2.5 5.

Article

Investigation of an Explosion at a
Styrene Plant with Alkylation
Reactor Feed Furnace

Yao-Chang Wu, Bin Laiwang and Chi-Min Shu

Special Issue
Selected Papers from IMETI 2018
Edited by
Prof. Dr. Wen-Hsiang Hsieh, Dr. Jia-Shing Sheu and Prof. Dr. Minvydas Ragulskis

https://doi.org/10.3390/app9030503
 applied
sciences
Article
Investigation of an Explosion at a Styrene Plant with
Alkylation Reactor Feed Furnace
Yao-Chang Wu 1 , Bin Laiwang 1 and Chi-Min Shu 1,2, *
1 Graduate School of Engineering Science and Technology, National Yunlin University of Science and
Technology, Yunlin 64002, Taiwan; [email protected] (Y.-C.W.); [email protected] (B.L.)
2 Center for Process Safety and Industrial Disaster Prevention, School of Engineering, National Yunlin
University of Science and Technology, Yunlin 64002, Taiwan
* Correspondence: [email protected]; Tel.: +886-5-534-2601 (ext. 4416)




Received: 9 December 2018; Accepted: 28 January 2019; Published: 1 February 2019


Abstract: To prevent and mitigate chemical risks in the petrochemical industry, such as fires and
spillage, process safety management (PSM), is essential, especially where flammable, corrosive,
explosive, toxic, or otherwise dangerous chemicals are used. We investigated process safety (PS)
between man–machine (material equipment) and environmental interfaces by using process hazard
analysis (PHA) and fault tree analysis (FTA). By analyzing the data obtained through machinery
and mechanical integrity (MI), pre-startup safety review (PSSR), current operating modes, and areal
locations of hazardous atmospheres (ALOHA) simulations of the disaster’s aftermath, the cause of the
styrene plant accident was found to be the fuel furnace (F101) switching process. Although the furnace
had been extinguished, fuel continued to enter the furnace, and it was exposed to a high-temperature
surface, resulting in the flashing ignition of the C4 fuel. The plan-do-check-act (PDCA) management
model can be used to forestall the system from accident, and it is used to improve the proposal and
develop countermeasures that would increase PSM performance and substantially lessen the impact
of the thermal hazard. Disasters are often attributable to the unsafe state of machinery, equipment,
or the environment, dangerous behaviors of the operator, and the lack of a thorough management
system. It is anticipated that the investigation and analysis of the accident would not only find the
real cause of the disaster but also lead to the establishment of better effective solutions for common
safety problems.

Keywords: process safety management; pre-startup safety review; simulation; aftermath; effective
solution

1. Introduction
Although the petrochemical industry has brought convenience to modern life, some accidents
sporadically occur during the production process. In general, the process of raw materials poses
enormous threats to human life. If improperly handled in preparation operation, management,
transportation, or storage, these chemicals can cause fires, explosions, or leakages of toxic gas.
In addition to causing loss of life and property damage, these kinds of industrial accidents could
also trigger public protest over these large-scale plants. Therefore, after complex disasters happen,
the social cost is often difficult to estimate. Nevertheless, the impact of a recent disaster was clear.
On the morning of 6 March 2017, an explosion thundered through a styrene monomer (SM) company
in Taiwan. In the aftermath of this incident, the public safety and environmental quality in the vicinity
of industrial park were seriously affected. In light of the growing frequency of industrial accidents,
this paper and its subject served as a relevant case study in the critical importance of process safety
management (PSM). We hope to assist manufacturers in dealing with their accident investigation

Appl. Sci. 2019, 9, 503; doi:10.3390/app9030503 www.mdpi.com/journal/applsci

Appl. Sci. 2019, 9, 503 2 of 17

reports and to help experts and scholars diagnose the causes of disasters to ensure that similar accidents
will not occur in the future. In this way, the lives of employees and the property of manufacturers
could be protected. Taiwan’s SM plant alkylation reactor feed heating furnace explosion (equipment
number: F101) accident on 6 March 2017 has been used as an example of PSM to provide analysis
information to substantially lessen the risks as well as severities to the petrochemical industry [1,2].
In chemical manufacturing, the heating furnace is a process equipment that exchanges heat energy
in furnace to heat the feed fluid to specific temperature range. The heating furnace is the main unit of
the chemical processing system, fluid flows through the tubes into the furnace. Outside the tube is the
heat generated by the combustion of fuel and air, so that the fluid is heated to a specific temperature.
After the fuel is ignited by a burner, a flame is generated, and the combustion air flowing in the furnace
is heated to exceptionally high temperature. Then, the combustion gas is transferred to heat the pipe
by means of radiation and convection; the fluid is heated to the setpoint of the temperature for the
process. Generally speaking, the gas in the heating furnace is not only extremely flammable, explosive,
but also corrosive and toxic. Since the furnace is operated under high temperature conditions, the
operating conditions may be poor; the gas generation period is short. If there is negligence or violation
of the operating procedures, it may cause a fire or an explosion.
The heating of raw materials in chemical plants mainly depends on the heating furnace. The
furnace temperature is sometimes as high as 700–800 ◦ C (or even higher). Once the material of the
furnace tube leaks, whether the temperature exceeds the self-ignition point or not, it is directly in
contact with the open flame, so the hazard is extremely serious. In addition, improper ignition could
also cause explosion. Once the furnace is damaged, it will directly lead to downtime and increase the
loss. Therefore, it is necessary to take measures to prevent and mitigate fire and explosion accident.
Before commencing with the study and investigation, a brief literature [3] search was implemented
to collect publicly available information about year’s explosion accidents in furnaces. Tables 2 and 3
report this information. It was observed that even with detailed design, these accidents still occur;
because they occurred at a high pressure within the system, they result in the loss of property and
life. The accidents reported in Tables 2 and 3 indicate that most of them are caused by human error,
improper maintenance, or faulty design. Accordingly, it is recognized that the accidents of heating
furnaces occur sporadically. Common accidents mainly include furnace explosion, coking in the
furnace, and explosion in the flue. The furnace tube is damaged, the material in the tube leaks into the
furnace, fire occurs, and the fire or explosion caused by overheating. In addition, the burner of heating
furnace is damaged, the refractory insulating material in the furnace is impaired. The environment is
seriously polluted, and the heating furnace has exploded. In numerous literature descriptions, furnace
explosion was the most common, destructive, and extremely serious accident. In the past, there have
been few literature studies on accident cases. Therefore, we used the furnace explosion accident as an
example to analyze and improve the understanding of the explosion accident of the heating furnace.

Table 1. Furnace explosion accident case summary in U.S.

Date of Specific Explosion Details Injuries/

Location
Incident Details Cause Consequence Fatalities
2 Weakened furnace wall (holes) 1 employee
Used for Explosion spewing hot
November Vernon, CA compromised the integrity of received 2nd and
smelting lead lead slag and dust
2002 the furnace 3rd degree burns
Attempted maintenance on a
Explosion resulting in 1 severe injury,
16 March Cucamonga, Electric arc water leak from the furnace,
hot steam and flying 2 minor burns
2004 CA furnace loud popping noise resulted
debris and cuts
in explosion
Furnace was tilted forward to Molten steel leaked out
10 March Midlothian, Melt shop begin tapping of slag door, slag pot 1 employee was
2006 TX furnace (steel) when the hydraulic hose overfilled and an killed
cylinder failed explosion occurred
Appl. Sci. 2019, 9, 503 3 of 17

Table 2. Furnace explosion accident case summary in U.S.

Date of Specific Explosion Details Injuries/

Location
Incident Details Cause Consequence Fatalities
Molten steel caused a water leak 1 employee
27 May Coatesville, Electric arc
to become superheated Explosion, molten steel killed, 2 seriously
2007 PA furnace
high-pressure vapor burned
Aluminum car rims were placed
29
Manchester, Used to melt into the furnace, moisture was 1 fatality,
November Explosion, molten metal
GA aluminum still on the rims and a violent 6 serious injuries
2007
explosion occurred
Explosion from
2 workers killed,
21 March Louisville, Large electric Water leaked into furnace which overpressure that sent
2 seriously
2011 KY arc furnace caused an overpressure event furnace contents
injured
spewing into air
Workers were opening/closing
21 a furnace valve that contained 2 workers killed,
Fairfield,
September N/A oxygen and hydrated lime, Fiery explosion 1 critically
AL
2014 while the furnace was injured
in operation

Table 3. Select cases on furnace explosion accident in Taiwan and Mainland China.

Date of Incident Location Equipment Explosion Details Injuries/Fatalities

During fuel gas into the furnace,
Mainland
7 November 2006 Furnace causing the gas pipe of heating Equipment damage
China
furnace to burst and explosion.
Mainland
3 May 2008 Furnace Heating furnace tube coking. Equipment damage
China
In the diesel plant, during starting
Taiwan 4 seriously
6 March2017 Furnace the furnace, the tube of furnace was
SM plant injured
cracked and led to explosion.
Taiwan In the diesel plant, during starting
No casualties
29 January 2018 Taoyuan Furnace the furnace, the tube of furnace was
Equipment damage
Refinery cracked and lead to explosion.

We integrated the furnace explosion case and analyzed the relevant factors comprehensively that
may cause fire and explosion during the operation of the furnace. The explosion factors of the heating
furnace system were established by fault tree analysis method, and qualitative analysis was carried out
to provide a basis for diagnosis and prevention of furnace explosion. The results of this study could
provide the operators and managers with further understanding of the elements of furnace explosion
and substantially diminish the chance of accidents in the future.

2. Accident Review

2.1. Accident Description

At 03:00 a.m. of 6 March 2017, one operator exchanged fuel gas for fuel oil in the alkylation reactor
feed heating furnace (F101) and found poor combustion occurring at 04:28 a.m. The operator began to
adjust accordingly, but at 04:31 a.m., a flash explosion occurred in the heating furnace, injuring four
operators. It also damaged the surrounding property, such as the ethylbenzene process area (EB1)
heating furnace (F101) being destroyed, both of which had to be shut down and repaired [4,5]. The
accident scene conditions are shown in Figure 1; the location of the accident in Figure 2. The heating
furnace (F101) accident’s time sequence is shown in Figure 3 [6].
Appl. Sci. 2019, 9, 503 4 of 17

(a) (b)

Figure 1. Accident scene photos: (a) Heating furnace (F101) shell cracking situation-part 1 and (b)
Heating furnace (F101) shell cracking situation-part 2.

Figure 2. Location of the accident site.

Figure 3. Time sequence for the heating furnace (F101) accident. DCS: distributed control system;
EB: ethylbenzene; SM: styrene monomer.
Appl. Sci. 2019, 9, 503 5 of 17

2.2. Process Description

In general, the styrene monomer (SM) manufacturing method is based on the alkylation of benzene
and ethylene. Ethylbenzene is formed and then dehydrogenated to produce a product–styrene, with
byproducts: Toluene and hydrogen [7] (Figure 4). According to the process flow chart, before the raw
materials enter the reactor, the materials (benzene) must be heated to reach the operating temperature
by the heating furnace (F101). The inner structure of heating furnace (F101) is shown in Figure 5.

Figure 4. Process diagram and reaction equation for the alkylation reaction.

Figure 5. Schematic diagram of F101 furnace.

SM is an important intermediate material for the petrochemical industry that can be used to
produce polystyrene (PS) plastic, acrylonitrile-butadiene-styrene (ABS) copolymers, styrene butadiene
rubber (SBR), and unsaturated polyester (UPS). Its products are commonly used in electrical,
mechanical, electronic, automotive, packaging, and other industries, all closely related to peoples’
livelihoods [8].
The accident was a section of the EB process. The raw material benzene was heated by a heating
furnace (F101), and then mixed to reaction with another raw material ethylene in the alkylation reactor.
The design of the heating furnace (F101) fuel operating system included fuel oil and fuel gas functions.
However, the furnace flash explosion occurred when fuel oil and fuel gas were switched [9].
Appl. Sci. 2019, 9, 503 6 of 17

3. Accident Investigation Analysis

3.1. Analysis Procedures

This section identifies and describes the accident which caused with current risk management
methods and principles. According to the risk assessment process, including risk analysis, control, and
four decision making steps [10], the implementation steps are as follows (Figure 6).

Figure 6. Accident case risk assessment process [10]. ALOHA: areal locations of hazardous
atmospheres; FTA: fault tree analysis; PDCA: plan-do-check-act; ALARP: risks reduced to levels
that are as low as reasonably practicable.

The accident analysis follows a four-step sequence:

• Hazard identification: First, identify all process activity and find the accident hazard causes and
consequences, then confirm existing protective measures.
• Assessment: Use qualitative or quantitative analysis method to implement risk assessment.
• Risk control: According to the risk assessment results compare with the critical risk value, then
take control measures for risk reduction.
• Decision making: According to the risk reduction control measures and principles, consider the
cost-effectiveness to formulate improvement strategies, and regularly implement supervision and
assessment to determine the risk reduction performance [11–14].

3.2. Fault Tree Analysis (FTA)

3.2.1. What is FTA

FTA is one of the most widely used methods in system reliability, maintainability, and safety
analysis. FTA was developed by Watson, at Bell Telephone Laboratories, in 1962 to apply the analytical
method of studying the causes of undesired event failures. As a false logic diagram, the graphical
structure of the model could be used as qualitative analysis and quantitative analysis [15].
The main purpose of the FTA is to identify potential causes of system failures before the failures
actually occur. It also can be used to evaluate the probability of the top event using analytical or
statistical methods. These calculations involve system quantitative reliability and maintainability
information, such as failure probability, failure rate, and repair rate. After completing an FTA, the focus
is on efforts for improving system safety and reliability. An FTA begins by defining the “undesirable”
Appl. Sci. 2019, 9, 503 7 of 17

top event, such as fire disaster, explosion, leakage, and out of control. Then, FTA can analyze the
top event to basic events through intermediate events. Once the FTA has been constructed, it can be
qualified and quantified [16,17].

3.2.2. FTA Effects

Effects of FTA are illustrated as follows:

• Applies deduction methods to figure out the possible cause of the system failure.
• Provides a clear graphical method, various easy ways to understand, and to count system failure.
• Points out the weaker links of the operating system.
• Renders system tools to evaluate system improvement strategies.

3.2.3. Fault Tree Establishment

In this study, the disaster investigation team summarized the accident investigation report for
the company and analyzed the causes of the accident by FTA to find numerous points of the accident.
The establishment of FTA is shown in Figure 7. The top event was a heating furnace (F101) explosion.
The heating furnace (F101) exploded and the fuel continued to enter and accumulate in the furnace.
Then, the fuel and air formed a mixed flammable substance (here, concentration was already within
the boundaries of the explosion) that was exposed to the furnace surface caused by high temperature
flashover (with enough energy) [18,19].

Figure 7. FTA diagram for this study.

Factors of the heating furnace explosion by FTA are as shown in Table 4. We discussed the causes
of heating furnace explosion disasters from the principle of burning triangles, including three elements,
fuel (flammable substances), oxygen (air), and heat (temperature). Then, we applied the FTA method
to find the reasons for the explosion accident; the meanings of the symbols are as follows:
T Heating furnace explosion: The result of the furnace accident

X1 Heating source: One of the possible causes of the furnace accident

X2 Air: One of the possible causes of the furnace accident
X3 Excessive combustibles: One of the possible causes of the furnace accident
Appl. Sci. 2019, 9, 503 8 of 17

X4 Furnace temperature: One of the possible causes of heating source

X5 Heated surface: One of the possible causes of heating source
X6 Heating furnace flame: One of the possible causes of heating source
X7 Electrical spark: One of the possible causes of heating source
X8 Abnormal fuel flow into the furnace: Excessive combustibles (oil) in the furnace
X9 Abnormal gas flow into the furnace: Excessive combustibles (gas) in the furnace
X10 Human-error: Did not operate emergency shutoff valve.

Indirect possible reasons are as the following:

• After the heating furnace (F101) explosion, the operator did not shut down the F101 according to
the standard operating procedure (SOP); then the atomization of fuel and gas continued into the
furnace, resulting in flash over under the high temperature.
• In the fuel adjusting period, the furnace produced black smoke. At this point, on-site staff
discovered the flame extinguished and performed the adjustment, then determined the fuel
gasification and caused the ignition, resulting in other burners flame extinguishing instantly.
• The on-site staff response time during the abnormal situation was too short. The staff lacked
awareness of the safety of the heating furnace operation and was unwilling to shut down the
heating furnace (F101), nor did the staff perform the emergency stop procedure.
• Another reason might be that the staff was unfamiliar with the operating environment
and equipment.

Table 4. Factors table of the heating furnace explosion (F101) fault tree.

Factor Meaning Factor Meaning

T Heating furnace explosion X1 Heating source
X2 Air X3 Excessive combustibles
X4 Furnace temperature X5 Heated surface
X6 Heating furnace flame X7 Electrical spark
X8 Into the furnace fuel oil abnormally X9 Into the furnace fuel gas abnormally
Human errors—do not operate
X10
emergency shutoff valve X11 Mechanical failure—oil
X12 Mechanical failure—gas

3.3. Qualitative Analysis

The goal of fault tree analysis was to find the basic cause of the accident. That is, the basic events
were found by Boolean algebra simplified calculation to obtain the minimal path sets, minimal cut
sets (MCSs) of fault tree and the structure importance coefficient of each basic event. To reduce the
probability of an accident, we could apply the consequence of analysis to take more preventive and
control measures [20–23].
The fault tree totally included top event equations illustrated as below:

T = X1·X2·X3 (1)

T = (X4 + X5 + X6 + X7)··X2·(X8 + X9) (2)

T = X2·(X4 + X5 + X6 + X7)··(X10 + X11 + X10 + X12) (3)

T = X2·(X4 + X5 + X6 + X7)··(X10 + X11 + X12) (4)

T = X2·X4·X10 + X2·X4·X11 + X2·X4·X12 + X2·X5·X10 + X2·X5·X11
+ X2·X5·X12 + X2·X6·X10 + X2·X6·X11 + X2·X6 (5)
X12 + X2·X7·X10 + X2·X7·X11 + X2·X7·X12
Appl. Sci. 2019, 9, 503 9 of 17

where T is the top event; the intermediate events are X1, X3, X8, and X9; basic events are X2, X4, X5,
X6, X7, X10, X11, and X12.
MCS is a combination of basic events which would lead to the top events occurring. It cannot be
simplified again and still ensure the occurrence of the top events. If this MCS is in the fail state, the
entire system is out of order state.

K1 = [X2 X4 X10], K2 = [X2 X4 X11], K3 = [X2 X4 X12] (6)

K4 = [X2 X5 X10], K5 = [X2 X5 X11], K6 = [X2 X5 X12] (7)

K7 = [X2 X6 X10], K8 = [X2 X6 X11], K9 = [X2 X6 X12] (8)

K10 = [X2 X7 X10], K11 = [X2 X7 X11], K12 = [X2 X7 X12] (9)

From the above result of analysis, we could find the key factors of top event are air, any heating
source, and human error or mechanical failure.

3.3.1. Importance (I) of the Basic Events

We performed a relative importance sorting of the base events and the MCSs. The purpose was to
find the basic events and MCS that contributed more to the top event. According to risk management,
if the basic event and the MCS contributed much to the top event, we proposed to lessen the probability
of the risk. The risk reduction would be more prominent. Based on the cost-effective consideration, we
established the priority time ranking for completing the improvement suggestions.
In general, an MCS containing which has only one basic event was more likely to cause system
failure than an MCS which contained two basic events. An MCS containing two basic events was easier
to make a system malfunction than a minimum cut which contained three basic events. By analogy,
the fewer basic events the MCS has, the more likely the system would fail. From the MCS, we can
acquire the basic events importance sorting (Table 5). According to Table 5, the importance of the basic
event is calculated and sorted as follows:

I(X2) > I(X10) = I(X11) = I(X12) > I(X4) = I(X5) = I(X6) = I(X7) (10)

Table 5. Basic events importance sorting. MCS: minimal cut sets.

Basic Event Code The Times of Occurrences in the MCS Importance Sorting
X2 12 1
X4 3 3
X5 3 3
X6 3 3
X7 3 3
X10 4 2
X11 4 2
X12 4 2

Importance sorting of the basic event provides the theoretical basis for preventing and controlling
measures of the fire and explosion accident in heating furnace. From a system security perspective,
to prevent the top event and promote the reliability of equipment, we could use the fault tree to convert
into the corresponding success tree. The success tree could be, in turn, simplified by using Boolean
algebra and obtaining the minimal path sets (MPSs):
The success tree includes:
T’ = (X1·X2·X3)’ (11)

T’ = X1’ + X2’ + X3’ (12)

Appl. Sci. 2019, 9, 503 10 of 17

T’ = (X4+X5+X6+X7)’ + X2’ + (X10+X11+X12)’ (13)

T’ = X4’·X5’·X6’·X7’ + X2’ + X10’·X11’·X12’ (14)

P1 = [X2] (15)

P2 = [X4, X5, X6, X7] (16)

P3 = [X10, X11, X12] (17)

Hence, the MPSs represent the degree of system safety. According to the above simplification
results, there are three groups of MPSs, which means that three possible ways could help prevent the
fire and explosion accident in the heating furnace. From the above analysis, we could find the key
factors of system security (the furnace would not explode) were neither air in the furnace, nor any
heating source and human error or mechanical failure.
From FTA and the success tree analysis, to prevent the explosion of the heating furnace and
improve the reliability of the equipment, the following aspects could be considered to develop
improvement measures.

• When abnormal status happens, reduce the air concentration in heating furnace by inert gas.
Then, the previous installed emergency shutdown valve would cut off fuel gas from entering the
heating furnace.
• Strengthen the management of heat sources, such as hot work and static electricity.
• Enhance the improvement of personnel operation capability and equipment reliability.

3.3.2. ALOHA Simulation Analysis

The results of ALOHA simulation analysis were immediately completed and provided for
reference by the business authority. According to the simulation results, there are injuries within about
56 m and fatal risks within 10 m, as illustrated in Figure 8.

Influence region
(1) Red: 10 m (10.0 kW/(m2) = potentially lethal within 60 s)
(2) Orange: 10 m (5.0 kW/(m2) = 2nd degree burns within 60 sec)
(3) Yellow: 56 m (2.0 kW/(m2) = pain within 60 sec)

Figure 8. ALOHA simulation results-C4 Gas for this study.

With reference to the simulation results, many people and equipment within the scope of
influence could be considered to establish second management strategies and feasible emergency
response mechanisms.
Appl. Sci. 2019, 9, 503 11 of 17

4. Results of Analysis and Discussion with the Method of Improvement

4.1. Results of Analysis

From the result of the qualitative analysis, this accident’s causes could be figured out. Therefore,
we could take measures to enhance the management, human behavior modes, operating procedures,
equipment reliability, and other environmental conditions. The detailed methods of improvement are
shown in Section 5.
It demonstrates a momentous point: When an abnormal operating situation occurred in the
heating furnace (F101), after the F101 burner was extinguished, the fuel continued to enter the furnace
and accumulated a high concentration of combustible particles. It was exposed to the high temperature
surface of the furnace and caused flashing; the operator did not start the emergency interlock to close
the fuel valve according to the SOP. Meanwhile, other conditions, such as air concentration and high
temperature heat source, led to combustion and explosion conditions, so an accident occurred.

4.2. Improvement Strategy

Based on the accident investigation and qualitative analysis results, we recommended the
following improvement (engineering or management) measures to prevent similar accidents.

4.2.1. Improvement of Engineering Approach

A. Mechanical and equipment integrity (MI)

• Remake the furnace body and check the status of the attached equipment as well
as pipelines.
• Monitor equipment integrity: Record maintenance history and regularly replace
equipment components.
• Add safety improvement advice, replace fuel gas cock valve with additional limited type
switch, regulate all gas burners into the burner, as all could not be reset when the fuel valve
was not fully closed.

B. Inherent safety design (ISD)

To prevent furnace abnormal combustion (combustion gas short-circuit, incomplete combustion,

and afterburning), we recommend installing a furnace combustion detection system and fuel gas
interception interlock device.
According to PSM, 14 items were reviewed to make the ISD more secure. The device installation
is shown in Figure 9:

(a) Oxygen detector

• In the original design, there was only one oxygen detector. When the process was running,
the offset anomaly could not be found immediately. One oxygen detector was added to
improve the O2 detection reliability of the equipment.
• Installed the other oxygen detector which was at the same height as the existing oxygen
detector location; immediately notified the operator to maintain and calibrate it if the oxygen
concentration difference was more than 1% original setting value.
• Checked the detector was vetted every month, which was using zero and full-scale
calibration by standard gas.

(b) Improvement of chimney baffle

• Made sure chimney baffle opening degree was controlled by distributed control system
(DCS) in the original design. The signal was transmitted to the site controller to actuate the
Appl. Sci. 2019, 9, 503 12 of 17

chimney baffle. The DCS could not detect the actual opening degree value of the chimney
baffle in the field.
• Installed a positioner in the chimney baffle and transfer the actual opening degree value
of the chimney baffle on site to the DCS. Then, compared the value of DCS to master the
opening degree with the actual value on site, whether DCS was consistent with the actual
opening degree value of the site.
• Improved flame detector setting position.

Figure 9. Strengthening inherent safety design (ISD) measures for the alkylation tower.

The flame detector scanned up from the bottom of the furnace in the original design. It was easy
to accumulate ash in the detector lens surface and generate a false signal scanning. We recommended
that the location of the flame detector could be changed to the side of the furnace wall; it scanned
down to improve the case of false signals due to fouling. The flame detector setting position is shown
in Figure 10.

Figure 10. Design improvement of flame detector position.

4.2.2. Improvement of Management Approach

• Rechecked the entire plant furnaces SOP, strengthen operators’ education and training.
• Established maintenance, check history, and check procedures.
• Pre-startup safety review (PSSR) for heating furnace (F101).
The PSSR is a systematic and thorough check to a process unit; the purpose of a PSSR is to help
confirm that adequate safety measures are in place and are operational before a hazardous chemical
is stored or into the process. Ensured that installations meet the original design criteria or operating
procedure, then found any potential hazard and modify facilities to meet the management of change
requirements. In other words, ensured the “Ready for Start-up” status of process facility/units.
Appl. Sci. 2019, 9, 503 13 of 17

The PSSR should confirm the following:

• Rechecked the PSSR of the heating furnace (F101). The PSSR covered four major aspects:
Process, safety, and environmental protection (safety/fire protection/environmental protection),
equipment (machinery) maintenance, and instrument equipment maintenance.
• Self-inspection by process and maintenance operators for the four major inspection items.
• A process hazard analysis for the heating furnace (F101). The recommendations should
be implemented before startup, and modified facilities must meet the management of
change requirements.
• Completed the training of each employee involved in operating a process.

4.2.3. Cost-effectiveness Analysis: As Low as Reasonably Practicable Sorting Analysis

If there is no industrial activity, it is completely free from risk. Many companies around the world
ask that risks be reduced to levels that are as low as reasonably practicable, or “ALARP”.
By definition, the ALARP principle is the minimum level of residual risk that needs to be reduced
to reasonable limits. ALARP is used to reduce potentially harmful risks as much as possible. When
deciding whether the ALARP point has been reached, the following factors should be seriously
considered:

• Safety and health guide

• Codes of practice
• Industry practice
• Manufacturer specifications and government regulations
• Compliance with international standards and codes
• Comparison with dangerous events in similar industries
• Further reduction of the cost of risk in proportion to the benefits obtained.

A cost benefit analysis is used to determine if the risk has fallen to ALARP. This included weighing
cost, time and technical sacrifices, and reducing risk levels. According to the above-mentioned
investigation analysis and improvement measures (engineering method or management method),
applying the cost effect analysis (ALARP principle) and sorting analysis (high risk and low cost)
was proposed, and the priority was improved. The principles of ranking risk and cost are listed as
Tables 6–9 [24,25]. The results of analysis are shown in Figure 11.

Table 6. Severity levels [26,27].

Severity of Consequences Description

Safety: One or more fatalities or permanent disabling injuries (PDIs)
5 Environmental: Major impact, making the national news
Economic: Losses greater than $10 million USD
Safety: PDIs, or serious injury to three or more people
4 Environmental: Continuous large impact, making the local news
Economic: Losses between $1 million and $10 million USD
Safety: Serious injury to one or two people, or minor injury to three or more
3 Environmental: Moderate impact, must be reported to environmental agency
Economic: Losses between $100,000 and $1 million USD
Safety: Minor injury to no more than two people. First aid.
2 Environmental: Minor impact
Economic: Losses less than $100,000 USD
Safety: No adverse health effects
1 Environmental: No detectable impact
Economic: Negligible economic impact
Appl. Sci. 2019, 9, 503 14 of 17

Table 7. Likelihood of occurrence [26,27].

Likelihood of Occurrence Description

The event has happened several times at the plant. Likelihood is more than
5
once in 1 year.
The event has occurred at the plant and frequently in industry. Likelihood is
4
between once in 1 year and once in 10 years.
Incident has occurred at the plant, but is not common in industry. Likelihood is
3
between once in 10 years and once in 100 years.
2 Incident has occurred in industry. Likelihood is less than once in 100 years
1 The event has a remote chance of happening and is unheard of in industry.

Table 8. Classification of risk level [26,27].

Consequences (Severity)
Frequency of Occurrence
(Likelihood) Catastrophic Major Serious Minor Incidental
(5) (4) (3) (2) (1)
Frequency (5) 5 4 4 3 2
Occasional (4) 4 4 3 2 2
Seldom (3) 4 3 3 2 1
Remote (2) 3 3 2 1 1
Unlikely (1) 3 2 2 r 1
Risk level Risk Control Measures Note
Reduction risk measures need to be taken immediately, and
5-extreme
tasks should not be started until the risk is reduced.
Risk control measures must be taken within a certain period Unacceptable
4-very high
of time, and tasks cannot be started until the risk is reduced. risk
Based on cost or financial considerations, risk reduction
3-high
measures should be taken gradually.
There is no need to take risk reduction measures at the
2-medium moment, but it is necessary to ensure the effectiveness of
existing protection facilities. Acceptable risk
No risk reduction measures are required, but the
1-low
effectiveness of existing safeguards must be ensured.

Table 9. Classification of cost level [26,27].

Cost Level Description

E Spending more than $10 million USD for risk reduction measures.
D Spending range from $1 million to $10 million USD for risk reduction measures.
C Spending range from $100,000 to $1 million USD for risk reduction measures.
B Spending range from $10,000 to $100,000 USD for risk reduction measures.
A Spending less than $10,000 USD for risk reduction measures.

Where risk level is from 1 to 5 in the vertical axis. Numbers 1 to 2 are acceptable risk. Numbers
3 to 5 are unacceptable risk. Cost level is from A to E in the horizontal axis. The cost is gradually
increasing from A to E. Numbers 1 to 7 are the order of reduction risk measures in the matrix. This
is a risk decision matrix, representing the relationship between cost and risk, numbers 6 to 7, We
can consider negligible or defer implementation. However, numbers 1 to 3 are listed as priorities for
improvement and numbers 4 to 5 as an ALARP.
Appl. Sci. 2019, 9, 503 15 of 17

Note: 1. Priority level ≤ 4 listed as a major security aspect.

Figure 11. Results of the cost-effective ALARP sorting analysis [26,27].

4.2.4. ALARP Sorting

The improvement measures were ranked according to the principle of ALARP (high risk and low
cost), the results were as follows:
• Rebuilt the heating furnace (F101) body and evaluated the damage status of the surrounding
ancillary equipment and pipelines.
• Rechecked the furnace SOP.
• Improved worker education and training.
• Monitored equipment integrity: Recorded maintenance history and regularly replaced
equipment components.
• Established maintenance, checked history, and followed procedures.
• Erected a gas feed interruption device.
• Installed a furnace combustion condition detection device.

5. Conclusions and Recommendations

We learned from the fault tree qualitative analysis and field investigation results that the accident
was caused by high temperature sources, mixed flammable gas, and human factors. Among them,
the human factors were the most important reason, such as the classification of human factors (i.e.,
misoperation, misjudgment, unmoved action, lack of training, or lack of skill). The key points of
human factors were on how people interact with tasks, with machine/equipment/technologies, and
with the environment, in order to comprehend and assess these interactions. The goals of human
factors were to optimize human and system efficiency and effectiveness, operation safety, health,
comfort, and quality of living. If the interface among human body, machine and environment were
not properly deployed, accordingly, the problem of human error should be solved from the heating
furnace design.
On the other hand, when human factors are not avoided, we should improve from existing
management systems and equipment.
Through FTA, cost-effectiveness analysis, and on-site investigation of the accident, the following
recommendations were drawn:
Appl. Sci. 2019, 9, 503 16 of 17

• Units with different specializations must cooperate with each other to investigate accidents:
Previous accident investigation reports biased in fire expertise should include chemical process
and other background experts to investigate. To find the real root cause of an accident, we should
establish an investigation team by accident procedure, and then take control measures to avoid
the accident from happening again.
• To improve the operation safety of the equipment, including ISD, mechanical and MI.
• To develop procedures of MOC (management of change), when the heating furnace (F101) uses
different fuels.
• To avoid increasing the risk of exchanging fuels (fuel oil and fuel gas), we suggest using a
single fuel.
• To enhance the operators’ ability of response when the heating furnace (F101) abnormal
situation occurs.
• To improve operator training and equipment regular maintenance.
• Before the heating furnace was opened, PSSR for F101 was divided into four major aspects: Process,
(industrial safety/firefighting/environmental protection), equipment (mechanical) maintenance,
and instrument as well as electrical section (safety) maintenance.
• Case exchanging experience should be promoted to the grassroots.
• Installing a system for nitrogen into the furnace to reduce the concentration of combustible gases.

Author Contributions: Y.-C.W. wrote this manuscript; B.L. provided idea; C.-M.S. rendered suggestions for the
manuscripts. All authors supplied comments on this theme.
Funding: This research received no external funding.
Acknowledgments: Thanks to Wei-Cheng Lin for his valuable suggestions and help.
Conflicts of Interest: The authors declare no competing interests.

References
1. Ng, H.D.; Lee, J.H.S. Comments on explosion problems for hydrogen safety. J. Loss Prev. Process Ind. 2008, 21,
136–146. [CrossRef]
2. Workplace Health, Safety and Compensation Commission (WHSCC) of New Brunswick. Accident
Investigation Report on the Explosion and Fire at the Irving Oil Refinery, New Brunswick; The Commission:
Fredericton, NB, Canada, 1999.
3. Ranganathan, S.; Gillis, S. Oven Furnace and Dryer Explosion Incidents; Fire Protection Research Foundation:
Quincy, MA, USA, 2016.
4. Zhou, Y.; Zhao, X.; Zhao, J.; Chen, D. Research on Fire and Explosion Accidents of Oil Depots. Chem. Eng.
Trans. 2016, 51, 163–168. [CrossRef]
5. Large-Scale Petrochemical Supervision Meeting; Bureau of Industry, Ministry of Economic Affairs: Taipei,
Taiwan, 2017.
6. Babrauskas, V. Explosions of ammonium nitrate fertilizer in storage or transportation are preventable
accidents. J. Hazard. Mater. 2016, 304, 134–149. [CrossRef] [PubMed]
7. Hu, K.H.; Kao, C.S.; Duh, Y.S. Studies on the runaway reaction of ABS polymerization process. J. Hazard.
Mater. 2008, 159, 25–34. [CrossRef] [PubMed]
8. Kao, C.S.; Duh, Y.S. Accident investigation of an ABS plant. J. Loss Prev. Process Ind. 2002, 15, 223–232.
[CrossRef]
9. ABS Powder Storage Tank (H701F) Flash Accident Report; Mai Lao PABS Plant of Taiwan Chemical Company:
Yunlin, Taiwan, 2017.
10. Kelly, B.D. Investigation of a hydrogen heater explosion. J. Loss Prev. Process Ind. 1997, 11, 257–259. [CrossRef]
11. Siwek, R. Determination of technical safety indices and factors influencing hazard evaluation of dusts. J. Loss
Prev. Process Ind. 1996, 9, 21–31. [CrossRef]
Appl. Sci. 2019, 9, 503 17 of 17

12. Nedkova, S.; Lybchev, L.; Lybcheva, M.; Kolev, N. Human factor analysis in the risk management in technological
processes and in A glass production plant. Sci. Technol. 2013, 3, 11–15. Available online: http://www.sustz.com/
journal/VolumeIII/Number4/Papers/SabinaNedkova.pdf (accessed on 10 August 2018).
13. Dugué, J. Fired equipment safety in the oil & gas industry. A review of changes in practices over the last
50 years. Energy Procedia 2017, 120, 2–19. [CrossRef]
14. Mirza, N.R.; Degenkolbe, S.; Witt, W. Analysis of hydrogen incidents to support risk assessment. Int. J.
Hydrogen Energy 2011, 36, 12068–12077. [CrossRef]
15. Xue, Z.M. Research on FTA of fire and explosion in the crude oil gatheringtransport combination station.
Procedia Eng. 2011, 11, 575–582. [CrossRef]
16. Abuswer, M.; Amyotte, P.; Khan, F. A quantitative risk management framework for dust and hybrid mixture
explosions. J. Loss Prev. Proc. Ind. 2013, 26, 283–289. [CrossRef]
17. Halme, J.; Aikala, A. Fault tree analysis for maintenance needs. In Proceedings of the 25th International
Congress on Condition Monitoring and Diagnostic Engineering, Huddersfield, UK, 18–20 June 2012;
pp. 1031–1041. [CrossRef]
18. Yasushi, N.; Rolf, L.; Philip, A. Fault tree analysis of system anomaly leading to red oil explosion in lutonium
evaporator. J. Nucl. Sci. Technol. 1994, 31, 850–860. [CrossRef]
19. Ruijters, E.; Stoelinga, M. Fault Tree Analysis: A survey of the state-of-the-art in modeling, analysis and
tools. Comput. Sci. Rev. 2015, 15–16, 29–62. [CrossRef]
20. Moreno, V.C.; Papasidero, S.; Scarponi, G.E.; Guglielmi, D.; Cozzani, V. Analysis of accidents in biogas
production and upgrading. Renew. Energy 2016, 96, 1127–1134. [CrossRef]
21. Fuentes-Bargues, J.L.; González-Cruz, M.C.; González-Gaya, C.; Baixauli-Pérez, M.P. Risk Analysis of a
Fuel Storage Terminal Using HAZOP and FTA. Int. J. Environ. Res. Public Health 2017, 14, 705. [CrossRef]
[PubMed]
22. Siu, N.; Herring, J.S.; Lee, C.; Reece, W.; Byers, J. Qualitative Risk Assessment for an LNG Refueling Station and
Review of Relevant Safety Issues; Idaho National Engineering and Environmental Laboratory: Idaho Falls, ID,
USA, 1999.
23. Mohammadfam, I.; Zarei, E. Safety risk modeling and major accidents analysis of hydrogen and natural gas
releases: A comprehensive risk analysis framework. Int. J. Hydrogen Energy 2015, 40, 13653–13663. [CrossRef]
24. Figueroa Jiménez, S.P.; Lombana Carmona, S.C.; Ruiz De La Cruz, I.L. Application of HAZOP, LOPA and SIL
to the Alkylation Unit Catalyzed with Hydrofluoric Acid at Ecopetrol Refinery in Cartagena–Colombia; Mary Kay
O’Connor Process Safety Center, Texas A&M University: College Station, TX, USA, 2015.
25. Moonis, M.; Wilday, A.J.; Wardman, M.J. Semi-quantitative risk assessment of commercial scale supply
chain of hydrogen fuel and implications for industry and society. Process Saf. Environ. Prot. 2010, 88, 97–108.
[CrossRef]
26. Linkov, I.; Bates, M.; Loney, D.; Sparrevik, M.; Bridges, T. Risk Management Practices-Cross-agency
Comparisons and Tolerable Risk. In Climate: Global Change and Local Adaptation; Springer: Dordrecht,
The Netherlands, 2011.
27. Kletz, T.A. Looking beyond ALARP: Overcoming its limitations. Process Saf. Environ. Prot. 2005, 83, 81–84.
[CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).