operation often results in refrigerant migration to
other parts of a system. Because liquid-phase refrigerant from one vessel can move to another vessel
within the system over the course of minutes or hours, it is important to determine liquid refrigerant
levels within vessels during periods of stable system operation and in a comparatively short snapshot in
time to avoid biasing (i.e. double accounting) system refrigerant inventory estimates that would result
from migration.
Determining Vessel Liquid Level
Determining the current level of liquid refrigerant within a vessel can be challenging. For insulated
vessels, the insulation system itself conceals visual cues that are indicative of the presence of low
temperature liquid-phase refrigerant, often, in the form of moisture condensation or frost appearing on
the outside of the vessel. Uninsulated vessels are commonly found on the high-pressure side of a
system and the comparatively higher temperature of liquid refrigerant in high-side vessels will not result
in the formation of condensation or frost on the outside of the vessel. Fortunately, most vessels are
equipped with some means of liquid level determination to aid site refrigeration personnel with system
operation. Examples of refrigerant liquid level indication include: fixed float devices, continuous level
sensors (capacitance probes), and other visual indicators such as bullseye sight glasses, flat-armored
sight indicators, and tubular level indicators. It is important to consider the operating characteristics
and limitations for each of these level sensing/indicating devices to ensure it is properly operating and
calibrated (if applicable) before using the measurements for operation or charge calculation. Once the
accuracy of the liquid level sensor (if exists) has been determined, the volume of liquid can be
Page 9 of 66
determined and, knowing the liquid density, the component’s refrigerant charge can be systematically
quantified.
Sight glasses
Sight glasses, sometimes referred to as “bullseye sight glasses” are widely used as means to provide an
indication of liquid level in a vessel. An example of these discrete sight glasses arranged on a level
column is shown in Figure 6. The level column itself can also contain float switches and electronic liquid
measurement devices such as a capacitance probe or guided wave technology. Sight glasses in level
columns with this setup are, typically, equipped with reflex lenses which turn dark in color in the
presence of liquid refrigerant and clear in the presence of vapor refrigerant as shown in Figure 7.
Page 10 of 66
When determining liquid level with sight glasses it is important to accurately measure the liquid height
corresponding to each glass. The bottom of the level column may, or may not, correspond with the
bottom of the vessel. Referring to Figure 7, the lower three sight glasses are dark indicating liquid is
present and the sight glasses above are transparent indicating that vapor is present at those locations.
This visual representation means that the actual level in the column is in a range that can span
anywhere from just above the top of the third sight glass to just below the bottom of the fourth sight
glass. Insulated vessels, such as that shown in Figure 6, also often utilize bullseye sight glasses on a level
column. Unless the sight glasses are equipped with a frost shield, the accumulation of frost on the
surface of the glass will occur as the cold liquid refrigerant fluctuates up and down the column. Frost
accumulation conceals the glass surface making it difficult to confirm if liquid is present at the time a
reading is sought.
Transparent glass
indicates presence of
vapor
Darkened glass
indicates presence of
liquid
Figure 6 – An insulated vessel (rear) with a Figure 7 – Column with reflex lens sight
sight glasses, float switch, and level probe. glasses on a high-pressure receiver.
Once an accurate measure of liquid level has been determined for each sight glass location, it is unlikely
the actual liquid level will stay within one discrete sight glass (a common sight glass in the industry is
1.6” in diameter); therefore, an estimation of the actual liquid level is required. The recommended
conservative estimate, if no other means of liquid level measurement is available, is to assume that the
liquid level is immediately below the lowest sight glass showing vapor (i.e. just below the 4th sight glass
from the bottom in Figure 7).
Float Switches
Mechanical floats are used for level control on smaller vessels such as surge drums attached to
individual evaporators as well as in older low-side vessels for level control. These devices function over
a relatively narrow range of liquid level (e.g. 2 inches). That is, when the liquid level within a vessel
drops 2 inches, the float will energize a solenoid valve for liquid make-up and when the liquid level
increases to the upper end of the float’s range (e.g. 2 inches), the solenoid is de-energized. The level of
Page 11 of 66
liquid within a vessel is assumed to coincide with the upper float chamber connection as shown in Figure
8.
Upper range of
liquid level
Figure 8 – Typical mechanical float showing the assumed upper range of liquid level as a frame
of reference for the attached vessel’s liquid level.
Capacitance Probes
Capacitance probes consist of long conductors inserted into a level column which are configured to
indicate the level of liquid ammonia in contact with it by monitoring electrical capacitance. This
capacitance is then translated into an electrical signal and calibrated to indicate the level of liquid
ammonia present. Capacitance probes are currently the most common electronic liquid level
measurement device used in industrial refrigeration systems and can be highly accurate, but accuracy
needs to be verified when conducting the refrigeration system charge calculation.
There are several common inaccuracies with capacitance probes to check for. First, it is important to
understand that the capacitance probe output indicates the proportion of the probe’s active length in
contact with liquid ammonia. Often, the control panel associated with most capacitance probes
indicates a “percent full” which raises additional questions. Does the displayed level value indicate the
percent of the probe exposed to liquid ammonia, percent of the vessel height with liquid ammonia, or
percent of the vessel volume containing liquid ammonia? Often the probe length is not equal to the
Page 12 of 66
overall length (or diameter for horizontal vessel orientations) of the vessel. Another consideration is
that any refrigeration oil present in the level column and in contact with the active capacity probe will
result in inaccurate level indications because the oil used in ammonia refrigeration systems has a sensed
capacitance closer to that of ammonia vapor. The dielectric constant of liquid ammonia ranges from 16-
24 under normal refrigeration conditions, as opposed to the vapor state which is near 1. Mineral oil on
the other hand has a dielectric constant of less than 4; therefore, any of the capacitance’s probe length
in contact with oil will yield an indicated liquid level lower than the actual liquid level present. Once the
capacitance probe’s indication of liquid ammonia level has been verified with sight glasses, floats, or
other means, they can be highly useful tools in tracking liquid ammonia levels in vessels.
Wave Sensors
Wave level sensors are a newer technology offering some advantages over more traditional electronic
level sensing technologies, such as flexible installation and higher accuracy. Wave sensors provide a
measurement of distance from the sensor to the liquid surface by measuring the time from when an
electromagnetic signal leaves the sensor head, reflects off the liquid surface, and returns to the sensor.
The sensor can then either directly output the measurement of vapor space length or convert that
measurement (if the correct vessel height is given) to liquid level.
The first type of wave sensor to be commonly seen in refrigeration systems is the wire guided wave
type. This type of sensor, shown in Figure 9, is used because it can be readily installed in a typical level
column attached to a vessel similar to a capacitance probe. The microwave signal must be guided
because without the wire there are many false signals and bounce backs resulting from the narrow
channel. The guide wire offers some installation and maintenance advantages over capacitance probes
since the wire can be inserted into the column in tighter quarters and the guide wire replacement cost is
low. There is also potential for shorter wavelength (≈80 GHz) sensors to be installed directly on vessels
through a ball type stop valve; however, this type of application is not commonly found in industrial
refrigeration systems.
Page 13 of 66
attached to the sensor and run plumb within the level column to provide an accurate reading. If
deposits or liquid movement disturb the wire, level indicating errors can occur due to signal reflection
off the walls of the float column. If the sensor directly outputs the vessel liquid level, the appropriate
vessel height must be programmed into the sensor electronics to provide an accurate level. If an
unguided wave sensor is utilized directly on the vessel, it must be oriented per the manufacturer’s
instructions and calibrated to indicate the actual liquid level while ignoring false peaks caused by
reflection signals. These sensors typically exhibit measurement “dead zones” of 5-12 inches at the top
and bottom of the range. If possible, it is important to verify the level readings with other means during
start up to ensure an accurate reading.
Figure 10 – The use of an infrared camera to locate the liquid level in a high-pressure receiver.
Page 14 of 66
Figure 11 – The use of an infrared temperature gun to locate the liquid level in a high-pressure receiver.
The higher surface temperature (left photo) is indicative of the presence of vapor and the transition to a
lower surface temperature (right photo) indicative of the presence of liquid.
Vertical Vessels
Vertical vessels typically operate with a liquid level between the vessel’s heads. This means one head is
liquid filled, and one head contains only vapor-phase refrigerant. In that case as shown in Figure 12, the
liquid volume can be determined by Equation 3.
2
𝜋𝜋 ∙ 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣
𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 = �𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 − �
4 6 ∙ 𝑅𝑅𝐷𝐷 Equation 3
𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣
𝑅𝑅𝐷𝐷 =
2 ∙ 𝐿𝐿𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻 Equation 4
Page 15 of 66
Figure 12 – Dimensions required to calculate the inventory of a vertical vessel.
In the case where the liquid level does not fill the bottom head, Equation 5 can be used to determine
the actual volume corresponding with liquid refrigerant in a partially-filled head.
2 3
𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 ∙ 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙
𝑉𝑉𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 ℎ𝑒𝑒𝑒𝑒𝑒𝑒 = 𝜋𝜋 ∙ 𝑅𝑅𝐷𝐷2 ∙ � − �
2 ∙ 𝑅𝑅𝐷𝐷 3 Equation 5
A vessel with an operating level in the top head would be rare, but in that case Equation 1 and Equation
5 can be reworked to provide the total charge.
Horizontal Vessels
The liquid volume in horizontal vessels can be determined using Equation 6.
𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
−1 2 ∙ 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙
2
𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 ⎛cos �1 − 𝐷𝐷
𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣
� 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 2
= 𝐷𝐷𝑣𝑣𝑣𝑣𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 �𝑂𝑂𝑂𝑂𝑂𝑂𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 − �⎜ −� −
𝑅𝑅𝐷𝐷 4 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 2
⎝
3 3 ∙ 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 2 2 ∙ 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 3
𝜋𝜋 ∙ 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 � 2 − 3 �
1 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ⎞ 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣
∙� − � +
2 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 ⎟ 6 ∙ 𝑅𝑅𝐷𝐷
⎠
Equation 6
Page 16 of 66
Figure 13 - Dimensions required to calculate the inventory of a horizontal vessel.
Total Vessel Charge
The vapor volume can be determined by subtracting the total vessel volume determined by Equation 1
from the calculated liquid volume. The density of liquid and vapor can be determined from property
sources such as Table 2. Equation 7 can be used to find the total vessel inventory.
For more detail on determining the liquid level in vessels, see the IRC Technote “Refrigerant Inventory
Determination” (IRC 2004).
Condensers
Operators of industrial refrigeration systems understand that evaporative condensers are capable of
“holding up” significant quantities of liquid refrigerant – particularly under abnormal operating
Page 17 of 66
conditions that often occur during cold weather conditions. During normal operation, condensers freely
drain condensed high-pressure liquid refrigerant out the bottom of the condenser’s heat exchanger
through a drain pipe connected to the heat exchanger’s header box. For the purposes of determining
refrigerant inventory in a system, we will focus on the normal operation and not the abnormal
operation; however, the inventory under abnormal operation should be considered in the context of risk
assessments and condenser siting.
Virtually all evaporative condenser manufacturers catalog two key pieces of information that aid in
determining refrigerant inventory: condenser heat exchanger volume and condenser normal operating
charge. Figure 14 below shows an excerpt from one manufacturer’s catalog for evaporative condensers.
The tenth column from the left notes the operating charge of ammonia for each condenser model. If
this type of operating charge information is not provided by the condenser manufacturer, two rules-of-
thumb can be used for estimating the condenser’s refrigerant charge. The first rule-of-thumb assumes
that the average density of refrigerant in the condenser heat exchanger is approximately 10 lbm/ft3. The
condenser inventory is then the product of the condenser’s gross coil volume (ft3) and the average
refrigerant density (10 lbm/ft3). The second rule-of-thumb was developed based on evaluating operating
charge data for a range of evaporative condenser types and sizes from which we concluded that a
reasonable average charge is 90 lbm per million Btu/hr (mmBh) of rated heat rejection at a
manufacturer’s specified rating condition of 105°F saturated condensing and 70°F ambient wet-bulb
temperature. In this case, the operating condenser operating charge can then be conservatively
estimated by:
𝑙𝑙𝑙𝑙𝑚𝑚 Equation 8
𝑀𝑀𝑟𝑟𝑟𝑟𝑟𝑟,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 ≈ 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 [𝑚𝑚𝑚𝑚𝑚𝑚ℎ] ∙ 90 � �
𝑚𝑚𝑚𝑚𝑚𝑚ℎ
If the rating of the condenser is expressed in “evaporator tons”, the corresponding rule-of-thumb is 1.85
lbm/evaporator ton (rated at 96.3°F saturated, 78°F wet-bulb and +20°F suction).
Figure 14 – Exemplar data for condenser refrigerant operating charge excerpted from a
condenser manufacturer’s product catalog (source: BAC, 2003).
Page 18 of 66
Evaporators
The evaporators used in industrial refrigeration systems come in all shapes, sizes, and styles. The
majority of evaporators in industrial refrigeration service are air-cooling plate-finned type heat
exchangers. Other heat exchanger designs in use include: shell-and-tube, plate-and-frame, corrugated
plate, plate-and-shell, scraped surface, and others. The refrigerant-side of evaporators for industrial
refrigeration systems are, generally, configured in one of three ways: liquid overfed, flooded, or direct-
expansion (DX). Unfortunately, most manufactures do not catalog the normal refrigerant operating
charge for their air-cooling evaporators. As a result, some basic guidelines, depending on evaporator
configuration and prevailing operating conditions, are used to estimate the normal operating charge of
refrigerant within the unit. In this case, the mass of refrigerant within an evaporator will be the product
of the heat exchanger’s internal volume and the average density of refrigerant resident within the
evaporator.
Most evaporator manufacturers do catalog internal volumes for each model or unit so methods for
estimating the density of refrigerant within the evaporator unit during normal operation are needed and
presented in this section.
Flooded and Liquid Overfeed Air Cooling Evaporators
For liquid overfed and flooded evaporators, the average refrigerant density can be approximated by the
following
where ρref is the average refrigerant density in the coil (lbm/ft3), C is an empirical constant provided in
Table 3, ρref,two-phase,outlet is the two-phase refrigerant density at the coil outlet, and ρref,sat,inlet is the density
of saturated refrigerant liquid at the evaporator inlet.
Refrigerant Feed C
Overfed 0.25
Flooded 0.5
The density of the two-phase refrigerant at the evaporator outlet is dependent on the quality of
refrigerant at the coil outlet and the quality will be dependent on the liquid overfed ratio, OR, for the
unit.
1
𝑥𝑥 =
(1 + 𝑂𝑂𝑂𝑂) Equation 11
where 𝑥𝑥 is the mass fraction of vapor (i.e. quality) at the evaporator outlet and 𝑂𝑂𝑂𝑂 is the overfeed ratio
(ratio of mass of liquid to mass of vapor leaving the evaporator). For liquid overfed evaporators, 𝑂𝑂𝑂𝑂 can
be approximated by using the manufacturer’s-recommended overfeed rate for each evaporator.
Page 19 of 66
Alternatively, 𝑂𝑂𝑂𝑂 can be estimated by calculating the ratio of the liquid refrigerant supply flow rate to
the evaporator over the minimum required flow rate just to meet the evaporator’s rated thermal
performance. In this case, the liquid supply flow rate can be estimated based on the pressure difference
across the hand-expansion valve and the valve’s 𝐶𝐶𝐶𝐶 for the given number of turns open on the valve.
For flooded evaporators, the 𝑂𝑂𝑂𝑂 is assumed to be 1. The specific volume of the mixed phase refrigerant
out of the evaporator and corresponding density are given by the following:
1 Equation 13
𝜌𝜌𝑟𝑟𝑟𝑟𝑟𝑟,𝑡𝑡𝑡𝑡𝑡𝑡−𝑝𝑝ℎ𝑎𝑎𝑎𝑎𝑎𝑎,𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 =
𝑣𝑣𝑟𝑟𝑟𝑟𝑟𝑟,𝑡𝑡𝑡𝑡𝑡𝑡−𝑝𝑝ℎ𝑎𝑎𝑎𝑎𝑎𝑎,𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
where 𝑥𝑥 is quality, 𝜌𝜌𝑟𝑟𝑟𝑟𝑟𝑟,𝑡𝑡𝑡𝑡𝑡𝑡−𝑝𝑝ℎ𝑎𝑎𝑎𝑎𝑎𝑎,𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 is the mixed phase density leaving the evaporator,
𝑣𝑣𝑟𝑟𝑟𝑟𝑟𝑟,𝑡𝑡𝑡𝑡𝑡𝑡−𝑝𝑝ℎ𝑎𝑎𝑎𝑎𝑎𝑎,𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 is the mixed phase specific volume leaving the evaporator, 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 is the vapor
specific volume and 𝑣𝑣𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 is the liquid phase specific volume both evaluated at saturation for the
pressure in the evaporator.
Direct Expansion Air Cooling Evaporators
For direct-expansion evaporators, the average density of refrigerant in the evaporator is approximated
as:
where 𝜌𝜌𝑟𝑟𝑟𝑟𝑟𝑟,𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 is the density of refrigerant downstream of the expansion device and 𝜌𝜌𝑟𝑟𝑟𝑟𝑟𝑟,𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 is the
refrigerant density at the outlet of the coil but upstream of any evaporator pressure regulator. The
refrigerant density at the coil inlet is given by
where 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓ℎ represents the flash gas (mass fraction of vapor) generated downstream of the expansion
device. The flash gas fraction is calculated by
where ℎℎ𝑖𝑖𝑖𝑖ℎ−𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 is the enthalpy of high pressure liquid upstream of the expansion device,
ℎ𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 is the enthalpy of saturated liquid at the evaporator pressure andℎ𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣,𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 is the enthalpy
of saturated vapor at the evaporator pressure.
Other Evaporator Types
For other evaporator types, such as shell-and-tube chiller heat exchangers, manufacturers often provide
operating charge data similar to that provided for evaporative condensers. As an example, Figure 15
shows a catalog sheet for a line of flooded shell-and-tube heat exchangers offered by one manufacturer.
The seventh column from the left shows operating charge data for each chiller size. Because the volume
of internal components in a chiller can vary significantly from manufacturer to manufacturer, the
specific chiller manufacturer should be contacted directly to determine whether they are able to provide
Page 20 of 66
refrigerant operating charge information for their particular product lines. In the absence of detailed
refrigerant operating charge data from the chiller manufacturer, a rule-of-thumb assumes the operating
refrigerant charge for a shell-and-tube chiller is 20 lbm per ton of refrigeration capacity per degree F
approach temperature as given by Equation 17.
20 ∙ 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 Equation 17
𝑀𝑀𝑐𝑐ℎ𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 ≈
(𝑇𝑇0 − 𝑆𝑆𝑆𝑆𝑆𝑆)
where Capacity is the chiller capacity in tons of refrigeration, 𝑇𝑇0 is the design leaving fluid temperature
from the chiller to the load [°F], and 𝑆𝑆𝑆𝑆𝑆𝑆 is the corresponding saturated evaporator temperature [°F] of
the ammonia in the chiller.
Welded (or nickel brazed) plate-pair heat exchangers may be approximated using the methods for an
air-cooling evaporator using the appropriate liquid feed type (gravity flooded or DX) with the volume
determined by obtaining the volume of the refrigerant containing passages in the plate pair and the
number of plate pairs. This information will require contact with the manufacturer of the plates.
Page 21 of 66
Figure 15 – Flooded shell-and-tube chiller data sheet (RVS).
Piping
A piping network carries refrigerant to and from components that make up the industrial refrigeration
system. When refrigerant is in a single phase (vapor or liquid), the total inventory in a pipe segment can
be accurately determined. Usually, uncertainties in refrigerant inventory for piping are associated with
uncertainties in the physical size of the piping segment and these are small for single-phase refrigerant
piping systems. Estimating the refrigerant inventory for piping with two-phase refrigerant flow is more
difficult.
In this section, guidance is provided for estimating the inventory of refrigerant for: liquid-only, vapor-
only, and two-phase liquid/vapor refrigerant in piping. The following table refers to the diagram shown
in Figure 4 which identifies the segment of piping by number, name, and typical state of refrigerant in
that piping segment.
Page 22 of 66
Table 4 – Refrigerant piping identification and state
Pipe Segment
Description Liquid Two-phase Vapor
(See Figure 4)
1 High stage suction
2 High stage discharge
3 High pressure liquid drain
4 Medium pressure wet return
4' “Medium pressure” liquid
4" “Medium pressure” pumped liquid
5 Low pressure wet return
5' “Low pressure” liquid
“Low pressure” pumped liquid
6 Low stage suction
7 Low stage discharge
8 High pressure liquid
Those piping segments identified in Table 4 as containing liquid become the highest priority to quantify
first followed by the two-phase piping segments. The inventory in vapor-containing piping will be
extremely small and should be considered last. Determining the inventory of refrigerant for a given
piping segment is as follows:
1. Determine internal cross-sectional area of pipe based on the pipe size (based on nominal pipe
size and schedule)
2. Estimate the pipe length
3. Calculate volume of pipe segment
4. Determine the state of refrigerant in the pipe segment
5. Calculate the refrigerant density
6. Calculate mass (product of segment volume and density)
Step 1 involves determining the cross-sectional area of the pipe segment in question. Refrigerant piping
will have a nominal pipe dimension and a corresponding wall thickness (standardized by its “schedule”).
Pipe sizing information is available from several sources including pipe suppliers, IIAR (2019), ASHRAE
(2017), and other reference handbooks. A subset of pipe sizes is given below in Table 5.
Page 23 of 66
The cross-sectional area of a pipe can be calculated by the following:
𝑑𝑑 2 Equation 18
𝜋𝜋 ∙ � 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 � 2
𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖
12
𝐴𝐴𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐−𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = =
4 183.3
where 𝐴𝐴𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐−𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 is the pipe cross-sectional area (ft2), 𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 is the inside pipe diameter in inches.
The volume of the pipe segment is calculated as
where 𝑉𝑉𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 is the pipe volume (ft3) and 𝐿𝐿𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 is the pipe segment length (ft). Note, Equation 18 and
Equation 19 should be applied only to those piping sections with constant cross-section. In situations
where pipe size changes, apply both equations consecutively for each respective pipe segment.
Liquid
Once the volume of piping is determined, the next step involves determining the density of the
refrigerant occupying each pipe segment. As previously discussed, the refrigerant density is dependent
on the pressure and temperature of the refrigerant. Properties of ammonia are available from sources
such as IRC (2000), ASHRAE (2017), and IIAR (2019). An excerpt from the IRC ammonia property tables is
given below in Table 6. For example, a high-pressure liquid line with saturated liquid at 95°F carries
refrigerant having a density of 36.67 lbm/ft3.
Page 24 of 66
Table 6 – Anhydrous ammonia properties at saturation conditions (IRC 2000).
The mass of refrigerant occupying the pipe segment is the product of the pipe segment volume (ft3) and
the refrigerant density (lbm/ft3).
For liquid lines, the following correlation can be used to estimate the refrigerant inventory given inside
pipe diameter and refrigerant saturation temperature.
2 Equation 21
𝑀𝑀𝑟𝑟𝑟𝑟𝑟𝑟,𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝,𝑒𝑒𝑒𝑒𝑒𝑒 = 4.441 − 3.353 ∙ 𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 + 23.0282 ∙ 𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖
2
+ 0.2142 ∙ 𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠 − 0.0002111 ∙ 𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠 − 0.16767
∙ 𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 ∙ 𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠
where 𝑀𝑀𝑟𝑟𝑟𝑟𝑟𝑟,𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝,𝑒𝑒𝑒𝑒𝑒𝑒 is the estimated mass (lbm) of liquid refrigerant per 100 ft of pipe, 𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 is the
inside diameter of the pipe (in), and 𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠 is the liquid saturation temperature (°F). This correlation is
valid for pipe sizes ranging from 1.5” – 2.5” (Schedule 80) and 3”-16” (Schedule 40). Applicable
refrigerant temperatures range from: -40°F to 100°F (saturated liquid). The correlation yields an
accuracy of -9.8% < 𝑀𝑀𝑟𝑟𝑟𝑟𝑟𝑟,𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝,𝑒𝑒𝑒𝑒𝑒𝑒 < 5% with an average error of 1%.
Page 25 of 66
Vapor
Significant mass does not typically reside in vapor-only piping; however, mains and long runs can add a
non-trivial amount of refrigerant to the charge calculation when added. In these cases, Equation 18,
Equation 19, and Equation 20 are used along with the vapor density of refrigerant within each vapor
section to arrive at the refrigerant quantity within those sections of piping.
Two Phase Suction
For piping segments that carry two-phase flow, determining the mixed phase density requires an
additional step. The mixed phase density of refrigerant in a recirculated liquid suction line (sometimes
referred to as a “wet suction return”) can be estimated by the following.
1 Equation 23
𝜌𝜌𝑟𝑟𝑟𝑟𝑟𝑟,𝑡𝑡𝑡𝑡𝑡𝑡−𝑝𝑝ℎ𝑎𝑎𝑎𝑎𝑎𝑎 =
𝑣𝑣𝑟𝑟𝑟𝑟𝑟𝑟,𝑡𝑡𝑡𝑡𝑡𝑡−𝑝𝑝ℎ𝑎𝑎𝑎𝑎𝑎𝑎
where 𝑥𝑥 is quality, 𝜌𝜌𝑟𝑟𝑒𝑒𝑒𝑒,𝑡𝑡𝑡𝑡𝑡𝑡−𝑝𝑝ℎ𝑎𝑎𝑎𝑎𝑎𝑎 is the mixed phase density, 𝑣𝑣𝑟𝑟𝑟𝑟𝑟𝑟,𝑡𝑡𝑡𝑡𝑡𝑡−𝑝𝑝ℎ𝑎𝑎𝑎𝑎𝑎𝑎 is the mixed phase specific
volume, 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 is the vapor specific volume, and 𝑣𝑣𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 is the liquid phase specific volume evaluated at
saturation conditions for the pressure in the pipe segment. The quality can be estimated based on the
overfeed ratio as follows.
1
𝑥𝑥 =
(1 + 𝑂𝑂𝑂𝑂) Equation 24
where 𝑂𝑂𝑂𝑂 is the average overfeed ratio for evaporators contributing return refrigerant to the
recirculated liquid suction (wet-suction return). As shown in Figure 16, the volume percentage of liquid
in the recirculated liquid suction line is typically low, and does not exceed 1% until 10°F. Liquid fractions
exceeding 3% would likely only occur in misadjusted or malfunctioning systems.
Page 26 of 66
3
2.75
2.5
Figure 16 – Recirculated liquid suction liquid refrigerant volume fraction for ammonia with 3:1
overfed evaporators.
Condenser Drain Piping
The amount of liquid in a condenser drain pipe will vary considerably depending on operating
conditions, piping arrangement, and ambient conditions. Some sections of drain piping will be liquid-
filled at all times, such as “P” traps, or piping below the receiver liquid level in a “surge arrangement.”
Infrared temperature measurements of the liquid drain piping provide a more precise indication of the
liquid level. The method is identical to that discussed above in the section on liquid level determination
for pressure vessels.
Compressors
Compressors do not contain a large refrigerant inventory. The compressor volume itself will not contain
any appreciable amount of refrigerant because it is relatively small; however, it is important to consider
the additional components that makeup the compressor “package.” The compressor-related
components that warrant consideration in a refrigerant inventory calculation include: oil separator, and
thermosiphon oil cooler and associated piping (if exists). the refrigerant inventory associated with each
of these key compressor package components is discussed in the sections that follow.
Oil Separators
The majority of screw compressors in industrial refrigeration systems utilize a vessel to separate oil from
compressor discharge vapor (Jekel et al., 2001). The state of refrigerant in the oil separator is
superheated vapor at the prevailing discharge pressure and temperature. Recall from the above
discussion of refrigerant properties that as the pressure of the vapor increases, the density of the vapor
will also increase; consequently, an oil separator operating in high stage duty will hold more mass of
vapor than an equal sized oil separator operating in booster service. A conservative estimate of the
refrigerant mass in an oil separator will neglect the volume associated with its internal components such
as coalescing filter elements, piping, baffles, heating elements, and the accumulated oil itself. These
internal components can reduce the gross internal volume of the oil separator by 20-30%.
Page 27 of 66
Assuming the separator as a simple cylinder, the following equation gives the gross volume of the oil
separator:
2
𝜋𝜋 ∙ 𝐷𝐷𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ∙ 𝐿𝐿 Equation 25
𝑉𝑉𝑜𝑜𝑜𝑜𝑜𝑜 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 =
4
The mass of refrigerant vapor in the separator can be conservatively estimated using:
where 𝜌𝜌𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 is the density of vapor refrigerant at the compressor discharge pressure and temperature.
Further refinement of the volume to account for dished heads and oil volume resident in the oil
separator can be done as previously discussed in the section on vessels.
Thermosiphon Oil Coolers
Thermosiphon oil coolers utilize high pressure saturated liquid as the heat sink for cooling oil on a screw
compressor package. The oil cooler usually consists of a shell-and-tube heat exchanger where oil
circulates on the shell-side while high pressure liquid refrigerant is supplied to the heat exchanger’s
tube-side. As the liquid refrigerant absorbs heat from the oil, it evaporates and the vapor returns to the
thermosiphon pilot receiver. During normal operation, the return vapor may entrain some liquid
refrigerant carried over from the cooler resulting in a mixture of liquid and vapor returning to the pilot.
Determining the refrigerant inventory of thermosiphon oil coolers is similar to evaporators. Most
compressor manufacturers publish data for both refrigerant and oil inventory for the oil coolers
integrated into their compressor packages as part of the technical documentation that is included with
the compressor purchase. Unlike evaporators, the thermosiphon oil cooler charge is there regardless of
whether there is an oil cooling load due to compressor operation.
Figure 17 is a technical manual excerpt from one compressor manufacturer’s product information. The
table shows the range of shell-and-tube oil coolers that are available for the entire series product line.
The refrigerant inventory for an oil cooler is found by multiplying the listed refrigerant volume for that
oil cooler model by the average density refrigerant in the oil cooler evaluated at the compressor’s
discharge pressure. The average density for the refrigerant is given by Equation 10 where C is
determined from Table 3 assuming a flooded heat exchanger. A maximum refrigerant charge for the oil
cooler would assume the entire refrigerant-side volume is filled with saturated liquid refrigerant at the
discharge pressure.
Page 28 of 66
Figure 17 – Thermosiphon oil cooler data (GEA 2004).
For a compressor, saturated liquid ammonia at 181 psig has a density of 36.67 lbm/ft3. The maximum
refrigerant inventory for a model 607 oil cooler would then be the product of the listed refrigerant
volume (0.5 ft3) and the refrigerant density resulting in an estimated operating charge of 18.24 lbm.
Recently, plate-type heat exchangers have found growing application for thermosiphon oil cooling. A
primary advantage of the plate-type heat exchanger is its compact size (which leads to low operating
charge) for a given oil cooling load. For example, a plate-type heat exchanger capable of rejecting 881
mBh of oil cooling load has a refrigerant-side internal volume of approximately 0.2 ft3. If the oil cooler in
this example is assumed to be completely filled with saturated liquid at a design condensing
temperature (181 psig), the refrigerant charge would be just over 7 lbm (the product of 0.2 ft3 and the
refrigerant density of 36.67 lbm/ft3).
Thermosiphon Piping
The methods presented in the piping portion should be applied to estimate inventory of thermosiphon
piping. The refrigerant state in the thermosiphon supply piping will be saturated liquid while the
thermosiphon return piping may contain a mixture of liquid and vapor.
The refrigerant inventory for the thermosiphon supply piping can be estimated using Equation 18-
Equation 20 or Equation 21. The refrigerant inventory for thermosiphon return piping can be estimated
by Equation 18-Equation 20 where the refrigerant density for Equation 20 is estimated using Equation
22 - Equation 24. The ratio of liquid mass to vapor mass in the return piping has been estimated at 3:1
(FES, 1998 and Welch, 2003); consequently, a value for 𝑂𝑂𝑂𝑂 in Equation 24 can be assumed to be 3 for
inventory estimating purposes.
Page 29 of 66
To illustrate component-level inventories for different types of system Table 7, Table 8, and Table 9
show the inventory breakdown by equipment of three industrial refrigeration systems of varying sizes.
Despite the total inventory of these systems ranging from 7,200 lbm to nearly 132,000 lbm, there are
notable similarities. Refrigerant in piping accounts for 16-28% of the inventory, vessels 47-49%,
evaporators 7-26%, and condensers 6-14%. It should be noted the air-cooling evaporators in the plant in
Table 9 consisted mostly of flooded units, which have the highest charge of the common evaporator
types.
Page 30 of 66
Table 9 – Charge breakdown by equipment of a large system.
Equipment Type Number Inventory Percent
Compressor 15 287 lbm 0%
Condenser 4 8,368 lbm 6%
Evaporator 400 33,642 lbm 26 %
Pipe Section 950 30,932 lbm 23 %
Horizontal Vessel 191 47,012 lbm 36 %
User Specified Inventory 7 128 lbm 0%
Vertical Vessel 147 11,521 lbm 9%
Some vessels are equipped only with a sight glass column for determining the level of liquid refrigerant.
For a vertical vessel, the charge difference between each sight glass is constant. Horizontal vessels, on
the other hand, vary depending on the level as shown in Table 10. For example, the system in Table 7
(charge of 7,200 lbm) has a 10 ft long, 4 ft diameter HPR. The difference in charge between two sight
glasses varies from under 300 lbm to nearly 700 lbm or 4%-10% of total system charge. In reality, the
distance between the very top of a glass to the bottom of the next is only about 4.5 inches, leaving a
worst-case inventory difference of 520 lbm which corresponds to 7% of total system charge! In this case
periodic checking of the HPR level could reveal a time when the vessel level falls within a sight glass.
Taking a system snapshot at this time would provide a less uncertain inventory.
For example, under typical conditions the 24” sight glass is consistently full of liquid and the 30” sight
glass is vapor. When a chiller with an estimated refrigerant charge of 300 lbm is not active and contains
only vapor, the HPR level rises to within the 30” glass. The entire system inventory can them be taken at
this condition with a more accurate HPR charge.
Table 10 – Refrigerant charge at each sight glass of a horizontal 10 ft long, 4 ft diameter vessel
with 2:1 ellipsoidal heads with 6 inch glass spacing.
Page 31 of 66
Vessels often have associated attachments (e.g. level columns, float switches, etc.) which should be
included in the charge calculation if possible. If these associated attachments are neglected, there will
be minimal changes in total system charge. Given the inherent uncertainty in calculating the internal
volume of the vessel due to internal features and components there is some balancing between external
liquid containing attachments and internal feature volume.
If desired, a great deal of time and effort can be expended measuring refrigerant piping in the course of
inventory calculation, particularly for vapor piping. For example, the system in Table 7 utilizes a flooded
chiller on the roof fed by a 1.5” recirculated liquid line and a 3” suction line. The lines are ≈150 ft long
with volumes and charges shown in Table 11.
In this case the physical vapor and liquid lines are different lengths due to field routing. For the
purposes of the charge calculation, the liquid line length was quantified using a measuring wheel, and
the same measurement was equally applied to the vapor line running in parallel with the liquid line.
Even if the vapor line was within 50% of the assumed length, the system inventory changes by less than
one hundredth of one percent. Similarly, this technique can be used in instances where there are
numerous similar evaporators in each area. Measure or closely estimate the liquid piping, and simply
apply that same length measurement for the vapor piping.
A large amount of vapor piping can be neglected outright. Examples include pump-out lines, purge
lines, equalizers, and small bypass piping will calculate out to tenths of a pound. For most systems,
lengths of piping containing less than 0.5 lbm of refrigerant can safely be neglected without a significant
sacrifice in overall estimate of the system charge. For liquid this equates to ≈0.013 ft3, for condensing
pressure vapor ≈0.4 ft3, and evaporator pressure vapor is ≈5 ft3. For reference, 5 ft3 is nearly 100 ft of 3”
sch 40 piping.
Page 32 of 66
Table 12 – Breakdown of refrigerant piping inventory for the system shown in Table 7.
Piping
Length 733.0 ft
Vapor Volume 32.4 ft³
Charge 7.3 lbm
Length 1,491.5 Ft
Liquid Volume 27.9 ft³
Charge 1,066.5 lbm
Length 30.0 ft
Condenser Drain Volume 1.5 ft³
Charge 14.5 lbm
Length 826.0 Ft
Recirculated Suction Return Volume 39.6 ft³
Charge 23.4 lbm
Length 3,080.5 ft
Totals Volume 101.4 ft³
Charge 1,111.7 lbm
For example, Table 12 shows the breakdown of refrigerant inventory for the system in Table 7. Nearly
96% of the total piping charge resides in liquid piping. Similarly, liquid piping for the system shown in
Table 8 comprises 93% of the piping charge.
The two largest factors when determining heat exchanger charge are type and internal volume. Type
(DX, overfed, flooded, condenser) is usually easy to determine visually. It is important to verify the feed
type of evaporators which may be in question as the type of liquid feed makes a large difference in the
estimated charge, as shown in Table 13.
Table 13 – Charge of a 5 ft3 evaporator with various liquid feed methods and 32°F ammonia
evaporating temperature.
Volume Operating Maximum
Descriptor Type
(ft3) (lbm) (lbm)
Direct Expansion Evaporator 5 8.6 199.3
Overfed 3:1 Evaporator 5 50.9 199.3
Flooded Evaporator 5 100.7 199.3
Determining the total volume of a newer evaporator can be as simple as looking the appropriate value
up in the manufacturer’s literature. For older equipment techniques can be used, such as counting U-
bends to estimate the total length of tubing in the evaporator, and then calculating the internal volume.
This is worth pursuing for overfed and flooded evaporators containing a large refrigerant charge.
Page 33 of 66
Similarly, working through product documentation for chillers can pay dividends with a better
calculation as the internal volume of chillers are hard to estimate, and they can have large operating
charges.
System Operation
The total inventory of a refrigeration system is the summation of the estimates of refrigerant inventory
for each component/subsystem that makes up the system. Keep in mind that refrigerant in the liquid
state will usually account for more than 98% of the total charge of the system with two-phase mixtures
of liquid and vapor accounting for an additional 1½%. In doing these calculations yourself, you will likely
find that vapor accounts for less than 1%.
Consider developing a summary table of refrigerant for your system. Table 14 provides a simple
example of the minimum information that should be included in a system summary of refrigerant
inventory. The “Description” column breaks out each of the main components for the system and
provides the component inventory estimate (second column) and a running total of cumulative
inventory (third column).
Page 34 of 66
Table 14 - Summary table of refrigerant inventory for a system.
ABC Foods, Madison, WI USA
Component Cumulative
Equipment Description
Inventory (lbm) Inventory (lbm)
Vessels
High pressure receiver (HPR-011) 6,200 6,200
Intercooler (IC-050) 4,250 10,450
Low pressure accumulator (LPR-150) 2,725 13,175
Low pressure accumulator (LPR-152) 2,840 16,015
Transfer station (TS-225) 600 16,615
Booster compressor oil separators (LS-101, LS-102) 30 16,645
High stage compressor oil separators (HS100-HS105) 570 17,215
Condensers
Evaporative condenser (EC-001) 2,640 19,855
Evaporative condenser (EC-002) 3,180 23,035
Evaporative condenser (EC-003) 1,200 24,235
Evaporators
Air-cooling evaporators (AC-205 thru AC-225) 840 25,075
Air-cooling evaporators (AC-300 thru AC-330) 990 26,065
Chiller (CH-510) 1,808 27,873
Chiller (CH-511) 2,480 30,353
Chiller (CH-512) 1,808 32,161
Piping
High pressure liquid piping (450 ft HPL) 341 32,502
Low temperature pumped liquid (300 ft @ -20 LTRL) 648 33,150
Low temperature pumped liquid (360 ft @ -40 LTRL) 1,372 34,522
Wet return piping (660 ft LTRS-1, LTRS-2) 280 34,802
Plant Total 34,802
The ideal inventory calculation is a snapshot in time of every vessel level, every piece of equipment state
(i.e. operational status), system pressures and temperatures, etc. Practically speaking, this can be
difficult to achieve so judgment must be exercised to arrive at an accurate system inventory calculation.
A good starting point is to measure the system piping and determine its inventory. The majority of
piping will be at a relatively constant inventory level. For example, many lines do not vary in refrigerant
phase and therefore operating charge, particularly the liquid lines. Focus attention on measuring the
length and size of liquid piping within your acceptable uncertainty. Often, vapor and two-phase piping
lengths can be approximated from the liquid line length measurements as they often run together.
Vessels with liquid level-controls will have nearly constant inventory in normal operation, those can be
added to the charge at any time. It is important to consider any conditions which could result in
controlled-level vessel liquid levels that are not at the fill location. Evaporator defrost, component
Page 35 of 66
pump-outs for maintenance, or load swings can result in liquid levels in excess of the fill location.
Vessels with controlled levels typically include pumped recirculators and evaporator surge drums.
The remaining inventory of the system will likely be dependent on individual components. For example,
if a large process evaporator is shut down and pumped out for cleaning the ammonia typically occupying
the unit for operation will end up in the high-pressure receiver. Therefore, for highest accuracy the
high-pressure receiver level should be taken when as many of the largest loads have a known status and
preferably operational. Computer control systems can simplify this process, allowing a complete
snapshot of process status and vessel levels. For systems lacking this instant snapshot the individual
must use their best judgement as to plant operation when the uncontrolled vessel liquid levels are
measured.
Once all piping and vessel charges have been determined, the final step is to determine the charge of
equipment as it was during the “snapshot” of the uncontrolled level vessel(s). Incorporating system
operating status while calculating inventory is an important factor in achieving the best possible result,
especially in systems with changing and unsteady operations.
The origin of refrigerant losses include accidental leaks & spills, intentional venting/disposal prior to
conducting planned maintenance, fugitive losses/emissions, and intentional theft. Jordan (2020)
compiled data on accidental releases of ammonia from a range of sources. One of Jordan’s conclusions
was that loss of mechanical integrity due to faulty or poorly maintained equipment was the enabling
factor in most incidents and accidents compiled as part of the author’s efforts. Interestingly, Jordan
found the second most frequent category of accidental releases related to line opening and oil draining
maintenance activities. Both categories of accidents can be blunted by implementing improved
mechanical integrity principles and practices as outlined by Reindl (2016) and IIAR (2019b).
When accidental leaks & spills do occur, the magnitude of refrigerant loss can be estimated using the
techniques such as those of Reindl and Jekel (2016). One type of accidental leak does occur and results
in comparatively large losses over time that go undetected are pinhole leaks in evaporative condenser
Page 36 of 66
tubes. In this case, water sprayed over the condenser tube readily absorbs the released ammonia so no
odor is evident downwind. When automated chemical treatment systems are used for condenser water
treatment, the automated system typically responds to the increased condenser water alkalinity by
injecting acid for pH balance. Increased consumption of water treatment chemicals is often a telltale
sign of condenser tube leaks.
In cases where refrigerant is intentionally discharged from the system during course of maintenance
activities, the quantity discharged is rarely calculated or estimated. The total quantity of ammonia
discharged from a given system on an annual basis will depend on the nature of maintenance being
conducted and the frequency. Ideally, refrigerant should be pumped out of the portion of the system
that is the target for maintenance and transferred to the portion of the system that will remain
operational. This is easily accomplished in cases where systems are designed with “hard-piped pump-
out systems” but not as easily accomplished when technicians need to make temporary connections to
transfer the refrigerant. Because ammonia is inexpensive, compared to fluorochemicals, most plants
take the path of safely discharging the refrigerant from the system.
By its very nature, fugitive emissions go undetected and, as such, plants do not quantify the losses
because they are unaware of the occurrence. Fugitive emissions can occur through valve stems,
gasketed connections, sight glasses, shaft seals, fittings, etc. These leaks can be below the odor
threshold and go unnoticed for extended periods of time, but these losses, collectively over time,
generally do not accumulate to an appreciable amount on an annual basis.
For more than a decade, plants using anhydrous ammonia as the refrigerant have been targets for
ammonia theft. Because ammonia is one of the key ingredients used in clandestine operations involving
the illegal manufacture of methamphetamine (N-methyl-1-phenylpropan-2-amine), it is sought after. If
substantial quantities of ammonia are “missing” from a given system and there has not been a history of
accidental releases or significant maintenance activities that might explain the loss, plants should
consider whether their ammonia refrigeration systems could be a victim of theft. Securing portions of
the system that may be accessible to the public and provide points for refrigerant removal will reduce
likelihood of theft from individuals outside the organization. More insidious is the theft of ammonia by
employees. Eliminating this as a possibility often involves surveillance of those locations where
refrigerant can be removed from the system and then transported off-site.
Apart from theft, it appears that the two greatest sources responsible for refrigerant losses from
industrial ammonia refrigeration systems over time are accidental releases and discharges associated
with maintenance activities. This conclusion is based on our findings in the field during the course of the
EPA-funded P2 project that explored fugitive emissions from industrial refrigeration systems (Reindl, et
al. 2020), where we identified comparatively infrequent instances of fugitive emissions and, when
discovered, quite low loss rates. The key to reducing substantial ammonia losses from these systems is
to first focus on managing the mechanical integrity of the system to reduce the frequency and
magnitude of uncontrolled leaks and spills and then take steps to eliminate intentional releases
associated with maintenance activities.
Page 37 of 66
serve as a basis for individual plants to compare their own performance annual loss rates. Plants having
annual losses in excess of the high end of the benchmark can critically review and assess their internal
practices to adjust their practices that may be contributing to excessive losses and revise their
procedures accordingly. Thereafter, they can proactively screen for fugitive emissions and repair them
as they are found. A systematic process for continuous improvement will drive annual refrigerant losses
down until they exceed the “best” benchmark threshold. This approach necessitates establishing an
accurate estimate of the system’s refrigerant inventory and this goal can be accomplished by applying
the principles and techniques provided in this guidance document.
Table 15 provides an example of refrigerant addition or “fill tracking” beginning with the initial system
charge as part of the system’s post-construction start-up and moving forward as additional refrigerant is
added to the system. If an initial refrigeration system refrigerant fill quantity is not known or records
don’t extend back far enough, the detailed refrigerant charge calculation using the methods presented
above can be used to establish the equivalent “System Initial Fill).”
Because industrial refrigeration systems frequently undergo changes that may include expansions or
equipment retirement, the fill tracking method has to be modified slightly to separate out the
refrigerant quantity associated with equipment additions or retirements. If ignored, the expansion of a
system will yield a “false positive” signal that refrigerant losses or emissions have increased. Conversely
if the refrigerant inventory associated with decommissioned equipment is not quantified, the top-off
tracking will yield a “false negative” signal that emission have remained constant or decreased.
Page 38 of 66
The best loss refrigerant estimate is made when refrigerant is added to the HPR to a consistent fill level.
For example, “add liquid ammonia to the HPR until the liquid height reaches the third sight glass with
more than three quarters of the loads running in the plant.” Then the amount of ammonia added to the
system is consistent, for example the amount needed to bring the HPR from 3 sight glasses to 5 sight
glasses. Situations where refrigerant is added at different HPR levels, and the amount added is not pre-
planned based on calculations can result in ammonia addition trends which do not reflect the amount of
ammonia leaving the system.
In the event of an expansion, a revised estimate of the system’s maximum intended refrigerant
inventory must be established. In the example shown in Table 16, the single fill of 4,560 lbm needs to be
broken into two components, one component that replenishes refrigerant due to losses and one
component to accommodate system expansion. In order to break up the refrigerant fill or addition, a
detailed required refrigerant charge estimate must be conducted for the expanded portion of the
system. Without this split-out, the apparent loss rate from the system increases from 569 lbm/yr to
1,385 lbm/yr. It is also important to update the calculated system inventory during system changes to
more accurately portray refrigerant losses from the system as a percentage of the system’s maximum
intended inventory (top-off amount), and break out loss estimates post-expansion separately.
Tracking refrigerant additions over time as a trailing measure of refrigerant losses has the potential to
provide plants information if an uptick in refrigerant loss seems to be occurring but this approach
requires data over more than a single refill event to establish the trend. Unfortunately, this process can
take many years which is undesirable because the goal is to begin better managing refrigerant losses
now.
Page 39 of 66
Estimating Aggregate Refrigerant Losses by Dynamic
Refrigerant Inventory Calculations
Another method of estimating total refrigerant loss with the potential to reduce the time lag associated
with fill tracking is to perform a “dynamic charge calculation” for the refrigeration system. A dynamic
charge calculation begins by dividing the entire refrigeration system into two parts: “controlled
refrigerant inventory” and “fluctuating refrigerant inventory” as shown in Figure 18. The “controlled
refrigerant inventory” portion of the system is assumed to be constant over time and assumes no
equipment is added or removed from this part of the system (or corrections are required). Plant
personnel then longitudinally track the refrigerant quantity in the “fluctuating refrigerant inventory”
portion of the system. Typically, this involves tracking the refrigerant inventory in the system’s high-
pressure receiver which is often the single largest refrigerant-containing subsystem in the “fluctuating
refrigerant inventory” portion of the system because its liquid level is not directly controlled. Because
the high-pressure receiver refrigerant inventory will naturally fluctuate, no conclusions are made on a
single level observation; rather, refrigerant inventory data for the high-pressure receiver (“fluctuating
refrigerant inventory” portion of the system) are gathered longitudinally over several days and trended.
Evaporative Evaporative
Condenser Condenser
Discharge line
Machinery Room
Plant/Process Area
Pumped
recirculator
4’
Compressor(s)
Page 40 of 66
loss rate based on the actual longitudinal fill quantity for Plant 1 was 496 lbm/yr while the estimated
loss rate of 550 lbm/yr developed by trending the dynamic charge calculation is within 11%. With a
maximum intended inventory of ammonia for Plant 1 at 7,500 lbm, the actual loss rate as a percentage of
the system charge is 6.6%. It is also noteworthy to observe that individual observations of the HPR
inventory can and will fluctuate so trending over time is essential.
2500
2000
Vessel Charge (lbm)
1500
1000
Average Losses: 550 lbm/yr
y = -1.548x + 68396
(7.3%/yr)
500
0
5/25/2017 8/8/2017 10/22/2017 1/5/2018 3/21/2018 6/4/2018
Figure 19 – Tracking HPR levels over time to determine total refrigerant losses (Plant 1).
Larger systems with multiple uncontrolled level vessels can also use this technique; however, the
inventory for these uncontrolled level vessels must be aggregated for the purpose of trending
refrigerant losses for the system. For example, Figure 20 thru Figure 22 shows plots of dynamic
inventory calculations from an actual plant that has two separate refrigeration systems, System 1 and
System 2. In this plant, each system has its own high-pressure receivers “HPR 1” and “HPR 2”,
respectively. Although the systems are capable of operating independently, they are interconnected
and share a common refrigerant charge.
The dynamic refrigerant inventory determination for System 1 using the observed inventory levels for
HPR 1 appears to show the system is actually gaining refrigerant at the rate of 1,000 lbm/year (shown in
the figure as a negative loss). Of course this trend, taken in isolation, is not technically feasible without
inbounding refrigerant for refill or “top-off.” In contrast, the trending of inventory for System 2 using
the fluctuating inventory of HPR 2 indicates the system is losing refrigerant at the rate of 5,000 lbm/yr.
Because these two systems are interconnected and share a common refrigerant charge, the correct
approach for trending is to aggregate the refrigerant inventory for both HPRs. When the inventory for
both HPRs are aggregated, the combined inventory paints a different picture and now shows an overall
total emissions estimate that is plausible at 4,100 lbm/yr as shown in Figure 22. With type of system
configuration, it is important to observe and log the inventory close together in time to avoid refrigerant
Page 41 of 66
migration between the system biasing the results. In the example of Figure 22, the information was
recorded by hand during the plants normal “rounds,” conducted twice per day. This information could
also be collected with a computer data logging system.
Figure 20 – Dynamic inventory calculation for System 1 using fluctuating level of HPR 1.
Figure 21 – Dynamic inventory calculation for System 2 using fluctuating level of HPR 2.
Page 42 of 66
Figure 22 – Dynamic inventory calculation that aggregates both HPR 1 and HPR2 for the
purposes of estimating refrigerant emissions.
Instead of aggregating the data point-by-point as shown in Figure 22, the two trendline slopes can be
combined to achieve the same net result. Taking these slopes together yields: -14.26 lbm/day + 3.05
lbm/day = 11.2 lbm/day. Over the course of a year this equates to 4,090 lbm/yr, which is then rounded to
the value of 4,100 lbm/yr as shown in the combined chart. For this particular plant, the average
ammonia fill rate was actually 2,350 lbm/yr over a 10-year period. As a result, the dynamic charge
calculation approach is overpredicting the apparent losses. There are several possible explanations for
the discrepancy. The plant could have added loads or condenser capacity over time, distributing more
refrigerant in the “controlled inventory” portion of the system and away from the “fluctuating
inventory” part of the system e.g. the high-pressure receiver. The plant could also be running its
existing equipment at higher refrigeration loads or more frequently which would tend to migrate a
greater amount of refrigerant to the controlled inventory part of the system, causing less average
refrigerant charge in the HPR. It is also possible the plant is simply replacing less refrigerant than they
are losing. Detailed knowledge of system changes would need to be analyzed to answer these
questions. It is easier to note these changes if and when they do occur, rather than attempt to analyze
them after the fact.
Depending on the rate of refrigerant losses and system load fluctuations, the general trend of losses can
take several weeks or months to provide a credible trend.
Page 43 of 66
Quantify Known Losses
By quantifying known losses and tracking refrigerant additions for “topping-off” the system, the facility
can better determine how much of the lost refrigerant is by fugitive emissions. Once those fugitive
emissions are quantified a program can be developed and implemented to reduce them.
Reindl & Jekel (2016) provide some basic methods to estimate refrigerant quantities released as a result
of incidents and accidents. For example, Table 17 illustrates the mass flow rates of refrigerant
discharging from a severed ½” schedule 80 pipe for both vapor-only and flashing liquid refrigerant over a
range of pressures. If a ½” schedule 80 hot gas (vapor) line is severed when operating at a pressure of
150 psig, the corresponding vapor mass flow rate would be 19.2 lbm/min. Clearly, the persistence of this
leak prior to mitigation can result in a substantial quantity of refrigerant being lost from the system. If
the same size pipe were carrying saturated liquid ammonia at 150 psig, the leak rate jumps to 170
lbm/min!
Table 17 - Refrigerant release rates (excerpted from Reindl & Jekel, 2016).
Opening Size Pressure Saturation Temp Vapor Flow Flashing Liquid Flow
Page 44 of 66
Refrigerant Releases into Liquids
Refrigerant emissions also have the potential to escape into water, or other secondary fluids. One
example note above is the evaporative condenser. Leaks of ammonia into the condenser water can go
unnoticed as the water flowing over the outside of the condenser tubes will quickly absorb the leaking
ammonia, eliminating any evidence (odor) that an active leak is occurring. In this case, monitoring for
unexpected increases in condenser water PH can identify potential leaks since ammonia and water
create a basic solution. If the water treatment system is equipped with an automated acid feed system
to balance PH, increased consumption of bulk acid should alert facility staff that a problem exists and a
potential ammonia leaks on the condenser(s) may be ongoing. In colder climates the condenser will
likely run without water during winter months, making this a good time for leak checking the
condensers.
Evaporative condensers spray water over a large serpentine coil heat exchanger (see Figure 23(a)) to
enhance the rejection of heat from the refrigerant to the ambient environment. If a pinhole leak
develops in the condenser heat exchanger’s tubing, the water circulating over the heat exchanger will
readily absorb the leaking ammonia all but eliminating an air release; thereby, masking any overt
indication (odor) that would otherwise be evident to refrigeration personnel conducting normal rounds
when they are downwind of the condenser. Figure 23 (b) and (c) show a field installation of an
evaporative condenser that developed a pinhole leak at the top of the top-row of tubes. The water
sprayed over the condenser’s heat exchanger readily absorbed the ammonia and not odor was sensed
downwind of the leaking condenser leading to the loss of ammonia that persisted for an extended time
period.
Depending on the site-specific condenser water treatment regimen, automated chemical feed systems
can further mask the refrigerant leak by neutralizing agents used for balancing increasing condenser
water pH. These factors can allow this type of accidental release to go undetected for days or weeks.
Because of the long period of time before plant personnel become aware of the leak, accurately
quantifying the refrigerant loss become nearly impossible yet the total quantity of refrigerant lost in this
type of accidental release can be significant. Preventing the occurrence of this unknown and
unquantified requires effectively managing the mechanical integrity of the condenser from its initial
installation through its operational life. The use of the dynamic system inventory calculation method
described previously is a resource that can potentially alert plant personnel to such a leak earlier.
Page 45 of 66
Water spray header
(a) (c)
(b)
Figure 23 - Evaporative condenser (a) newly fabricated heat exchanger, (b) field-installed
condense with suspected tube leak (c) actual tube leak location.
Similarly, ammonia leaking into a secondary fluid such as brine or glycol can also be identified by
sampling and analyzing the secondary fluid for ammonia. Evidence of leaking should be investigated
and addressed expediently.
Quantifying these emissions is best done by calculations based on the operating conditions and hole size
as discussed by Reindl and Jekel (2016).
Ideally, the technicians will “pump down” the portion of the system being serviced and recover or return
the ammonia refrigerant back into the operating part of the system; however, it is quite common for
technicians to simply vent the ammonia directly to atmosphere or discharge the ammonia into a
Page 46 of 66
reservoir of water that for later treatment/disposal. The combination of low refrigerant cost, low
environmental footprint (zero ODP and zero GWP), and the added time to recover/return refrigerant to
the system are factors that tend to prevent reversing the practice of simply venting the ammonia during
servicing.
How much ammonia refrigerant is vented during maintenance? It depends on the subsystem being
accessed including the volume of the subsystem, state of refrigerant within the subsystem, and final
pressure in the subsystem prior to initiating venting. Figure 23 shows an illustration of an overfed
evaporator with two valve groups that would be typical for requiring periodic maintenance: liquid supply
and suction.
hot gas
Suction Stop
Valve
Recirculated Suction
Liquid Feed
Solenoid
Pumped
Liquid Supply
Defrost condensate
Defrost relief
regulator
Recirculated liquid/vapor
suction
Figure 24 - Illustration of an overfed evaporator with key valve trains highlighting key
components associated with the liquid supply valve train and suction valve train.
Figure 24 shows a close-up of the liquid feed valve train that consists of an upstream isolation valve,
strainer, solenoid valve, hand-expansion valve, check valve, and downstream isolation valve. A pump-
out connection is shown between the hand-expansion and isolation valve on the downstream side of
the valve group. Also shown in Figure 24 is the internal volume for each component for 3/4"-1” valves
as well as an allotted 1 lineal foot of 1” schedule 80 pipe interconnecting the valves in the group.
Assuming the valve group is isolated and pumped down to a pressure of 25 psig and a residual quantity
(5% by volume) of liquid ammonia remains just prior to venting, the mass of ammonia that would be
vented from the system for a given service instance is estimated at 0.13 lbm.
Page 47 of 66
Figure 25 - Illustration of the liquid feed valve group and associated refrigerant quantity that
would be vented during service.
Similarly, Figure 25 shows the suction valve group for the evaporator illustration shown in Figure 23. On
the suction side of the evaporator, there are fewer components that comprise the valve group. Also,
the pipe/valve size is significantly larger due to the refrigerant being, principally, in a vapor state.
Assuming the valve group is pumped down to 25 psig and a residual quantity of overfed liquid at 1%, the
mass of ammonia that would be vented from the system during a given service instance is estimated at
0.39 lbm. The quantity of these two valve groups combined would vent just over a half-pound of
refrigerant in each service instance.
Page 48 of 66
Figure 26 - Illustration of the suctions-side of an overfed evaporator and the associated
refrigerant quantity that would be vented during service.
Although losses of refrigerant through venting during maintenance activities are comparatively small,
those losses become increasingly significant as the number of units in each system increase and the
frequency of maintenance activity for those units increases. Further, it is possible to apply these simple
first principles to estimate in advance the quantity of refrigerant that would be vented during planned
service. Accounting for refrigerant losses due to maintenance will provide needed data for plants to
justify recovering/re-using the refrigerant rather than venting.
Table 18 shows a further example of losses with from an evaporator valve group. The liquid valve group
is comprised of a strainer, solenoid, hand-expansion valve, check valve, and 12” of piping between the
two isolation valves on either end of the valve group. The suction valve group is comprised of a
regulator, suction stop valve, and 8” of piping between isolation valves. In both cases, assumed the
valve group volume has been pumped out to a 25 psig suction pressure. Since no pump-out is perfect,
the remaining liquid fraction is assumed to be 0.1 and 0.01 in the liquid line and suction line,
respectively.
Page 49 of 66
Table 18 – Ammonia losses of various line sizes for evaporator liquid feed and suction when
pumped out to 25psig suction.
Evaporator valve Liquid line Suction line
group size purge loss purge loss
(nominal inch) (lbm) (lbm)
1/2"-5/8" 0.06 de minimis
3/4"-1" 0.26 0.03
1-1/4" 0.51 0.05
1-1/2"-2" 1.40 0.14
2-1/2" 2.41 0.24
3" 3.29 0.33
4" 5.08 0.53
5" 6.08 0.62
As noted above, we recommend facility staff maintain documentation and a running total of
“maintenance losses” that is reset when refrigerant is added to the system for top-off. The running
total maintenance loss can be subtracted from the top-off amount to provide a better estimate of the
refrigerant loss from accidental releases, fugitive emissions or other refrigerant loss pathways. By
evaluating known refrigerant losses, procedures can be evaluated and modified as needed to reduce
maintenance and other losses.
Page 50 of 66
condensing pressure during normal operation, and the purge cycle, when non-condensables and
ammonia are sent to the water bubbler, should last between 10 and 45 seconds. Possibly longer if there
is a large amount of air in the system, but this should be rare.
0.3
Ammonia Mass Fraction of Purge Gas
0.2
0.15
0.1
Page 51 of 66
1
Ammonia Mass Fraction of Purge Gas 45 Psig Suction Malfunctioning Purger (0.025 lbm nh3/min)
0.9
0.8
0.7
0.6
0.5
Page 52 of 66
Figure 29 – A high ammonia concentration purge is shown on the left, an normal non-
condensable gas purge on the right.
Once the frequency and persistence of these small “noticeable odor” leaks have been reduced, the next
step is to implement strategies to finding and quantifying small “fugitive” leaks as outlined below. These
methods are similar to those routinely used in the petrochemical industry, and are detailed in
documents such as the “Protocol for Equipment Leak Emission Estimates” (EPA, 1995).
Screening Components
The first, most basic method of identifying small refrigerant emissions is by screening components.
Screening involves using an ammonia detector, preferably with an onboard sampling pump and probe,
held closely to potential leak sites (gasketed connections, screwed connections, stem packing, etc.). This
approach is intended to identify any location with detectable ammonia levels. An example setup is
shown in Figure 26 where screening a sight glass is taking place by carefully traversing the detector’s
sampling probe around the face and then retaining ring of the sight glass.
Page 53 of 66
Figure 30 – Screening a sight glass with an ammonia detector.
Figure 27 shows the detector’s sampling probe being moved slowly along the gasketed portion of a
compressor housing in an effort to pinpoint a suspected leak sites. Refrigerant detectors vary, in terms
of response time, so care must be taken to avoid moving the probe too quickly and overrunning an
active leak site. Response time is defined as the amount of time for the detector to read 90% of a
known concentration value after initial exposure (EPA 1995). Based on experience, a detector with a
response time of 30 seconds drawing gas flow at a rate of 1 ft3/hr can detect most small leak rates
(≈0.01 lbm/yr) while the sampling probe is traveling ½ inch per second. It is also important to recognize
that bulk air flow in the neighborhood of screening a component can mask small leaks. Each refrigerant
detector is unique, and experimentation with a planned screening setup is required. The screening
process can be quite slow for larger items. Sometimes when a reading appears on the detector, the
probe must be moved back along the path of travel to better pinpoint the actual origin of the leak site
before confirming the final reading. To determine the final screening reading, stop the probe on the
suspected leak sight and remain motionless for the detector’s stated response time; the highest
indicated ammonia reading is then recorded.
Page 54 of 66
Figure 31 – Screening a compressor housing.
The screening method can also provide a lower-bound estimate of the leak rate. Using the method
outlined in Table 4-1 of “Protocol for Equipment Leak Emission Estimates” Equation 27 can be used if
the volumetric flow through the detector is consistent and known.
where:
Page 55 of 66
14.75
14.5 95°F
75°F
14.25 55°F
Local Pressure [psia]
35°F
14 15°F
13.75
13.5
13.25
13
12.75
12.5
12.25
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Local Altitude [ft]
12.5
12.25 95°F
75°F
12 55°F
Local Pressure [psia]
35°F
11.75 15°F
11.5
11.25
11
10.75
10.5
10.25
10
5000 5500 6000 6500 7000 7500 8000 8500 9000 9500 10000
Local Altitude [ft]
Figure 32 - Fundamental relationship between elevation and ambient air pressure.
Page 56 of 66
The atmospheric pressure at different elevations can be determined from the plots in Figure 28 or from
Equation 28.
5.257
𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙
𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 (𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝) = 𝑃𝑃𝑠𝑠𝑠𝑠𝑠𝑠 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ∙ �1 − 500 �
(𝑇𝑇𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 − 32) ∙ 5 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 Equation 28
9 + 273.2 +
500
where:
Psea level is the atmospheric pressure at sea level, psia (typically 14.7 psia)
Example:
The ammonia detector shown in Figure 27 draws a gas flow rate of 1 ft3/hr through the probe and the
measured concentration of ammonia emanating from the compressor housing is 117 ppm. The local
atmospheric pressure is 14.3 psia and the ambient temperature is 75°F. The estimated fugitive emission
is then:
0.000817 ∙ 1 ∙ 17.03 ∙ 117 ∙ 14.3 𝑙𝑙𝑙𝑙𝑚𝑚 Calculated fugitive emission
= 0.044
75 + 459.7 𝑦𝑦𝑦𝑦
In this case, the estimated leak rate from the compressor housing gasket is 0.044 lbm/yr. Clearly, the
leak rate is small and a very large number of similar sources would be needed to total a loss rate in the
single digit percentages for most refrigeration systems.
While the odor threshold for ammonia varies between individuals, most can detect its presence at 5
ppm and this concentration would correspond to an ammonia leak rate as low as 0.05 lbm/yr assuming
the person is within a 2 ft radius for an indoor source. Using the equipment and conditions in the above
example, this leak rate would equate to a screening reading of ≈135 ppm (keeping in mind this high
concentration because the probe is very close to the leak source while the odor being sensed by the
individual may be several feet away from the leak source). This information can be used as a means of
collaborating screening findings, since higher screening readings should be expected when ammonia
odor is present. Conversely, larger leaks on the order of 1 lbm/yr or greater can easily go unnoticed
outdoors, particularly when operators may be doing their rounds in locations such as on the roof during
windy conditions. Oftentimes simply walking by a leak site is not enough to detect it, and periodic
screening can assist with locating refrigerant leaks. For areas that are not within an easy reach for
periodic screening, tools such as the retractable pole shown in Figure 29 can be used to bring the
ammonia detector to the proximate location being screened to identify if refrigerant is actively leaking.
Page 57 of 66
In this case, the detector is configured to log data every 5 seconds, which can be read later on a
computer. Done periodically, this practice can find leaks when they are few and small.
Bagging Components
The most practical means of measuring the fugitive emission rates from a piece of equipment is by
bagging. This method allows establishing estimates of the fugitive emission leak rate by completely
enclosing the expected or suspected leak site origin using an impermeable bag and then drawing gas
from within the microenvironment created by the across an ammonia detector as illustrated in Figure
30. The gas flow rate and sensed ammonia concentration are monitored as a function of time. For small
leak rates, the on-board sampling pump of the ammonia detector can be used directly. Because many
hand-held ammonia detectors have an upper detection limit of 1,000 ppm, this “direct bag sampling
method” is limited to leak rates on the order of ≈0.2 lbm/yr to avoid exceeding the sensor’s upper limit
of detection. If the detector is outfitted with an LFL sensor capable of indicating ammonia
Page 58 of 66
concentrations to 15%, leak rate estimations can be made up to 72 lbm/yr with somewhat larger
uncertainty. Figure 31 shows the use of bagging applied to a shutoff valve to determine if there is any
ammonia leakage from the bonnet gasket or valve stem packing and to quantify the leak rate.
Figure 34 – Drawing from a bagged leak location for small leak rates.
Figure 35 – Photo showing a shutoff valve on the high-side of an ammonia refrigeration system
(left) and bagged to determine if there is any refrigerant leakage from the bonnet gasket or
stem packing.
Page 59 of 66
Because higher ammonia leak rates lead to higher sensed bag concentrations, higher gas flow rates of
fresh air through the bag are needed to dilute the concentration of ammonia in the bag’s
microenvironment in order to dilute the ammonia concentration to a threshold below the refrigerant
detector’s upper detection limit. This dilution can be achieved using a vacuum pump with additional
tubing and a control valve as shown in Figure 32.
Figure 36 - Drawing from a bagged leak location for higher leak rates.
The control valve shown in the top of the illustration enables adjusting the gas flow through the entire
sampling system in order to ensure the ammonia concentration remains within a range acceptable to
the ammonia detector. In this case, ambient (fresh) air flows into the bag before scavenging across the
leak site and mixing with leaked ammonia prior to entering the supply tubing connected to the
downstream refrigerant detector. The bagging system must be set up to accommodate this fresh air
flow for dilution. A micromanometer can be used to compare the pressure at the inlet of the ammonia
detector with ambient. An increase in manometer reading is an indication of a clog or other obstruction
in the sampling setup. The tape seal used to attach the sampling tubing to the bag does not need to be
hermetic as long as there is airflow across the leak site from the air entrance to the tubing. It is
important to ensure that the ambient air being used to dilute the ammonia within the bag does not
short-circuit to the detector however.
Page 60 of 66
Figure 37 - Connection diagram for a dual testing setup.
Depending on the type of gas flow meter used, multiple gas flow meters may be necessary to measure
the full range of expected gas flow rates in the dilution arrangement. In the system shown in Figure 33,
a small rotameter is used for quantifying fresh air flow rates for small leaks using only the ammonia
detector. A larger rotameter is used for larger leaks when the larger vacuum pump is needed for greater
levels of dilution. This type of bag dilution setup enables measuring ammonia leak rates from 0.0009 –
11.5 lbm/year with fresh airflows ranging from 1 to 30 ft3/hr. If the detector is outfitted with an LFL
sensor capable of indicating ammonia concentrations to 15%, the leak measurement range can extend
up to 2,000 lbm/year.
When using a bagging setup, the refrigerant detection system response rate must be measured to
ensure the bagging sample is held for a long enough period of time to reach steady state. The bag
sampling system’s response time is a function of the: volume of the bagging setup, gas flow rate through
the bagging setup, and response time of the ammonia detector. There will be a gas volume within the
bag itself affecting the response rate, but this volume will vary depending on the size of the item being
bagged, accessibility, and other factors. For larger bags it is best increase the dwell time by an
additional 10% to account for a zero reading. To measure the best-case system response time, expose
the system to a known source, such as calibration gas and measure the amount of time for the ammonia
detector to read 90% of the known concentration. Once the response time is known, any bagged item
with an ammonia concentration reading of 0 ppm after running for the response time can be recorded
as a 0 reading (no leakage). Bags typically cannot be re-used since residual ammonia is difficult to
remove from them and not readily purged as with the tubing. Even a modest amount of tubing and bag
volume will produce a system with 2 or more minutes of response time. This can make bagging a time-
intensive activity more suited for gauging the state of a facility by limiting bag sampling to a small
sample of prospective leak points rather than bagging every individual component.
Page 61 of 66
Fugitive Emissions Field-Survey Results
Reindl et al. (2020) conducted field surveys across five (5) plants with six (6) industrial ammonia
refrigeration systems of varying sizes and ages to evaluate the prevalence of fugitive refrigerant
emissions. A total of 175 different refrigeration system components were surveyed, including 159
screened and 110 bagged. Of the 175 components, a total of 34 items had detectable refrigerant
emissions and they comprised 21 sight glasses, 12 system operating valves, and a compressor housing.
Interestingly, a number of investigated items resulted in no refrigerant emissions found and these
included threaded connections, unions, flare fittings, flanges, check valves, plugs and pressure relief
valves. It is known fugitive emissions and accidental releases can originate from these items, particularly
pressure relief valves, however none were found during the field investigation held in conjunction with
the present survey. Items with no detectable emissions were assumed to be a release rate below the
lowest detectable level for this set-up at 0.001 lbm/yr.
The average leak rates found during the present study are summarized below in Table 18. Although
somewhat arbitrary, a pressure of 80 psig was chosen as the transition from “low” pressure to “high”
pressure. Nominally, the qualitative flags “low” and “high” for pressure correspond to the refrigeration
system’s “high-side” and “low-side.” In an ammonia refrigeration system during normal operations,
condensing pressures below and refrigeration loads above 80 psig are quite rare. During the present
field survey, the highest evaporator pressure observed was 60 psig, while the lowest condensing
pressure observed was 110 psig. The average leak rate corresponds to the 34 components that had
detectable refrigerant emissions. The “Zero Odor Rate” corresponded to bagged components where
ammonia was detected during the bagging process but no ammonia odor was previously sensed by the
staff member conducting the screening/bagging. The “Zero Screen Rate” is a threshold level above
which the refrigerant-containing component would need to be leaking for the screening process to
indicate a detectable amount of refrigerant. Clearly, the leak rates on a per-equipment basis are quite
low and, collectively, they did not approach the actual total refrigerant loss rate from each of the five
facilities where field work was conducted.
Page 62 of 66
Table 19 – Summary of fugitive emissions survey results.
During the present survey, a compressor housing leak was found after initially noticing an odor in the
area but the component could not be bagged due to physical restrictions. The estimated release rate,
based on the highest screening reading, was estimated to be 0.043 lbm/yr. The other four compressors
within the same machinery room were similarly screened and indicated 0 ppm, providing the average
listed in Table 18. Because only one compressor screened exhibited a leak, there was too little data to
determine zero-odor and zero-screening average release rates.
Since emission sampling by bagging is more sensitive than screening, there are instances when a leak
will be revealed via bagging even when the screening value of “0 ppm” is obtained. Therefore, a non-
zero leak rate could be assumed for items, even when screened at 0 ppm. Since ammonia has a distinct
odor, the smell of ammonia becomes an inherent tool in screening. As noted above, odor detection for
is less sensitive than an ammonia detector because of the closer proximity of the detector’s sampling
tube to the component being screened. As a result, the zero-screen emissions rate is slightly higher than
when screening with a detector but less than the average leak rate. Equipment where plant or other
personnel routinely have close contact with components throughout a given industrial ammonia
refrigeration system could be considered regularly screened and assumed to be the “Zero Screen Rate”
in Table 18 for fugitive emission estimating purposes.
To relate screening values to actual emissions rates, a least squares regression can be used (EPA, 1995).
Ideally, this regression is prepared for each type or category of equipment; however, the equipment-
specific instances of fugitive emissions found only produced enough data for regression of refrigerant
sight glasses found on the high-side of refrigeration systems. These sight glasses then dominated the
regressions for both “all equipment” and “high side equipment” categories.
Page 63 of 66
The regression for screening all equipment screening using an ammonia detector drawing ≈1ft3/hr
through the sampling pump is shown in Figure 34. It would be expected that refrigerant detectors with
higher gas flow rates would yield lower screening values for the same leak point, and higher screening
values for lower flow detectors. The actual relationship would need to be investigated further if a
screening/bagging relationship is needed for other ammonia detectors or sampling rates. Most
screening was conducted by contact between the detector probe tip and the leak site while still allowing
for full sampling airflow. Retracting the detector probe tip any distance away from the leak site creates
lower screening readings for a given leak rate due to dilution of the sample drawn.
Page 64 of 66
Least Squares Regression of All Components
0.60
0.50
Leak Estimate (lbm/year)
0.40
0.30
0.20
0.10
0.00
0 25 50 75 100 125 150 175 200 225 250 275 300
Screening Value (ppm at 1 ft3/hr ammonia detector flow rate)
(a)
1.50
1.25
1.00
0.75
0.50
0.25
0.00
0 100 200 300 400 500 600 700 800 900 1000
Screening Value (ppm at 1 ft3/hr ammonia detector flow rate)
(b)
Page 65 of 66
Least Squares Regression of All Components
1000
900
800
700
Leak Estimate (lbm/year)
600
500
400
300
200
100
0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Screening Value (% LFL at 1 ft3/hr ammonia detector flow rate)
(c)
Figure 38 – Least squares regression of the screening and bagging values of all equipment
including (a) ammonia concentration ranging from 0-300 ppm screening values, (b) ammonia
concentration from 0-1000 ppm - the full range of the PID ammonia detector, and (c) ammonia
concentration range from 0-150,000 ppm the range of the LFL detector.
In ammonia refrigerated facilities, the periodic search for fugitive emissions is not widely practiced at-
present. At best, technicians will respond to odor complaints by plant personnel in order to locate and
repair leak points or in cases where a refrigerant technician detects an odor during their normal rounds.
Even more rare is a facility that would conduct bagging of components to actually quantify the leak rate.
Table 19 show an estimate of the annual loss of refrigerant via fugitive emissions using the equipment-
specific average leak rates from the present study (Reindl, et al. 2020) for a representative count of each
equipment type actively leaking. In this case, a total of 129 components are assumed to be releasing
refrigerant at fugitive emission rates with the total annual loss reaching 16.3 lbm which is a quantity
substantially less by two to three orders of magnitude refrigerant losses from the systems evaluated in
this project. Based on the average emission rates in the survey conducted and summarized in Table 19,
a system’s total refrigerant emission rate calculated with this method is fairly insignificant compared to
the cost and environmental impact of the ammonia refrigerant.
Page 66 of 66
Table 20 – An example of using the average emission factors to estimate total system emissions.
Average Estimated
Equipment Typical
Equipment leak rate emissions
count points / unit
(lbm/yr) (lbm/yr)
Evaporator Valves (Low) 50 14 0.001 0.97
Evaporator Valves (High) 50 4 0.053 10.64
Compressors 5 1 0.009 0.05
Vessel Sight Glasses (High) 2 6 0.090 1.08
Vessel Sight Glasses (Low) 2 6 0.001 0.01
Other Leak Points 20 5 0.035 3.54
Total 16.3
At a facility outside of the present survey, a highly noticeable odor was encountered in a well-ventilated
machinery room and tracked to a leaking valve stem packing. The item was screened at 934 ppm, but
the bagging concentration exceeded the refrigerant detector’s PID sensor limit of 1000 ppm but did not
reach the detector’s secondary LEL (catalytic bead) sensor detection threshold of 4,800 ppm. Therefore,
the leak rate was estimated to range between 0.37 lbm/yr and 1.76 lbm/yr. The distinctive odor of
ammonia assures leaks of even this relatively small magnitude are typically not allowed to persist for
long periods, and promptly repaired by plant personnel.
Given the low concentration threshold for ammonia odor, screening is only likely to be utilized for
equipment not readily accessible for inspection. For example, attaching some tubing to a long rod to
sample the area near overhead evaporators for screening. Even if every potential leak point in the
above example was near the odor threshold of 0.05 lbm/yr, the total annual emission estimate for the
system would be just over 50 lbm/yr. Even for a small refrigeration system, that level of annual
refrigerant loss pales in comparison to the amount of refrigerant most facilities purchase on an
annualized basis. This clearly indicates that fugitive emissions, as a category of refrigerant loss, is minor
in its contribution to total refrigerant losses from ammonia refrigeration systems. After completing
their fugitive emissions project, Reindl et al. (2020) concluded that fugitive emissions of ammonia was
not a significant component that contributed the total refrigerant loss from the systems evaluated.
Rather, two categories more likely responsible for the majority of refrigerant losses from the
ammonia refrigeration systems included (1) accidental releases and spills (including relief valves
actuating) and (2) discharging ammonia during maintenance activities that involve pumping out one
or more subsystems or components.
Page 67 of 66
unnoticed. In the event a facility cannot effectively maintain a zero-odor environment a screening
program could be developed to estimate losses, though it would make more sense deploying those
resources to locate and repair noticeable leaks.
In a facility where odors are noted and promptly addressed, annual losses from these fugitive emissions
can be estimated using the “Zero Odor Rate” in Table 18. System maintenance staff can then focus on
utilizing the methods in this guide to track total emissions using ammonia purchase logs and dynamic
charge calculations. The refrigerant in/out balance can then be reconciled by identifying and quantifying
other sources of emissions such as maintenance, refrigerant leaking into secondary fluids, and
accidental releases from incidents. Reducing these other sources of refrigerant loss depends on
developing and executing robust mechanical integrity program, which is the best tool for controlling
refrigerant emissions in industrial refrigeration systems using ammonia as the refrigerant.
Ideally, clear benchmarks or thresholds of refrigerant losses for industrial ammonia refrigeration
systems could be developed to trigger repairs when ammonia purchase amounts exceed these
benchmarks. Using EPA Section 608 as a basis for comparison, it requires owners/operators with
stationary refrigeration systems containing 50 lbm or more of Class I or Class II ozone depleting
substances to initiate corrective action to repair refrigerant leaks whenever annual refrigerant losses
reach or exceed a 30% threshold (40 CFR Part 82, Subpart F). Although ammonia is exempt from these
requirements because it is not an ozone depleting substance, it does raise a question of appropriateness
to directly apply 30% to ammonia refrigeration systems as well.
Before providing specific recommendations on threshold refrigerant losses for industrial ammonia
refrigeration systems that should trigger repairs, concept of a single leakage rate and its applicability as
an individual metric warrants discussion. Table 20 shows refrigerant loss rates for three different scales
of refrigeration systems with nominal inventories ranging from 400 lbm to 400,000 lbm. The 40,000 lbm
system size actually corresponds to the mean of all 2,738 filtered from EPA’s RMP database for
ammonia refrigeration systems as RMP-covered processes under 311, 312, and 493 NAICS codes (EPA
2020). For small systems, a leak rate of 10% corresponds to a somewhat modest loss of 40 lbm of
ammonia annually. As the system size increase, the quantity of ammonia lost increases significantly.
For a 40,000 lbm system, a 10% annual loss is 2 tons (4,000 lbm). For a 400,000 lbm system, that same
10% loss is 40,000 lbm on an annual basis. At this quantity, the plant would need to experience an
average daily loss of 109.6 lbm, which is greater than the reportable quantity (RQ) of 100 lbm for
anhydrous ammonia under section 304 of EPCRA (see also 40 CFR part 355). Clearly, a singular value of
annual loss rate expressed as a percentage of the system’s normal inventory is not appropriate.
Page 68 of 66
Table 21 – Comparison of refrigerant loss rates for refrigeration systems of varying sizes.
Daily [Annual loss]* quantity (lbm)
Annual loss rate
400 lbm system 40,000 lbm system 400,000 lbm system
1% 0.01 [4] 1.1 [400] 11.0 [4,000]
5% 0.05 [20] 5.5 [2,000] 54.8 [20,000]
10% 0.11 [40] 11.0 [4,000] 109.6 [40,000]
20% 0.22 [80] 22.0 [8,000] 219.2 [80,000]
30% 0.33 [120] 33.0 [12,000] 328.8 [120,000]
Based on our findings during the course of the present fugitive emissions project as well as our
experiences during more than two decades in the industry, we believe the threshold annual refrigerant
losses for ammonia refrigeration system to active intervention and repair can be expressed on a
percentage of the system’s normal refrigerant inventory with a cap on the total refrigerant loss.
For plants that have unknown and unquantified refrigerant losses in excess of 5% of the normal
system charge or 2,000 lbm (whichever is less), action should be taken to identify root causes for the
refrigerant losses and appropriate repairs to decrease the refrigerant losses.
Page 69 of 66
References
ASHRAE, Fundamentals Handbook, American Society of Heating, Refrigerating, and Air conditioning
Engineers, Atlanta, GA (2017).
Commission for Instruments and Methods of Observation. CIMO Expert Team on Standardization,
Pressure Reduction Formula. Geneva, Switzerland: World Meteorological Organization, November
(2012).
EPA, Office of Air Quality Planning and Standards, “Protocol for Equipment Leak Emission Estimates”,
Research Triangle Park, NC, U.S. Environmental Protection Agency, November (1995).
EPA, Air Emission Measurement Center (EMC), “Method 21 Determination of Volatile Organic
Compound Leaks”, https://www.epa.gov/emc/method-21-volatile-organic-compound-leaks, August 3
(2017).
EPA, “Risk Management Plan Database”, obtained from EPA via. FOIA request, April (2020).
FES, “S Series Screw Compressor Technical Manual”, Bulletin 209(S)-TM, February (1998).
GEA, Excerpted from the FES compressor Technical Manual, personal communication, (2004).
IIAR, Ammonia Data Book, International Institute of Ammonia Refrigeration, Arlington, VA (1992).
IIAR, Standard 6 “Standard for Inspecting, Testing, and Maintenance of Closed-Circuit Ammonia
Refrigeration Systems”, Arlington, VA, (2019b)
Jordan, P., “Case History: A Study of Incidents in the Ammonia Refrigeration Industry”, Technical Paper
#7, Proceedings of the 42nd Annual Meeting of the International Institute of Ammonia Refrigeration,
March 15–18, (2020).
Perkyns, J. S., Kusalik, P. G., & Patey, G. N., “On the dielectric constant of liquid ammonia”, Chemical
Physics Letters, Vol. 129, No. 3, pp. 258–261. https://doi.org/10.1016/0009-2614(86)80207-x (1986).
Reindl, D. T. and Jekel, T. B., “Revisiting Refrigerant Release Estimates”, IIAR Conference Proceedings,
March (2016).
Reindl, D. T., Principles and Practices of Mechanical Integrity Guidebook for Industrial Refrigeration
Systems, University of Wisconsin-Madison, Industrial Refrigeration Consortium, ISBN: ISBN: 978-0-
0074258-0-2, (2016).
Page 70 of 66
Reindl, D., Claas, M., Davis, J., Jekel,T. and Grayless, J., “Reduction of Fugitive Emissions from Industrial
Ammonia Refrigeration Systems”, Final Report, EPA Grant Number: 00E02366, (2020).
Spohner, M. "Study of dielectric properties of mineral oils and natural oils and methyl esters of natural
oils," 2017 IEEE 19th International Conference on Dielectric Liquids (ICDL), Manchester, pp. 1-4, doi:
10.1109/ICDL.2017.8124717, (2017)
Welch, J., “Thermosiphon System Design”, Proceedings of the International Institute of Ammonia
Refrigeration, Albuquerque, pp. 283-313, (2003).
Page 71 of 66