You are on page 1of 8

GCPS 2015 __________________________________________________________________________

Calculating Facility Siting Study Leak Sizes One Size Does Not Fit
All

Gary A Fitzgerald
ABS Consulting
140 Heimer Rd, Suite 300
San Antonio, TX 78232
gfitzgerald@absconsulting.com

Copyright 2015 ABSG Consulting Inc. All rights reserved.

Prepared for Presentation at


American Institute of Chemical Engineers
2015 Spring Meeting
11th Global Congress on Process Safety
Austin, Texas
April 27-29, 2015

UNPUBLISHED

AIChE shall not be responsible for statements or opinions contained


in papers or printed in its publications
GCPS 2015
__________________________________________________________________________

Calculating Facility Siting Study Leak Sizes One Size Does


Not Fit All

Gary A Fitzgerald
ABS Consulting
140 Heimer Rd, Suite 300
San Antonio, TX 78232
gfitzgerald@absconsulting.com

Keywords: Facility Siting, Risk, Risk Assessment, Risk Management, PHA, Leak Size,
Maximum Design Leak, MDL

Abstract
Consequence-based Facility Siting Studies (FSSs) typically requires the user assume a
credible leak size to use in the evaluation of potential releases, which is often up to a 2
diameter leak. Many facilities tend to be less complex in comparison to large refineries or
petrochemical plants, leading operators at the less complex facilities to ask why they should
assume the same leak sizes as more complex facilities. Other facilities have unique
processes with safety systems and factors they would like to quantify in a FSS. One
solution would be to perform a Quantitative Risk Assessment (QRA) to capture the risks
from all potential release locations and release sizes. However, many companies have not
defined risk tolerance criteria and are resistant to do so for various reasons or do not want
to invest in a QRA, which are more costly than a consequence-based study. A unique
approach developed by ABS Consulting and first presented in 2011 is called the Maximum
Design Leak (MDL) approach [1]. This approach calculates frequency-based leak sizes and
then applies the leak size that exceeds a frequency criterion (events/year) in a consequence-
based FSS instead of assuming a given leak size as credible. This avoids having to establish
risk criteria in terms of fatalities/year and having to model a large number of scenarios yet
takes advantage of many features in a QRA. This paper presents three case studies as
examples of how the MDL has been applied and illustrates the advantages of calculating
leak sizes specific to scenarios being evaluated for low complexity and low risk facilities.

1. Introduction

Facility Siting has matured over the last 20 years. Initially, only the facilities with the
greatest hazards performed hazard studies to calculate potential consequences of accidental
releases. As the practice grew in popularity, more and more facilities would evaluate their
hazards even if they were perceived to be at relatively low-risk. Regulatory agencies have
also come to expect hazard studies for any facility within their regulatory jurisdiction.
Thus, now all facilities covered under OSHA PSM and EPA RMP are expected to have
hazard studies performed to quantify potential consequences from accidental releases.

Note: Do not add page numbers. Do not refer to page numbers when referencing different portions of
the paper
GCPS 2015
__________________________________________________________________________

A typical consequence-based FSS will evaluate leaks from locations determined to have
the greatest consequences, regardless of the likelihood of a leak at that location other than
a qualitative assessment of the maximum leak size. This has the potential to over-predict
consequences from one location in lieu of evaluating leaks at other locations which may
be more frequent but less severe. Several guidance documents are available for selecting
leak size that tend to agree in either limiting leaks to a maximum diameter of 2 inches or
only using a portion of the pipe cross-section as their assumed leak size. Industry has
generally adopted the 2 inch maximum leak size for FSS. While typically considered a
conservative approach, it is an accepted level of conservatism for most hazardous facilities
that allows for hazards evaluations without considering numerous sources or calculating
fatalities.

The MDL approach was proposed as a means to calculate leak sizes based on published
equipment failure rates. It is recommended to adjust these published frequencies as
appropriate to reflect a specific site location, company operating practices and accident
history. ABS Consulting has performed MDL studies at over a dozen facilities and found
them to be most beneficial to the smaller and less complex facilities where a 2 leak size
is believed to be overly conservative for most leaks and an assumption less than 2 is
difficult to defend.

2. Case Study 1

A specialty-chemical company performed a consequence-based facility siting study for a


new unit assuming 2 leak sizes. Results were severe and did not make sense in comparison
to company or industry experience given the high level of mechanical integrity and
numerous safety systems. As a result, the company decided to perform a MDL study for
the new process.

The process was first segmented into hazardous streams with similar operating conditions
such that a leak anywhere in that segment would have similar consequences. A total of
fifteen segments were identified. Four segments were very low hazard and were excluded.
The remaining eleven segments were evaluated in a MDL study. The company performed
a parts count for these eleven segments.

Frequency tolerance criteria were developed based on company maximum individual risk
criteria. Factors such as building occupancy, number of hazardous segments, number of
independent hazards, ignition likelihood, wind directions, acceptable occupant
vulnerability and an additional factor of safety were used to generate a tolerable single
event frequency. Additionally, some segments had Safety Instrumented Systems (SISs)
with a Safety Integrity Level 1 (SIL 1) assigned. For those segments, a 10% chance of
safety system failure was taken into account such that those segments were evaluated as
two scenarios (with and without safety system operation) with different tolerable
frequencies. The subject of developing frequency tolerance criteria for use in a MDL is the
subject of a future paper and details will not be discussed here.
GCPS 2015
__________________________________________________________________________

A count of equipment within each defined segment was performed and published
equipment failure rates were applied for various leak sizes (corrected for plant specific
operations). Results are plotted in the following figure.

Figure 1. Case Study 1 Parts Count

Curve fits were performed for each segment and the equations solved for the leak size at
the frequency tolerance criteria. Resulting leak sizes are shown in the following table
(rounding up to the next 0.05 with a minimum size of 0.1).

Table 1. Case Study 1 Leak Sizes


Leak Size (inches)
Segment No SIS SIS Installed SIS Installed
Installed and Operates and Fails
1 N/A 0.85 0.10
2 N/A 0.65 0.10
3 N/A 0.65 0.10
4 N/A 0.75 0.15
5 0.15 N/A N/A
6 0.15 N/A N/A
9 N/A 0.35 0.10
10 N/A 0.35 0.10
11 N/A 0.50 0.10
12 N/A 0.80 0.15
13 0.15 N/A N/A
GCPS 2015
__________________________________________________________________________

Consequences were then evaluated and blast loads on buildings from the new process were
reduced to about half of their initial pressure and impulse as a result of the MDL study.

3. Case Study 2
A small oil and gas terminal not covered under OSHA PSM with no processing facilities
wished to perform a FSS but recognized their facility presented far lower risks than other
large facilities. However, they did not have a defendable basis to assume a leak size smaller
than typical industry practice for large facilities. Thus, they decided to use the MDL as the
basis for leak sizes in the FSS. The company had a single event frequency tolerance already
defined in LOPA procedures which was applied in this evaluation. Their process was
segmented into 4 sections with potentially hazardous consequences. A parts count was
performed for each segment and frequencies were corrected for plant specific operations.
The resulting plot of leak size versus frequency is shown in the following figure.

Figure 2. Case Study 2 Parts Count

Curve fits were performed for each segment and the equations solved for the frequency
tolerance criteria. Resulting leak sizes are shown in the following table (rounding up to the
next 0.05 with a minimum size of 0.25 and a maximum size of 2).
GCPS 2015
__________________________________________________________________________

Table 2. Case Study 2 Leak Sizes

Comparing the leak sizes to a typical 2 leak size assumption shows that three scenarios
would be expected to have much reduced consequences. Since it is typical practice to use
a maximum leak size of 2 in a consequence-based FSS where leak sizes are assumed, the
company decided to use that as a maximum leak size because it would have been the
assumption if the MDL study were not performed.

Note that this illustrates a flaw in the leak size assumption practice that the risks from all
leaks are equal. If the company had performed a typical FSS assuming a leak size less than
2 based on the perceived simplicity of operations (e.g. 1 leak size assumption),
consequences from one of these scenarios may have been underestimated.

4. Case Study 3
An asphalt terminal decided to perform a MDL to use as a basis or leak sizes in a FSS since
they believed their risks to be less than that typical for most refineries and petrochemical
plants. The company had no risk tolerance criteria or LOPA single event frequency criteria
so frequency tolerance criterion was developed based on common industry maximum
individual risk criteria of 1 E-4 fatalities per year. Factors such as building occupancy,
number of hazardous segments, number of independent hazards, ignition likelihood, wind
directions, acceptable occupant vulnerability and an additional factor of safety were used
to generate a single event frequency tolerance.

Their process was segmented into 4 sections with potentially hazardous consequences. A
parts count was performed for each segment and frequencies were corrected for plant
specific operations. The resulting plot of leak size versus frequency is shown in the
following figure.
GCPS 2015
__________________________________________________________________________

Figure 3. Case Study 3 Parts Count

Curve fits were performed for each segment and the equations solved for the frequency
tolerance criteria. Resulting leak sizes are shown in the following table (rounding up to the
next 0.05).

Table 3. Case Study 3 Leak Sizes


Segment Leak Size (in)
1 1.90
2 0.25
3 0.15
4 0.20

Had the company used 2 leaks for all leak sizes, they would have greatly overestimated
the risk from three of their scenarios and may have spent money remediating risks that
werent present. Had the company used a smaller leak size such as 1 or less, they would
have underestimated the risk from one of their scenarios and still overestimated the risk
from three scenarios.

5 Conclusion
In Facility Siting, no approach can predict maximum risk unless many scenarios with a
range of leak sizes are evaluated and resulting vulnerabilities used with failure frequencies
GCPS 2015
__________________________________________________________________________

to calculate occupant fatalities for comparison with a risk tolerance criteria (i.e. QRA).
Many facilities have not determined acceptable risk tolerance nor wish to spend the
resources on a QRA so they chose to perform consequence-based FSSs where a leak size
is assumed and only the scenario locations with the potential maximum consequences are
evaluated. One leak size (usually smaller than full bore rupture) is typically used in FSSs
because the likelihood from full bore rupture scenarios is assumed to be low.

Typical consequence-based approaches have an inherent flaw in their leak size


assumptions. This single leak size assumption equates the frequency of all leaks, which is
not realistic. The MDL approach provides a means to evaluate leaks from sources with
varying frequency such that higher frequency leak sources are modeled using larger leak
sizes than lower frequency leaks.

This paper shows real-world application of the MDL to illustrate the advantage of using
risk calculations to determine leak size frequencies for facilities that are not prepared to
document risk in terms of acceptable fatalities per year and which are not complex or not
high risk (e.g. not refineries or large petrochemical facilities). It illustrates there may be
some facilities where a single leak size would be a valid assumption for some scenarios but
there may be other scenarios where it would be overly conservative and not a good
allocation of limited resources.

6. References.

[1] G.A. Fitzgerald, M.W. Stahl, D.J. Campbell, F. Nouri, R.L. Montgomery, The
Maximum Design Leak (MDL) Approach to Leak Size Selection, presented at the
2011 Mary Kay OConnor Process Safety Center International Symposium,
October 25, 2011.

You might also like