Proceedings of ASME Turbo Expo 2016: Turbomachinery Technical Conference and Exposition

GT2016
June 13 – 17, 2016, Seoul, South Korea

GT2016-57076

ANALYSIS OF HISTORICAL DRY GAS SEAL FAILURE DATA

Meera Day Tim Allison, Ph.D.

Southwest Research Institute Southwest Research Institute
San Antonio, Texas, USA San Antonio, Texas, USA
[email protected] [email protected]

ABSTRACT While many design and operational recommendations

are provided based on reported failures, these recommendations
While proper design and maintenance of dry gas seals are not intended to be a comprehensive set of guidelines for all
is a well-understood topic, dry gas seal failures are still applications. In addition, the purpose of this paper is not
relatively common. These failures can result in frequent repairs intended to assign blame to any party for any particular type of
and costly downtime. Although several case studies of failure, but rather to examine and highlight the technical
individual failures are available, relatively few large-scale dry challenges and issues that most often result in seal failures.
gas seal failure studies exist. Based on a review of existing This report first presents an overview of dry gas seals
literature, very little has been published on failure statistics and their associated support systems. Next, it describes the data
aimed at improving seal reliability. As a part of an industry- collection process and outlines a classification system for root
funded dry gas seal reliability project, a failure database has causes. Finally, it presents a statistical analysis of the failure
been populated with information provided by both end users data collected and presents suggestions for avoiding common
and original equipment manufacturers. The database includes failure modes based on information received.
details on dry gas seal and separation seal configuration, seal
gas supply, operating history, conditions at time of failure, and DRY GAS SEALS OVERVIEW
failure symptoms, including any results from failure analyses
performed by the survey respondent. In total, eight companies OPERATION
contributed 194 failures. Of these, 144 cases had root causes The purpose of dry gas seals is to provide reliable
provided. shaft sealing on turbomachinery to reduce the amount of
process gas escaping from the machine’s primary flowpath to an
From this database of failures, statistical analysis is
acceptable amount. According to Stahley [1], dry gas seals first
used to determine common reasons behind dry gas seal failures
emerged around 1970, in response to an industry need for a less
in gas compression service. Failure trends are identified based
complex and hazardous system than floating ring oil seals. By
on data collected, and corresponding recommendations are
comparison, dry gas seals are simpler and require fewer support
provided for improving dry gas seal reliability.
systems that wet seal systems.
There are a large variety of dry gas seal configurations,
INTRODUCTION including single seals, tandem seals, and tandem seals with an
intermediate labyrinth seal. While most of the seals examined
The aim of this study is to improve reliability of dry were tandem seals or tandem seals with intermediate labyrinth
gas seals in natural gas service through a better understanding seals, single seals will be first explained due to their simplicity.
of failure trends. While API 692, currently under development, Figure 1 shows a single dry gas seal configuration.
aims to provide comprehensive recommendations for dry gas The rotating face (shown in red) and the stator ring (shown in
seal systems design and operation in order to improve dry gas orange) are held in place by a spring (shown in grey). The
seal operation, this paper aims to specifically study reported rotating face has a grooved pattern on the face intended to
failure data in order to examine common failure modes and optimize seal performance [1]. The stator ring is held in contact
draw conclusions about reliability based on failure trends. with the mating ring by the spring retainer assembly when the
machine is not spinning or during pressurization. When the

1 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 faces are in contact and not rotating, there is very little or no stiffness is equal to the derivative of opening force with respect
leakage. As shown in Figure 1, O-rings (labeled as “dynamic to the running gap and increases nonlinearly as the running gap
sealing element”) are also used in the seal assembly to provide decreases [1]. Thus to minimize leakage, ideally the gap would
seal face alignment. be as narrow as possible, and the film would have a high gas
stiffness [1].
The rotating face can be designed for unidirectional
rotation or bidirectional rotation.

Figure 1. A Single Dry Gas Seal [2] Figure 3. Drawing of Groove Patterns of Unidirectional Seal Designs
[2]
There are two ways for the rotating face and primary
ring to lift-off: static and dynamic. Static lift-off may occur
Unidirectional seals (several examples shown in Figure
when hydraulic balancing within the seal assembly causes the
3) are generally better suited for applications with axial
rings to separate slightly. Dynamic lift-off occurs during
movement since they have a wider running gap that reduces the
rotation when flow enters the grooves on the rotating face. The
risk of contact and a stiffer gas film making the seal able to
volume reduction at the tips of the grooves causes the gas to
accommodate some disturbance [1].
compress and form a pressure dam (an area of slightly higher
pressure) [1]. The pressure dam causes lift-off between the
faces to a running gap of 3 to 10 µm [1]. There is a wide
variety of groove patterns intended to accomplish this; a few of
these are shown in Figure 3 and Figure 4.
During operation, distance between the rotating face
and primary ring is self-regulating to accommodate axial
movement during operation. If axial movement causes the gap Figure 4. Drawing of Groove Patterns of Bidirectional Seals [2]
to increase, the pressure dam will be reduced due to the
increased volume, and the faces will be drawn back together to Bidirectional seals, shown in Figure 4, are generally
their intended running gap [1]. If the gap decreases, the only recommended for applications where the ability to
pressure dam will increase and the faces will be pushed apart. withstand counter-rotation is necessary [1]. However,
Figure 2 graphically explains this behavior. The spring force is maintaining a stock of bidirectional replacement seals is a
only noticeable at lower pressures. At higher pressures, the common end user practice because it reduces the required
spring force is negligible compared to forces due to pressure. inventory of spare seals, since a bidirectional seal can be used
on either end of a rotor. Each seal is very application-specific;
therefore it may be possible that conditions exist where either a
unidirectional or bidirectional seal will function equally well. It
is recommended that the original equipment manufacturer be
contacted to ensure that the seal being selected matches the
application and system.

FLOW PATH
For a single seal configuration such as the one in
Figure 5, seal gas supply comes from an external supply system,
which is often taken from the compressor discharge. Most of
the flow will move into the process, while a very small amount
Figure 2. Representation of Self-Regulating Gap [2] will flow through the seal faces and out through the primary
vent. The amount of flow between the faces will be quite small
The rotating face and primary ring should not come compared to the flow into the process. The exact amount of
into contact when spinning at high speeds, as this will rapidly flow will depend on the seal design and application. The
degrade the faces and lead to seal failure. Seal leakage is figures presented in this section are of the seal only; they do not
proportional to the cube of the running gap and directly show the vent lines. The outer portion of the seal exit is not
proportional to the sealing pressure and seal diameter [1]. Film directly exposed to atmosphere.

2 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 In some applications where sufficient supply pressure
is not constantly available from the process (e.g. during startup
or pressurized hold), it is recommended that a booster system or
alternate gas supply be provided to ensure sufficient supply
pressure at all times.

Figure 7. Tandem Dry Gas Seal Configuration with Intermediate

Labyrinth Seal (Circled in Yellow) and Flow Path [2]

SEPARATION SEALS OVERVIEW

Figure 5. Single Dry Gas Seal Flow Path [2]
Separation seals are outboard of the dry gas seals and
While there are a few different configurations for are intended to prevent bearing lube oil from migrating into the
seals, all but a few of the seals studied for this report were dry gas seal. Separation seals can be labyrinth seals, contacting
tandem seals, shown in Figure 6. Tandem seals, or tandem seals segmented carbon ring seals, and noncontacting carbon ring
with an intermediate labyrinth (see Figure 7), are most seals. Each type of separation seal has its own set of
commonly used for compressor pipeline applications. Tandem advantages and disadvantages.
seals have two seals (a primary and secondary seal) in series. Labyrinth seals were the original separation seals used
Similar to single seals, most of the leakage goes into the process with dry gas seals. They are inexpensive and highly reliable
side, but there is a small amount of gas that will go through the [1]. However, they have the highest leakage rate of all the
primary and secondary seals and vents. separation seals, which will increase cost of separation gas
required.
Segmented carbon ring seals consist of two sets of
circumferential rings within a housing. The rings are held
together with springs, and separation gas is injected between the
rings. Segmented carbon rings seals use significantly less
separation gas due to the smaller shaft clearance, which can
reduce the cost of separation seal gas. Segmented carbon ring
seals can be contacting or noncontacting. The contacting shaft
seal has no clearance between the seal inner diameter and shaft,
while the noncontacting seal has a small clearance. Contacting
segmented carbon ring seals have a service life of 3-5 years due
to wear, while noncontacting segmented carbon have a nearly
Figure 6. Tandem Dry Gas Seal Configuration and Flow Path [2] indefinite service life [1]. Figure 8 shows a schematic of a
noncontacting segmented carbon ring seal, and Figure 9 shows
The addition of an intermediate labyrinth seal ensures a contacting segmented ring seal.
that no primary seal leakage will flow into the secondary vent.
This is useful in applications where harmful process gas is not
allowed in the secondary vent, as it may vent to atmosphere.
This configuration is more complex because it requires the
addition of a secondary seal gas supply, labelled “inert buffer
gas” in Figure 7.

3 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 A few other sources provided information about failure
case studies, and this information was incorporated into the
database. This includes case studies from Stahley’s book [1]
and a paper by Stahley [3]. An article from CompressorTech2
[5] with a case study was included, as well as results from an
API survey on dry gas seal reliability [6].
It should be noted that this data was analyzed “as-is.”
This means that root causes provided by companies were
assumed to be correct. Care was taken to ensure that the root
cause was reasonable given the diagnostic findings and
conditions. Whenever possible, clarifying questions were asked
Figure 8. Representation of Noncontacting Segmented Carbon Ring of companies providing data to increase the fidelity of the
Seal [2] database. In some cases, multiple root causes were provided by
the company. Therefore, the percentages provided below will
not necessarily add up to 100%.

DATA CLASSIFICATION
The root causes of the failures were broken into the
following categories: process contamination, supply
contamination, lube oil contamination, and
installation/geometry problems. Within each of these
categories, further categories were identified based on the
specific root cause. The resulting distribution of root causes
and subcategories are shown in Table 1 and Figure 10. It
should be re-emphasized that some failures had multiple root
causes; therefore the percentages below do not add up to 100%.
Figure 9. Contacting Segmented Carbon Rings [2]
Details regarding failure data for each of these categories are
discussed further in the following subsections. Percentages of
While the separation seals presented here are the more
subcategories are based on the data set of the subcategory only
commonly used seals, a variety of other separation seal designs
(i.e. 64% of failures in the supply contamination category were
do exist, including noncontacting coaxial face seals.
due to liquid contamination).
STATISTICAL ANALYSIS Root Cause
Supply Installation / Process Lube Oil
DATA COLLECTION Contamination Geometry Contamination Contamination
Analysis of dry gas seal failures was completed by (43%) (26%) (34%) (21%)
approaching 11 companies and asking if they would be willing • Liquid • Geometry • Heavy (not broken into
Contamination 62%) Hydrocarbon subcategories)
to contribute failure data for the project. The companies (64%) • Installation Contamination
approached included both end users and original equipment • Other (38%) (25%)
manufacturers. Care was taken to cross check all data collected Contamination • Process Gas
to ensure that failures were not double counted in cases where (28%) Contamination
both the end user and OEM reported on the same failed seal. • Filter Overload (41%)
(8%) • Insufficient
To protect the companies involved, all failure data was placed Seal Gas
into an anonymous database for analysis. A few other reported (25%)
cases of failures were taken from available literature, previous
API surveys, and publications. Table 1. Breakdown of Root Cause Categories and Subcategories
Eight companies agreed to participate in the study.
Data was collected from these companies via phone calls,
failure reports, and internal failure databases put together by the
participating companies. The goal was to gather as much
information as possible, while minimizing the work of the
participating companies. Data for 194 failures was collected
into the database, created a sample set many times larger than
any other previously published study. Of this database, 144
provided root causes.

4 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 analyzed did not have a heater, which may have contributed to
the condensation problem.
Visible signs of process gas on the seal face is a
common indication of a failure due to liquid contamination.
For sour gas, this could be a white sulfur powder build-up.
Water lines may be visible on the seal if part of the seal was
immersed in fluid for extended periods of time. Liquid on the
seal faces can increase the necessary lift-off pressure, causing
the seals to show signs of rubbing or contact.
To avoid liquid contamination in systems prone to
liquid dropout, the addition and proper operation and
maintenance of a seal gas heater may be recommended. It is
particularly important to make sure that the seal gas is allowed
to heat sufficiently before starting the compressor. A few
Figure 10. Percent Failures by Category reported cases suggested that due to a rush to restart the
compressor and a lack of experience or training, compressors
The following sections will outline the findings of each were started prior to the seal gas reaching its required
category and subcategory, followed by a study of other temperature. To avoid this, a minimum seal gas temperature
parameters, such as application and state at time of failure. could be specified as a start permissive. Additionally, alarms
and trips at low temperatures could warn of a heater
SUPPLY CONTAMINATION malfunction or insufficient heating.
The results show that supply contamination is by far Another cause of liquid contamination is from liquids
the highest cause of failures in the dry gas seals studied. Of the in the vent line not being drained. These liquids may build up
collected failures, 43% were a result of supply contamination. in the vent line and flow back into the seal. It is imperative that
Supply contamination consists of several types of contaminants the vent line has a low point somewhere before the flare, and
introduced to the seal through the supply system. Failures are that the operator checks it regularly. The operator should also
broken down into subcategories for liquid contamination, filter check the seal drains to ensure no liquid build up. Furthermore,
overload, and other contamination, which will be discussed in if the seal gas supply comes from the compressor discharge, the
detail below. take-off should not be from the bottom of the casing, as this
could allow liquids into the supply gas.
LIQUID CONTAMINATION It should be noted that some companies reported seals
Supply contamination includes liquid contamination, operating successfully for extended periods of time with liquid
which is where liquid from the seal supply gas gets between the continuously between the rotating face and mating ring,
rotating face and mating ring, and eventually causes seal covering most of the contact area. If the liquid is sufficiently
failure. Of the supply contamination failures, 64% of the lubricating, it can possibly help in seal operation. However,
failures were due to liquid contamination. Liquid can drop out given that the chance for failure is high, it is recommended that
of the gas due to the Joule-Thomson effect; as the seals gas seal supply systems be designed to consider eliminating the
expands in volume and drops in pressure over the seal face, the chance of liquid entering the seal.
temperature will also drop. This causes condensation to drop
out of the seal gas, which may result in seal failure due to the FILTER OVERLOAD
liquid between the rotating face and mating ring causing Another cause of supply contamination is filter
rubbing and possibly heat buildup. overload or degradation. 8% of reported supply contamination
To avoid condensation, API 614 [4] recommends that failures were due to filter overload or degradation. Filter
seal gas supply must be at least 20°C above the dew point of the overload is a quick failure, where the filter becomes too full and
gas at all points throughout the seal (not simply at the supply collapses, sending a buildup of debris through the lines and into
conditions). It is possible that the 20°C requirement is more the seal. This can be seen as a large, rapid jump in vent
than necessary, which could lead to an oversized heater. pressure and vibrations. Filter degradation, on the other hand,
Stahley [1] recommends a computer model of the temperature is where the filter gradually becomes too full and starts to allow
and pressure drops across the system be used to determine the larger and larger particles through. This will manifest itself as a
required inlet temperature of the supply gas to both avoid liquid slow increase in vent pressure over many days, leading to a
dropout and not oversize the heater, if one is needed. In cases gradual seal failure.
where liquid dropout may occur, it is highly recommended to Filter failure can be caused by the operator not heeding
add a heater to the seal gas panel to ensure that the gas entering alarms to change the filter. A differential pressure meter should
the seal is going to be warm enough. Many of the failures be placed across the filter, and it should alarm on high DP. The
operator must not ignore this alarm. Filters for continuous

5 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 service applications should always have a duplex setup that
allows the filter to be changed without taking the compressor
offline, and the operator should be trained in filter replacement
to ensure that debris does not enter the line during the filter
change.
Coalescing filters are recommended to catch any liquid
mist that may be present in the gas. However, it should be
noted that a liquid slug coming through the lines will not be
caught by a coalescing filter and will likely result in seal failure.

OTHER CONTAMINATION
Of the supply contamination failures reported, 28%
Figure 11. Drive Pin Alignment Feature [2]
were due to other contaminants not included in the above
subcategories. This contamination could come from anything
To avoid drive pin issues, the drive pin dimensions can
upstream of the seal getting blown into the seal or debris being
be checked prior to installation of the seal. Since this may
in the seal at time of installation. Contamination could result
impractical, it may be easier to perform a few checks during
from a process where the system was opened for cleaning.
installation suggested by Stahley instead. These test can verify
During transients, contamination in the lines could be stirred up
that the seal is able rotate properly and move axially [1].
and forced into the seal. It is also possible to over-pressurize
Several failures occurred when oil-trapping features in
valves or other components upstream of the seal, causing parts
the compressor flow path allowed oil to drain intro the process
to break off and flow into the seal.
labyrinth during pressurized hold and/or shutdown, even when a
Symptoms of impending failure due to other
normal seal gas supply pressure was maintained. This proved
contamination may include high primary vent pressure. Solid
to be an incredibly challenging problem to detect, since
contamination may be visible upon seal removal and there may
everything on the supply line appeared functional.
be pitting or signs of rubbing on the faces. If the seal gas is
supplied by a separate reciprocating compressor, it is possible
for lube oil to make its way into the seal. This type of INSTALLATION
contamination usually results in contact and visible marks on Of reported geometry and installation failures, 38%
the seal faces. It can often lead to seal face break up. were due to installation issues. In some cases, the seal was not
To avoid failure by supply contamination, it is installed with all of the necessary components. This would
recommended that all upstream pipes are blown out after the result in immediate seal failure. Missing parts included bolts
loop is opened for cleaning. Also, using stainless steel pipes and labyrinth snap rings. There were a few cases where the seal
rather than carbon steel can minimize the risk of corrosion was not aligned properly during installation and was therefore
getting into the seal. Operating procedures should be checked not able to rotate or move axially. For this reason, Stahley [1]
to ensure that components are not over-pressurized at any time, recommends checking the axial travel, and rotation if possible,
particularly during transients. Valves and components upstream of a seal manually prior to installation. A couple of cases
of the seal should be specified for the maximum pressure they reported sheared O-rings, or seals containing debris prior to
could possibly experience to reduce the risk of them breaking startup.
during operation.
PROCESS CONTAMINATION
INSTALLATION AND GEOMETRY PROBLEMS Of the total failures, 34% were due to process
The next highest cause of seal failure is installation contamination, which includes heavy hydrocarbon
and geometry problems, with 26% of the total failures reported. contamination (25%), process gas contamination (41%), and
insufficient seal gas (25%). Heavy hydrocarbon contamination
is when condensates from the process drop out and contaminate
GEOMETRY PROBLEMS
the seal. Process gas contamination is when solids drop out of
Of geometry and installation failures, 62% were due to
the process gas, contaminating the seal. For sour gas
geometry issues, most of which were related to the drive pin
applications, this usually results in a buildup of sulfur along the
(shown in Figure 11). In many cases, the drive pin showed
seal faces. Failures were only classified as having insufficient
visible signs of being too long, which may restrict axial
seal gas when it was explicitly stated as such. A surprising
movement on the seal faces. This results in immediate failure
number of cases reported insufficient or no seal gas
on startup due to excessive swash (axial runout) and potential
continuously or intermittently during operation. This is almost
axial misalignment between the rotating face and mating ring.
guaranteed to let the process gas flow into the seal and will
This swash will likely degrade the O-rings quickly and cause
likely result in failure. Indications of insufficient seal gas
them to break. The mating ring retainer may show physical
include sulfur residue on the seal face (for sour gas operations)
damage from the excessive movement.

6 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 or sticky black or clear liquids for heavy hydrocarbons. Seal reduce the likelihood of failure. Additionally, there are some
face contact is nearly guaranteed, which may be severe enough specific seals designed for gasses with particularly low dew
to result in broken rotating faces. points.
To reduce the likelihood of failure from process
contamination, it is recommended that alarms alert the operator SEAL LIFESPAN
when there is not enough supply gas pressure. A trip is advised Although not all reported failure cases have associated
for the supply gas to prevent operation without seal gas. If running hours, the breakdown of the 21% of cases with
needed, a booster system can be added to ensure that that there operating life data is shown in Figure 12. The failure data in
is sufficient seal gas pressure during transients or if something the plot below are expressed as percentages of failure cases
causes the seal to have inadequate supply pressure. with operating hour information.
Furthermore, supply flow should be automatically regulated via
a flow control or differential pressure control scheme, as it is
more reliable than manual adjustment.
A highly important cause of seal failure due to
insufficient seal gas is slow roll. It was found that 14.7% of
process contamination occurred during slow roll. If the seal gas
supply is taken off of the compressor discharge, during slow roll
there may not be enough pressure rise across the compressor to
supply adequate pressure to the seal gas supply. To avoid this,
slow roll speed should be adjusted to be high enough to
maintain sufficient seal gas supply pressure or a booster system
should be used for seal gas supply.
Also of interest were several outer seal failures that
failed due to insufficient supply gas. While this may seem
counterintuitive, it was determined that moisture passed through
the inner seal (or formed along the inner seal), and subsequently Figure 12. Hours of Seal Operation Prior to Failure
contaminated the outer seal. The outer seal has a smaller
running gap due to the lower differential pressure, which may Geometry and installation issues usually manifest
make it more susceptible to small contaminants that the primary themselves within half an hour of startup. While this 30 minute
seal is able to handle. rule cannot be applied universally, it does appear to suggest that
a failure within the first 30 minutes is very likely a result of
LUBE OIL CONTAMINATION geometry or installation.
Twenty-one percent of reported failures were due to Also of note is supply contamination. In the data set,
lube oil contamination, which is when lube oil from the supply contamination does not occur until after a span of
bearings flows into the seal between the rotating face and running time. While seals are designed to last indefinitely when
mating ring. This is caused by insufficient separation seal given perfect operating conditions, operating conditions in the
supply gas or lube oil flow that is excessively higher than field are almost never perfect. The data suggests that seals
design. Seals that failed due to lube oil contamination often had generally do not fail from supply contamination until after at
wet lube oil on the seal faces. Excessively dry nitrogen as the least 50 hours of running time. This suggests that seals do not
separation gas can increase the likelihood of lube oil always fail the moment they become contaminated; rather, seals
contamination because the separation seal will become dried are sufficiently well-designed to handle a certain amount of
out and fail prematurely. A few reported failures were caused contamination to account for the less-than-ideal situations
by the separation seal not being adequate for the lube oil encountered in the field.
conditions present. In an effort to investigate failure causes, rather than
A solution to avoid lube oil contamination is to failure rates, this study only collected data from cases where
increase the separation gas pressure to the upper design limit of failures occurred, and therefore this dataset cannot be used to
the seal, as specified by the seal manufacturer. However, determine a mean time between failures for the entire dry gas
excessive differential pressure across the separation seal can seal population. Seals operating in a properly designed and
accelerate wear of the separation seal and subsequently maintained system can exhibit exceptional reliability, as the O-
decrease its reliability. It is also recommended to have a ring is the only element of the seal subject to normal wear in
minimum separation gas pressure as a start permissive for the perfect conditions. Because the data set collected consists
lube oil system to ensure that the lube oil is not allowed into the entirely of failed seals, it is clear that the cases studied are
system without separation gas present and to shutdown on very representative of less than ideal operating conditions, as it is not
low or no separation gas pressure. In the case of excessively representative of all seals collectively. To emphasize, these data
dry nitrogen, experience suggests that adding humidity will

7 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 are NOT stating, for example, that 15% of ALL seals will fail in database, have been sorted into categories and associated
1-5 hours of operation. failures.
Based on verbal feedback from participating
companies, a good seal in normal conditions generally lasts
30,000 to 50,000 hours or more, or 5+ years. This is not an
absolute, however, as there are occasional anecdotes about seals
lasting for over 20 years.

OPERATIONAL STATUS AT TIME OF FAILURE

Since Figure 12 suggests that startup and
commissioning are common times for seals to fail, it is worth
investigating at what point during operation the seal failed. The
seal can potentially fail at normal operating conditions, startup,
shutdown, or commissioning. It should be noted that Figure 13
represents only failures where operation state was specified,
which was 40% of the failures collected.
Figure 14. Failures Grouped by Application

From Figure 14, supply contamination appears to be a

large problem for injection, pipeline, and sour gas compressors.
Sour wet gas applications may be dirtier than the other
applications studied, which could explain why there are more
reported supply gas contamination failures in that category.

FAILURES BY YEAR
Of the data collected, 68% of the failures collected
included information on the year the failure occurred.

Figure 13. Failures Sorted by Operation State at Time of Failure

In Figure 13, geometry and installation problems are

generally seen during commissioning, which includes factory
acceptance testing, and startup. Supply and process
contamination is most commonly seen during normal operating
conditions. Seals are unlikely to be reported as failing during
shutdown. When combined, commissioning and startup make
up a significant portion of the failures. This is likely due to off-
design conditions during transients, which could provide
inadequate seal supply flow. Most of the failures reported are Figure 15. Failures by Year
from normal operating conditions, which may be due to seals
spending more time in normal operation than in startup, It must be noted that Figure 15 should not be
shutdown, or commissioning. interpreted as a reliability trend for the entire dry gas seal
population; it is simply representative of data gathered (e.g. it
FAILURES GROUPED BY APPLICATION cannot be assumed that there were no dry gas seal failures
Communication with participants has indicated there is anywhere in 2012). However, some trends may still be
a wide variety of seal sizes (both in physical size and power of identified. Supply contamination appears to be a smaller
the associated machinery) in the data reported, but there is problem in 2007 and onwards, suggesting that strategies to
insufficient specific data on this to break failures down into size avoid supply contamination were developed at that time.
groups. However, 84% of the data provided included a specific Reducing the chance of supply contamination is arguably one of
application. These applications, as provided in the failure the easiest ways of improving dry gas seal reliability.
Subsequent years saw a significant decrease in seal failures.

8 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 It can also be seen that geometry and installation issues provide valuable information on which separation seal is
have become less common through the years. This suggests the most reliable and economical. Additionally, there is
that dry gas seals are an evolving technology, meaning that opportunity for improving separation seal systems
continued collaboration between end users and original through a study of separation seal and separation gas
equipment manufacturers can provide valuable information to interaction, given that the type of gas and the humidity of
continue improving seal reliability. Studies such as this are gas can greatly impact the separation seal performance.
invaluable to all interested parties in providing data-based
guidance for improved reliability. Operators have noted that older wet gas seal systems are
still fairly reliable and cost effective. Since dry gas seals
CONCLUSIONS were intended to be cheaper and more reliable, it would
be worth comparing dry gas seals to wet seals. A cost-
This report has presented a statistical analysis of
benefit analysis could study various types of wet seals
failure data from 194 dry gas seal failures. The data span
and dry gas seals to quantify the advantages and
failures since 2003 and include failures that occur at a variety of
disadvantages of both types of systems. With new
running times and operational states. Supply contamination,
methane emissions regulations possibly requiring
particularly liquid contamination, was identified as the leading
monitoring systems for wet seals, it may be less costly
cause for reported dry gas seal failures. Supply contamination
and challenging to retrofit a dry gas seal instead of
is most commonly found in applications prone to potentially
continuing to use wet seals.
higher levels of contamination, particularly sour wet gas
applications. Other major causes of failures included
Secondary seal failures in tandem dry gas seals can be
installation/geometry errors, process gas contamination, and
challenging to detect. If a seal operates with a damaged
lube oil contamination.
secondary seal, in the event of a primary seal failure,
The reported failures also suggest that the reliability
harmful gases could be released to the atmosphere. To
and common failure modes of dry gas seals change as the
avoid this, a control system for detecting failure, both on
technology improves over time. Thus it is ever important that
primary and secondary seals, would be highly useful.
studies such as this continue to provide statistical analysis to
improve the body of knowledge of seal operation with the goal
ACKNOWLEDGMENTS
of improving overall dry gas seal reliability.
The authors would like to thank the Gas Machinery
FUTURE WORK
Research Council for sponsoring this study, as well as the
This study suggests that there is much future research industry committee that reviewed this project and provided
that can be done on dry gas seals to improve understanding and guidance along the way. Furthermore, the authors are highly
reliability. Future work might include the following: appreciative of the companies that contributed data and
knowledge for populating the database.
While some standards exist for the supply gas inlet
temperature, it is challenging to properly determine the REFERENCES
needed supply gas inlet temperature since it may be
affected by a number of factors, including leakage, [1] J. S. Stahley, Dry Gas Seals Handbook, Tulsa: PennWell
allowable dropout at standstill, and heat generation Corporation, 2005.
within the seal. As a result, standards for supply gas
temperatures may be overestimated and could potentially [2] M. Klosek, "Short Course 4 on Dry Gas Seals," in
be too conservative and expensive. There have also been Turbomachinery Symposium, Houston, 2003.
some cases where the existing temperature
recommendations are not enough to prevent liquid [3] J. S. Stahley, "Design, Operation, and Maintenance
dropout. A study could be done to investigate this further Considerations for Improved Dry Gas Seal Reliability in
and determine an efficient way to minimize the required Centrifugal Compressors," in Proceedings of the 30th
heater size while reducing the likelihood of failure due to Turbomachinery Symposium, Houston, 2001.
liquid dropout. Additionally, an investigation on how
hydrocarbon condensates that build up in the seal affect [4] American Petroleum Institute, API Standard 614:
startup could lend some useful results. Lubrication, Shaft-Sealing, and Control-Oil Systems and
Auxiliaries for Petroleum, Chemical, and Gas Industry Services,
Labyrinth seals are generally considered to be the most 5th ed., Washington D.C.: CSSinfo, 2008.
reliable separation seal, but they have the highest
required separation seal flow rate. A reliability study [5] S. Vidal, "CompressorTech2," June 2015. [Online].
similar to this one with a focus on separation seals could

9 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo

 Available: http://www.ct2-
digital.com/ct2/june_2015?pg=18#pg18. [Accessed 2
December 2015].

[6] J.M. Thorp, “Improving Dry Gas Seal Reliability”, API

Spring Refining Meeting, New Orleans, April 15, 2008.

10 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/89527/ on 05/17/2017 Terms of Use: http://www.asme.org/abo