Risk Based Assessment of a 25,000 Ton

Ammonia Storage Tank

Cytec's ammonia storage tank can be used f or at least five more years of service without internal inspection.
A risk based inspection by Capstone Engineering Services and inspection reports support this conclusion.

D. Daly, Gregory J. Deis, D. Mclntyre, and R. Smallwood

Cytec, LA

Ronald E. Frishmuth
Consultant

Doug Paneitz
Capstone

Introduction The tank was inspected in 1989 after 11 years of service.

(AH and Smallwood, 1991) An AE test was performed dur-

C
ytec Industries' Portier Complex is an 800 acre, ing filling for a hydrotest (48 ft, 4 in. fill height) that locat-
continuous process facility located just outside of ed one "C" Monpac source emission.
New Orleans on the Mississippi River. The plant The tank was entered and WFMT testing was completed
employs more than 440 people in six operating units, which on all side wall and bottom weld seams.
use ammonia or ammonia derivatives to manufacture other One 10 in. long weld defect was found in the floor to
chemicals. Ammonia is inventoried in a 25 K atmospheric side-wall seam at the location of the "C" Monpac AE emis-
tank. sion. Three small (7 mm long) crack-like indications were
Our ammonia storage tank, with a nameplate capacity of found on the upper shell rings where clips had been weld-
25,825 tons, was designed to API 620, 5th edition, ed and removed. 25 sets of cracks were found in and adja-
Supplement No. 3 and commissioned in 1978. The ammo- cent to the bottom plate seam welds. Two penetrated the
nia storage tank is elevated on pile caps, and is 148 ft hi floor.
diameter with a straight side-wall height of 70 ft, 8 in. The No environmental damage was found in the side-wall
side-wall is made up of nine courses of plate steel. The tank during the inspections. The cracks found in the floor were
has a self-supporting dome roof. The maximum product fill attributed to the startup conditions. The acoustic emission
height is 68 ft 5 in. Tank nozzles located in the bottom side- test completed during the inspection in 1989 should have
wall course were furnace stress relieved prior to installa- identified any fabrication defects that would cause cracking
tion. in the tank.

AMMONIA TECHNICAL MANUAL 39 2002

 Since the last tank inspection, Cytec had incorporated the defect is significant and must be addressed. FFS was used
practice of setting internal inspections every 10 years, to determine when an evolving subcritical failure mecha-
based on the recommendation of API 510 "API 510 nism would become significant. Once it is known how and
Pressure Vessel Inspection Code: Maintenance Inspection, when a failure may occur and what the consequences of the
Rating, Repair, and Alteration." However, recent changes failure are, then managing the risks becomes relatively
to API 510 allow the maximum inspection cycle to be mod- straightforward.
ified, predicated on (1) quantifying the risks associated to
operating a given vessel and (2) balancing that risk against RBI Overview
the owner/operator's risk tolerance. Cytec used the RBI
process to extend internal inspection cycle. RBI defines the risk of operating equipment using a pri-
oritization process that combines the likelihood and the
Background consequence of failure. The prioritization process can vary
from qualitative to quantitative, but the intent is to charac-
The purpose of internal tank inspections is to detect and, terize each equipment item with a likelihood and conse-
where necessary, repair corrosive or mechanical damage quence of failure. A semi-quantitative method for risk rat-
before such damage results in a tank failure. The ammonia ing, using a 5 x 5 matrix as shown in Figure 1, which has
industry has long recognized the potential to crack steel been demonstrated to have an appropriate balance between
tanks in the presence of oxygen and ammonia during the being simple to implement, while still providing adequate
commissioning process. Hypothetically, environmentally discrimination between the evaluated items.
induced cracks from the commissioning, remnant construc- The likelihood of a specific piece of equipment failing
tion defects, or fatigue induced cracks could propagate and from inspectable causes is a function of the identified fail-
lead to a tank failure. The acceptance that these hypotheti- ure mechanisms, the rate of deterioration, and the effec-
cal problems were active would warrant the ten year tiveness of the prior inspection history in detecting and
inspection cycle. The 1989 inspection of Cytec's ammonia properly characterizing the damage to the equipment.
storage tank did find blatant construction flaws and some The safety consequence of failure of the item is a func-
indications of cracking, but no flaws were found to threat- tion of the type of fluid the equipment contains, how much
en the integrity of the tank. might be released in the event of a failure, and the effect of
In today's market, it makes no sense to make an expen- such a release. Well-defined methods have been developed
sive inspection based on an arbitrary deadline. The 1989 to characterize the safety consequence of a release.
post construction inspection findings mitigated the con- Once the Risk Rating has been identified for a piece of
cerns of hypothetical problems today. The key question equipment, an appropriate inspection and maintenance
was: "Is the tank fit to continue in service without entry?" strategy can be developed. An understanding of the identi-
The validity of deferring the inspection of the tank was fied failure mechanisms and their severity will allow the
provided by two changes in the API codes: API 510 was selection of appropriate inspection methods. The inspec-
recently modified to allow the practices of API 581, "Risk tion frequency and coverage can also be adjusted, based on
Based Inspection" (RBI) to determine the internal inspec- the results of the risk analysis.
tion cycle, and the new API 579, "Fitness For Service" The Risk Rating estimates the probability of failure along
(FFS), which provided the technical basis for evaluating with the failure consequence. For fixed equipment and pip-
hypothetical concerns. These two standards supplemented ing systems, the most significant failure mode is usually
and complemented each other. A team of engineers was loss of containment, which was the basis for this study.
formed to evaluate the course of action for Cytec. Using the method described, both the consequence and
RBI quantifies risk by (1) requiring owner knowledge probability rankings are calibrated in order-of-magnitude
and equipment data to be compiled, (2) identifying known steps. This approach is considered a semi-quantitative
failure mechanisms, (3) using science and engineering fun- method.
damentals to define and document the tank limits, and (4)
predicting failure consequences.
FFS is intended to provide a guide to determine when a

AMMONIA TECHNICAL MANUAL 40 2002

 plexity of analysis being performed. Level 1 is the most
conservative, but is easiest to use. Practitioners usually pro-
ceed sequentially from a Level 1 to a Level 3 analysis
(unless otherwise directed by the assessment techniques) if
A-CMaMrophle
B-Vtiy S«ftou»
the current assessment level does not provide an acceptable
C-Sirious result, or a clear course of action cannot be determined
D-Significant
E-MInor (such as rating, repair or replacement of the component is
PraluhlIMy required). A general definition of each assessment level fol-
1-V«ry High
2-High lows.
3-Mod«rat*
«-Low
• Level 1: Conservative screening criteria that can be uti-
6-V«ylow
E D C B A lized with a minimum amount of inspection or component
Consequence Ranking information.
• Level 2: A more detailed evaluation which produces
results that are more accurate than those from a Level 1
Figure 1. Reliability Based Mechanical Integrity rank
assessment. However, more data and more detailed calcu-
rating.
lations are used in the evaluation.
• Level 3: the most detailed and accurate evaluation,
Fitness for Service Overview requiring the most detailed inspection and component
information, and typically based on numerical analysis
Fitness for Service (FFS) can be defined as a methodolo- techniques such as "Finite Element Analysis" (FEA).
gy to demonstrate equipment integrity by evaluating in- An example of a simplified Failure Assessment Diagram
service components containing known or suspected (FAD) is shown in Figure 2. This is an example of an analy-
defects. It includes quantitative engineering evaluation to sis of a circumferential flaw in a pressure vessel. The Level
determine suitability of equipment for continued use and 1 assessment of this hypothetical indicates that additional
predict remaining life. The intent of FFS is to ensure older analysis is required. Generally, when a defect fails a Level
equipment continues to operate safely using technically 1 analysis, it can be modeled with a Level 2 analysis, which
sound and consistent assessment procedures to maintain will normally require more data and a more sophisticated
availability and long-term economic viability. analysis, but generally will be less conservative. Normally,
API 579 Recommended Practice for FSS was issued in going to a Level 2 or Level 3 analysis requires specialized
January 2000. Many companies are using the methodolo- computer analysis that can either be developed by the user
gy on a regular basis. The assessment procedures are or purchased as a commercial software package.
included for the following types of flaws or damage mech-
anisms: brittle fracture, general metal loss, local metal loss,
pitting corrosion, blisters and laminations, weld misalign-
ment and shell distortions, crack-like flaws, high tempera-
ture operation and creep, and fire damage. More than one
flaw type or damage mechanism may need to be evaluated
to determine the FSS of a component or piece of equip-
ment.

Level of detail

Three Levels of assessment are provided in each evalua- O.2 0.3 O.4 0.5 O.6 O.7 0.8
tion section of the FFS document. In general each assess- NORMALIZED FLAW LENGTH (9/it)
ment level provides a balance between conservatism, the
amount of information required for the evaluation, the skill Figure 2. Level assessment of circumferential flaws in
of the personnel performing the assessment, and the com- cylindrical shells.

AMMONIA TECHNICAL MANUAL 41 2002

 Figure 3 shows the results of a Level 2 FFS analysis of tank." In practice, "knowledge of the tank" means "docu-
the same flaw that was illustrated in the Level 1 example, mentation of your tank." If there is no supporting docu-
and provides the results, including the location of the mentation, then the RBI process becomes more complicat-
assessment point on a FAD. ed as the requirements to gain equivalent documented
The graphical results indicate the location of the analysis information. For example, if the mill reports are missing
point on a FAD. The diagram plots the assessment point on for the steel plate, then boat samples of the steel may have
a graph of the Toughness Ratio (Kr) vs. the Load Ratio to be collected and tested to complete a fracture mechanics
(Lr)2. (The Lr is the ratio of applied load or stress to the study.
load or stress required to have a fully plastic limit load fail- What was the steel that the tank was built from? How
ure. The Kr axis is the ratio of the applied stress intensity was it built? What are the stresses experienced by the tank?
to the stress intensity required to generate a brittle fracture.) Has water been added to the tank to inhibit oxygen stress
cracking? What has the fill/empty frequency been? What
was the condition of the tank the last time it was inspected?
How is it likely to react to the stresses that it experiences?
BéSC
These are some of the questions that we answered as we
developed our knowledge base for the RBI assessment.
Our knowledge of the storage tank was collated and used
N to build a finite-element (FEA) mathematical model of the
\ tank (a commercial finite-element software package by
£ A
\ ANSYS was used to develop the model). Areas of high
stress were identified, and a fracture mechanics study of the
steel at these locations was performed. Areas that did not
meet the "leak before failure" criteria, that is, the crack
length that would initiate a brittle failure was less than the
thickness of steel, were further evaluated against a crack
Figure 3. Fracture graphic failure assessment diagram. growth model.
A previous inspection of the tank gave us the confidence
to accept the crack model, that is, no active cracks were
In this example, the analysis indicates the vessel can con- found, and we had confidence that we had no cracks greater
tinue to be operated safely (assuming the crack is not grow- than 0.1 hi. in length, providing the initial conditions for
ing), but it is approaching the boundary where brittle fail- the crack growth model.
ure would be predicted. This detailed work allowed us to reduce the probability
API 579 recommends a Level 3 analysis for atmospheric of a failure occurring, which allowed us to assign a "relia-
or low-pressure storage tanks that contain a refrigerated bility risk rating" of our storage tank to "medium high."
product. A Level 3 assessment was used in the evaluation
of the tank in this study. The assessment included detailed Risk based inspection assessment
evaluation of the factors that control the susceptibility to
brittle fracture: stress, flaw size, and material toughness. The RBI assessment was completed using software
The risk analysis, which considers likelihood and conse- developed by Capstone Engineering Services, Inc. The
quences of a brittle fracture, is also part of the Level 3 RBI model for the tank evaluated the roof, each shell
assessment. course, and the bottom as separate components.
Due to the extreme consequences of a tank failure, the
Engineering Evaluation Process lowest "risk rating" that the tank would rate is a "medium
high," as per Figure 1. The essence of the RBI study
The objective of the engineering study was to refine the revolved around reducing the probability of failure by
probability of failure to the RBI matrix. The first step in increasing the knowledge of active failure mechanisms in
the process can be summarized by the phrase "know your the tank; that is, eliminating the failure modes that aren't

AMMONIA TECHNICAL MANUAL 42 2002

 Table 1. Present Tank Criticality Ranking Summary

Consequence Probability Criticality

Category Category Category

Roof C 4 M
Course No. 9 A 4 MH
Course No. 8 A 4 MH
Course No. 7 A 4 MH
Course No. 6 A 4 MH
Course No. 5 A 4 MH
Course No. 4 A 4 MH
Course No. 3 A 4 MH
Course No. 2 A 4 MH
Course No. 1 A 4 MH Figure 4. FEA model of ammonia tank.
Floor B 4 M
then further evaluation, which may include an internal
present from the RBI evaluation and quantifying those fail- inspection, will be warranted.
ure modes that are present.
The analysis of the tank included an assessment of the Consequence analysis for the tank
potential for internal corrosion and ammonia stress corro-
sion cracking. The potential for environmental cracking in For this analysis, liquid releases were assumed to be
the sidewalls was calculated in the environmental model by toxic and was assumed to leak from the lower shell cours-
setting the potential for cracking to high for the first year es and the floor. Vapor was assumed to leak from the roof
and reducing it to very low after one year. An internal cor- and the top two shell courses.
rosion rate of 0.0005 in. per year was assumed for the Since the bottom of the tank does not sit on the ground,
analysis. the tank bottom was not modeled as a persistent leak to the
A summary of the results of the analysis of the tank is ground, but as environmental release.
shown in Table 1.
The criticality was also calculated for the tank at five and Tank FEA model
ten years in the future. There was no change in any of the
categories at five years. At ten years, the probability cate- The finite-element analysis was used to identify the
gory of all of the shell courses changed to a "3" category stresses in the tank. The southern half of the tank was mod-
due to the potential for stress corrosion cracking (SCC) and eled and is represented in Figure 4. (The tank was modeled
their criticality category changed to the "high" category. as an 180° section of the tank as shown in Figure 4. The cut
This part of the study validated the Cytec 10 year internal edges of the side wall, top, bottom and the roof support
Standard. beams were set to be symmetry planes, resulting in the
The potential for SCC cracking in the side-walls was missing half of the model beings identical to the half
evaluated by the RBI environmental model which set the shown. All bottom nodes of the tank were constrained in
potential for cracking to "high" for the first year and reduc- the vertical direction.) The model contains 5,324 shell ele-
ing it to "very low" after one year. Based on that conserva- ment nodes and 220 3-D elastic beam elements. The self-
tive model, the probability category of the tank due to supporting roof structure is made of beam elements. The
Ammonia SCC will increase to a "3" twenty years after the model also includes three sidewall nozzles. A 4-in. suction
last inspection. The criticality rating for the tank will nozzle, as will be shown in Figure 6, has piping loads
remain at medium-high until that time, and then it will applied.
become "high." When the criticality rating rates "high," The model was run with various loadings. The loading
scenario that illustrates the salient findings of the study are:

AMMONIA TECHNICAL MANUAL 43 2002

 Reliability Technology, Inc. of Boulder, CO) fracture
(1) High ammonia level (hydrostatic loading) in the tank. mechanics computer code was used to answer two critical
(2) The metal temperature was set at -28°F at the ammo- questions: "Could the tank be expected to leak before a cat-
nia level and +260°F above the ammonia level. This sharp astrophic brittle fracture will occur?" and "Will a subcriti-
change in metal temperature is a worst-case assumption. cal crack grow to a critical size during the expected time
(3) 1 psi over pressure plus pipe loading on the 4-in. suc- the vessel will be in service?"
tion nozzle.
(4) 100 mph wind from the west. Leak Before Break Analysis: Fracture
Mechanics Analysis
Finite-element analysis results
The FM analysis of the three high stress areas was per-
The three highest stresses were: formed using commercial fracture mechanic computer pro-
(a) A high membrane stress at the circumferential welds gram.
between the first and second shell courses located in the Side wall girth weld between first and second cours-
tank side wall is shown in Figure 5. This is an important es (Figure 5). In this case, the residual stresses from
location because it was not thermally stress relieved. welding are assumed present, because no stress relief was
(b) In the hoop direction, located at the top of the man- performed. As a result, the distribution of the residual
way cylinder where it intersects with the ID of the repad on stresses parallel to the weld must be established.
the OD of the tank, shown in Figure 6. The hoop stress at Fortunately, during the 1989 internal inspection, surface
this location is primarily membrane tension. weld residual stresses were determined in actual welds.
(c) Bending stresses through the vessel wall on the ID of This data allowed the through wall residual stress field to
the tank located at the bottom intersection of the suction be estimated.
nozzle repad (Figure 7). The FG analysis for this case included an extra step. The
These three critical areas merited a Level 2 FM analysis. FG code indicates that a short deep surface defect cannot
Fracturegraphic (sold and maintained by Structural reach the depth of 0.784 in., the side-wall thickness. Thus,
a leak before break condition cannot be attained. To further

AKSÏS 5.7
AD6 S 2001 ASSÏS S..7
09:38:55 ADC 5 2001
XODAL SOH.TIOB 08:37:36
«ODAIi SOIiUTIOB
STEP=4
STEP=4
SUB =1
SUB =1
TIME=4
ST (AVG)
SI (AVGl
BOTTOM TOP
RSYS=11 DMX -.859202
DHX =.859202 SMH =-6593
SM =-23772 SHX =32820
SMX =27805 A =-4403
=-20906 B =-23.791
=-15176 C =4355
=-9445 D =8735
=-3714 E =13114
=2016 I =17493
=7747 G =21872
=13478 H =26251
=19209 I -=30631
=24939

TEMP <• 1 PSI PRESSURE + PIPE LOAD + 62.8 IT KM * 100 MPH W WIBD

Figure 5. Hoop direction max. stress on tank ID located Figure 6. Max. principal stress on tank OD oriented
at side wall girth between 1st and 2nd course near in noop direction at manvvay cylinder to repad
manway on left side of plot. intersection.

AMMONIA TECHNICAL MANUAL 44 2002

 AK3Ï3 5.7
The assumed surface defect size for this case was 0.95 in.
ADS 5 2001
08:43:01 long by 0.475 in. in depth, which fell just inside the FAD
SOBAL SGLDITiOH
STE8=4 limits, very close to the critical size. Since the depth of the
sira =1
TIM£=4 assumed defect is only 0.475 in compared to the plate
31 (AYG)
BÓÏTCM thickness of 0.784 in, a defect at this location is predicted
DUX =.859202
swi =-323.925 to break prior to development of a leak in the tank. That is,
SMX -51949
.A =2580 the leak before break criteria is not met.
S -8338 .
C =14196
D =20005

* =Ï16Z1
Fatigue Crack Growth Analysis
G =37429
B -43231
I =49045 (The crack growth mechanism is assumed to be cycle-
TÏPH JlüH
dependent. If both cyclic and time-dependent mechanism
AKEA3
play a role, then both should be considered. In the case of
an ammonia tank, it has been established that the presence
of oxygen plays a key role in determining if SCC will
occur. If oxygen is present with the ammonia in the tank,
Figure 7. Max. principal stress on tank ID (oriented in SCC is likely to occur very quickly and result in severe
global Z direction) at edge of suction nozzle repad. cracking. For this tank, it is highly unlikely that any oxygen
exists in the tank; any time-dependent SCC, if it were to
investigate the situation, another FG ran was made where a exist, would have stopped a short time after the tank was
2 to 1 aspect ratio (length to depth) was assumed and a FAD placed in service. The probability of time-dependent SCC
analysis performed. The assumed surface defect is 1.0 in. is believed to be quite low.)
in surface length by 0.5 in. in depth, close to the limits that Two of the three regions selected for leak before break
predict brittle fracture. analysis failed to meet the leak before break criteria, and
Side wall at bottom of suction nozzle repad (Figure were evaluated for cyclic fatigue crack growth, propagated
6). In this case, various FG analyses were used to find by filling and emptying the tank. For this analysis, crack
defect sizes that fall just on or inside the FAD limits. The growth following Paris law behavior was assumed (No
defect was assumed to be 1.85 in. in length, completely crack growth data is available for the steels used in the
through the wall of the tank and it lies on the FAD limits. tank. A default, worst case, assumption was made using
Since the wall thickness of the shell side wall at this loca- guidelines from the British Standards Institution Published
tion is 0.784 in, the assumed through wall crack is longer Document (PD) 6493-1991.)
than 2 times the 0.784 or 1.568 in. Therefore, in this case, ( 1) Side wall girth weld between 1st and 2nd courses that
leak before break is predicted. did not meet the leak before break criteria was a transverse
Manwoy cylinder to repad ID intersection. (See Figure crack forming on the girth weld between the 1st and 2nd
7. j (The crack growth mechanism is assumed to be cycle- courses. The initial assumed defect at the start of the load
dependent. If stress corrosion cracking (SCC) is the mech- cycling was 0.2 in. long by 0.1 in. depth, the smallest defect
anism for crack growth, the crack extension would be time- that would not have been found during the previous inspec-
dependent. If both a cyclic and time-dependent mechanism tion. Fatigue analysis shows that it would require about
play a role, then both should be considered. In the case of 32,000 cycles to grow the assumed crack to a critical size.
an ammonia tank, it has been established that the presence This translates to 2,666 years of service life at a rate of 12
of oxygen plays a key role in determining if stress corro- cycles per year.
sion cracking will occur. If oxygen is present with the (2) Manway cyclinder to repad ID intersection (Figure
ammonia in the tank, SCC is likely to occur very quickly 6). In this case, the initial assumed defect was assumed to
and results in severe cracking. For this tank, it is highly be 0.1 in. deep by 0.2 in. in surface length, the same as the
unlikely that any oxygen exists in the tank; any time- previous case. For this size of initial defect, it was predict-
dependent SCC, if it were to exist, would have stopped a ed that it would require nearly 25,000 cycles to grow this
short time after the tank was placed in service. The proba- defect to a critical size. If the cyclic loading history from
bility of time-dependent SCC is believed to be quite low.) the alternate filling and emptying the tank is 12 cycles per

AMMONIA TECHNICAL MANUAL 45 2002

 year, it would require 2,083 years to grow the assumed has demonstrated that small subcritical defects in these crit-
defect to a critical size. ical areas can be expected to grow at a very slow rate, even
The analysis results clearly show that the tank does not under the worst case conditions. The tank is expected to be
meet required criteria for leak before break conditions. In safe throughout its expected lifetime.
spite of this, even in the most highly high stressed areas of The future integrity of the tank can be evaluated by a
the tank, the calculated crack growth rate of any existing properly conducted acoustic emission test instead of an
subcritical defects in the tank would be expected that grow internal inspection. Within five years (a maximum of 15
at a very small rate, and not threaten the integrity of the years from the previous internal inspection), an acoustic
tank over its service life. emission study will be conducted and the criticality analy-
sis will be re-run to incorporate the findings. If no audible
Summary and Conclusions defects are detected, continued operation of the tank should
be warranted for an additional period of time, without fur-
Cytec's ammonia storage tank can be used for at least ther internal inspection.
five more years of service without internal inspection. A
risk based inspection documented by Capstone Literature Cited
Engineering Services and inspection reports supports this
conclusion. All, S. B., and R. E. Smallwood, "Inspection of
The fracture mechanics study added two salient pieces of Anhydrous Ammonia Atmospheric Pressure Storage
knowledge about the tank, which reduced the probability of Tank," Ammonia Plant Safety & Related Facilities, Vol.
failure: (1) the tank does not meet leak before break crite- 31, AIChE, New York (1991).
ria in at least two critical areas; (2) crack growth analysis

QUESTIONS AND ANSWERS

Najib I. Alwabli, Al-Jubail Fertilizer Co. (SAMAD): How Do we have to build a spare tank?
long it takes to do an inspection of a 20,000 ton ammonia Deis: We receive ammonia by barges on the Mississippi
storage tank starting from de-commissioning to commis- River. Our plan would have required multiple barges rotat-
sioning? ing in and out of dock facilities.
Gregory J. Deis, Cytec: Cytec last inspected our ammo- Moustafa A. M. Abdou, Abu Qir Fertilizer & Chemical
nia storage tank in 1989. The tank was out of service for Industries Co,: Please give a brief idea about the acoustic
approximately 2.5 months. Based on my conversations emission test.
with others who have made internal inspections, the Deis: Acoustic emission (AE) testing is based on the fact
decommissioning, inspection and recommissioning could that cracks make noise if they are active and growing. In
probably be done within 5-6 weeks. However, work other practice, several sensitive microphones are attached to the
than the inspection usually extend the turnaround to 2-3 shell of the tank and the tank stressed, usually by an
months. empty/fill cycle. Noise from growing cracks is detected by
Alwabli: What are the options to continue running the the microphones and analyzed by a computer to determine
ammonia plant without the storage tank during inspection? where the crack is.

AMMONIA TECHNICAL MANUAL 46 2002