development of a relational chemical process safety … · abstract development of a relational...

DEVELOPMENT OF A RELATIONAL CHEMICAL PROCESS SAFETY

DATABASE AND APPLICATIONS TO SAFETY IMPROVEMENTS

A Thesis

by

FAHAD AL-QURASHI

Submitted to the Office of Graduate Studies ofTexas A&M University

In partial fulfillment of the requirements for the degree of

MASTER OF SCIENCE

December 2000

Major Subject: Chemical Engineering

ABSTRACT

Development of a Relational Chemical Process Safety Database and Applications to

Safety Improvements. (December 2000)

Fahad Al-Qurashi, B.S., Tulsa University

Chair of Advisory Committee: Dr. Sam. Mannan

Industrial accidents still show a major concern to both the public and the environment.

It has been a governmental objective to minimize these accidents. Several rules and

regulations have emerged to reduce the impacts of chemical releases on people and

environment. As a result of these rules, many databases were developed to record

incidents in an attempt to learn from previous mistakes and hence to reduce accidents.

Most of these databases are maintained by federal agencies. However, the taxonomy

inconsistencies of these databases make it difficult to develop a national picture of the

problem of accidental release.

Part of this research presents an analysis of the RMP*Info database, the latest EPA

database, to determine the most significant chemicals released and other trends.

According to this analysis, 85% of the releases in the chemical industry are due to

twelve chemicals. The sources of those releases and their consequences are presented.

In addition, the effects of the chemical type, toxic or flammable, and the number of full

time employees in the facilities are discussed.

To increase the value of the lessons learned from this database, proposed links with

failure rate databases and reactive chemical databases were discussed. The objective of

the relationship among these databases is to bring all relevant information of both

equipment and chemicals into one database. As a result, the new database will make

possible a better understanding by plant personnel about the reliability of plant

equipment and the danger of the chemicals they are dealing with. Consequently,

accidents will be reduced.

This research shows that relationships can be established among the three databases.

Examples were given to demonstrate the procedure of establishing these relationships.

This research is one step in this regard and should be followed by applying the proposed

procedure in a development of a more developed and beneficial relational database that

can help improve the safety performance of industry.

ACKNOWLEDGMENTS

I would like to express my sincere gratitude to my advisor Dr. Sam Mannan for giving

me this opportunity to work on such an interesting topic and all his support and

motivation during my graduate study. I am grateful to Dr. Kenneth Hall and Dr. John

Wagner for serving on my committee. I would like also to express my appreciation to

Dr. William Rogers for his help throughout the course of this research. This work would

not have been completed without all the help and support I received from my wife. I am

also thankful to my wonderful kids, Arwa and Abdulrahman, for their continuous

question “Are you going to school today?” This question inspired me to finish my

research in this short time period. Finally, I would like to thank my parents, especially

my mother, for their prayers and support.

TABLE OF CONTENTS

PageABSTRACT iii

ACKNOWLEDGMENTS v

TABLE OF CONTENTS vi

CHAPTER I Introduction… … … … … … … … … … … … … … … … … … … … .. 1

Accidental Databases… … … … … … … … … … … … … … … … … … . 4 Emergency Response Notification System Database (ERNS) .......... 4 Accidental Release Information Program Database (ARIP) ............. 5 Major Accident Reporting System (MARS) … .....................… ...... 5 Risk Management Program Database (RMP*Info) .......................... 6

CHAPTER II Analysis of the EPA RMP*Info Database .................................. 8

Introduction to the RMP*Info Database ........................................... 8 Analysis of the Top Twelve Chemicals Released ............................. 11 Number of Releases per Year ........................................................... 12 Release Sources ................................................................................. 13 Initiating Events and Contributing Factors ....................................... 15 Toxic vs. Flammable Chemicals ....................................................... 15 Number of Releases per Facility Size ............................................... 16 Conclusions ....................................................................................... 16

CHAPTER III Relationship Between Accidental Databases and Failure Rate Databases… … … … … … … … … … … … … … … … … … … … .. 18

Example ............................................................................................ 22 Problems with the Proposed Relationship ........................................ 23 Modifications to the Current Accidental Databases … ...................... 24

CHAPTER IV Relation Between Accidental Databases and Reactive Chemical Data … … … … … … … … … … … … … … … … … … .. 26

Unstable Molecular Structures .......................................................... 27 Enthalpy of Decomposition/Reaction ............................................... 28 Compatibility with Process Material ................................................. 28 Example ............................................................................................ 29 Problems with the Proposed Relationship ........................................ 30

Modification to Current Accident Databases … ................................ 31

CHAPTER V CONCLUSIONS AND RECOMMENDATIONS ..............… ... 32

LITRUTURE CITED ........................................................................................... 34

APPENDIX A ...................................................................................................... 37

APPENDIX B … .................................................................................................. 41

VITA … … … … … … … … … … … … … … … … … … … … ...… … … … … … … … . 62

CHAPTER I

INTRODUCTION

Accident databases are used to record where, when, and how an accident may have

occurred. This information is being collected by several agencies, both public and

private. This collection of chemical release data is voluntary in some cases and

mandated in others. In many cases, federal laws (e.g., the Clean Air Act) require that

chemical release information be reported to governmental authorities. As a result,

numerous databases have evolved over the years. Most of these databases are

maintained by federal agencies including the Department of Transportation (DOT),

Occupational Safety and Health Administration (OSHA), and Environmental Protection

Agency (EPA).

The inconsistencies of different taxonomy of these databases make it difficult to develop

a national picture of the problem of accidental release. Additional discussion about

these troubles is given in an EPA report to the U.S. Congress[1].

Other problems with current databases include

• They have been developed using inconsistent and faulty data-collecting methods, and the terms

used to describe accidents are often ambiguous.

• The ever-changing reporting requirements prevent comparisons from year to year, making trends

impossible to identify. '

• They provide little to no information about the specifics of accidents, and many accidents are

incorrectly lumped under the heading of chemical accidents[2].

This thesis follows the style of Process Safety Progress

However, that is not to say that the current databases offer no useful information.

In fact, lessons can be learned form current databases to improve the industry safety

practices. See, for example, Accident History Database: An Opportunity by Mannan et

al[3].

The study of accident databases allow for development of realistic programs that are

sensitive to public, environmental, and employee issues, as well as to the chemical

industry’s objectives. Analysis of the current databases will lead to the development of

national safety goals, and will help to refine databases to provide more accurate data for

future evaluation of industry’s safety progress.

The ability to learn from previous incidents has long been regarded as an essential aspect

of any research to reduce the frequency and severity of future incidents. Nonetheless,

many major accidents, which capture media attention, continue to implicate “failure to

learn from previous losses” as a major contributor[4]. Therefore, incidents, once

captured in a database, should be analyzed effectively to raise flags around certain

industry practices to prevent reoccurrence of similar incidents. These effective analyses

cannot be done unless the data included in the database are accurate and truly describe

the incident. For this reason, the database itself must be structured to accomplish this

objective.

The most useful incident databases include both factual and analytical information.

Factual information includes a description of the incident and specific data associated

with the incident such as equipment type, process unit, etc. Analytical information

captures the results of analyses to determine root causes[4]. “Major chemical accidents

cannot be prevented solely through command and control regulatory requirements;

understanding the fundamental root causes of accidents, widely disseminating the

lessons learned, and integrating these lessons learned into safe operations are also

required”[5].

One way to increase the value of a database is to compare its information to that of

another database. Furthermore, the more common facts in different databases, the more

important and valuable these facts are. Therefore, the objective of this project is to

explore three different databases and attempt to link them to increase their value and

hence improve the industry safety performance. These databases are,

• Accidental Databases

• Failure Rate Databases

• Reactive Chemical Databases

Accidental Databases

Several databases have been searched for the linkage. These databases are

• Emergency Response Notification System Database (ERNS)

• Accidental Release Information Program Database (ARIP)

• Major Accidents Reporting System (MARS)

• Risk Management Program Database (RMP*Info)

Emergency Response Notification System Database (ERNS)

Emergency Response Notification System (ERNS) database is an EPA database that was

established in 1986. It contains data on oil and hazardous substances releases. It is a

database of initial notifications made during or immediately after a release occurs. The

quality of the data is limited by four major factors. First, since reports are made during

or immediately after the release, exact details of the release are often unknown, and

therefore the information available in the database may be incomplete or inaccurate.

Second, most of the information in the database have not been verified or validated.

Reports are usually not updated. Third, some of the reports in the database are

duplicates, which may happen when more than one person is reporting a release. And

finally, since reports are taken over the phone, transcription errors may limit the quality

of some data[6].

Accidental Release Information Program Database (ARIP)

Accidental Release Information Program Database (ARIP) is another EPA subsidiary

project. The EPA selects certain significant releases from the ERNS database to be

subjected to the ARIP 23-question survey. Unlike ERNS, data in ARIP database are

considered accurate because information are provided by facilities several months

following the incident. In addition, the database is periodically reviewed for data quality

control. However, the ARIP questionnaire was revised several times and therefore some

information were added and some were deleted. As a result, analysis on some fields of

the database cannot be performed over the entire database. Also, the data collection was

dependent on uneven collection efforts of the EPA regions; therefore, the data do not

reflect the geographical distribution of releases nor represent release trends over time.

Furthermore, because ARIP focuses on significant releases, it is not statistically

representative of all industry releases[6].

Major Accident Reporting System (MARS)

Major Accident Reporting System (MARS) database contains about 400 major accidents

from all of the European Union (EU). The database is well structured and very detailed.

The database may give a good understanding of safety performance if analyzed

collectively. However, analyzing the database per chemical provide few or no useful

conclusions. This is due to the fact that one chemical may result in too few accidents to

produce a general trend.

Risk Management Program Database (RMP*Info)

The United States government formulated new requirements for facilities in an attempt

to prevent, or at least minimize, the consequences of chemical releases. These new

requirements were mandated by the 1990 Clean Air Act Amendments (CAAA) section

112(r)[7]. This law covers facilities that produce, process, handle, or store regulated

substances at or above specified threshold quantities [8]. EPA translated this law into a

rule called Risk Management Programs for Chemical Accidental Release Prevention.

This rule requires relevant facilities to develop a Risk Management Plan (RMP) which

includes three elements; a hazard assessment program, a prevention program, and an

emergency response program.

The hazard assessment program analyzes the consequences associated with an

anticipated worst-case scenario and alternative-case scenario that has higher probability

of occurrence. The hazard assessment program also includes a five-year history of the

facility’s accidental releases.

The prevention program includes safety, maintenance, training, and monitoring

measures to prevent accidents from happening.

The emergency response program describes emergency health care measures and the

procedures of informing the public and response agencies should an accident occur.

This is the latest database and it represents potentially large and useful information that

could be used to analyze and subsequently reduce releases. For this reason, this database

was analyzed and the results were incorporated in chapter II of this study.

CHAPTER II

ANALYSIS OF THE EPA RMP*INFO DATABASE

Introduction to the RMP Database

In 1996, the United States Environmental Protection Agency promulgated the final rule

for Risk Management Programs for Chemical Accident Release Prevention (40 CFR

68)[9,10]. This federal rule is mandated by section 112(r) of the Clean Air Act

Amendments of 1990 and requires regulated facilities to develop and implement

appropriate risk management programs to minimize the frequency and severity of

chemical plant accidents. The EPA rule also requires regulated facilities to develop a

Risk Management Plan (RMP), which includes a description of a hazard assessment, a

prevention program, and an emergency response program. The hazard assessment in

turn must include worst-case scenarios, alternative release scenarios, and a five-year

accident history for the regulated chemicals.

The RMP rule covers facilities that produce, process, handle, or store regulated

substances at or above specified threshold quantities. Notwithstanding the RMP rule

requirements, these facilities “have a general duty to identify hazards which may result

from such releases using appropriate hazard assessment techniques to design and

maintain a safe facility taking such steps as are necessary to prevent releases and

minimize the consequences of accidental releases which do occur[7]”. The regulated

substances are those that if released could cause adverse effects on human health or the

environment.

The hazard assessment program analyzes the consequences associated with an

anticipated worst-case scenario and an alternative release scenario that has higher

probability of occurrence. The hazard assessment program also includes a five-year

history of the facility’s accidental releases.

The prevention program includes safety, maintenance, training, and monitoring

measures to prevent accidents from happening.

The emergency response program describes emergency health care measures and the

procedures for informing the public and response agencies should an accident occur.

This information about the risk management program must be submitted to EPA through

the RMP*Info Database. This database is divided into nine sections as follows:

Section 1: Identification of the facility and its process

Section 2 and 3: Description of worst-case scenarios for toxic and flammable

substances, respectively

Section 4 and 5: Description of alternative release scenarios for toxic and

flammable substances, respectively

Section 6: Five-year accident history

Section 7 and 8: Prevention program description for program 3 and 2 processes,

respectively

Section 9: Emergency response program

Also included in the database is an executive summary of the nature of the facility and

its safety policies.

Since different processes have different levels of risk and impact on human health or the

environment, EPA has adopted three program levels to cover three types of processes.

Program 1 applies to processes that have no off-site accident history and there is no

danger to the public in the case of the worst-case scenarios. Program 3 applies to

processes of higher risk and are already subject to OSHA’s PSM standard. Program 2 is

for processes that are not eligible for Program 1 or 3. It should be noted that each

facility may have more than one process, and each of these processes could be

categorized in program levels depending on the risk involved and the complexity of

operations.

This database is, therefore, more than an accident database. It gives a general overview

of the facility’s risk assessment and helps promotes an effective prevention of accidents.

It also helps reduce chemical risk at the local level and gives response agencies the

necessary information in case of an accident. The general public can benefit from this

database in understanding the chemical hazards in their localities.

As noted above, this database is based on the past 5-year history of each facility. If a

facility has no accident during the 5-year period, then that facility has no entry in

Section 6 of the database; however, they must submit the other information.

Although this database has fewer number of accidents compared to other databases, it

has been submitted to an extensive data review to assure its accuracy, so lessons learned

and conclusions drawn from this database are more meaningful.

Analysis of the Top Twelve Chemicals Released

The following sections discuss an analysis of the top twelve chemicals released

according to the RMP*Info database. These chemicals contributed to 85% of the

releases reported in the database. Lessons learned from the releases of these chemicals

can help improve the safety performance of the chemical industry. These chemicals are

listed below in the order of number of releases, where ammonia has the highest number

of releases.

Ammonia (anhydrous)

Chlorine

Flammable Mixture

Hydrofluoric Acid

Chlorine Dioxide

Propane

Sulfur Dioxide

Ammonia (conc. 20% or greater)

Hydrochloric Acid

Hydrogen

Methane

Butane

Similar to some of the other databases, the EPA 5-year accident history database also

shows ammonia and chlorine as the top two released chemicals. The RMP*Info

database includes a total of 1869 releases from mid-1994 to mid-1999. However, some

releases involve more that one chemical. If each chemical is considered separately, the

total number of releases would be 2004.

Number of Releases per Year

Figure 1 shows the number of accidents per year for all chemicals reported in the

RMP*info database. It should be noted here that the 1994 and 1999 numbers are lower

because these years were only partly within the reporting window[8]. However, there is a

general upward trend in the number of accidents during the period 1995 to 1998. The

same trend is followed if we looked at the number of releases for each chemical per year.

Ammonia and chlorine are shown in Figure 2 as examples.

Release Sources

Looking at the release sources, Figure 3, we see that approximately 50% of all-chemical

releases are associated with failures in valves and piping. More than one-third of the

ammonia releases are associated with valves, whereas piping is the largest source of

releases for chlorine, as shown in Table 1.

Furthermore, if we investigate the initiating event of the ammonia releases due to valves,

we find that 57% of them are due to equipment failure compared to 36% due to human

error, as shown in Table 2. This result suggests that ammonia facilities should give

higher priority to the maintenance and inspection of ammonia valves. This fact is further

emphasized by the knowledge that valves and piping are the sources of most of the

ammonia release deaths, 88% of which is due to valves, as shown in Table 3.

Although the higher number of chlorine releases was due to piping, all chlorine deaths

were due to failures in valves, as shown in Table 3. However, the initiating event of the

valve releases was distributed between equipment failure and human error, 50%

compared to 44%, respectively, as shown in Table 2.

Piping and valves were responsible for more than 70% of the ammonia injuries.

However, process vessels in addition to piping and valves are responsible for 66% of the

chlorine injuries, as shown in Table 5.

If we examine the hydrofluoric releases, we find a similar trend: 50% of the hydrofluoric

acid releases are due to valves (21%) and piping (31%), as shown in Table 1. However,

unlike ammonia and chlorine, 67% of the valve releases were due to human error and the

remainder was due to equipment failure, as shown in Table 2. Although there were no

deaths as a result of hydrofluoric acid releases, 57% of the injuries are due to piping and

valve releases, as shown in Table 5.

The other chemicals have generally similar trends with regard to the importance of

proper maintenance and operation of valves and piping. Therefore, facilities should give

more emphasis to adequate maintenance and operating procedures for valves and piping.

For the top twelve chemical releases, the category “other,” in the release sources section,

ranges from 11-33% depending on the chemical, as shown in Table 1. This indicates

that the multiple choices given for the source of the release works better for some

chemicals than others. For example, 28% of the hydrogen releases fall under the

category “other”. Upon investigation, we see that 50% of the releases under that

category “other” are associated with compressors. Therefore, for hydrogen, the category

compressor is more important than storage vessel, transfer hose, or pump. This suggests

that the choices given should be revised to reflect the wide range of chemicals and

facilities involved. Although the database was designed to include a wide range of

facilities, the choices given are tailored more for manufacturing plants.

Initiating Events and Contributing Factors

More than 90% of the initiating events for each chemical of the top twelve was due

either to equipment failure or human error, which agrees with the general trend of all

chemicals, as shown in Table 6. Equipment failure was the primary cause of most

releases. According to this database, equipment failure resulted in 52 – 77% of the

releases, depending on the chemical released. Mannan et al [3] reported a similar

conclusion from an analysis of the EPA’s Accidental Release Information Program

database.

Equipment failure and human error are listed as an initiating event and contributing

factor. This may lead to confusion for the reporter because of the inability to

differentiate between the initiating event and contributing factor. In fact if we look at

Table 7, we find that the highest percentage occurred where the contributing factor was

the same as the initiating event.

Toxic vs. Flammable Chemicals

The RMP*info database has 77 toxic chemicals and 63 flammables. 81% of the releases

reported to RMP*Info database are from toxic substances. Although most of the

releases reported are toxic, the probability that a toxic release will result in a fatality is

3.7%. However, this percentage increases to 4.5% for the case of a flammable release,

as shown in Table 8.

Looking at the subject chemicals, we find that all flammable chemicals, except butane,

have fatalities; however, three of the toxic chemicals released resulted in no deaths.

These chemicals are hydrofluoric acid, chlorine dioxide, and hydrochloric acid, as

shown in Table 9. It is also interesting to note that all fatalities and almost all injuries

due to flammable substance releases affected on-site workers or contractors, as shown in

Table 4.

Number of Releases per Facility Size

Table 10 shows the relationship between the size of the facility and the number of

releases in which for eleven out of the twelve chemicals, the highest number of releases

were in large facilities, where the number of full-time employees was greater than one

hundred.

Conclusions

1. Analysis of the EPA RMP*Info database indicates that 85% of the releases were

caused by twelve chemicals.

2. 49% of the releases reported in the database are associated with failures in valves

and piping. Valves are responsible for most of the fatalities in the ammonia

environment and all fatalities in the chlorine environment.

3. The category “other” has high number of releases, because it is believed that the

database does not reflect the wide range of chemicals reported and facilities

reporting.

4. The primary initiating event for valve and piping releases is either equipment

failure or human error depending on the chemical released. For example,

equipment failure was the primary initiating event for ammonia and chlorine

releases. On the other hand, human error is the primary initiating event for

hydrofluoric acid releases, which were associated with piping and valves.

5. Although most of the releases reported in the database are toxic, flammable

substances showed a higher probability that a release would result in a fatality.

All of these fatalities were on-site workers and contractors.

6. Most releases occurred in facilities where the number of full-time employees was

100 – 1000.

CHAPTER III

RELATIONSHIP BETWEEN ACCIDENT DATABASES AND FAILURE RATE

DATABASES

As mentioned before, one way to increase the value of a database is to compare its

information to that of another database. Here, the author proposes a relationship

between the information found in the accident databases with those found in the failure

rate databases. A brief introduction to failure rate data is given first.

There are two types of failure rate data available: time-related failure rates and demand-

related failure rates. Time-related failure rates, presented as failures per million hours,

are for equipment that is normally functioning. The failure rate is calculated based on

the following equation,

hours exposure totalequipment failuresequipment related- timeofnumber totalRate Failure = (1)

Demand-related failure rates are presented in failures per one thousand demands and are

for equipment that is normally static but is called upon to operate at indeterminate

intervals. The following equation is used to calculated demand-related failure rate,

equipment upon the demand totalfailuresequipment related-demand ofnumber totalRate Failure = (2)

There are two sources for failure rate data, plant-specific data and generic data. Plant-

specific failure rate data are generated from equipment failure experience at a plant. A

characteristic of plant-specific data is that they reflect the plant’s process, environment,

maintenance practices, and choice and operation of equipment. Generic failure rate data,

on the other hand, are accumulated and aggregated from a variety of plants and

industries, such as nuclear power plants, chemical process industries or offshore

petroleum platforms. Hence, generic data are derived from equipment of many

manufacturers, a number of processes, and many plants with various operating

strategies. Consequently, they are much less specific and detailed, but they are

frequently adequate to identify the major risk contributors in a process or plant. [11]

Failure rate data can be obtained from the following resources:

AIChE CCPS publication, “Guidelines for Process Equipment Reliability Data

with Data Tables”

Offshore Reliability Data (OREDA)

Government Industry Data Exchange Program (GIDEP)

IEEE Standard 500-1984

United Kingdom Atomic Energy Authority (UKAEA)

Since the AIChE CCPS failure rate data are the primary data for the chemical process

industry, it is the one that was used in the study to illustrate the proposed relationship.

Table 11 shows the CCPS failure rate data input form.

It is always preferred, when using failure rate data, to have valid historical data from the

identical equipment in the same application. In most cases, however, plant-specific data

are unavailable or may be of low confidence level to be used without supporting data.

The supporting data in this case are the generic data. [12]

There are many variables that affect the confidence level of plant-specific data. Process

conditions and/or maintenance practices can fluctuate throughout a study period and

have a major influence on the results: “intensified preventive maintenance can lower and

eliminate failures; changes in process conditions may severely aggravate fouling

tendencies or corrosion rates; equipment may be upgraded or even replaced thereby

extending operating life during the study; many failures may have been missed; many

failures may be wrongly recorded such as a reported pump failure when the push button

was really at fault” [12]. Plant-specific data may be heavily biased by the variations noted

and result in data of little statistical confidence.

There are several factors that are not incorporated in failure rate data. Some of these

factors are

1. External environment such as vibration, temperature, humidity, etc

2. Materials of construction

3. Maintenance and inspection strategies

4. Process internal temperature or pressure

5. Design standards

6. Equipment manufacturer [12]

Information in failure rate data can help plant engineers identify the plant major

deficiencies and focus on them. For example, if certain facility have a pump failure rate

above the upper value reported in failure rate data and a valve failure rate below the

mean value, then this facility should give more attention to pump reliability than valves.

Similarly, if this facility showed a higher number of incidents attributed to pumps in

accident databases, then this fact is further emphasized.

One way to do this relationship is as follows:

1. Pick a facility from the RMP*Info database

2. Count the number of failures for certain equipment during 5 years

3. Convert (2) to failures per million hours

4. Compare (3) with the reported failure rates in failure rate databases for that

specific equipment

5. If (3) is below the mean, this equipment is not a major source of accidents

6. Otherwise, if (3) is above the mean, this equipment is a major source of accidents

and the facility should do something to lower this rate.

Facilities should strive to meet the mean failure rate for all of their equipment. This goal

is more achievable and practical than the “industry best practice” idea implemented by

industry. As facilities meet the mean of the equipment failure rate, the mean will keep

going down. Consequently, equipment reliability should improve and failure rates

should decrease.

The steps above can be summarized in the following equation.

Q * Z* Y

1000,000 * X hoursMillion per Failures = (3)

Where,

X = Number of equipment failures

Y = Number of years covered

Z = Number of days the equipment is operational per year

Q = Number of working hours per day

Example

Facility ID = 1118

Facility Name = Harcros Chemicals Inc. -- St. Gabriel

Number of years covered = 5 years

Number of valve failures = 1

Assumptions

1. Number of days the equipment is operational = 365 days

2. Number of working hours per day = 24 hours

Therefore, Failures per million hours = 22.83

However, when we compare this number with valve failure rate data, we encounter

several problems. These problems are listed below.

Problems with the Proposed Relationship

1. To accurately determine the failure rate for any equipment in a plant, the equipment

population in that facility is necessary. We need to divide by the population number

to calculate the failures per million hours per equipment. Unfortunately, this

information is not available in the RMP*Info database.

2. Failures in accident databases do not have the same subdivisions as in the failure rate

databases. For example, accidents may be reported under valves in accident

databases. However in failure rate databases, valves are divided into check valves,

manual valves, and operated valves. In this case, the analyst does not know under

which type of valves that accident falls.

3. Another problem is that the analyst cannot know if the equipment that caused the

incident in accident databases was being used continuously or intermittently. This

distinction is important if the failure is time-related or demand-related.

4. Failure mode is another important category in failure rate data. Catastrophic,

degraded, and incipient are the types of failure modes in failure rate databases.

Failure data are reported based on their failure modes, but this information is not

included in accidental databases.

5. Comparing the failure rates of equipment with failure rate databases should be done

with caution. The analyst may choose to use the data if the equipment design,

process condition, properties of the chemicals being processed, severity of duty or

quality of the facility maintenance regime, etc. are similar to the equipment being

studied. Since most failure rate data are generic, they are derived from equipment of

many manufacturers, a number of processes, and many plants with various operating

strategies, these factors will usually be different. To account for these differences,

the analyst should use his or her own expertise to adjust failure rate data to his or her

needs.

Because of these problems, it is difficult to compare the valve failures in the above

example with failure rate data. We do not know what type of valve it is, what is the

population of similar valves is in that facility, whether this valve is being used

continuously or intermittently, etc. These data are necessary for accurate comparisons.

To account for the population of equipment in a facility, equation 3 can be modified as

follows:

N*Q * Z* Y

1000,000 * X hoursMillion per Failures = (4)

Where N is the population of that particular equipment in the facility.

Modifications to the Current Accidental Databases

For a more efficient application of this relationship, the following modifications to

accident databases are proposed:

1. Introduce new fields in accidental databases to indicate the type of equipment that

failed. These new fields should be subdivided based on the subdivisions found in

failure rate data. Table 12 shows an example of such subdivisions.

2. Accident databases should also include a new field to specify the type of operation of

the equipment (i.e. continuous or per demand). This segregation is important for

comparing accidental databases with failure rate data.

3. Accident databases should include a brief description of the accident. The

description should include the mode of failure of the equipment. For example, in

case of valve failure, the description should indicate what type of valve it is and how

it failed (e.g. fails to close, fails to open, external leakage, etc).

4. A field should also be added for the total number of similar equipment in use at the

facility at the time of the incident.

CHAPTER IV

RELATIONSHIP BETWEEN ACCIDENT DATABASES AND REACTIVE

CHEMICAL DATA

In chapter III we discussed how accidental databases can be linked to failure rate

databases. Here, we consider another proposed linkage, i.e. relating accidental

databases to reactivity data. Knowledge of the chemical reactivity and applying this

knowledge in the design, operation, and maintenance of chemical processes can help

prevent many accidents.

The most catastrophic accident in the history of industry occurred in December 1984

in Bhopal, India. More than 6000 civilians were killed as a result of a release of a

reactive, toxic, and volatile chemical, methyl isocyanate (MIC). MIC is an

intermediate step of pesticide production in that plant and it demonstrates a number

of potentially dangerous physical properties. Its boiling point at atmospheric

pressure is 39.1 oC, its vapor pressure is 348 mmHg at 20 oC, its vapor is heavier

than air, it reacts exothermally with water, and the maximum exposure concentration

for an eight-hour period is only 0.02 ppm. Since we know these properties of MIC,

an alternative reaction scheme could have been used in that plant with a less

dangerous chloroformate intermediate[13]. If this knowledge was applied in the

design of that plant, the consequences of the accident may have been less

catastrophic.

“Hazardous chemical reactivity is any chemical reaction with the potential to exhibit

rates of increase in temperature and/or pressure too high to be absorbed by the

environment surrounding the system.” [14]

There are several ways potential hazards of a reactive chemical can be identified.

Some of these are,

1. Literature reviews of chemical properties and past incidents

2. Unstable molecular structure

3. Enthalpy of decomposition/reaction

4. Compatibility with process materials [14]

Initially, literature should be searched for relevant data on the chemical in question.

These data should include physical-chemical properties, thermodynamics, past

incidents, case studies, etc. If available information is insufficient, a parallel

investigation of the remaining three approaches should be carried out for the

chemical under investigation.

Unstable Molecular Structures

Substances with a positive enthalpy of formation will always release energy during

their decomposition. Energetic substances can in general be identified by the

presence of hazardous molecular structures. Bretherick [15] published a compilation

of energetic groups that can release substantial energy upon cleavage. However,

there are two deficiencies with these listings:

1. The presence of one of the mentioned groups in the molecule does not

necessarily imply that the substance is hazardous. The presence of such

active groups in a high molecular weight compound reduces the explosive

potential.

2. In addition, the absence of these active groups is no guarantee for long-term

stability of the compound. Some aldehydes, for example, which are not

present in these tables, can be converted to peroxides, which are unstable, by

reacting with oxygen [14,16].

Enthalpy of Decomposition/Reaction

Correlating the computations of about 200 different compounds of various hazard

potential concluded that a reaction with a maximum change in enthalpy less than

- 0.7 kcal/g can be potentially explosive [16]. The 200 chemicals included substances

known to have caused explosions and mixtures considered to be non-hazardous.

Table 13 summarizes these findings.

Compatibility with the Process Material

In case a substance is stable in itself, it can be dangerous when mixed with other

process materials. Information about the possibility of such incompatibilities

problem is available in the National Fire Protection Association (NFPA) Manual [17].

Table 14 gives an example of such reactions.

Even if the substance is considered to be a non-explosion hazard, both non-energetic

and compatible with the process-common materials, it still should be subjected to

screening tests to ensure the absence of instabilities and the potential of a runaway

reaction [14]. Also, the presence of catalysts and contaminants or changes in process

conditions can alter the reaction and lead to dangerous by-products.

Example

Acetylene is one of the compounds listed in the RMP*Info database. This

compound will be used to illustrate the above procedures. Its formula is C2H2 and its

molecular weight is 26.04.

• Molecular Structure:

This molecular structure is one of the high-energy molecular structures listed in

the reactive chemical data.

• Enthalpy of Decomposition/Reaction

Let us consider the following reaction,

I2 (g) + C2H2 (g) à C2H2I2 (g)

This reaction has an enthalpy of reaction, at standard conditions, equals to

– 0.764 kcal/g. Since the enthalpy of reaction is less than - 0.7 kcal/g, it is

H - C C - H

considered violently exothermic and can lead to catastrophic accidents if not

handled properly.

• Compatibility with Process Material

We have indicated above that acetylene can react explosively with iodine.

Furthermore, looking at Table 13, we find that acetylene in contact with copper will

produce copper acetylide, which is sensitive to shock and friction.

The above example illustrates the fact that if this information was incorporated in a

database, plant personnel would be more acquainted of the danger of acetylene and

strive to minimize its impacts.

Problems with the Proposed Relationship

1. The required information, especially enthalpy of reaction and compatibility, is

not available for every chemical. Experiments and chemical computations must

be conducted to determine the missing data.

2. This proposal is valid under normal plant operation conditions. A more

dangerous situation could result in case of abnormal conditions, i.e. high

temperature, high pressure, presence of contaminants, etc., where

experimentation or chemical computation to obtain the missing data could be

more difficult.

Modification to Current Accidental Databases

New fields can be added to accident databases to incorporate reactivity data. The

needed fields are the chemical molecular structure, known enthalpy of formations

and reactions, and known compatibility problems. Of course, it is impossible to

include all possible reactions for each chemical in a database, but reactions that

industry consider important or known to have caused accidents in the past must be

included.

CHAPTER V

CONCLUSIONS AND RECOMMENDATIONS

1. Current accident databases do not represent the safety performance and safety

trend of the chemical industry.

2. Although there are still areas of improvements, there has been much effort

given to the RMP*Info database, which makes it a good candidate for safety

performance analysis.

3. Equipment failures constitute the primary initiating event of industrial

accidents.

4. Facilities are urged to compare their equipment failure rates with equipment

failure rate data and develop procedure for improvement.

5. Accident database can be linked to both failure rate data and reactivity data.

This paper is a step forward in this regard.

6. Available failure rate data are not comprehensive. There are several types of

data that are not available.

7. To better utilize the relational aspects between accident databases and failure

rate data, the procedure proposed here is recommended.

8. Reactivity data are not available for every chemical. As a result, to optimize

the linkage between accidental databases and reactivity data, the missing

information must be made available. Experiments and chemical computation

should be conducted to achieve this goal.

9. If linkages are established among the three databases, it will give a better

understanding to plant personnel about the reliability of plant equipment and

the danger of the chemicals they are dealing with. Consequently, accidents

will be reduced.

LITERATURE CITED

1. US EPA, “A Review of Federal Authorities for Hazardous Materials Accident

Safety”, US EPA-550-R-93-002, (December 1993).

2. McCray, E.T; “ Chemical Accident Databases: What They Tell Us and How They

Can Be Improved to Establish National Safety Goals.” MS Thesis, Texas A&M

University, (May 2000).

3. Mannan, M. S., O’Connor, M., West, H. H., “Accidental History Databases: An Opportunity”,

Environmental Progress, 18, No. 1, pp 1 – 6, (Spring 1999).

4. Vaughan, R., Kelly, B.,“CCPS Process Safety Incident Database (PSID)”, online, available:

www.aiche.org/ccps/lldb.htm. Accessed: Feb 2000.

5. US EPA, “Prevention of Reactive Chemical Explosions”, www.epa.gov, (April 2000).

6. US EPA, “User’s Guide to Federal Accidental Release Databases”, EPA-550-B-95-001, (September

1995).

7. Clean Air Act, Title III-section 112 (r): Hazardous Air Pollutants, Prevention of Accidental Releases,

online, available: www.epa.gov. Accessed: March 2000.

8. Kleindorfer, P. R.; Feldman, H.; Lowe, R. A., “Accident Epidemiology and U.S. Chemical

Industry: Preliminary results from RMP*Info”, Working Paper 00-01-University of Pennsylvania,

(March 6, 2000).

9. US EPA, “Risk Management Programs for Chemical Accidental Release Prevention; Final Rule”, (40

CFR Part 68), Federal Register, 61, No. 120, pp. 31667-31732, Washington, DC, (June 20, 1996)

10. US EPA, “List of Regulated Substances and Thresholds for Accidental Release Prevention and Risk

Management Programs for chemical Accidental Release Prevention; Final Rule and Notice”, (40 CFR

Parts 9 and 68), Federal Register, 59, No. 20, pp. 4478-4501, Washington, DC, (January 31, 1994)

11. Center for Chemical Process Safety: “Guidelines for Process Equipment Reliability Data with Data

Tables”, American Institute of Chemical Engineers, (1989).

12. Greenberg, H. R. and Cramer, J. J., “Risk Assessment and Risk Management for the Chemical

Process Industry”, Van Nostrand Reinhold, New York, (1991).

13. Daniel A. Crowl and Joseph F. Louvar, “Chemical Process Safety: Fundamentals with

Applications”, Prentice Hall PTR, Englewood Cliffs, NJ, (1990).

14. Center for Chemical Process Safety, “Guidelines for Chemical Reactivity Evaluation and

Application to Process Design”, American Institute of Chemical Engineers, (1995).

15. Bretherick, L., “Handbook of Reactive Chemical Hazard”, Butterworth-Heinemann, Stoneham, MA,

(1990).

16. Howard H. Fawcett and William S. Wood, “Safety and Accident Prevention in Chemical

Operations”, John Wiley & Sons, New York, (1982).

17. National Fire Protection Association, “Manual of Hazardous Chemical Reactions”, NFPA, 491M,

Quincy, MA, (1991).

APPENDIX A

Table 1. Release Sources per Chemical

Ammonia (Anhhydrous)Number Percent

Storage Vessel 55 9Piping 141 22Process Vessel 62 10Transfer Hose 67 11Valve 213 34Pump 41 7Joint 18 3Other 76 12Total 630 107

ChlorineNumber Percent


Flammable MixtureNumber Percent


HydrofloricAcid

Number PercentStorage Vessel 6 6Piping 30 31Process Vessel 9 9Transfer Hose 6 6Valve 21 21Pump 17 17Joint 3 3Other 12 12Total 98 106

Chlorine DioxideNumber Percent


PropaneNumber Percent


HydrogenNumber Percent


MethaneNumber Percent


ButaneNumber Percent


AllNumber Percent


Percent adds to more than 100% due to multiple causes.

Table 4. Consequnces of Releases per Chemical

Ammonia (Anhydrouse)Workers/Contractors 7

Deaths Public Responders 0Public 0Workers/Contractors 629

Injuries Public Responders 17Public 18

On-Site Property Damage $341,765,401Deaths 0Hospitalization 48

Off-site Other Medical Treatments 415Evacuation 7,977Sheltered in Place 10,779Property Damage $1,653,014

ChlorineWorkers/Contractors 0




Off-site Other Medical Treatments 109Evacuation 6,214Sheltered in Place 107,416Property Damage $99,976

Flammable MixtureWorkers/Contractors 8




Off-site Other Medical Treatments 219Evacuation 30Sheltered in Place 2,500Property Damage $7,125,132

Hydrofloric AcidWorkers/Contractors 0




Off-site Other Medical Treatments 34Evacuation 675Sheltered in Place 1,500Property Damage $420,000

Chlorine dioxideWorkers/Contractors 0



On-Site Property Damage $202,000Deaths 0Hospitalization 0

Off-site Other Medical Treatments 1Evacuation 0Sheltered in Place 0Property Damage $0

PropaneWorkers/Contractors 1





Sulfur dioxideWorkers/Contractors 1




Off-site Other Medical Treatments 89Evacuation 1,150Sheltered in Place 2,430Property Damage $1,725

Ammonia (conc. 20% or greater)Workers/Contractors 0



On-Site Property Damage $992,000Deaths 0Hospitalization 0

Off-site Other Medical Treatments 3Evacuation 121Sheltered in Place 2,215Property Damage $0

Hydrochloric AcidWorkers/Contractors 0




Off-site Other Medical Treatments 112Evacuation 655Sheltered in Place 20,864Property Damage $467,973

HydrogenWorkers/Contractors 2




Off-site Other Medical Treatments 0Evacuation 0Sheltered in Place 10Property Damage $200,000

MethaneWorkers/Contractors 1




Off-site Other Medical Treatments 1Evacuation 300Sheltered in Place 0Property Damage $2,300

ButaneWorkers/Contractors 0





AllWorkers/Contractors 26

Deaths Public Responders 0Public 0Workers/Contractors 1,715


On-Site Property Damage $1,000,877,807Deaths 0Hospitalization 192

Off-site Other Medical Treatments 5,580Evacuation 22,011Sheltered in Place 196,570Property Damage $10,871,363

Table 6. Releases Initiating Events per Chemical

Ammonia (Anhydrous)Number Percent

Equipment Failure 353 56Human Error 223 35Natural 21 3Unknown 32 5Total 630 100

ChlorineNumber Percent


Flammable MixtureNumber Percent


Hydrofloric AcidNumber Percent


Chlorine DioxideNumber Percent


PropaneNumber Percent


Sulfur DioxideNumber Percent


Ammonia (conc. 20% or greater)Number Percent


Hydrochloric AcidNumber Percent


HydrogenNumber Percent


MethaneNumber Percent


ButaneNumber Percent


AllNumber Percent


Table 7. Releases Primary and Contributing Causes

Ammonia (Anhydrous)Contributing Factor Initiating Event

Equipment Failure Percent Human Error PercentEquipment Failure 254 40 28 4Human Error 56 9 152 24Improper Procedure 28 4 96 15Overpressurization 43 7 23 4Upset Condition 16 3 2 0By pass 3 0 1 0Maintenance 43 7 44 7Process Design Failure 22 3 14 2Unsuitable Equipment 15 2 8 1Unusual Weather 4 1 0 0Management Error 9 1 7 1Other 33 5 20 3Total Releases 630

ChlorineContributing Factor Initiating Event


Flammable MixtureContributing Factor Initiating Event


Hydrofloric AcidContributing Factor Initiating Event


Chlorine DioxideContributing Factor Initiating Event


PropaneContributing Factor Initiating Event


Sulfure DioxideContributing Factor Initiating Event


Ammonia (conc. 20% or greater)Contributing Factor Initiating Event


Hydrochloric AcidContributing Factor Initiating Event


HydrogenContributing Factor Initiating Event


MethaneContributing Factor Initiating Event


ButaneContributing Factor Initiating Event


AllContributing Factor Initiating Event

Equipment Failure Percent Human Error PercentEquipment Failure 724 39 96 5Human Error 204 11 478 26Improper Procedure 131 7 293 16Overpressurization 104 6 50 3Upset Condition 99 5 25 1By pass 32 2 10 1Maintenance 162 9 157 8Process Design Failure 83 4 46 2Unsuitable Equipment 71 4 28 1Unusual Weather 18 1 8 0Management Error 29 2 40 2Other 111 6 64 3Total Releases 1869Percent adds to more than 100% due to multiple causes.

Table 10. Relationship Between the Size of Facilities and Number of Releases

Ammonia (Anhydrous)Full Time Employees Number Percent1 - 10 115 1811 - 100 128 20101 - 1000 295 471001 + 85 13

630 99

ChlorineFull Time Employees Number Percent1 - 10 63 1211 - 100 216 42101 - 1000 171 331001 + 53 10

511 98

Flammable MixtureFull Time Employees Number Percent1 - 10 6 611 - 100 19 19101 - 1000 59 591001 + 15 15

100 99

Hydrofloric AcidFull Time Employees Number Percent1 - 10 1 111 - 100 6 6101 - 1000 81 831001 + 10 10

98 100

ChlorineDioxide

Full Time Employees Number Percent1 - 10 0 011 - 100 0 0101 - 1000 31 531001 + 27 47

58 100

PropaneFull Time Employees Number Percent1 - 10 3 611 - 100 13 25101 - 1000 33 631001 + 3 6

52 100

Sulfur DioxideFull Time Employees Number Percent1 - 10 2 411 - 100 15 31101 - 1000 27 561001 + 4 8

48 100

Ammonia (conc. 20% or greater)Full Time Employees Number Percent1 - 10 12 3111 - 100 9 23101 - 1000 15 381001 + 3 8

39 100

Hydrochloric AcidFull Time Employees Number Percent1 - 10 0 011 - 100 11 33101 - 1000 17 521001 + 5 15

33 100

HydrogenFull Time Employees Number Percent1 - 10 0 011 - 100 2 6101 - 1000 20 631001 + 10 31

32 100

MethaneFull Time Employees Number Percent1 - 10 5 1911 - 100 6 22101 - 1000 12 441001 + 4 15

27 100

ButaneFull Time Employees Number Percent1 - 10 0 011 - 100 3 12101 - 1000 19 761001 + 3 12

25 100

AllFull Time Employees Number Percent1 - 10 209 1111 - 100 494 26101 - 1000 897 481001 + 254 14

1869 99

Figure 1. Number of Releases per Year

0

50

100

150

200

250

300

350

400

450

500

1994* 1995 1996 1997 1998 1999*

Year

Num

ber o

f Rel

ease

s

* incomplete data

Ref: Kleindorfer, P. R. et al, "Accident Epidemiology and U.S. Chemical Industry: Preliminary results from RMP*Info"

Figure 2. Number of Ammonia and Chlorine Releases per Year

0

20

40

60

80

100

120

140

160

180

1994* 1995 1996 1997 1998 1999*

Year

Num

ber o

f Rel

ease

s

Ammonia (anhydrous) Chlorine

* incomplete data

Figure 3. Percent of Total Releases vs. Release Source

0

5

10

15

20

25

30

StorageVessel

Piping ProcessVessel

TransferHose

Valve Pump Joint Other

Release Source

Per

cent

of T

otal

Rel

ease

s

APPENDIX B

Table 2. The Primary Initiating Event for Valve Releases per Chemical

Equipment Failure (%) Human Error (%)Ammonia (Anhydrouse) 57 36

Chlorine 50 44Flammable Mixture 71 29

Hydrofloric Acid 33 67Chlorine Dioxide 75 25

Propane 50 44Sulfur Dioxide 20 80

Ammonia (conc. 20% or greater) 50 50Hydrochloric Acid 100 0

Hydrogen 100 0Methane 75 25Butane 50 50

Table 3. Percent Deaths due to Valve and Piping Releases

Valve (%) Piping (%) Process Vessel (%) Sum (%)Ammonia (Anhydrouse) 88 28 13 128

Chlorine 100 0 0 100Flammable Mixture 13 75 13 100

Hydrofloric Acid 0 0 0 0Chlorine Dioxide 0 0 0 0

Propane 0 100 100 200Sulfur Dioxide 0 0 100 100

Ammonia (conc. 20% or greater) 100 0 0 100Hydrochloric Acid 0 0 0 0

Hydrogen 0 100 100 200Methane 0 100 100 200Butane 0 0 0 0


Table 5. Percent Injuries due to Valve, Piping and Process VesselReleases

Valve (%) Piping (%) Process Vessle (%) Sum (%)Ammonia (Anhydrouse) 30 41 9 80

Chlorine 18 29 18 66Flammable Mixture 22 47 19 88

Hydrofloric Acid 19 38 9 66Chlorine Dioxide 10 23 50 83

Propane 32 49 41 122Sulfur Dioxide 32 32 28 92

Ammonia (conc. 20% or greater) 15 15 15 46Hydrochloric Acid 0 3 11 14

Hydrogen 30 48 55 132Methane 26 46 44 116Butane 34 34 46 114


Table 8. Percent of Occurrence of Fatality per Chemical Type

Chemical Type Number of Releases Number of Deaths Percent of Death OccurrenceToxic ( T ) 1624 60 3.7%Flammable (F) 380 17 4.5%

Table 9. The Effect of the ChemicalType on the Number of Releases and Number of Deaths

Type Number of Releases Number of DeathsAmmonia (Anhydrous) T 630 37

Chlorine T 511 9Flammable Mixture F 100 8

Hydrofloric Acid T 98 0Chlorine Dioxide T 58 0

Propane F 52 1Sulfur Dioxide T 48 1

Ammonia (conc. 20% or greater) T 39 5Hydrochloric Acid T 33 0

Hydrogen F 32 2Methane F 27 1Butane F 25 0

VITA

Fahad Alqurashi was born in Taif, Kingdom of Saudi Arabia, on July 31, 1973.

After completing his schooling in Dhahran High School, he pursued his B. S. degree in

Chemical Engineering at Tulsa University from 1991 to 1995. Prior to continuing his

education, he worked for Saudi Aramco Oil Company. In August 1999, he entered the

Chemical Engineering Department at Texas A&M University to pursue his master’s

degree. He will be working for Saudi Aramco Oil Company, Saudi Arabia on January 1,

2001.

Permanent address:

Kingdom of Saudi ArabiaSaudi Aramco Oil CompanyUthmaniyah Gas Plant DepartmentEngineering DivisionProcess Unit

development of a relational chemical process safety … · abstract development of a relational...

Documents