U.S.

C HEMICAL S AFETY AND HAZARD INVESTIGATION B OARD

INVESTIGATION REPORT

THERMAL DECOMPOSITION INCIDENT

(3 Killed)

BP AMOCO POLYMERS, INC.

AUGUSTA, GEORGIA
MARCH 13, 2001

KEY ISSUES
RECOGNITION OF REACTIVE HAZARDS
LEARNING FROM NEAR-MISS INCIDENTS
OPENING OF PROCESS EQUIPMENT

REPORT NO. 2001-03-I-GA

ISSUE DATE: JUNE 2002
2
 Abstract

T his investigation report examines the thermal decomposition

incident that occurred on March 13, 2001, at the BP Amoco
Polymers, Inc., plant in Augusta, Georgia. Three workers were killed.
This report identifies the root and contributing causes of the incident
and makes recommendations for recognizing reactive hazards, learning
from near-miss incidents, and opening process equipment.
The U.S. Chemical Safety and Hazard Investigation Board (CSB) is an
independent Federal agency whose mission is to ensure the safety of
workers and the public by preventing or minimizing the effects of
chemical incidents. CSB is a scientific investigative organization; it is not
an enforcement or regulatory body. Established by the Clean Air Act
Amendments of 1990, CSB is responsible for determining the root and
contributing causes of accidents, issuing safety recommendations,
studying chemical safety issues, and evaluating the effectiveness of other
government agencies involved in chemical safety. No part of the
conclusions, findings, or recommendations of CSB relating to any
chemical incident may be admitted as evidence or used in any action or Information about available
suit for damages arising out of any matter mentioned in an investigation publications may be obtained by
contacting:
report (see 42 U.S.C. § 7412(r)(6)(G)). CSB makes public its actions
U.S. Chemical Safety and Hazard
and decisions through investigation reports, summary reports, safety Investigation Board
bulletins, safety recommendations, special technical publications, and Office of Prevention, Outreach,
statistical reviews. More information about CSB may be found on the and Policy
World Wide Web at http://www.chemsafety.gov. 2175 K Street NW, Suite 400
Washington, DC 20037
(202) 261-7600
CSB publications may be
purchased from:
National Technical Information
Service
5285 Port Royal Road
Springfield, VA 22161
(800) 553-NTIS or
(703) 487-4600
Email: info@ntis.fedworld.gov
For international orders, see:
http://www.ntis.gov/support/
cooperat.htm.
Salus Populi Est Lex Suprema For this report, refer to NTIS
People’s Safety is the Highest Law number PB2002-107365.

3
4
 Contents

EXECUTIVE SUMMARY ............................................................. 9

1.0 INTRODUCTION ...............................................................19

1.1 Background ...........................................................................19
1.2 Investigative Process ...............................................................19
1.3 BP Amoco Polymers Augusta Facility .......................................19
1.4 History of Process Development .............................................20

2.0 PROCESS DESCRIPTION ...................................................21

2.1 Chemistry.............................................................................21
2.2 Process Configuration ............................................................22
2.3 Polymer Catch Tank and Reactor Knockout Pot ........................23

3.0 DESCRIPTION OF INCIDENT ...........................................27

3.1 Pre-Incident Events ................................................................27
3.1.1 Attempted Startup ......................................................27
3.1.2 Aborted Startup ..........................................................28
3.1.3 Preparation for Removal of Polymer
Catch Tank Cover ......................................................29
3.2 Incident Description ...............................................................30
3.2.1 Cover Failure and Expulsion of Hot Plastic ..................30
3.2.2 Subsequent Fire .........................................................30
3.2.3 Emergency Response .................................................31
3.3 Reconstructive Analysis ..........................................................32
3.3.1 Extruder Failure .........................................................32
3.3.2 Polymer Catch Tank ...................................................32
3.3.3 Physical Testing and Chemistry ...................................34
3.3.4 Vessel Pressurization ...................................................35

4.0 ANALYSIS OF INCIDENT...................................................37

4.1 Reactive Chemical Hazard Management ................................37
4.2 Process Safety Information .....................................................39
4.3 Management of Change ........................................................41
4.4 Process Hazard Analysis ........................................................42
4.5 Design Deficiencies ................................................................44
4.6 Procedures for Safe Opening of Process Equipment .................47
4.7 Incident and Near-Miss Investigations .....................................48
4.7.1 Exploding Polymer Pods ..............................................48
4.7.2 Waste Polymer Fires ..................................................49
4.7.3 Inadequate Review of Cumulative Incidents .................50
4.8 Review of Similar Incident .....................................................51

5
Contents (cont’d)

5.0 REGULATORY ANALYSIS ...................................................53

5.1 Process Safety Management ...................................................53
5.2 Lockout/Tagout .....................................................................55

6.0 ROOT AND CONTRIBUTING CAUSES ............................57

6.1 Root Causes ..........................................................................57
6.2 Contributing Causes ..............................................................58

7.0 RECOMMENDATIONS ......................................................61

8.0 REFERENCES .....................................................................65

APPENDIX A: RECONSTRUCTION OF INCIDENT ..............67

A.1 Post-Incident Scene ...............................................................67
A.2 Extruder Failure ....................................................................71
A.3 Polymer Catch Tank ...............................................................72
A.4 Physical Testing and Chemistry...............................................78
A.5 Vessel Pressurization ..............................................................84

APPENDIX B: TIMELINE ..........................................................85

APPENDIX C: LOGIC DIAGRAM ..............................................93

6
 Figures

1 Highly Simplified Flow Diagram for Amodel Production .........22

2 KD-502 Polymer Catch Tank (simplified representation) ..........24
3 KD-502 Polymer Catch Tank (actual vessel with one
cover removed) ....................................................................24
4 Simplified Process Diagram Showing Polymer
Catch Tank and Reactor Knockout Pot ...................................25
5 Damaged Girder ..................................................................30
6 Polymer Catch Tank Cover ...................................................31
7 Polymer Plugging Polymer Catch Tank Vent Nozzle ................33
A-1 Polymer Plugging Polymer Catch Tank Vent Nozzle ................67
A-2 Failed Bolts From Polymer Catch Tank Cover ........................68
A-3 Reactor Knockout Pot With Internal Frame Visible .................68
A-4 Vent Line Obstructed by Polymer Layers ................................69
A-5 Polymer Scattered Throughout Work Area ............................70
A-6 Material Extracted From Polymer Catch Tank on
March 11, Prior to Incident ...................................................73
A-7 Normal Operation–Settlement of Particles .............................75
A-8 Polymer Catch Tank Filled Beyond Working Capacity–
Particles Entrained in Vent Vapor ...........................................75
A-9 Sample of Material Expelled From Polymer Catch Tank ..........79
A-10 Layer of Hardened Polymer Taken From Inside
Perimeter of Polymer Catch Tank ...........................................79
A-11 ARC Data for Amodel ..........................................................82

7
Acronyms and Abbreviations

AIChE American Institute of Chemical Engineers

ARC Accelerating rate calorimeter
ASTM American Society for Testing and Materials
BP British Petroleum
°C Degrees Celsius
CCPS Center for Chemical Process Safety
CFR Code of Federal Regulations
CPR Cardiopulmonary resuscitation
CSB U.S. Chemical Safety and Hazard Investigation Board
FEA Finite element analysis
HAZOP Hazard and operability
HSE Health and Safety Executive (United Kingdom)
IChemE Institution of Chemical Engineers (United Kingdom)
lb/gal Pounds per gallon
L.L.C. Limited liability company
LTA Less than adequate
MOC Management of change
MSDS Material safety data sheet
OSHA Occupational Safety and Health Administration
PCR Process change request
PHA Process hazard analysis
psia Pounds per square inch absolute
psig Pounds per square inch gage
PSM Process safety management
R&D Research and development
SA Societé anonyme

8
 Executive Summary

ES.1 Introduction
O n March 13, 2001, three people were killed as they opened a
process vessel containing hot plastic at the BP Amoco Polymers
(now SolvayAdvancedPolymers,L.L.C. 1
) plant in Augusta, Georgia.
They were unaware that the vessel was pressurized.
The workers were killed when the partially unbolted cover blew off the
vessel, expelling hot plastic. The force of the release caused some
nearby tubing to break. Hot fluid from the tubing ignited, resulting in a
fire. Because of the serious nature of this incident, the U.S. Chemical
Safety and Hazard Investigation Board (CSB) initiated an investigation
to determine the root and contributing causes and to issue recommen-
dations to help prevent similar occurrences.

ES.2 Incident
The BP Amoco Polymers Augusta facility produced plastics, including
Amodel, a hard but moldable high-performance nylon. Amodel is
manufactured by passing a solution of di-amines and di-carboxylic acids
through a series of reactors. The reaction is completed in an extruder, 2
and the material is then formed and cooled into solid pellets.
Workers were attempting to open a cover on a process vessel when the Twelve hours prior to the incident,
incident occurred. The vessel–referred to as the polymer catch tank an attempt was made to start the
(KD-502)–was designed to receive partially reacted waste Amodel production unit. After approximately
diverted from a chemical reactor during periods of startup and shut- 1 hour, the startup was aborted due
down. to problems with the extruder
Twelve hours prior to the incident, an attempt was made to start the downstream of the reactor–but not
production unit. After approximately 1 hour, the startup was aborted before an unusually large amount of
due to problems with the extruder downstream of the reactor–but not partially reacted material had been
before an unusually large amount of partially reacted material had been sent to the polymer catch tank.
sent to the polymer catch tank.
Hot molten plastic inside the polymer catch tank continued to react and
also began to slowly decompose, thereby generating gases and causing
the contents to foam. The material expanded as foaming continued,
and eventually the entire tank was filled. The material then forced its
way into connecting pipes, including the normal and emergency vents.

————————–
1
Ownership of the facility was transferred from BP Amoco Polymers, Inc., to Solvay
Advanced Polymers, L.L.C., on November 1, 2001, seven months after the incident.
2
The extruder is a machine composed of turning screws within a stationary barrel.
It is used to simultaneously heat, knead, and pump plastic.

9
 Once in the pipes, the plastic solidified as it cooled. A hardened layer
of plastic 3 to 5 inches thick also formed around the entire inner wall
of the tank. However, the core of the plastic mass remained hot and
molten, and likely continued to decompose over several hours,
generating gases that pressurized the vessel.
Before opening the polymer catch tank, personnel may have relied on
a pressure gauge and a transmitter on the vent piping from the vessel to
ascertain whether it was under pressure. They also knew that the
process was shut down. However, any reading from the pressure
gauge would likely have been unreliable because plastic had entered the
vent line and solidified.
On previous occasions, the polymer catch tank contained no pressure
when it was opened. Varying amounts of plastic were found inside;
sometimes the plastic was hot, but it was always solid. Expecting that
to be the case again, the workers proceeded to remove the 44 bolts
from the cover. When half of the bolts had been removed, the cover
suddenly blew off. Hot plastic spewed throughout the area, traveling as
far as 70 feet. The cover and the expelled plastic struck the workers,
killing them.
The force created by the ejection of gas and plastic propelled the
polymer catch tank backward and bent the attached piping. A section
of hot oil supply tubing (370 degrees Celsius [°C]) for a heating jacket
on the inlet line from the reactor knockout pot to the catch tank broke,
and the fluid spilled into the area. A flammable vapor cloud formed
and ignited within a few minutes. Several hours of firefighting were
required to extinguish the fire.

ES.3 Key Findings

1. Operations and technical support staff at the manufacturing site
were unaware that Amodel could decompose and generate high
pressure when held at elevated temperatures for an extended time.
2. Product performance testing conducted by the company’s research
and development (R&D) group demonstrated that the plastic was
susceptible to thermal decomposition at processing temperatures.
However, the manufacturing process was not subjected to a
specialized design review to identify hazards from unintended and

10
 uncontrolled reactions; and the risks posed by decomposition of
the plastic were not recognized.
3. More than the normal amount of hot plastic entered the polymer
catch tank during the aborted startup. Reactions and decomposi-
tion of the material produced gases, which caused the plastic to
foam and expand. The expanded plastic forced its way into Operations and technical
connecting pipes, where it solidified and plugged the inlet to the support staff at the manufacturing
vent line. Once this occurred, the gases could not escape and the site were unaware that [the plastic]
vessel became pressurized. could decompose and generate high
4. Since 1993, there had been several near-miss incidents involving pressure when held at elevated
both the polymer catch tank and the waste plastic. Had these temperatures for an extended time.
incidents been more thoroughly investigated, they could have
provided insight into the hazards associated with the operation,
as noted below:
a. In earlier years, large lumps of solidifying waste plastic had
burst, hurtling fragments a considerable distance. Investigations
were not thorough enough to determine that the hot, molten
material within the lumps was most likely continuing to react
and decompose, creating gas and pressure. The expanded plastic forced its way
into connecting pipes, where it
b. On one occasion after the polymer catch tank was opened, the
solidified and plugged the inlet to the
waste plastic inside spontaneously caught fire. This also
vent line. Once this occurred, the
happened when a companion vessel was opened. On two
gases could not escape and
other occasions, waste plastic removed from these vessels
the vessel became pressurized.
spontaneously caught fire after being placed in a dumpster.
Investigations did not identify that the fires were likely related
to the formation of volatile and flammable substances from
thermal decomposition of the plastic.
c. The polymer catch tank had been overfilled on prior occasions,
resulting in plugging of connected piping and the pressure relief
device. No adequate measures were developed to prevent
recurrence.
d. During inspections, the pressure relief device was found to be
fouled with solid plastic, which could have rendered it inopera-
tive. The potential consequences of such fouling were not
analyzed, and no adequate measures were developed to
prevent recurrence.

11
 5. Operating experience revealed that the design of the polymer
catch tank did not afford a practical and reliable method for
workers to check for hazards before opening the vessel. Drains
. . . the design of the polymer catch were often plugged with solidified plastic, making it impossible to
tank did not afford a practical and verify the absence of pressure or hazardous chemicals in accor-
reliable method for workers to check dance with Occupational Safety and Health Administration
for hazards before opening the vessel. (OSHA) lockout/tagout regulations and standard company safety
procedures for opening process equipment. Nevertheless, these
design deficiencies were not corrected, and no standard practice
was developed for preparing the vessel for opening.
6. Process safety information inadequately described the design basis
and operating principles for the polymer catch tank. There was no
discussion of the means by which overfilling could occur and its
consequences.
7. The possibility of overfilling was increased when the original
startup procedures were revised; the diversion of output from the
reactor to the polymer catch tank was extended from 30 to 50
The consequences of the extruder
minutes.3 This modification of procedures was not subjected to a
failing to start or shutting down
management of change (MOC) review.
during routine operation
were not considered. 8. Process hazard analyses were inadequate. For example:
n Credible scenarios for overfilling or pressurizing the polymer
catch tank were not recognized.
n The consequences of the extruder failing to start or shutting
down during routine operation were not considered.
n Reactivity and decomposition hazards were treated
inadequately or not at all.
9. Operators had no direct measure of the extent to which the
polymer catch tank had been filled. The design for the level
detector on the vessel was inadequate.

————————–
3
A 1999 training document specified a startup time of 40 to 45 minutes prior to
production through the extruder. However, CSB interviews of operators and supervi-
sors indicate with consistency that the practice at the time of the incident was to run for
50 minutes before directing the process flow to the extruder. In prior years, it had been
30 minutes.

12
10. The failure of the extruder to run forced the startup to be aborted.
Although the prestartup checklist called for verification of the
operability of the extruder, that procedural check was omitted on
this occasion.

ES.4 Root Causes

1. Amoco, the developer of the Amodel process, did not
adequately review the conceptual process design to identify
chemical reaction hazards.
Neither Amoco’s R&D department nor the process design depart-
ment had a systematic procedure specifically for identifying and
controlling hazards from unintended or uncontrolled chemical
reactions.
2. The Augusta facility did not have an adequate review process Neither Amoco’s R&D department
for correcting design deficiencies. nor the process design department had
The Augusta site was the first and only commercial manufacturing a systematic procedure specifically for
facility for Amodel. Several problems in design of the polymer identifying and controlling hazards
catch tank became apparent with operating experience. Opera- from unintended or uncontrolled
tions management did not ensure that deficiencies were corrected chemical reactions.
in a timely manner.
n Workers were unable to follow established company policies
for lockout/tagout and equipment opening because the plugged
drains on the polymer catch tank prevented them from verifying
the absence of pressure in the tank.
n Previous occurrences of overfilling and plastic entrainment into
connected piping indicated that the polymer catch tank was too
small to handle foreseeable process upsets. Workers were unable to follow
established company policies for
n The level indicating device for the polymer catch tank was
lockout/tagout and equipment opening
unreliable.
because the plugged drains on the
3. The Augusta site system for investigating incidents and near- polymer catch tank prevented them
miss incidents did not adequately identify causes or related from verifying the absence
hazards. This information was needed to correct the design of pressure in the tank.
and operating deficiencies that led to the recurrence of
incidents.

13
 n Sound technical theories were not developed to explain the
spontaneous ignition of waste plastic or the phenomenon
Incidents and near misses tended to be
whereby lumps of waste plastic burst.
treated as isolated events.
Management did not have a review n Incidents and near misses tended to be treated as isolated
system to detect trends and patterns events. Management did not have a review system to detect
among incidents. trends and patterns among incidents.
n The polymer catch tank had been overfilled and the vent lines
plugged on other occasions. No effective measures were
developed to prevent recurrence.
n Fires occurred at the extruder on numerous occasions. No
effective countermeasures were developed.

ES.5 Contributing
1. Hazard analyses of the Amodel process were inadequate and
Causes incomplete.
n Reactivity hazards, such as unintended reactions, were not
examined in the final design-phase hazard analysis.
n The extruder operation and its overall impact on the rest of
Reactivity hazards, such as unintended the process were not adequately reviewed during the formal
reactions, were not examined in the hazard analysis conducted during design and construction.
final design-phase hazard analysis.
n Credible scenarios by which the polymer catch tank could
become overfilled were not identified.
2. Design documentation did not adequately describe the Amodel
process.
n The process description did not adequately explain the design
basis and operating principles for the polymer catch tank; as a
result, misunderstandings of these characteristics developed.
The maximum fill level was not clearly specified. No warn-
ings were provided about the consequences of overfilling.
n Operations management did not update the documentation to
reflect changes in procedures and practices.
3. Equipment opening procedures did not specify what actions to
take when safety precautions could not be met.

14
 On the day of the incident and frequently during the life of the
Amodel process, it was not possible to verify the absence of On the day of the incident and
pressure in the polymer catch tank, as required by company frequently during the life of the Amodel
procedures and OSHA lockout/tagout regulations. Solid polymer process, it was not possible to verify
plugged the drain nozzles, which should have been used to verify the absence of pressure in the polymer
the absence of pressure. Without a policy requiring management catch tank, as required by company
review and authorization in such circumstances, workers pro- procedures and OSHA
ceeded to open the vessel despite the lack of positive verification lockout/tagout regulations.
that it was not pressurized.
4. Revisions to operating procedures were not subjected to
management of change reviews to evaluate safety effects.
Flow was originally directed to the polymer catch tank for 30
minutes during startup. Without adequate consideration of
possible adverse consequences, this time was later extended to
50 minutes–which increased the amount of material to be
disposed of to the vessel.

ES.6 Recommendations
Solvay Advanced Polymers, L.L.C.
1. Examine the manufacturing businesses acquired from BP Amoco
Performance Polymers and ensure that a systematic safety review
procedure is developed and implemented for identifying and
controlling hazards from unintended chemical reactions. Addition-
ally, ensure that reactive hazards are identified and evaluated:
n During product R&D, during conceptual design of a new
process, and during detailed design of a new process.
n Before changes are made to existing equipment or process
chemistry.
Communicate the results of this review to the workforce.
2. Ensure that a program is in place at facilities acquired from BP
Amoco Performance Polymers to systematically review the hazards
associated with new and modified processes and equipment as

15
 operating experience accrues. Ensure that facilities correct all
identified design, operation, and maintenance deficiencies.
Verify that operating experience does not invalidate the design
basis for equipment.
3. Revise the Material Safety Data Sheet (MSDS) for Amodel to
warn of the hazards of accumulating large masses of molten
polymer. Communicate the MSDS changes to current and past
customers (who may retain inventories of this product).

Solvay Advanced Polymers, L.L.C.,

Augusta Facility
1. Implement a program to conduct periodic management reviews
of incidents and near-miss incidents. Look for trends and
patterns among incidents. Address root causes and implement
and track corrective measures.
2. Revise process safety information to include:
n Information regarding the decomposition reactions of
Amodel.
n Design intent, basis, capacity, and limitations of equipment.
n Hazards and consequences of deviations from design intent
and operating limits.
3. Revalidate hazard analyses for the Amodel process to address:
n Credible deviations from process intent and their
consequences.
n Hazards associated with startup and shutdown operations.
n Prevention of accumulations of potentially hazardous masses
of polymer.

16
4. Revise your lockout/tagout program to ensure that equipment is
rendered safe prior to opening for maintenance. At a minimum,
ensure that equipment opening procedures contain a stop work
provision that requires higher levels of management review and
approval when safe opening conditions, such as equipment
depressurization, cannot be verified.
5. Ensure that your management of change policy applies to opera-
tional and procedural modifications.

BP Chemicals Group
Communicate the findings of this report to your chemical and plastics
manufacturing facilities in North America.

American Chemistry Council

Society of Plastics Engineers
Communicate the findings of this report to your membership.

17
18
 1.0 Introduction

1.1 Background
O n March 13, 2001, at approximately 2:40 am, three workers
were fatally injured when pressurized material was suddenly
expelled from the polymer catch tank (KD-502) at the BP Amoco
Polymers plant in Augusta, Georgia. The two operators and one
maintenance technician were attempting to remove a cover from the
catch tank, expecting to extract hardened plastic. After half of the 44
bolts had been removed, pressure inside the vessel caused the remain-
ing bolts to break and the cover to fly off. The workers were struck by
the cover and the expelled plastic. All three men were employees of
BP Amoco.
The U.S. Chemical Safety and Hazard Investigation Board (CSB)
launched an investigation to determine root and contributing causes
and to issue recommendations to help prevent similar occurrences.

1.2 Investigative
CSB investigators examined physical evidence at the scene, took
samples, conducted interviews with current and former employees of Process
Amoco and BP Amoco, and reviewed company documents and scien-
tific literature. CSB also arranged for experiments to be performed at
two plastic testing laboratories.

1.3 BP Amoco
The BP Amoco facility is located on a 100-acre site in Augusta,
Georgia, and produces various specialty plastics. It employs Polymers
approximately 250 full-time personnel. Amodel–the trade name for Augusta Facility
the product involved in this incident–is a high-performance nylon
thermoplastic 1 used for automotive, electronic, consumer, and medical
device applications.
The manufacturing site was initially developed by Dartco Manufactur-
ing in 1984. Amoco purchased the facility in 1988 and constructed
the first Amodel process there in 1992. In 1998, Amoco and British
Petroleum merged to form BP Amoco. In November 2001, BP Amoco
and Belgium-based Solvay SA completed an asset exchange; the
Augusta site and associated business were transferred to Solvay
Advanced Polymers, L.L.C.

————————–
1
Nylon is a type of plastic characterized by the chemistry of repeating amide groups.
Thermoplastic is a plastic that melts when heated.

19
1.4 History of Amoco initiated development of the process for making Amodel in the
Process Development early 1980s. A pilot plant2 built in Naperville, Illinois, was the
company’s first manufacturing effort for a nylon plastic.
In 1987, a semiworks3 production facility (also referred to as the
“experimental unit”) was built in Greenville, South Carolina. In 1990,
the advanced polymers and fibers research and development (R&D)
organization and the pilot plant were relocated from Naperville to
Alpharetta, Georgia. A semiworks unit was then constructed on the
Augusta site. It commenced operation in 1992, and was followed a
year later by the company’s first and only commercial production unit.
Amoco was the recipient of several patents for the Amodel
manufacturing process.

————————–
2
A pilot plant is a small-scale manufacturing plant used to evaluate a new process
technology.
3
A semiworks production facility is larger than a pilot plant, but smaller than a commer-
cial production unit. The purpose of a semiworks facility is to produce product samples
for commercial purposes.

20
 2.0 Process Description

2.1 Chemistry
A model is a polymer4 prepared by the reaction of an aliphatic
di-amine with aromatic and aliphatic di-carboxylic acids.
It is referred to as a polyphthalamide nylon because it is derived from
isophthalic and terephthalic acids.
When dissolved and heated in the presence of a catalyst, the raw
materials react and combine to form a molecular chain. With each
addition to the chain, a molecule of water is formed as a byproduct.
An equilibrium exists between the forward and reverse reactions,
represented as:
Di-amine + Di-carboxylic acid ⇔ Polymer chain + Water
The polymerization reaction occurs in a liquid solution and is mildly
exothermic.5 The water created is ultimately converted to steam
because the reaction is conducted at high temperature. The vaporizing
water absorbs energy released by the reaction. If the energy required to
form the steam is taken into account, the overall reaction is considered
to be slightly endothermic.6

————————–
4
A polymer is a natural or synthetic chainlike molecule formed by the union of at least
five identical smaller molecules (monomers). Polymers usually contain many more than
five monomers, and some may contain hundreds or thousands of monomers in each
chain.
5
An exothermic reaction is characterized by the release of energy.
6
An endothermic reaction is characterized by the consumption of energy.

21
2.2 Process Figure 1 is a highly simplified sketch of the process for making Amodel.
Configuration

Figure 1. Highly simplified flow diagram for Amodel production.

Because the chemicals used to make the plastic are solids at ambient
temperature, the first step in the process is to dissolve them. The result
is a salt solution of raw materials that is then fed to a pressurized
prereaction vessel, where the temperature is increased to between 190
and 230 degrees Celsius (°C). The reaction proceeds partially at this
stage to form a prepolymer. 7
A high-pressure pump sends the prepolymer forward to the preheater,
where the temperature is raised to between 290° and 340°C. The
increasing temperature further advances the reaction. The solution
then passes through a flow orifice to decrease pressure, which causes
some of the water in the solution to rapidly vaporize to steam.

————————–
7
Prepolymer has a molecular weight lower than the finished product.

22
The dispersion of prepolymer and steam is then fed to a tubular reactor,
where the temperature is increased still further. The process fluid
remains in the reactor for less than a minute. When it exits, the overall
reaction is more than 80 percent complete.
An extruder8 finishes the reaction to a polymer of the desired molecular
weight. The product from the extruder is then pelletized, cooled, and
stored.

2.3 Polymer Catch

During normal process operation, the prepolymer produced in the
reactor is fed to the extruder. During periods of startup and shutdown,
Tank and Reactor
and when there are mechanical difficulties with the extruder, the Knockout Pot
reactor effluent is diverted to a waste collection vessel–referred to as the
polymer catch tank (Figures 2 and 3). The catch tank is a 750-gallon
horizontal cylindrical steel vessel with a flat cover bolted to each end.
Before startup, it is empty except for a removable branched metal frame
that serves as a surface on which polymer agglomerates.
The polymer catch tank acts as a chamber for separating prepolymer
from vapor. The rapidly flowing process stream from the reactor enters
the catch tank through a 4-inch pipe nozzle located on one of the end During periods of startup and shut-
covers. The velocity of the stream decreases as it enters, causing liquid down, and when there are
droplets of polymer to fall and settle inside the vessel. The prepolymer mechanical difficulties with the
accumulates in the vessel; the vapor portion of flow exits through a 6- extruder, the reactor effluent is
inch vent nozzle located on top of the cylindrical shell. diverted to a waste collection vessel–
referred to as the polymer catch tank.
Three other piping connections are located on top of the polymer catch
tank. One is used for the pressure relief device. The other two are
utility connections for nitrogen and service water, which can be used
for purging and flushing the vessel.

————————–
8
An extruder is a machine consisting of turning screws within a stationary barrel. It is
used to simultaneously heat, knead, and pump plastic.

23
 Figure 2. KD-502 polymer catch tank (simplified representation).

Figure 3. KD-502 polymer catch tank (actual vessel with one cover removed).

24
The material that enters the polymer catch tank from the reactor is hot;
because the tank is neither heated nor cooled, the rate of cooling
depends on a number of factors, including the total amount of material
present and ambient conditions.
Typically, during startup of the process, the effluent from the reactor is
directedtothepolymercatchtankfor50minut 9
ebefore
s being di-
verted to the extruder (Figure 4). Prepolymer that accumulates inside
the catch tank during startup normally cools and solidifies.

Figure 4. Simplified process diagram showing polymer catch tank and reactor knockout pot.

————————–
9
A 1999 training document specified a startup time of 40 to 45 minutes prior to
production through the extruder. However, CSB interviews of operators and supervi-
sors indicate with consistency that the practice at the time of the incident was to run for
50 minutes before directing the process flow to the extruder.

25
When the manufacturing process is shut down, the flow of raw
materials is discontinued; a solvent is used to displace and dissolve the
prepolymer. Early in this stage, the flow from the outlet of the reactor
continues to be directed to the extruder. Once solvent is detected at the
extruder, the flow from the reactor is diverted to the polymer catch tank.
The solvent is capable of dissolving polymer residues in the reactor,
piping, and associated systems. As the solution of polymer approaches
the polymer catch tank, the pressure decreases and the solvent vapor-
izes almost completely upon reaching the tank. Inside, prepolymer
separates from the vaporized solvent and settles as a liquid on top of the
typically solid residue of startup material already present. The solvent
vapors exit through a vent line on top of the tank. The amount of
prepolymer added to the vessel during a normal shutdown is small
compared to the amount sent during startup.
Further flushing of the equipment with water and steam completes the
shutdown. Both the solvent flush and the water flush also remove heat
from the polymer inside the polymer catch tank; the plastic typically
solidifies. Later, one of the covers is removed from the tank, and the
accumulated solid waste on and around the metal frame is pulled out
and disposed of.
The reactor knockout pot (KD-501) is a waste vessel very similar in
design to the polymer catch tank. Figure 4 depicts its arrangement and
connectivity to the process. The emergency pressure relief devices at
the outlet of the reactor discharge to the knockout pot. The reactor
effluent can be directed to the knockout pot when the catch tank is
too full.

26
 3.0 Description of Incident

3.1 Pre-Incident Events

T he Amodel unit operated normally during the week prior to the
incident. However, it was shut down on Saturday, March 10,
2001, for repair of an equipment failure in the extruder. Restart was
scheduled for March 12.
As was typical following a shutdown, during the early morning hours of
March 11, the polymer catch tank was prepared for opening and the
cover was removed. The frame inside was extracted, and the tank was
cleaned of accumulated polymer. Maintenance personnel reported that
the vessel appeared to be clean and in good condition. The cover was Although the polymer catch tank was
subsequently closed and rebolted. designed to have a level detector, the
device was often damaged during
Although the polymer catch tank was designed to have a level detector, cleanout of hardened polymer. There
the device was often damaged during cleanout of hardened polymer. was no level detector in the vessel
There was no level detector in the vessel when it was closed on when it was closed on March 11.
March 11.

3.1.1 Attempted Startup

Operators commenced prestartup checks on March 12 at 6:45 am.
The lead operator was told that the extruder had been run the night
before with a purge material to clean the screws. Data confirm that the
extruder was purged at 2:19 am on March 12. The lead operator
thought that it was not necessary to repurge the extruder, and his
supervisor agreed.
As part of the normal startup checklist, the extruder was to be prerun
for at least 1 to 2 minutes to verify operability before startup of the
reaction section of the process. This activity was not identified as
complete on the checklist, and control system logs show no data
indicating that the extruder was prerun, in accordance with procedure,
from the time it was purged early on the morning of March 12 until the
failed startup attempt.
Raw material feed was introduced to the reaction section of the unit
at 1:29 pm on March 12. Temperatures and pressures were within
normal ranges. As specified by standard procedure, the initial polymer
produced was directed to–and began to accumulate in–the polymer
catch tank. The procedure called for this mode of operation to con-
tinue for 50 minutes before the flow of material from the reactor was

27
 diverted from the catch tank to the extruder. To finalize startup, opera-
tors attempted to start the extruder at 2:17 pm, and it was discovered
that the screws would not turn.

Over the next 25 minutes, personnel Unit supervisors and the production coordinator were present at the
made several attempts to diagnose unit during the attempted startup, and they were immediately notified of
and resolve the problem [with the the extruder problem. Maintenance was called to assist. Over the next
extruder]. During this time, molten 25 minutes, personnel made several attempts to diagnose and resolve
polymer continued to be directed to the problem. During this time, molten polymer continued to be directed
the polymer catch tank. to the polymer catch tank.

3.1.2 Aborted Startup

At 2:35 pm on March 12, with the extruder malfunction still unre-
solved, supervision made a decision to abort the startup. Operators
stopped the flow of raw materials to the reaction line. Flushing solvent
was injected at 2:41 pm to begin cleaning the piping and equipment of
accumulated polymer. Because the extruder was not operating, the
molten polymer continued to accumulate in the polymer catch tank.
The flushing operation continued normally until approximately
3:45 pm, at which time an engineer noticed vapor leaking from the top
of the bolted cover on the polymer catch tank between the 11 and
2 o’clock positions. Plant personnel described the leaking vapors as
characteristic of the solvent.
Supervision was made aware of the leak, and a decision was made to
divert the process flow from the polymer catch tank to the adjacent
reactor knockout pot. This was accomplished at approximately
3:55 pm. From this point until 2:40 am on March 13–when the
incident occurred–no additional material entered the catch tank.

28
Shortly after flow was diverted to the reactor knockout pot, it too began
to leak vapor from its cover. At 6:53 pm on March 12, the solvent flush
was completed, and water was circulated through the unit. The water
passed through the knockout pot in the form of steam. Because the
polymer catch tank had been bypassed earlier, the hot polymer inside
did not receive any cooling that the steam flush could have provided.
Operators continued to monitor temperatures and pressures for the next
several hours. At 11:21 pm, the water flush was discontinued, com-
pleting the shutdown.

3.1.3 Preparation for Removal

of Polymer Catch Tank Cover
Instructions were left for the night shift to empty the polymer catch tank
and the reactor knockout pot of polymer that had accumulated during
the aborted startup. A maintenance technician arrived at 2:15 am on Other than the nitrogen line, no
March 13 to do the work. connections to the polymer
catch tank were locked or tagged.
Operators closed the valves that supplied nitrogen to the two vessels
and placed energy isolation tags on them. Other than the nitrogen line,
no connections to the polymer catch tank were locked or tagged. It is
unknown whether an attempt was made to use the drain valves on the
tank to ascertain the absence of pressure or liquid inside.
An equipmentisolationlis twas filled out by the operator and signed
10

by the operator and the maintenance technician. Two operators

accompanied the maintenance technician to the catch tank to assist with
opening and cleaning it.

————————–
10
The equipment isolation list is a written record describing the devices used to de-
energize and secure a machine, equipment item, or system prior to the commencement
of maintenance activities. The devices may include valves, controls, electrical
switchgear, locks, and warning tags.

29
3.2 Incident Description

3.2.1 Cover Failure and

Expulsion of Hot Plastic
The 1,750-pound cover blew off the
polymer catch tank . . . A mass of hot Forty-four 1 1/8-inch bolts11 were secured around the circumference
plastic was ejected from the of the flanged cover of the polymer catch tank. At 2:25 am on
48-inch-diameter tank opening. March 13, with the assistance of two operators, the maintenance
technician began removing bolts with a pneumatic impact wrench.
The removal of each bolt gradually reduced the restraining capacity of
the cover. At 2:36 am, after 22 of the bolts had been removed–ranging
from the 1:30 to 7:30 o’clock position on the cover–the remaining bolts
suddenly broke, resulting in an explosive release of pressure.
The 1,750-pound cover blew off the polymer catch tank, flying upward
and to the north. After striking a girder12 on the overhead rain canopy,
the cover landed approximately 14 feet from its original position
(Figures 5 and 6). A mass of hot plastic was ejected from the 48-inch-
diameter tank opening. The material struck the three workers and
scattered widely.
Witnesses described the subsequent formation of a plume of white
smoke or steam in the Amodel unit, but no visible fire after the first
explosion. The maintenance technician and both operators suffered
severe impact trauma. Two men died at the scene, and the other was
pronounced dead on arrival at the Medical College of Georgia.

3.2.2 Subsequent Fire

The force created by the rapid expulsion of material from the polymer
catch tank propelled it backward, causing the attached piping to bend.
A section of hot oil supply tubing for a heating jacket on the inlet line to
the catch tank broke; the hot oil escaped and formed a vapor cloud,
Figure 5. Damaged girder. which ignited at 2:42 am.

————————–
11
The bolts were actually 1 1/8-inch threaded studs with nuts welded on one end.
12
A girder is a horizontal structural member used to stiffen the frame of a building.

30
 Figure 6. Polymer catch tank cover.

Witnesses who responded to the sound of the first explosion described

the second event as a large orange fireball. Following the initial flash,
the release of hot oil continued to feed the fire, which burned for
several hours in an area behind the polymer catch tank.

3.2.3 Emergency Response

Following the initial explosion, the plant emergency response team,
the plant fire brigade, and the Richmond County Fire Department
responded to the incident.
The area water deluge system and manual hose streams were used to
control the fire. The victims were recovered, and cardiopulmonary
resuscitation (CPR) was administered to one of them. After emergency
responders isolated the hot oil system, the fire was extinguished at about
8:15 am on March 13.

31
3.3 Reconstructive
Analysis
3.3.1 Extruder Failure
Inspection of the extruder after the incident revealed that significant
quantities of ash had accumulated inside it. As described in Appendix
A, the ash was most likely the result of an internal fire involving residual
purging material. (A small fire had been observed at the extruder during
the early morning hours of March 12.) Because the extruder was not
When the extruder experienced designed to convey powders, the accumulated ash probably caused the
certain mechanical difficulties, screws to bind when attempts were made to start the machine later that
it was necessary to divert the reactor afternoon.
effluent and to accumulate polymer in Statements from several operators indicated that this was the first time
the polymer catch tank or the reactor the extruder failed to run coincidentally with a startup of the reaction
knockout pot. On at least two prior section. However, the machine or its components were known to
occasions, this resulted in overfilling occasionally fail during routine operation. It was an extruder compo-
one of the two vessels and plugging nent failure during a production run that resulted in the plant shutdown
their common vent line. on March 10. In early February 2001, a 2-day series of instrument
failures on the extruder motor also culminated in shutdown of the
process.
BP Amoco personnel were aware of the possibility of extruder malfunc-
tions during production, whatever the cause. When the extruder
experienced certain mechanical difficulties, it was necessary to divert
the reactor effluent and to accumulate polymer in the polymer catch
tank or the reactor knockout pot. On at least two prior occasions, this
resulted in overfilling one of the two vessels and plugging their common
vent line.

. . . a solid layer of polymer 3 to 5

inches thick remained in the vessel
and coated the entire inside surface, 3.3.2 Polymer Catch Tank
including the rear cover. The vessel
nozzles, including the vent nozzle, Investigators found that most of the contents of the polymer catch tank
were completely plugged with were expelled during the incident. However, a solid layer of polymer
polymer. 3 to 5 inches thick remained in the vessel and coated the entire inside
surface, including the rear cover (see cover photograph). The vessel
nozzles, including the vent nozzle, were completely plugged with
polymer (Figure 7). Significant amounts of polymer were found in the
vent system. Appendix A provides a more extensive discussion of the
physical evidence.

32
 Figure 7. Polymer plugging the polymer
catch tank vent nozzle.

As a result of the aborted startup on

March 12, more than twice the normal
amount of hot prepolymer was
During a typical startup, the amount of prepolymer sent to the polymer directed to the tank . . .
catch tank filled less than half the vessel. As a result of the aborted
startup on March 12, more than twice the normal amount of hot
prepolymer was directed to the tank;13 this had not occurred uninter-
ruptedly on any previous occasion. Consequently, the amount of
thermal energy inside the vessel was most likely greater than it had ever
been before.
The properties of Amodel make it an effective insulator. 14 Although
the plastic near the walls of the polymer catch tank cooled and solidi- Although the plastic near the walls of
fied, the core likely remained hot and molten for an extended time. the polymer catch tank cooled and
As the vessel filled beyond its working capacity, its capability to separate solidified, the core likely remained hot
the incoming vapor from molten prepolymer decreased. Polymer and molten for an extended time. As
droplets likely became entrained within the vapor stream, left the vessel, the vessel filled beyond its working
and escaped into the vent system,15 where they could be expected to capacity, its capability to separate the
deposit on the pipe walls and harden. incoming vapor from molten
prepolymer decreased.

————————–
13
This included the prepolymer made during startup and produced during extruder
troubleshooting, plus the hot material displaced from the reaction section of the plant
during the solvent flush following the decision to abort the startup.
14
Amodel readily solidifies; the solid polymer resists the transmission of heat.
15
Appendix A provides further details on the mechanism of polymer entrainment.

33
 Continued accumulation of polymer in the piping likely created a partial
blockage that ultimately restricted the flow of vapor through the vent
system. This restriction probably caused the polymer catch tank to
become pressurized during the solvent flush. The accumulated pres-
sure in the tank exceeded the sealing capability of the bolted joint
between the 4-foot-diameter cover and the vessel flange, which ex-
plains why solvent vapor leaked from the top of the cover at 3:45 pm
on March 12.
Although flow was restricted, it is unlikely that the vent line was com-
pletely plugged off at this time because solvent continued to flow
through the system and the polymer catch tank. However, plugging in
the vent system was extensive and continued to worsen even several
hours after the solvent flush was discontinued.
The emergency pressure relief valve on the reactor knockout pot finally
activated16 during water flushing operations at approximately 8:10 pm
on March 12. It is probable that increased plugging in the common
vent piping eventually created an obstruction substantial enough to
cause the back pressure in the knockout pot to exceed the set pressure
of the relief valve.

. . . these reaction characteristics led

CSB investigators to conclude that the 3.3.3 Physical Testing and Chemistry
source of pressure inside the polymer
catch tank was gas formation due to The public literature on nylon plastics describes them as susceptible to
unintended chemical reactions. thermal degradation and side reactions. Both mechanisms produce
gases, which can include carbon dioxide, ammonia, and water vapor.
Together with other evidence, these reaction characteristics led CSB
investigators to conclude that the source of pressure inside the polymer
catch tank was gas formation due to unintended chemical reactions.
Appendix A provides further details on the chemical testing conducted
by CSB.

————————–
16
From 6:53 to 11:21 pm on March 12, water was circulated through the unit and
steam was vented through the reactor knockout pot, which shares vent piping with the
polymer catch tank.

34
As part of its investigation, CSB arranged for testing of typical extruded
Amodel. One sample was tested by thermogravimetry17 under condi-
tions similar to those expected in the interior of the polymer catch tank Further testing using an accelerating
(i.e., 315°C for up to 12 hours). The tests revealed that Amodel rate calorimeter provided other
undergoes a significant weight loss at these conditions. CSB concludes significant evidence of the thermal
that the weight loss is due partly to decomposition of the plastic and decomposition potential of the plastic.
partly to side reactions. Further testing using an accelerating rate
calorimeter (ARC)18 provided other significant evidence of the thermal
decomposition potential of the plastic (Appendix A).
Samples were also taken from the layer of hardened plastic found on
the inside wall of the polymer catch tank. Tests carried out by BP
Amoco indicated that the polymer chains were highly branched, which
is strong evidence that side reactions occurred inside the vessel in Significant concentrations of carbon
addition to decomposition. monoxide, carbon dioxide, ammonia,
and water were found–all of which
The company also analyzed the vapors generated when Amodel was
are consistent with decomposition or
heated to 340°C. Significant concentrations of carbon monoxide,
side reactions. These substances are
carbon dioxide, ammonia, and water were found–all of which are
gases at moderate temperatures and
consistent with decomposition or side reactions. These substances are
would develop pressure in a
gases at moderate temperatures and would develop pressure in a closed
closed vessel.
vessel.

3.3.4 Vessel Pressurization

No additional material flowed into the polymer catch tank after
3:55 pm on March 12. Over the next several hours, reactions in the No additional material flowed into the
core of the vessel most likely generated vapor–thus causing the viscous, polymer catch tank after 3:55 pm on
molten contents to foam and expand, occupying the entire volume of March 12. Over the next several
the catch tank. Further expansion likely pushed the material through the hours, reactions in the core of the
tank nozzles, and into the vent line and the emergency pressure relief vessel most likely generated vapor–
inlet–fully blocking them. As polymer reached the sides and upper thus causing the viscous, molten
contents to foam and expand . . .

————————–
17
Thermogravimetry is a method of measuring the change in mass of a substance on
heating. Losses of mass may be due to dehydration, decomposition, desorption,
vaporization, or chemical reactions that create gaseous products.
18
ARC is a trademark of Arthur D. Little, Inc.

35
surfaces of the vessel, it tended to cool and solidify. 19 It also cooled
and solidified in the nozzles and connected pipes. Once this occurred,
the evolving vapor had no escape pathway and it pressurized the catch
tank.

————————–
19
Although the polymer catch tank was insulated, it was not heated. Heat losses to the
environment would cause the polymer to solidify.

36
 4.0 Analysis of Incident

4.1 Reactive Chemical

B P Amoco Augusta site operations and technical support
personnel were unaware of the potential for a hazardous
decomposition reaction in the Amodel process. It was widely believed
Hazard Management
that no additional reactions occurred once material was sent to the
polymer catch tank.
Following the March 13, 2001, incident, however, investigators theo-
rized that chemical reactions likely occurred in the polymer catch tank,
though it was not readily apparent which specific reaction mechanisms
led to the accumulation of hazardous pressure. Chemical testing
conducted after the incident confirmed that crosslinking and decomposi-
tion reactions likely caused the pressure accumulation.
BP Amoco Augusta site operations
CSB investigators interviewed R&D chemists and engineers involved in
and technical support personnel
the Amodel project from its inception in the early 1980s. The develop-
were unaware of the potential for a
ment team did not conduct any research into the hazards of normal or
hazardous decomposition reaction in
unanticipated reactions in the Amodel process.20 Amoco did not have a
the Amodel process. It was widely
policy or procedure for such reviews in product research, process
believed that no additional reactions
development, or engineering design of a manufacturing process.21
occurred once material was sent to
The design of the Amodel process predates publication of the Institution the polymer catch tank.
of Chemical Engineers’ (IChemE) Chemical Reaction Hazards, A
Guide to Safety. However, this guidebook states:
. . . it is essential to allocate responsibilities and establish
procedures for the following key stages of assessing chemical
reaction hazards:
– defining process chemistry/operating conditions;
– investigating potential chemical reaction hazards;
– selecting and specifying safety measures; Chemical testing conducted after the
– implementing and maintaining safety measures incident confirmed that crosslinking
(Barton and Rogers, 1997; p. 14). and decomposition reactions likely
caused the pressure accumulation.

————————–
20
Such research typically includes literature search, screening calorimetry, and–as
needed–more sophisticated calorimetry of both intended and unintended reactions to
examine potentially hazardous releases of energy or gaseous byproducts.
21
Procedures for evaluating reactive hazards are described in Guidelines for Chemical
Reactivity Evaluation and Application to Process Design (CCPS, 1995a), Guidelines for
Safe Storage and Handling of Reactive Materials (CCPS, 1995b), Chemical Reaction
Hazards, A Guide to Safety (Barton and Rogers, 1997), and Designing and Operating
Safe Chemical Reaction Processes (HSE, 2000). Note: Although these guidelines were
not available at the time the Amodel process was designed, they are based, in part, on
industry practices established in the 1980s.

37
 As discussed in Appendix A, publicly available literature describes
potential decomposition and crosslinking reactions of nylon-family
polymers. If such literature had been carefully studied and a more
extensive hazards identification review undertaken, BP Amoco would
have likely discovered that Amodel decomposes and produces signifi-
. . . Amodel decomposes and produces
cant quantities of gaseous byproducts when held at or near its melting
significant quantities of gaseous
temperature for an extended time.
byproducts when held at or near
its melting temperature for an Had this information been known to the process development team,
extended time. they may have designed the Amodel process in such a way as to
minimize or eliminate the large-scale accumulation of molten polymer,
thus preventing this type of reactive incident. Such accumulation
contributed not only to the March 13 incident, but also to other inci-
dents involving burning or exploding waste polymer.
Accelerating rate calorimetry (Appendix A) shows that Amodel
decomposes and forms gaseous byproducts at temperatures below
349°C, albeit slowly and in an endothermic manner (i.e., heat
absorbing). Guidelines for Safe Storage and Handling of Reactive
. . . on BP Amoco’s larger commercial Materials describes the hazards of slow gas-forming reactions:
production scale, a buildup of gaseous
Numerous process incidents have occurred in closed systems
byproducts led to the catastrophic
where evolution of gas over time has caused pressure to build
March 13 incident, which up in the system to the point of containment failure (CCPS,
demonstrates that even slow, 1995b; p. 22).
endothermic reactions can result in
This phenomenon is not likely to represent a serious hazard when only
serious hazards.
small quantities of polymer are handled. However, on BP Amoco’s
larger commercial production scale, a buildup of gaseous byproducts
led to the catastrophic March 13 incident, which demonstrates that
even slow, endothermic reactions can result in serious hazards.
In 1990 and again in 1994, Amoco’s R&D function conducted
thermogravimetric tests of Amodel in support of product application
technology. The purpose of the tests was to examine ways to minimize
plate-out,22 a phenomenon in injection molding machines.23 The 1994
study demonstrated that, when heated to 335°C for 8 minutes, Amodel

————————–
22
Plate-out refers to unwanted material accumulation on the open faces of extrusion
dies. Because this buildup can interfere with molding, the processing line must be shut
down periodically to clean the die (Gander and Giacomin, 1997).
23
Injection molding machines are used to extrude and mold raw plastic into finished
parts.

38
polymers slowly degrade24 and produce volatile compounds that can
interfere with product quality.
The significance of this information with respect to process safety was
not recognized. Amoco did not apply its findings beyond product
application bulletins–except for the Material Safety Data Sheet (MSDS)
for Amodel (various grades), which states that the product is stable to
349°C and recommends avoiding higher temperatures to prevent
thermal decomposition. This threshold is slightly higher than the
highest temperature in the manufacturing process.
In 1990, an Amoco corporate engineer at the Naperville, Illinois,
research center convinced management of the need for a
thermophysical properties laboratory to conduct sophisticated testing
on chemical reactions. Although Amoco made a commitment to the
personnel and equipment needed to evaluate reactive hazards, no
complementary supporting policies and programs were developed to
guide business units.
The laboratory ultimately conducted little or no work on Amoco
processes and products. When the engineer retired in 1995, Amoco
donated the testing equipment to a university research institute.

4.2 Process Safety

The Augusta workforce was unaware of important principles of
operation for the polymer catch tank, which were absent from the Information
documented process safety information. For example:
n Design information did not explain that the catch tank was re-
quired to serve as a separator. Critical parameters that affect The Augusta workforce was unaware
separator performance, such as maximum operating level, were of important principles of operation for
not identified. the polymer catch tank, which were
n The design documentation stated that the catch tank would be absent from the documented
used during flushing operations, but did not describe its other process safety information.
required roles for startup and process upsets. It was stated that
the vessel would fill up in about an hour at a specified flow rate,
but there was no explanation of how much material was expected

————————–
24
The degradation rate was less than 0.05 percent per minute or 3 percent per hour.

39
 to collect during each stage of operation. There was no guidance
on how much space to leave in the vessel to ensure adequate
capacity for upset conditions.
n No information described the potential for foaming, even though
the solidified waste plastic routinely removed from the catch tank
appeared to be foamy. Foaming altered the density of the material.
Investigators noted that most plant technical personnel were
surprised that the vessel overfilled during the attempted and
aborted startup. When offering potential explanations to investiga-
tors, they referred to the density of pure Amodel when accounting
for their calculations. They did not recognize that foaming could
Operating limits for a maximum level
in the catch tank were not provided, lower the density of the plastic. (At a constant mass flow rate, a
nor were the consequences of lower density material more quickly fills a vessel.)
operating at abnormal conditions. n Operating limits for a maximum level in the catch tank were not
Because the vessel also acted as a provided, nor were the consequences of operating at abnormal
separator, operation beyond a certain conditions. Because the vessel also acted as a separator, operation
level entrained liquid plastic into the beyond a certain level entrained liquid plastic into the vent line.
vent line.
On the day of the incident, the polymer catch tank was filled beyond its
working capacity. Its performance as a separator diminished, allowing
liquid plastic to be carried along with vapors into the vent line. Al-
though foaming of the hot molten polymer in the core of the vessel
ultimately caused the vent and emergency lines to plug off completely,
overfilling was a contributing factor.
The Center for Chemical Process Safety (CCPS), an industry-
[Process safety information] preserves sponsored affiliate of the American Institute of Chemical Engineers
a record of design conditions and (AIChE), identifies process safety information (also referred to by CCPS
materials of construction for existing as “documentation of process knowledge”) as one of the 12 elements of
equipment, which helps ensure that chemical process safety management. This element includes such
operation and maintenance remains information as process and equipment design criteria.
faithful to the original intent.
CCPS (1989) cites several reasons for preserving process safety infor-
mation and making it available within the company, as noted below:
n It provides a basis for understanding how the process should be
operated and why it should be run in a given way.

40
 n It preserves a record of design conditions and materials of con-
struction for existing equipment, which helps ensure that operation
and maintenance remains faithful to the original intent.
n It allows recall of the rationale for key design decisions during The limited information provided in
major capital projects, which is useful for a variety of reasons, such Amoco’s process safety information for
as future projects and modifications. the polymer catch tank was inadequate
n It offers a “baseline” for evaluating process change. to inform personnel of the
consequences of operating outside
The limited information provided in Amoco’s process safety information
the original design intent.
for the polymer catch tank was inadequate to inform personnel of the
consequences of operating outside the original design intent. The
excessive accumulation of molten polymer, which contributed to the
March 13 incident, would have been less likely if process safety infor-
mation had been more complete.

4.3 Management of
During the mid to late 1990s, the time that startup material was directed
to the polymer catch tank was increased from 30 to 50 minutes. For Change
the first 20 minutes, the startup material primarily consists of water that
exits the catch tank as vapor through the vent line. The rate of polymer
accumulation is substantial for the remaining 30 minutes.
The extension of startup time to 50 minutes actually increased approxi-
mately threefold the amount of polymer deposited in the polymer catch
tank during startup. Correspondingly, it decreased the capability of the
vessel to hold material that might arrive if there were problems with the
extruder, thus increasing the possibility of overfilling. The extension of startup time to
50 minutes actually increased
The Augusta facility had a management system for evaluating the safety
approximately threefold the amount
consequences of process changes, referred to as the “process change
of polymer deposited in the polymer
request procedure” (PCR). It was applied to hardware changes but not
catch tank during startup.
necessarily to modifications to operating procedures and practices.
Chemical Process Safety: Learning From Case Histories states the
following about process change:

41
 A change requiring a process safety risk analysis before
implementing is any change (except ‘replacement in kind’) of
process chemicals, technology, equipment and procedures.
The risk analysis must ensure that the technical basis of the
change and the impact of the change on safety and health are
addressed (Sanders, 1999; p. 223).
No management of change (MOC) documents were available for the
procedural change that extended the startup time of the polymer catch
tank from 30 to 50 minutes.

4.4 Process Hazard

During design in 1990 and again in 1999, after several years of operat-
Analysis ing experience, Amoco conducted process hazard analyses25 of the
Amodel process using the hazard and operability (HAZOP)26 technique.
Both the polymer catch tank and the reactor knockout pot were consid-
ered during the analyses, but credible scenarios that could lead to excess
pressure or level were not identified.

The polymer catch tank and the In the 1990 HAZOP, the team identified failure of the extruder drive as
reactor knockout pot were the only a condition that could create a “no flow” situation, in which case it was
possible destinations to which the flow recommended that the polymer flow be stopped. The polymer catch
could be diverted. However, the tank and the reactor knockout pot were the only possible destinations to
HAZOP team did not consider this which the flow could be diverted. However, the HAZOP team did not
situation as a possible cause of excess consider this situation as a possible cause of excess polymer accumula-
polymer accumulation and level in tion and level in either vessel.
either vessel. The 1990 HAZOP study did not completely evaluate the extruder.
The team noted that insufficient design information was available to
conduct a full analysis and recommended a followup HAZOP of the
extruder once the engineering drawings were finalized. This analysis
was never conducted.

————————–
25
Process hazard analysis (PHA) is a structured examination of a chemical process to
identify factors that have the potential to create hazards; to uncover credible sequences
of events (scenarios) that could result in undesired consequences; to evaluate the
consequences of these scenarios should they occur; and to propose improvements, as
warranted, to equipment, procedures, and management systems that reduce or eliminate
the hazards, prevent the scenarios from occurring, or mitigate their consequences.
26
HAZOP makes use of guidewords to help identify deviations from normal, intended
operation that could result in potential hazards or operating problems.

42
In a 1993 incident, the polymer catch tank was overfilled when the
extruder malfunctioned. Polymer was carried into the vent line and
solidified, and the line had to be cut. Nevertheless, the 1999 HAZOP
still failed to identify the means by which an excess level could occur in
the vessel.
Overfilling contributed to the March 13, 2001, incident because it was
partly responsible for plugging of the vent and relief piping–which
confined the mass of plastic in the polymer catch tank. If the HAZOP
studies had identified credible scenarios involving vessel overfilling and
overpressurization due to extruder malfunctions, additional safeguards
could have been recommended to reduce the probability or severity of
the hazards. If overfilling had been effectively controlled, the sequence
of events that led to the March 13 incident would have been less likely–
even without knowledge of the decomposition hazard.
During the 1990 HAZOP, the team recognized that high pressure could
occur in the reactor knockout pot if the emergency pressure relief
system discharge line was plugged with solidified polymer. A recom-
mendation was made to provide a system to ensure that the line was
clear during operation, but no such system was established.
In the 1999 HAZOP, the team determined that the emergency pressure
relief system was an adequate safeguard in the event of plugging of the
normal vent. They did not recognize the credible scenario that both the
normal and emergency vents could simultaneously plug with polymer,
as occurred on March 13.
The 1990 HAZOP team recognized the possibility that personnel could
be injured when opening the polymer catch tank cover if pressure had
accumulated. A pressure gauge was installed on the vent line to
address this concern, as recommended. However, neither the 1990 . . . the HAZOP documentation
nor the 1999 HAZOP studies identified the possibility that polymer discusses no reactivity issues related
plugging in the line could render the gauge useless. to the polymer catch tank or
The potential hazards associated with startup and shutdown operations the reactor knockout pot.
were not addressed because this was not a requirement of the HAZOP
protocols.

43
 Neither analysis documented any concerns with reactivity of material in
the vessels.27 The 1990 study made no reference to chemical reactivity.
Although the 1999 study included the guidewords “properties of
chemicals”–which may have been intended to encompass reactivity–the
HAZOP documentation discusses no reactivity issues related to the
polymer catch tank or the reactor knockout pot.

4.5 Design Prior to startup of the Amodel unit, Amoco had no experience in full-
Deficiencies scale production of nylon plastics. The design for the commercial
manufacturing facility was based on several years of experience in pilot-
plant and semiworks operations. Problems with the polymer catch tank
became evident as full-scale operating experience accrued.
The polymer catch tank level indicating instrument was unreliable and
Prior to startup of the Amodel unit, prone to false indications. Additionally, it often broke when waste
Amoco had no experience in full-scale plastic was removed from the vessel, and frequently it was not replaced
production of nylon plastics. before restart. Although several alternative configurations were studied
and tested, no reliable method was identified to indicate the level of
polymer in the catch tank.
There were several instances when either the polymer catch tank or the
reactor knockout pot was completely filled. This occurred as early as
1993; during each occurrence, polymer was carried into the vent
Although several alternative
system. Because the plastic tends to rapidly solidify upon cooling,
configurations were studied and tested,
overfilling could render inoperative both the vent piping and the emer-
no reliable method was identified to
gency pressure relief devices–a potentially hazardous situation because
indicate the level of polymer in the
the upstream process is capable of applying pressure far higher than the
catch tank.
vessels’ maximum allowable working pressure. Operations manage-
ment recognized that the catch tank was too small as early as 1993, but
did not recognize all the consequences of overfilling.

————————–
27
Although HAZOP can be a useful aid in recognizing hazards, it is not a primary
mechanism to identify hazardous chemical properties. It is not a substitute for chemical
testing. PHA is only one component of effective reactive chemical hazard management.

44
Spring-operated pressure relief valves on the polymer catch tank and
the reactor knockout pot were intended to protect the vessels from
overpressure. However, neither relief valve was shielded from the
process fluid by a rupture disk28 upstream of the inlet. It is typical Because the plastic tends to rapidly
engineering practice to provide such protection where the process fluid solidify upon cooling, overfilling could
may solidify and foul the valve inlet. Rupture disks were used to protect render inoperative both the vent
relief valves on other upstream equipment. piping and the emergency pressure
relief devices . . .
The IChemE Relief Systems Handbook discusses the need for protect-
ing pressure relief valves with rupture disks. It states:
. . . the objective here is to protect the safety valve against
conditions in the pressurized system which may be corrosive,
fouling or arduous in some other way (Parry, 1998; p. 30).
Maintenance records show that the relief valve on the polymer catch
tank was machined and repaired in June 1993 because of polymer
fouling. The valve was put back in service, but it required repair again
just 2 months later. Similar damage occurred in 1995. The valve was
reconditioned more often than any other relief valve in the Amodel unit.
The relief valve for the reactor knockout pot was reconditioned twice in
the same period.
Despite this damage history, neither valve had been shop inspected
from 1995 until the time of the incident. Problems with the relief
system were symptomatic of design deficiencies with the polymer catch
tank; however, there is no conclusive evidence that inadequate pressure
relief capability directly contributed to the March 13 incident.29
Double block and bleed valves30 on the inlet line and vent line of the
polymer catch tank were intended to provide secure isolation of the
vessel from the operating process during maintenance. Each of the

————————–
28
A rupture disk is a thin disk composed of metal or composite material that is designed
to break at a predetermined pressure.
29
As discussed in Appendix A, there is considerable uncertainty as to the pressure inside
the polymer catch tank at the time of failure; it is not known whether it exceeded the
vessel’s maximum allowable working pressure.
30
Double block and bleed refers to an arrangement where two inline valves are placed
in series with a drain valve between them. To provide isolation, the two inline valves
are closed and the drain is opened. Leakage across the first inline valve is apparent as
flow through the drain.

45
 vessel end covers was equipped with a ram-type31 valve for use as a
drain or vent. However, soon after the Augusta plant was commis-
sioned, it became apparent that the double block valves were ineffec-
tive. Solidified plastic seized the internal mechanisms, and the valves
would not close. Similarly, plastic inside the catch tank sometimes
obstructed nozzles leading to the ram-type valves, which were intended
. . . plastic inside the catch tank to positively verify that the vessel was depressurized and safe to open.
sometimes obstructed nozzles leading On the day of the incident, solid polymer had accumulated in the
to the ram-type valves, which were nozzles leading to the drain valves, which likely rendered them
intended to positively verify that the inoperative.
vessel was depressurized and Waste plastic tenaciously adhered to the inside of the polymer catch
safe to open. tank. It was typical practice to attach a cable to the metal frame in the
vessel and use a forklift to pull and extract the plastic. Augusta opera-
tions management had expressed concern about possible occupational
injuries to maintenance workers during this manual cleaning operation.
Because of these problems, operations management considered various
alternative designs for the polymer catch tank. It was not until late 2000
that an engineering request for a complete redesign was submitted.
Each problem noted above warranted a reconsideration of the hazards
of the waste handling system by Augusta facility management and, in
some cases, implementation of incremental engineering or procedural
safeguards. In total, these deficiencies were indications that the system
design was flawed; a fundamental re-evaluation was needed to
effectively correct the deficiencies.

————————–
31
A ram-type valve is designed to resist plugging.

46
The Augusta site policy for control of hazardous energy and lockout/
4.6 Procedures for
tagout required specific written procedures for preparing individual Safe Opening of
pieces of process equipment for opening. Although a written guideline
Process Equipment
existed for the polymer catch tank, it could not be followed because it
called for the use of drain valves, which had proven to be ineffective.
Consequently, the guideline was not used, and there was no standard
practice among the workforce for ensuring that the vessel was depres-
surized prior to opening. . . . there was no standard practice
One maintenance technician told investigators that it was his practice to among the workforce for ensuring
open one of the drain valves and check whether vacuum applied to the that the vessel was depressurized
polymer catch tank vent line drew in ambient air. If it did, he inferred prior to opening.
that the vessel was partly clear inside, depressurized, and safe to open.
Other employees stated that they did not use the drain valves because
of plugging.
Investigators examined previously completed equipment isolation lists
and confirmed that drain valves were seldom used to verify that the
polymer catch tank was depressurized. Some personnel asserted that
they relied on the pressure gauge on the vent line to check for pressure
in the vessel, which was its intended use.
The Augusta facility policy for control of hazardous energy and lockout/
tagout required the completion of a “try step,”32 which was intended to
verify the effectiveness of equipment de-energization. For the polymer . . . workers most likely relied on
catch tank, the logical means of conducting the try step was to open a knowledge that the process was shut
drain valve on the vessel to observe the absence of gas, liquid, or down–and possibly on a reading from
pressure. the pressure gauge on the plugged vent
line–to conclude that there was no
There were no initials under the try step on the equipment isolation list
pressure in the polymer catch tank.
issued on the day of the incident, and it is not known whether the drain
No positive verification occurred.
valves were used. However, it is unlikely because they were found in
the closed position after the incident.
Investigators believe that, on March 13, workers most likely relied on
knowledge that the process was shut down–and possibly on a reading
from the pressure gauge on the plugged vent line–to conclude that there
was no pressure in the polymer catch tank. No positive verification
occurred.

————————–
32
The try step is a positive means of verifying that equipment is de-energized. It might
include pushing the start button on a machine to confirm that it does not start or–for
process equipment–opening a vent or drain to ensure that there is no pressure.

47
 A petrochemical industry consensus standard, The Safe Isolation of
Plants and Equipment, warns about the potential hazard of reliance on
pressure gauges:
Pressure gauges are reliable indicators of the existence of
pressure but not of complete depressurization. Final
confirmation of zero pressure before opening must always
The control of hazardous energy be by checking [an] open vent (HSE, 1997; p. 27).
policy for the Augusta site did not The control of hazardous energy policy for the Augusta site did not
advise the workforce when to advise the workforce when to suspend activities if problems occurred
suspend activities if problems and safe equipment opening precautions could not be met. In such
occurred and safe equipment circumstances, stop work provisions–which trigger higher level manage-
opening precautions could ment review and authorization of alternate work procedures–can
not be met. increase safety.
For example, failure to obtain a positive result during the try step might
trigger a stop work provision. Had such a policy been in force, the
procedure for isolating and verifying de-energization of the polymer
catch tank may have been reviewed by management years before the
incident, when problems with plugging of the valves first became
apparent. This review might have led to redesign of the equipment.

4.7 Incident The commercial Amodel unit experienced polymer reaction incidents
and Near-Miss dating from its initial startup in late 1993. These incidents presented
Investigations the opportunity to recognize a reactive hazard that was not identified
during process design.

By one estimate, 500 pods were

made during the first night of 4.7.1 Exploding Polymer Pods
startup; the next morning the pods
began to explode. During initial startup of the commercial unit, the startup team ran the
reaction system and extruder for an extended time while the pelletizing
system was inoperative. Polymer from the extruder discharge was
diverted from the pelletizer and manually collected in wheelbarrows.
It was then cooled by water spray, which caused it to harden on the
outside. The results were “pods” of polymer roughly the shape of the
wheelbarrow, which were dumped and left to cool for later disposal.
By one estimate, 500 pods were made during the first night of startup;

48
the next morning the pods began to explode. Large pieces of the
hardened outer shells blew off and traveled 30 feet or more. One
fragment weighed 9 pounds.
The pods were formed from molten material with an initial tempera-
ture of approximately 315°C. Because solid Amodel is a good thermal
insulator, the inner core of a pod is increasingly shielded from heat
losses as the outer shell cools, hardens, and thickens. Witnesses
Because solid Amodel is a good
described the exploded pods as having molten cores.
thermal insulator, the inner core of a
A company investigation concluded that the pods exploded because pod is increasingly shielded from
uneven cooling resulted in large stresses in the hardened outer shells, heat losses as the outer shell cools,
which led to fracturing and ejection of fragments. To correct this hardens, and thickens. . . . Pod
problem, Amoco installed a system to parcel the waste into smaller explosions were partly caused by
pieces and quickly cool it when the polymer could not be extruded internal pressurization due to gas-
through the pelletizing die. forming decomposition reactions.
Fewer pods were made during subsequent startups, and they were
taken to a more remote area to cool. These actions–combined with
other operational changes–significantly reduced the occurrence of
exploding pods.
As described in Appendix A, further reactions and decomposition may
occur if nylon plastics are held at elevated temperatures in a molten
state. Such situations can produce water vapor and gases, which–if
confined–create pressure.33 Witnesses described the molten cores of
exploded pods as discolored, which is consistent with decomposition.
Pod explosions were partly caused by internal pressurization due to gas-
forming decomposition reactions. Company investigations did not
recognize this element of causation.

4.7.2 Waste Polymer Fires

Prior to the March 13 incident, there were also numerous fires involving
the extruder and its associated equipment. CSB investigators reviewed
21 near-miss incident reports since 1997 in which the description of fire

————————–
33
The saturation pressure of steam (produced by side reactions of the polymer) at
315°C is about 1,500 pounds per square inch gage (psig). Decomposition would form
other gases, further increasing the potential for high pressure.

49
 was consistent with chemical decomposition of polymer in the extruder.
Most fires were small and caused little or no damage; they typically
occurred when air was introduced into the equipment. However, in
July 2000, a fire inside the extruder was severe enough to turn the
extruder vent system ducting “cherry red” and to ignite external insula-
tion. Although each incident was reported and documented, none
None of the investigations into
were adequately investigated to determine the cause/source of flam-
these four ignition incidents
mable or combustible materials. Product decomposition was not
recognized that they may have
identified as a contributing factor.
been caused by decomposition of
the plastic and subsequent In August 2000, a fire occurred when the extruder was being purged
formation of volatile and with a polyethylene-based cleaning material. As a result of the incident
flammable substances. investigation, an action was identified to take necessary measures to
eliminate fires from the extruder. Although a different type of cleaning
material was selected, fires continued to occur. No subsequent actions
were taken.
On March 12, 2001, a similar fire involving purge material caused the
extruder system to malfunction, which led to the aborted startup. The
fire was extinguished, but no incident report was filed.
In addition, spontaneous fires occurred on two occasions when the
polymer catch tank and the reactor knockout pot were opened. On
two other occasions, waste polymer extracted from these vessels
spontaneously caught fire after being disposed of in a dumpster. Investi-
gations incorrectly attributed the dumpster fires to spontaneous combus-
tion of extraneous materials. None of the investigations into these four
Beyond the near-miss reporting ignition incidents recognized that they may have been caused by
system, the Augusta site had no decomposition of the plastic and subsequent formation of volatile and
system in place to collectively flammable substances.
analyze incidents for recurring
problems and identify issues for
additional study. 4.7.3 Inadequate Review of
Cumulative Incidents
These incidents are evidence of a pattern of near-miss events in which
hazards were not successfully identified and recurring issues were not
adequately resolved. Beyond the near-miss reporting system, the
Augusta site had no system in place to collectively analyze incidents for
recurring problems and identify issues for additional study.

50
The reactive hazards of large masses of Amodel could have been
identified if the company had recognized recurring patterns among
incidents and more intensively investigated the causes of exploding pods
and waste polymer fires. Detailed investigations would likely have led
to recognition that a significant hazard was associated with accumulat-
ing large quantities of molten polymer. Other points at which molten
polymer accumulated in the Amodel process could have been identified
as having a similar hazard. The polymer catch tank and the reactor
knockout pot were the most obvious of these possibilities.

4.8 Review of
CSB investigated a similar incident at the Equilon Enterprises oil
refinery delayed coking unit34 in Anacortes, Washington (CSB, 2001). Similar Incident
On November 25, 1998, hot petroleum liquid and vapor rapidly
discharged from the bottom head of a coke drum when it was opened.
Six persons working close to the drum were fatally injured in the
ensuing fire.
A loss of electric power and steam supply prior to the fire had resulted
in abnormal process conditions, which interrupted completion of the
normal cooling cycle required to produce a batch of coke. A buildup of
tarry oil in piping prevented the injection of steam and water to cool the
contents of the coke drum, and it was necessary to empty the partially
filled drum.
The process supervisor and crew assumed that the contents had cooled
to safe levels during the 2 days since the power failure. In addition, the
relatively cool temperature of the bottom flange erroneously suggested
that the temperature inside the drum was cool when only the material
adjacent to the inside walls had actually cooled.
Although unknown to the workers present, the core material in the
drum was insulated from heat loss. Within the core, residual heat

————————–
34
A delayed coking unit converts heavy tar-like oil to lighter petroleum products, such as
gasoline and fuel oil. Petroleum coke is a byproduct of the process. Petroleum coke is
produced in batches contained in large drums. A full coke drum contains a tarry mass,
which solidifies to a coal-like substance (coke) when cooled by the addition of steam and
water. The top and bottom of the drum are opened at the completion of the cooling
cycle, and the solid mass of coke is cut into pieces and removed from the vessel.

51
 allowed petroleum to continue to react and break down, creating a
pocket of hot, pressurized hydrocarbons. When the drum was opened,
the hydrocarbons rushed out and formed a flammable vapor cloud that
subsequently ignited.
The BP Amoco and Equilon incidents are similar because each
involved:
n Opening of equipment when there were false indications or
misunderstandings of temperature and pressure in the vessel core.
In both incidents, unanticipated
reactions in a vessel created gas and n Endothermic decomposition reactions that produced gaseous
pressure. A layer of material near the byproducts.
vessel surface provided an insulating n Hazards created by abnormal startup or shutdown operations.
effect and pressure-containing
n Manual opening of hot, pressurized equipment.
boundary.
In both incidents, unanticipated reactions in a vessel created gas and
pressure. A layer of material near the vessel surface provided an
insulating effect and pressure-containing boundary. Both process
instrumentation and opening practices failed to provide adequate
warnings of the state of the material inside.
The hazards of exothermic reactions–such as rapid energy release due
to self propagation–are extensively documented in the literature; there is
Although endothermic reactions
less documentation on the hazards of endothermic reactions. Although
absorb energy, they can proceed as
endothermic reactions absorb energy, they can proceed as long as a
long as a reservoir of heat is available
reservoir of heat is available to sustain the reaction. If the temperature
to sustain the reaction. If the
is sufficiently high, the energy in the materials themselves serves as a
temperature is sufficiently high, the
reservoir.
energy in the materials themselves
serves as a reservoir. In these two incidents, the reaction rate was probably slow; however,
endothermic reactions can also proceed at high rates. Furthermore,
they may pose other hazards such as gas production or the formation of
toxic substances.
Following each incident, the process was redesigned so that personnel
did not have to directly open the equipment. For the coke drums, a
new remotely controlled station allowed workers to place themselves
farther from the drum while the head was being lowered. The Amodel
process was redesigned to eliminate large accumulations of molten
plastic.

52
 5.0 Regulatory Analysis

5.1 Process Safety

T he Occupational Safety and Health Administration (OSHA) has
issued a process safety management (PSM) standard intended to
prevent or minimize the consequences of catastrophic releases of toxic,
Management
reactive, flammable, or explosive chemicals.35 The standard applies to
processes36 containing more than a threshold quantity of any one of
137 OSHA-listed “highly hazardous chemicals.” Substances are listed
based on their toxic or reactive properties. Flammable substances are
also covered as a class.
Certain manufacturing processes at the Augusta facility are required to
comply with the OSHA PSM standard because they meet the criteria
noted above. However, the Amodel process does not contain any of
the listed chemicals, nor does it contain substances that meet the
flammability criteria. It is not subject to the standard.
To implement OSHA requirements, BP Amoco developed a PSM
program based on the 14 safety management elements of the OSHA
standard. The site PSM plan specified that the program be applied to
processes covered by the OSHA standard. However, BP Amoco
voluntarily applied much of the program to the Amodel process as well.
For example, the company had compiled process safety information,
conducted PHAs, developed standard operating procedures, conducted
training, applied MOC policies, and investigated incidents. Interviews
with Augusta facility personnel indicated that–though most of the PSM
program elements were implemented in the Amodel process–some
were applied less rigorously than others.
With regard to the Amodel process, CSB has identified the following
weaknesses in BP Amoco’s PSM program implementation as having
some causal relationship to the incident:

————————–
35
29 CFR 1910.119, Process Safety Management of Highly Hazardous Chemicals.
36
OSHA defines a process as any activity involving a highly hazardous chemical,
including any use, storage, manufacturing, handling, or onsite movement of such
chemicals, or any combination of these activities. For purposes of this definition, any
group of vessels that is interconnected and separate vessels that are located such that a
highly hazardous chemical could be involved in a potential release are considered to be
a single process.

53
 n The program required that a PHA be completed prior to initial
startup of a manufacturing process. However, because the
extruder was not included in the PHA before startup of the
Amodel process, some extruder malfunctions were not identified
as potential causes of overfilling the polymer catch tank (see
The procedure for preparing to Section 4.4).
open the polymer catch tank became n The program called for an MOC review of safety consequences
invalid after it was discovered that before changes were made to key operating procedures. No
the isolation and drain valves MOC review was conducted when the Amodel process startup
were ineffective. time was changed from 30 to 50 minutes. The implications of this
change were not fully identified (see Section 4.3).
n The procedure for preparing to open the polymer catch tank
became invalid after it was discovered that the isolation and drain
valves were ineffective. The site PSM program required regular
revalidation of operating procedures. Had this requirement been
more rigorously applied, the need to make design changes to the
polymer catch tank may have been identified earlier.
One of the elements of the OSHA
PSM standard requires the compilation n One of the elements of the OSHA PSM standard37 requires the
of thermal stability data for process compilation of thermal stability data for process chemicals. Al-
chemicals. though the BP Amoco R&D department developed some thermal
stability data for Amodel because of concerns for product quality,
the results of this work were not recognized as applicable to
process safety.

————————–
37
29 CFR 1910.119, (d) (1) (vi).

54
OSHA has also established a standard on controlling hazardous energy
5.2 Lockout/Tagout
(lockout/tagout). 38 Its purpose is to prevent unexpected energization,
startup, or release of stored energy when personnel interact with
machinery or equipment. The standard requires employers to establish
an energy control program and follow procedures to apply lockout/
tagout39 devices or energy isolation devices, or to otherwise disable
such equipment.
The Augusta facility had a policy in place to comply with OSHA
lockout/tagout requirements. However, opening the polymer catch tank
while it was pressurized represented a failure to control hazardous . . . opening the polymer catch tank
energy. while it was pressurized represented a
failure to control hazardous energy.

————————–
38
29 CFR 1910.147, The Control of Hazardous Energy (lockout/tagout).
39
OSHA defines “lockout” as the placement of a lockout device on an energy isolating
device, in accordance with an established procedure, to ensure that the energy isolating
device and the equipment being controlled cannot be operated until the lockout device
is removed. OSHA defines “tagout” as the placement of a tagout device on an energy
isolating device, in accordance with an established procedure, to indicate that the energy
isolating device and the equipment being controlled cannot be operated until the tagout
device is removed (29 CFR 1910.147(b)).

55
56
 6.0 Root and Contributing Causes

1. Amoco, the developer of the Amodel process, did not

6.1 Root Causes
adequately review the conceptual process design to identify
chemical reaction hazards.
Neither Amoco’s R&D department nor the process design depart-
ment had a systematic procedure specifically for identifying and
controlling hazards from unintended or uncontrolled chemical
reactions.
2. The Augusta facility did not have an adequate review process
for correcting design deficiencies.
The Augusta site was the first and only commercial manufacturing
facility for Amodel. Several problems in design of the polymer
catch tank became apparent with operating experience. Opera-
tions management did not ensure that deficiencies were corrected
in a timely manner.
n Workers were unable to follow established company policies
for lockout/tagout and equipment opening because the plugged
drains on the polymer catch tank prevented them from verifying
the absence of pressure in the tank.
n Previous occurrences of overfilling and plastic entrainment into
connected piping indicated that the polymer catch tank was too
small to handle foreseeable process upsets.
n The level indicating device for the polymer catch tank was
unreliable.
3. The Augusta site system for investigating incidents and near-
miss incidents did not adequately identify causes or related
hazards. This information was needed to correct the design
and operating deficiencies that led to the recurrence of
incidents.
n Sound technical theories were not developed to explain the
spontaneous ignition of waste plastic or the phenomenon
whereby lumps of waste plastic burst.
n Incidents and near misses tended to be treated as isolated
events. Management did not have a review system to detect
trends and patterns among incidents.

57
 n The polymer catch tank had been overfilled and the vent lines
plugged on other occasions. No effective measures were
developed to prevent recurrence.
n Fires had occurred at the extruder on numerous occasions.
No effective countermeasures were developed.

6.2 Contributing
1. Hazard analyses of the Amodel process were inadequate and
Causes incomplete.
n Reactivity hazards, such as unintended reactions, were not
examined in the final design-phase hazard analysis.
n The extruder operation and its overall impact on the rest of the
process were not adequately reviewed during the formal
hazard analysis conducted during design and construction.
n Credible scenarios by which the polymer catch tank could
become overfilled were not identified.
2. Design documentation did not adequately describe the Amodel
process.
n The process description did not adequately explain the design
basis and operating principles for the polymer catch tank; as a
result, misunderstandings of these characteristics developed.
The maximum fill level was not clearly specified. No warn-
ings were provided about the consequences of overfilling.
n Operations management did not update the documentation to
reflect changes in procedures and practices.
3. Equipment opening procedures did not specify what actions to
take when safety precautions could not be met.
On the day of the incident and frequently during the life of the
Amodel process, it was not possible to verify the absence of
pressure in the polymer catch tank, as required by company
procedures and OSHA lockout/tagout regulations. Solid polymer

58
 plugged the drain nozzles, which should have been used to verify
the absence of pressure. Without a policy requiring management
review and authorization in such circumstances, workers pro-
ceeded to open the vessel despite the lack of positive verification
that it was not pressurized.
4. Revisions to operating procedures were not subjected to
management of change reviews to evaluate safety effects.
Flow was originally directed to the polymer catch tank for
30 minutes during startup. Without adequate consideration of
possible adverse consequences, this time was later extended to
50 minutes–which increased the amount of material to be disposed
of to the vessel.

59
60
 7.0 Recommendations

1. Examine the manufacturing businesses acquired from

Solvay Advanced
BP Amoco Performance Polymers and ensure that a systematic Polymers, L.L.C.
safety review procedure is developed and implemented for
identifying and controlling hazards from unintended chemical
reactions. Additionally, ensure that reactive hazards are identified
and evaluated:
n During product R&D, during conceptual design of a new
process, and during detailed design of a new process.
n Before changes are made to existing equipment or process
chemistry.
Communicate the results of this review to the workforce.
(2001-03-I-GA-R1)
2. Ensure that a program is in place at facilities acquired from
BP Amoco Performance Polymers to systematically review the
hazards associated with new and modified processes and equip-
ment as operating experience accrues. Ensure that facilities correct
all identified design, operation, and maintenance deficiencies.
Verify that operating experience does not invalidate the design basis
for equipment. (2001-03-I-GA-R2)
3. Revise the Material Safety Data Sheet (MSDS) for Amodel to warn
of the hazards of accumulating large masses of molten polymer.
Communicate the MSDS changes to current and past customers
(who may retain inventories of this product). (2001-03-I-GA-R3)

61
Solvay Advanced 1. Implement a program to conduct periodic management reviews of
Polymers, L.L.C., incidents and near-miss incidents. Look for trends and patterns
Augusta Facility among incidents. Address root causes and implement and track
corrective measures. (2001-03-I-GA-R4)
2. Revise process safety information to include (2001-03-I-GA-R5):
n Information regarding the decomposition reactions of Amodel.
n Design intent, basis, capacity, and limitations of equipment.
n Hazards and consequences of deviations from design intent and
operating limits.
3. Revalidate hazard analyses for the Amodel process to address
(2001-03-I-GA-R6):
n Credible deviations from process intent and their
consequences.
n Hazards associated with startup and shutdown operations.
n Prevention of accumulations of potentially hazardous masses of
polymer.
4. Revise your lockout/tagout program to ensure that equipment is
rendered safe prior to opening for maintenance. At a minimum,
ensure that equipment opening procedures contain a stop work
provision that requires higher levels of management review and
approval when safe opening conditions, such as equipment
depressurization, cannot be verified. (2001-03-I-GA-R7)
5. Ensure that your management of change policy applies to opera-
tional and procedural modifications. (2001-03-I-GA-R8)

BP Chemicals Group
Communicate the findings of this report to your chemical and plastics
manufacturing facilities in North America. (2001-03-I-GA-R9)

62
Communicate the findings of this report to your membership. American Chemistry
(2001-03-I-GA-R10) Council

Communicate the findings of this report to your membership.

Society of
(2001-03-I-GA-R11) Plastics Engineers

By the
U.S. CHEMICAL SAFETY AND HAZARD INVESTIGATION BOARD

Gerald V. Poje, Ph.D.

Member

Isadore Rosenthal, Ph.D.

Member

Andrea Kidd Taylor, Dr. P.H.

Member

May 20, 2002

63
64
 8.0 References

Ballistreri, A., D. Garozzo, M. Giuffrida, and G. Montaudo, 1987.

“Mechanism of Thermal Decomposition of Nylon 66,”
Macromolecules, Vol. 20, pp. 2991-2997.
Ballistreri, A., D. Garazzo, P. Maravigna, M. Giuffrida, and
G. Montaudo, 1987. “Thermal Decomposition Processes in
Aromatic-Aliphatic Polyamides Investigated by Mass Spectrom-
etry,” Journal of Polymer Science, Part A: Polymer Chemistry,
Vol. 25, pp. 1049-1063.
Barton, John, and Richard Rogers, 1997. Chemical Reaction Hazards,
A Guide to Safety, Rugby, U.K.: Institution of Chemical Engi-
neers (IChemE).
Center for Chemical Process Safety (CCPS), 1995a. Guidelines for
Chemical Reactivity Evaluation and Application to Process
Design, American Institute of Chemical Engineers (AIChE).
CCPS, 1995b. Guidelines for Safe Storage and Handling of Reactive
Materials, AIChE.
CCPS, 1989. Technical Management of Chemical Process Safety,
AIChE.
Gander, Jesse D., and A. Jeffrey Giacomin, 1997. “Review of Die Lip
Buildup in Plastics Extrusion,” Polymer Engineering & Science,
July 1997.
Health and Safety Executive (HSE), 2000. Designing and Operating
Safe Chemical Reaction Processes, Norwich, U.K.: HSE Books.
HSE, 1997. The Safe Isolation of Plants and Equipment, Oil Industry
Advisory Committee, Norwich, U.K.: HSE Books.
Kamerbeek, B., G. H. Kroes, and W. Grolle, 1961. Thermal Degrada-
tion of Polymers, SCI Monograph No. 13, London, U.K., pp.
357-391.
Kohan, Melvin, 1995. Nylon Plastics Handbook, Hanser Publishers.
Kohan, Melvin, 1973. Nylon Plastics, Wiley Interscience.
Ohtani, H., T. Nagaya, Y. Sugimura, and S. Tsuge, 1982. “Studies on
Thermal Degradation of Aliphatic Polyamides by Pyrolysis-Glass
Capillary Gas Chromatography,” Journal of Analytical and
Applied Pyrolysis, Vol. 4, pp. 117-131.

65
Parry, Cyril F., 1998. Relief Systems Handbook, Rugby, U.K.:
IChemE.
Sanders, Roy F., 1999. Chemical Process Safety: Learning From Case
Histories, Butterworth-Heinemann.
Schulten, H., and B. Plage, 1988. “Thermal Degradation of Aliphatic
Polyamides Studied by Field Ionization and Field Desorption
Mass Spectrometry,” Journal of Polymer Science, Part A:
Polymer Chemistry, Vol. 26, pp. 2381-2394.
U.S. Chemical Safety and Hazard Investigation Board (CSB), 2001.
Safety Bulletin, Management of Change, No. 2001-04-SB,
August 2001.

66
 APPENDIX A: Reconstruction of Incident

A.1 Post-Incident
Scene
A.1.1 Suspension of Site Operations
Following the March 13, 2001, incident, BP Amoco shut down all
manufacturing units at the Augusta site to review safety programs and
conduct additional training. Operations were restored at one process
unit on April 8 and at others on April 13. The Amodel unit did not
restart until July 27, following the implementation of major design
changes.

A.1.2 Polymer Catch Tank

Most of the polymer catch tank contents was expelled, including the
metal frame. A solid layer of polymer 3 to 5 inches thick, predominately
yellow, remained in the vessel and coated the entire inside surface (see A solid layer of polymer 3 to 5 inches
cover photo). Although the layer contained some bubbles and voids, it thick, predominately yellow, remained
was composed of relatively dense material. Polymer was extruded into in the vessel and coated the entire
all the vessel nozzles, and they were completely plugged at the inlet inside surface.
(Figure A-1).
The upper portion of the flange, to which the polymer catch tank cover
had been bolted, was bent. The thrust created by the sudden expulsion
displaced the catch tank in an opposite direction. This movement bent
the vessel inlet line and broke the hot oil tubing behind the vessel.
The cover was found 14 feet from the polymer catch tank. A girder
for a canopy that sheltered the catch tank and extruder was torn from
its structural members. The cover probably struck the girder as it
blew off the tank.
Examination of the cover suggested that it was bent by the pressure
that caused the bolts to break. Half of the 44 bolts were still in place
at the time of failure; all were recovered (Figure A-2). Examination
of the bolts revealed a combination of failure modes–ductile (i.e.,
thinning before breaking) and shear (i.e., rapid overload). None of
Figure A-1. Polymer plugging the polymer
the studs separated from the nuts. The drain valve attached to the catch tank vent nozzle.
cover was found in the closed position.

67
 Figure A-2. Failed bolts from polymer catch tank cover.

A.1.3 Reactor Knockout Pot

The reactor knockout pot was not opened until April 17, 2001, to
allow time for a hazard analysis and development of a safe work plan.
Once opened, the vessel was found to be largely empty except for a 6-
inch layer of material on the bottom (Figure A-3).

Figure A-3. Reactor knockout pot with internal frame visible.

68
 A.1.4 Vent Piping
The outlet nozzle for the 6-inch vent line on the polymer catch tank
was plugged with solid polymer (Figure A-4). The vent line itself was
plugged except for a small hole through the center of the obstruction.
A total of 22 feet of 6-inch pipe was plugged with polymer, including
the part of the line where the pressure gauge was installed. The
plugging extended to the common portion of the vent line shared with
the reactor knockout pot, as outlined in Figure 4 (Section 2.3). The
polymer inside the pipe had an appearance characteristic of several
annular layers, each distinct in color and separated by discontinuities
(Figure A-4).
The tap1 to the pressure gauge on the 6-inch vent line was observed
to be clear despite extensive plugging of the line itself.
Figure A-4. Vent line obstructed by
polymer layers.

A.1.5 Pressure Relief System

The pressure relief valve and associated piping were removed for
examination. The inlet line to the relief valve was plugged with poly-
mer up to the device. The discharge line was clear beyond the relief
valve. CSB concludes that it is unlikely that the pressure relief valve
activated prior to the incident.
The polymer inside the pipe had an
appearance characteristic of several
annular layers, each distinct in color
A.1.6 Extruder and separated by discontinuities.

The extruder was disassembled and inspected. The gearbox was

found to be out of synchronization, and a bearing for one of the screws
was cracked. The screws appeared to have collided with each other
and were fouled by an ash-like material.
A maintenance technician had first noticed the ash at approximately The screws appeared to have collided
3:45 pm on March 12, during the aborted startup, while attempts were with each other and were fouled
made to diagnose the extruder malfunction (see Sections 3.1.2 and by an ash-like material.
3.3.1). This was the first observation of ash on the screws; there was
no explanation for it at the time.

————————–
1
A tap is a small-diameter branch pipe connected to a larger pipe.

69
 A.1.7 General Area
Expelled polymer was scattered northward throughout the area, up to
70 feet from the polymer catch tank (Figure A-5). Some of the polymer
was charred by fire. Much of the area at grade was covered with a
fine silica ash, which was determined to be a product of combustion of
the heat transfer fluid that had fueled the fire.
Much of the area at grade was Part of the metal rack expelled from the tank was found 40 feet from the
covered with a fine silica ash, which vessel, imbedded into insulation on piping to a control valve station.
was determined to be a product of The control valve and piping were displaced more than 6 inches in the
combustion of the heat transfer fluid direction of impact.
that had fueled the fire.

Figure A-5. Polymer scattered throughout work area.

70
Inspection of the extruder barrels after the incident revealed a large
A.2 Extruder Failure
amount of ash, which was very likely the result of a fire in the extruder.
In the early morning of March 12–about 12 hours before the aborted
startup attempt–the extruder temperature was raised to 315°C, and the
machine was briefly prerun on a purging material. This activity was Temperature-measuring instruments on
standard procedure; it was always done just before the reaction section the extruder recorded abnormally high
of the plant was started up. values during this period; it is likely that
However, on this occasion, a mechanical problem in the reaction the residual purge material inside the
section delayed startup until the afternoon of March 12. Air had likely extruder had ignited.
entered an open divert valve on one of the extruder barrels and flowed
across the screws. Temperature-measuring instruments on the extruder
recorded abnormally high values during this period; it is likely that the
residual purge material inside the extruder had ignited.
The observation of higher-than-normal temperatures is consistent with
a release of heat from combustion of the purge material. Gaseous
byproducts would have escaped through vents on top of the extruder, These multiple attempts to start the
leaving ash behind. motor applied torque to the seized
Because the extruder was not designed to convey powder, the accumu- screws of the machine, which likely
lated ash probably caused the screws to bind when attempts were explains how the bearing became
made to start the machine at 2:16 pm on March 12. Although the cracked and the gearbox
motor start push button was activated more than 30 times, interlocks unsynchronized.
caused the motor to repeatedly shut down. These multiple attempts to
start the motor applied torque to the seized screws of the machine,
which likely explains how the bearing became cracked and the gearbox
unsynchronized.
Statements from several operators indicated that this was the first time
the extruder failed to run coincidentally with startup of the reaction
. . . an extruder component failure
section. However, the machine or components were known to occa- during a production run . . . resulted in
sionally fail during routine operation. It was an extruder component the plant shutdown on March 10. In
failure during a production run that resulted in the plant shutdown on early February 2001, a series of
March 10. In early February 2001, a series of instrument failures on instrument failures on the extruder
the extruder motor over 2 days culminated in shutdown of the process. motor over 2 days culminated in
BP Amoco personnel were aware that various malfunctions could shutdown of the process.
interrupt operation of the extruder during production. When the
extruder experienced certain mechanical difficulties, it was necessary

71
 to divert the reactor effluent and accumulate polymer in the polymer
catch tank or the reactor knockout pot. On prior occasions, this had
resulted in overfilling one of the vessels and plugging their common vent
line.

A.3 Polymer
Catch Tank
A.3.1 Thermal Considerations
During a typical startup, the amount of prepolymer sent to the polymer
catch tank fills almost half the vessel. Because the process usually runs
for an extended time, the prepolymer accumulated during startup has
the opportunity to cool and solidify before the injection of additional hot
material during shutdown and solvent flushing.
As a result of the aborted startup attempt on March 12, more than twice
the normal mass of hot prepolymer was continuously directed to the
polymer catch tank. This included prepolymer made during startup and
Because the process usually runs for extruder troubleshooting, plus hot material displaced from the reaction
an extended time, the prepolymer section of the plant during the solvent flush after the aborted startup.
accumulated during startup has the This much prepolymer had not been continuously fed to the catch tank
opportunity to cool and solidify before on any previous occasions. The amount of thermal energy inside the
the injection of additional hot material vessel was probably greater than it had ever been before.
during shutdown and solvent flushing.
Amodel readily solidifies, becoming an effective insulator. Heat losses
through the shell of the polymer catch tank ultimately resulted in the
formation of a solid layer of plastic adjacent to the walls. Once formed,
this layer created an insulating effect that allowed the core of material in
the vessel to remain hot and molten for an extended time.

72
 A.3.2 Filling Beyond Working Capacity
CSB investigators determined that significant amounts of polymer had
escaped into the vent system for the polymer catch tank and the reactor
knockout pot–evidence that the catch tank was overloaded (i.e., run at
conditions outside its design limit). Although the quantity of accumu-
lated material is reasonably well known, the liquid level in the catch
tank is unknown due to uncertainties with regard to density. 2
Waste polymer cleaned from the catch tank on March 11, following the
previous production run, was composed of multiple layers of varying
density, with considerable voids between layers (Figure A-6). The
waste polymer occupied 55 percent of the vessel and had a bulk density
of only 3 lb/gal.

Figure A-6. Material extracted from polymer catch tank on

March 11, prior to incident.

————————–
2
BP Amoco plant engineering personnel estimated that the bulk density of material in the
polymer catch tank ranged from 7 to 10 pounds per gallon (lb/gal), with the latter being
the density of the final Amodel product. Material taken from the vessel post-incident had
a measured bulk density varying from 5 to 7 lb/gal; however, this may not represent the
density inside the vessel prior to the release of pressure.

73
 The reactor effluent enters the polymer catch tank as a two-phase
mixture of polymer with a significant quantity of vapor. Vapor separates
As the polymer catch tank filled from the liquid inside the vessel, as depicted in Figure A-7. Additional
beyond its working capacity, its ability vapor is generated when water and excess reactants evaporate from the
to separate vapor from molten melt.
prepolymer degraded.
As the polymer catch tank filled beyond its working capacity, its ability to
separate vapor from molten prepolymer degraded. Polymer droplets
were entrained with the vapor stream into the vent system (Figure A-8).
Once inside the vent system, polymer deposited on the piping, hard-
ened, and restricted the flow of vapor. This activity caused the catch
tank to become pressurized.
The accumulated pressure eventually exceeded the sealing capability of
the cover gasket, and solvent vapor leaked from the top of the 4-foot-
The accumulated pressure eventually diameter bolted cover. Upon discovery of the leak, the solvent flow
exceeded the sealing capability of the was diverted from the polymer catch tank to the reactor knockout pot–
cover gasket, and solvent vapor leaked which also began to leak because significant quantities of polymer had
from the top of the . . . cover. accumulated in the common vent piping, causing the pot to pressurize.
Although the vent line was restricted, it did not completely plug off at
this time; flow continued through the unit. Process pressures upstream
were within normal operating ranges.

There was no direct indication of the

level of material in the polymer catch A.3.3 Safeguards to Prevent Overfilling
tank because the level detection device
was not installed. There was no direct indication of the level of material in the polymer
catch tank because the level detection device was not installed. A
capacitance-type level probe had been used; however, it was unreliable
because it was often fouled with polymer.
A timer mechanism was installed for the polymer catch tank in 1994 to
indicate when accumulated prepolymer reduced the vessel’s spare
The timer . . . did not account for capacity to the extent that a production shutdown was required. The
prepolymer that entered during the timer tracked the time that prepolymer entered the catch tank during
solvent flush phase of the shutdown startup or upset conditions, but it did not account for prepolymer that
operation. entered during the solvent flush phase of the shutdown operation.
The basis of the timer setting was to provide sufficient capacity in the
polymer catch tank, plus an additional unspecified safety margin. There
is no evidence to suggest that the timer mechanism took into account the

74
Figure A-7. Normal operation–settlement of particles.

Figure A-8. Polymer catch tank filled beyond working

capacity–particles entrained in vent vapor.

75
 density variations of the material3 or the added requirement for vapor-
liquid disengagement. Considering these limitations–and the fact that it
did not account for prepolymer that arrived during shutdown–the timer
was an unreliable indicator of spare capacity in the vessel.

A.3.4 Analysis of Cover Bolts

The cover was attached to the polymer catch tank with 44 carbon steel
grade SA-193 B16 bolts, each measuring 1 1/8 inches in diameter. The
22 bolts that failed were examined to determine the presence of
damage or defects. Two of the intact bolts were analyzed by ASTM
A370, Standard Test Methods and Definitions for Mechanical Testing of
Steel Products, and found to be within specification for yield and tensile
strength.
No brittle fracture characteristics were observed when the bolts were
tested; they failed in a ductile mode. There was little variability among
the specimens. From these test results, it was concluded that the bolts
No brittle fracture characteristics
contained no prior defects–nor did they exhibit damage from fatigue,
were observed when the bolts were
previous overloading, or environmental conditions.
tested; they failed in a ductile
mode. . . . it was concluded that the
bolts contained no prior defects–nor
did they exhibit damage from fatigue, A.3.5 Determination of Internal Pressure
previous overloading, or
environmental conditions. The polymer catch tank had a maximum allowable working pressure of
108 pounds per square inch gage (psig). There was no direct indica-
tion of pressure in the tank at the time of the incident. Estimates were
made of the force and pressure required to cause failure of the 22 cover
bolts. However, this analysis was complicated by the fact that the
restraining force applied by the remaining 22 bolts was not concentric
to the cover; the bolts were removed in a sequential pattern from the
1:30 to 7:30 o’clock position.

————————–
3
The timer algorithm was based on a polymer density of 10 lb/gal, which is equivalent
to the finished polymer.

76
At the request of BP Amoco, M&M Engineering of Shaker Heights,
Ohio, conducted a finite element analysis (FEA) to estimate pressure on
the cover at the time of failure. A three-dimensional model was created
for the shell of the polymer catch tank, mating flange, flange gasket,
cover, and bolts. The static properties of the bolts were taken from the
testing results described above, and the tensile properties were in-
creased to account for dynamic effects (i.e., a rapid rate of bolt strain).
The results of the FEA indicate that the best estimate of the required
pressure to fail the 22 remaining bolts is 50 psig (assuming static tensile
properties) to 70 psig (assuming a 30 percent dynamic increase over
static strength). Considering the uncertainties in dynamic loading of the
bolts, 110 psig is provided as a conservative upper-bound estimate of
failure pressure. This value corresponds to an assumed 100 percent
increase in tensile properties due to dynamic effects. The analysis The analysis results indicate a
results indicate a significant decrease in the capability of the 22 remain- significant decrease in the capability
ing bolts due to the eccentric (i.e., not at or in the geometric center) of the 22 remaining bolts due to
nature of the restraining force. the eccentric . . . nature of
The FEA did not attempt to account for the additional restraining force the restraining force.
created by the 3- to 5-inch layer of hardened polymer that coated the
inside wall of the polymer catch tank. Molded Amodel is a high-
strength polymer with a tensile strength of 12,000 psig at ambient
temperature. There is considerable uncertainty as to the strength of the
polymer layer at the time of failure due to its differences in composition
from neat Amodel and its elevated temperature.
Although the polymer coating is potentially capable of containing There is considerable uncertainty
significant pressure beyond the 50- to 70-psig range estimated by FEA, as to the strength of the polymer layer
any analysis of this additional pressure is highly speculative considering at the time of failure due to its
the aforementioned uncertainties. CSB concludes that the pressure differences in composition from neat
inside the polymer catch tank at the time of failure was at least 50 psig. Amodel and its elevated temperature.

77
A.4 Physical Testing
and Chemistry
A.4.1 Potential Chemical Reactions
The public literature on nylon plastics describes them as susceptible to
thermal degradation. The Nylon Plastics Handbook states that two
general types of thermal reactions are expected (Kohan, 1995). One
involves breakdown of the amide groups of the polymer; the other
relates to instability inherent in the raw materials themselves. One
. . . a source of pressure inside the example of the latter is the thermal breakdown of carboxylic acids to
polymer catch tank was gas formation form gaseous carbon dioxide. A nylon polymer created from a di-
due to thermal degradation. carboxylic acid may be susceptible to this type of decomposition.
The literature describes other more complex possibilities for
decomposition of the monomeric units of the polymer. These
possibilities, among other evidence, led CSB investigators to theorize
that a source of pressure inside the polymer catch tank was gas
formation due to thermal degradation.
Side reactions are another mechanism by which molten nylon polymers
can generate gas. Kohan (1995) describes how two amine ends from
separate nylon polymer molecules can combine in a condensation
Side reactions are another mechanism reaction, producing a single molecule with a triamine group and
by which molten nylon polymers releasing an ammonia molecule. The triamine entity of the newly
can generate gas. formed molecule may, in turn, react by a second condensation with
another polymer molecule that has a carboxylic acid end group,
releasing a water molecule. When both of these reactions occur,
polymer molecules become chemically linked through branches.4
One consequence of this phenomenon is referred to as “gelation.”
The branched molecules have extremely high molecular weight and are
referred to as gels. The BP Amoco documentation for the Amodel
process includes a brief discussion of these potential branching reactions
and states that limiting the time the reacting mix is subject to elevated
temperatures will minimize their occurrence.
The Augusta site produces various grades of Amodel. A principal
distinction among grades is how various di-carboxylic acids are appor-
tioned among the ingredients. Amodel is made from isophthalic and

————————–
4
The interconnection of polymer molecules through side branches is known as
crosslinking.

78
adipic acids. At the time of the aborted startup, a grade with a lower
percentage of aromatic acids was being produced. Fundamental
principles of organic chemistry suggest that thermal stability should
decrease with a reduction in aromatic acid content. This grade of
Amodel had a relatively high melting point compared to other grades.

A.4.2 Characterization of Expelled Plastic

Investigators collected remnants of the plastic that was expelled from
the polymer catch tank; the material was hard, brittle, and filled with
bubbles (Figure A-9). However, the plastic was partially molten and
sticky at the time it was expelled because some of it stuck to nearby Figure A-9. Sample of material
expelled from polymer catch tank (note
pipes, steel columns, and the side of a storage shanty located almost
bubbles and closed cells).
70 feet from the vessel.
Bubbles within the expelled material caused it to have a cellular struc-
ture. Consequently, the bulk density of the material was low compared
to product samples of extruded Amodel. However, the plastic encasing
the individual cells was dense. It had a specific gravity of 1.1, typical of . . . the plastic was partially molten
normal extruded Amodel product. The presence of bubbles within the and sticky at the time it
expelled material and its cellular structure are consistent with the was expelled . . .
generation and entrapment of gas while material is in a molten state.
These characteristics support the theory that decomposition and side
reactions–occurring while the material was molten inside the polymer
catch tank–produced gas and pressure.
The expelled plastic was yellow, though typical extruded Amodel is
cream colored. The discoloration is also consistent with exposure to
high temperature for an extended time.

A.4.3 Characterization of Plastic

Remaining in Vessel
A ring of plastic 3 to 5 inches thick remained inside the polymer catch
tank after the molten core blew out (Figure A-10). The ring was
Figure A-10. Layer of hardened
generally yellow, though some of the outermost portions close to the polymer taken from inside perimeter of
vessel wall were cream colored. The plastic in the ring contained far polymer catch tank

79
 fewer bubbles than the expelled material. The concentration of
bubbles was greatest farther from the vessel wall, toward the core.
Heat losses through the wall of the polymer catch tank from the time it
The presence of bubbles within the was filled until it was opened likely explain why the material adjacent to
expelled material and its cellular the vessel wall solidified by the time workers started removing the cover.
structure are consistent with the This explanation is also consistent with the observation that cream-
generation and entrapment of gas colored polymer was prevalent along the outermost perimeter of the
while material is in a molten state. ring. Because it was close to the vessel wall, this material cooled sooner
and was less susceptible to side reactions and thermal decomposition.
BP Amoco analysis of the plastic ring in the polymer catch tank showed
the outer white material to be similar to typical Amodel, while the
yellow material farther from the vessel wall contained gelled entities.
Because it was close to the vessel wall, It also had unusually high levels of triamine groups and atypically low
[the cream-colored polymer] cooled levels of carboxylic acid groups. These differences indicate that the
sooner and was less susceptible material farthest from the vessel wall had probably reacted by means of
to side reactions and thermal the branching side reactions described above. Such reactions generate
decomposition. gas, consistent with the observation that the maximum concentration of
bubbles in the ring was farthest from the wall.

A.4.4 Simplified Thermal Testing

Following the incident, CSB arranged for testing of samples of typical
extruded Amodel product pellets. Samples were sent to Polyhedron
Laboratories, Inc., in Houston, Texas. Two grams of the cream-colored
pelletsw eredriedandplacedinthebottom ofanautocla5vwhich e,
was evacuated to vacuum and then pressurized three times with 60-psig
nitrogen. Finally, the material was heated to 315°C (the approximate
condition in the molten core of the polymer catch tank) and held at that
temperature for 12 hours. After cooling to room temperature, the
autoclave was depressurized and opened. A solid brown residue was
present on the bottom, and the entire inner surface of the autoclave
was coated with a brown liquid. From the results of this experiment,
CSB concludes that Amodel decomposes at the test temperature.

————————–
5
An autoclave is an airtight vessel for heating and sometimes agitating its contents under
high pressure.

80
Following this testing, a portion of the pellet sample was tested by
thermogravimetry6 at Chemir/Polytech Laboratories, in Maryland
Heights, Missouri. The pellets were ground to a powder and vacuum
Given the published degradation
dried at 100°C for 12 hours. Two conditions were chosen for testing–
theory for nylon polymers, CSB
holding at 315°C for 2 hours and for 12 hours. The 2-hour test
concludes that the weight loss is due
resulted in a weight loss of 2.2 percent. The 12-hour test was con-
partly to decomposition of the plastic
ducted twice, with weight losses of 4.9 percent and 5.8 percent.
and partly to side reactions.
Given the published degradation theory for nylon polymers, CSB
concludes that the weight loss is due partly to decomposition of the
plastic and partly to side reactions.

A.4.5 Advanced Thermal Analysis

BP Amoco retained Kinetica, Inc., of Franklin, Ohio, to conduct thermal
stability testing of the finished plastic. Kinetica used an accelerating rate
calorimeter (ARC) for this purpose.
When one grade of Amodel was heated in the calorimeter, pressure in
the vapor space above the sample began to increase at 350°C. The
rate of increase became rapid at 380°C and peaked at a value 4.3 times
the initial pressure. The reaction process was observed to be endother-
mic, particularly during the period of pressure increase.
Two runs were also performed on a different grade of Amodel. In both
of these tests, pressure initially increased at 330°C. In one test, there
was a rapid rise in pressure at 370°C; in the other, pressure increased
suddenly at that temperature but less rapidly. The pressure peaked at The reaction process was
4.5 times the initial pressure in one test and at 4.4 in the other. The endothermic. . . . The ARC data
reaction process was endothermic. Figure A-11–from the company’s fully support the conclusion that the
internal investigation report–depicts temperature and pressure data from plastic decomposes.
the first test.
The ARC data fully support the conclusion that the plastic decomposes.
The lower onset temperature for the pressure rise observed in the tests
on one grade of Amodel indicates that it is more susceptible to decom-
position than the other grade.

————————–
6
Thermogravimetry measures the change in mass of a substance on heating. Losses of
mass may be due to dehydration, decomposition, desorption, vaporization, or chemical
reactions that create gaseous products.

81
 Figure A-11. ARC data for Amodel.

A.4.6 Analytical Chemistry

BP Amoco contracted with the Thermal Analysis Laboratory at
Western Kentucky University for quantitative chemical analyses of the
vapors generated from thermal degradation of Amodel. A thermo-
gravimetric analyzer was used to heat the Amodel. The resulting
vapors were captured for analysis by mass spectrometry and Fourier
transform infrared spectrometry.
As reported in the company’s internal investigation report, the labora-
tory provided the following weight conversions of the plastic for certain
gases:

82
 Weight Conversions of Amodel to Certain Gaseous Products
During Decomposition

Species Wt % generated
Carbon monoxide 0.86
Carbon dioxide 1.03
Ammonia 3.98
Water 0.52

From these data, CSB concludes

The above species are all gases at moderate temperatures. From these that pressure would develop if the
data, CSB concludes that pressure would develop if the gases were gases were generated in a
generated in a confined environment. confined environment.
The BP Amoco Performance Chemicals Laboratory in Naperville,
Illinois, heated a sample of plastic extracted from the residual ring in the
polymer catch tank to 340°C. The vapors that evolved were captured
in a cryogenic trap for analysis by gas chromatography/mass spectrom-
etry. A large number of organic chemicals with molecular weights
below 200 were identified. Nitrile compounds were among the
dominant species.
Kohan (1995) proposes that nitrile compounds are among the decom-
position products from the cleavage of amide groups in certain types of Given the relatively low molecular
nylons. Given the relatively low molecular weight of the organic weight of the organic compounds
compounds produced, as noted above, they could be expected to exert produced, . . . they could be expected
significant vapor pressures at the temperatures in the polymer catch to exert significant vapor pressures
tank. In light of these analyses, the nitrile compounds are expected to at the temperatures in the
contribute to additional pressurization. polymer catch tank.

83
A.5 Vessel No additional material flowed into the polymer catch tank after
Pressurization 3:45 pm on March 12. However, over the next several hours,
reactions in the core of the vessel generated vapor that caused the
viscous, molten contents to foam and expand. Eventually, the foaming
material occupied the entire volume of the catch tank.
Pressure pushed the material into the vessel nozzles, including the
6-inch vent line and the emergency pressure relief inlet. As polymer
reached the sides and upper surfaces of the vessel, it cooled and
solidified.7 The material also cooled and solidified in the vent line and
emergency inlet. Once this occurred, there was no pathway for gas to
The material also cooled and solidified
escape from the vessel.
in the vent line and emergency inlet.
Once this occurred, there was no Plugging in the vent system was extensive and continued to worsen over
pathway for gas to escape from the several hours as polymer was pushed from the polymer catch tank into
vessel. the system. Eventually, the pathway through the common vent piping
degraded to the extent that the emergency pressure relief system on the
reactor knockout pot activated at approximately 8:10 pm on March 12.8
As heat from the polymer catch tank was gradually lost to the environ-
ment, further solidification occurred around the entire inner wall of the
vessel. Although a hardened layer of polymer 3 to 5 inches thick
developed, the core continued to react and decompose over several
hours, generating gas. The trapped gas pressurized the vessel.

————————–
7
Although the polymer catch tank was insulated, it was not heated with a hot oil jacket.
Heat losses to the environment would cause the polymer to solidify.
8
From 6:53 to 11:21 pm on March 12, water was circulated through the Amodel unit
and steam was vented through the reactor knockout put. The knockout pot and the
polymer catch tank use common vent piping.

84
 APPENDIX B: Timeline

Process safety information

development
LTA

Detailed documentation for

Design review for reactive potential volume (startup/
hazards LTA shutdown/upset) never
generated for KD-502

No review to identify Designers intended

reactive hazards during vessel to be no more
product development than 50-60% full

1980s
1981 1987 July 1989 1990 Polymer catch tank KD-502
Amodel formulations
First Amodel pilot-plant developed and commercial Amodel experimental unit HAZOP completed on Amodel semiworks built in designed for Amodel 2
production in Naperville, IL process design built in Greenville, SC Augusta semiworks Augusta commercial unit in Augusta
conceptualized

KEY

Black box =
management system
defficiency
White box above
timeline = causal
factor

Gray box = main

timeline event

White box below

timeline = editorial KD-501 = reactor knockout pot
comment KD-502 = polymer catch tank
LTA = less than adequate
Double arrow represents
an event that took place MOC = management of change
over time PHA = process hazard analysis

85
 PHA
LTA

KD-502 overloading
scenarios not identified

Emergency pressure relief Incident investigation

No PHA on extruder
design LTA LTA

Relief system not

appropriate for service: Incident investigation
PHA did not address
– no rupture disk concludes that pods
– no provision to ensure startup, shutdown, or
ruptured due to stress
polymer would not harden reactivity
cracking
near relief valve inlet

1990 December 1990 1991

KD-502 emergency 1993
Thermal degradation of Design phase PHA Corporate R&D establishes Pods generated during
2 pressure relief system Amodel commercial unit 3
Amodel observed in complete for Amodel lab with capability to study startup explode
designed commissioned in Augusta
product performance tests commercial unit reactive chemistry hazards

Pods are startup material

Lab is commissioned, but that has been extruded and
no program is established collected in small
wheelbarrows

Pods likely exploded due to

Lab decommissioned and
a mechanism similar to that
transferred to third party in
which occurred in KD-502
1995
during incident

Amodel commercial
process designed

86
 No action taken when
personnel realized that
Incident investigation Process safety information MOC
vessel could not be
LTA LTA LTA
locked, tagged, and
verified

No MOC conducted to
No documentation Procedure for opening
examine impact of change
No incident investigation Inspection followup LTA generated on basis for vessel for maintenance not
on polymer accumulation in
timing algorithm developed
KD-502

August 1994 Level indicating device in

1993 August 1993 No uniform practice
Further degradation data KD-502 not reliable; timing
developed to verify KD-502 Startup time extended from
3 KD-502 overfilled when Fouling of relief valve on Amodel obtained for algorithm developed to 4
was depressured prior 30 to 50 minutes
extruder shut down found during inspection product application estimate polymer
to opening
purposes only accumulation/level

Plugging of drain valves

Fewer pods produced,
Same occurrence prevents positive
more material sent to
happended in 1995 verification of
KD-502 during startup
depressurization

Incident investigation
LTA

Pluggage in vent line not

investigated

Entrainment causes
partial plugging
of vent line

87
 PHA
LTA

Cummulative effects and

trends from previous
incidents not considered

Credible deviations from

normal operation not
identified

Startup and shutdown not Incident investigation Incident investigation Incident investigation
considered in PHA LTA LTA LTA

Investigation does not

PHA overlooks issues from Investigation concludes Investigation does not
examine issue of limits on
1990 PHA, such as cause is combustible examine potential
polymer accumulation in
overfilling vessel materials in dumpster decomposion of polymer
KD-501 or KD-502

February 2001 February 2001

1999 Material extracted from March 10, 2001 March 10, 2001
KD-501 reactor knockout
4 KD-502 burns after placed Fire occurrs while 5
New PHA completed on pot filled during plant Unit shutdown, extruder KD-502 opened, cleaned,
in dumpster opening KD-502 for upset caused by
Amodel process malfunction and closed
maintenance extruder problems

Similiar event occurred in

August 1994

Incident investigation for

extruder fires
LTA

Decomposition of polymer
not examined as
possible cause of some
extruder fires

Fires occur in extruder

88
 March 12, 2001 March 12, 2001 March 12, 2001
2:45 am 6:45 am 9:03 am
Extruder purged; purge fire Operators begin plant S9.2(1)rcn0.07026T8n0Tc-0.t
occurs startup checklist

89
 PHA
LTA

Inadequate recognition
of consequences of
overloading KD-502

KD-502 has more hot

material in it than normal

Devolatilization in
Startup aborted because of
KD-502
extruder malfunction
causes foaming

March 12, 2001 March 12,2001

2:41pm March 12, 2001 March 12, 2001 7
Polymer entrainment 3:55 pm
6 3:45 pm 4:00 pm
Solvent flush commences occurs in vent line
as part of normal shutdown Diverted flow from
Leak noted on KD-502 Leak noted on KD-501 8
procedure KD-502 to KD-501

Pressure on vent line rises

Vent line becomes temporarily (from switch to
partially obstructed KD-501 until the water
flush commences)

Personnel do not
understand relationship
between filling vessel
and entrainment

90
 Vent system plugging
continues to worsen
during water flush

Procedure for isolation

and verification
Emergency pressure
LTA
relief system on
KD-501 activates

Drain valve used

March 12, 2001 March 12, 2001 sometimes, but often not
March 12, 2001
6:53 pm 8:00 pm found to be effective
7 11:21 pm
Water flush commences High temperature in relief
Water flush discontinued
just after shift change catch tank

Main isolation valves for

Decomposition and cross KD-502 were unreliable
linking reactions occur in and not used
hot core, producing gas,
causing foaming
and swelling
Nitrogen to KD-502 valved
out and tagged; isolation
KD-502 core expands and form completed
All nozzles in KD-502
extrudes into nozzles,
obstructed
where material solidifies

Material near KD-502 March 13, 2001

Internal pressure Not proven that KD-502
8 perimeter cools and Prepared KD-502 for 9
develops in KD-502 has no pressure
solidifies opening

Personnel do not Effective means of

understand potential for verification of
decomposition in mass of depressurization defeated
hot, molten polymer by polymer blockage

Understanding of
reaction chemistry
LTA

Occurs once KD-502 is taken out of service

91
 March 13, 2001 Pressure in vessel March 13, 2001
Release of pressure March 13, 2001
2:20 am escapes, breaking 2:36 am Fluid released from piping
causes recoil, which 2:38 am
9 22 of 44 bolts removed remaining bolts; vessel 3 employees fatally injured damaged during recoil
displaces vessel Fire deluge system is
sequentially from cover and contents when KD-502 cover and flashes and burns
backward manually activated
KD-502 cover expelled contents are expelled

KD-502 vessel is
on rollers

92
APPENDIX C: Logic Diagram

93
94
95
96
97