Risk Assessment and Reduction Strategies for Acrylonitrile Production Plants

Journal of Loss Prevention in the Process Industries 63 2020 104015 Available online 22 November 2019 09504230 2019 Elsevier Ltd All rights reserved Risk assessment and risk reduction of an acrylonitrile production plant Kazuhiko Sano a Yusuke Koshiba b Hideo Ohtani c a Department of Risk Management Graduate school of Environment and Information Sciences Yokohama National University 797 Tokiwadai Hodogayaku Yokohama 2408501 Japan b Department of Materials Science and Chemical Engineering Faculty of Engineering Yokohama National University 795 Tokiwadai Hodogayaku Yokohama 240 8501 Japan c Department of Safety Management Faculty of Environment and Information Sciences Yokohama National University 797 Tokiwadai Hodogayaku Yokohama 240 8501 Japan A R T I C L E I N F O Keywords Acrylonitrile Fire explosion index Process safety metrics Fault tree analysis Risk identification Safety investment A B S T R A C T Reducing accident occurrence in petrochemical plants is crucial thus appropriately allocating management resources to safety investment is a vital issue for corporate management as international competition intensifies Understanding the priority of safety investment in a rational way helps achieve this objective In this study we targeted an acrylonitrile plant First Dow Chemicals Fire and Explosion Index FEI identified the reaction process as having the greatest physical risk We evaluated the severity of accidents in the reaction process using the Process Safety Metrics advocated by the Center for Chemical Process Safety CCPS however this index does not express damages a company actually experience To solve this problem we pro posed a new metric that adds indirect cost to CCPS metrics We adopted fault tree analysis FTA as a risk assessment method In identifying top events and basic events we attempted to improve the completeness of risk identification by considering accidents from the past actual plant operation and equipment characteristics natural disasters and cyberattacks and terrorist attacks Consequently we identified the top events with high priority in handling because of serious accidents as fireexplosion outside the reactor fireexplosion inside the reactor and reactor destruction The new CCPS evaluation index proposed in this study found that fire and explosion outside the reactor has the highest severity We considered the creation of the fault tree FT diagram of the top event estimating the occurrence probability and identifying the risk reduction part and capital in vestment aimed at risk reduction As an economically feasible selection method for risk reduction investment using the difference in loss amounts before and after safety investments indicated investment priority 1 Introduction Petrochemical plants are complex technical systems with tightly coupled operation equipment and human factors and are places where accumulated technological knowhow is applied Unfortunately serious accidents such as fires and explosions have occurred at petrochemical plants Such accidents can be caused by various factors such as design and manufacturing defects construction and inspection management failures deficient operations standards and information transmission insufficient technical prediction aging facilities or reduced personnel and maintenance When accidents or unexpected failures occur it is sometimes said that there was some negligence Therefore engineers working in petrochemical plants must identify and reduce risks Several methods risk prediction and analysis methods have been developed and have contributed to a reduction in accidentrelated fatalities and losses For example Baybutt 2018 reported that although risk management using a risk matrix is useful there are pitfalls and provided guidelines for constructing risk matrices that address these pitfalls Cox 2008 re ported that flaws have been identified in the underlying theoretical framework of risk matrices Additionally Curcurù et al 2013 devel oped an imprecise fault tree analysis FTA method to characterize systems affected by a lack of reliability data They estimated rate of occurrence of the top event not probability of occurrence Ferdous et al 2009 reported on methodology for a fuzzybased computeraided FTA tool Their goal was to handle cases in which the overall result may be questionable because of imprecise basic failure data There has also been a lot of study on the Hazard and Operability Study HAZOP which is Corresponding author79Tokiwadai Hodogayaku Yokohama 2408501 Japan Email address sanokhomasahikaseicojp K Sano Contents lists available at ScienceDirect Journal of Loss Prevention in the Process Industries journal homepage httpwwwelseviercomlocatejlp httpsdoiorg101016jjlp2019104015 Received 16 May 2019 Received in revised form 23 October 2019 Accepted 19 November 2019 Journal of Loss Prevention in the Process Industries 63 2020 104015 2 the standard for process risk analysis in the chemical process industry To assist HAZOP Kang and Guo 2016 introduced sensitivity evalua tion into HAZOP deviation analysis to measure the effect degree of each cause on the corresponding deviation Guo and Kang 2015 proposed an extended HAZOP analysis approach using a dynamic fault tree FT to identify potential hazards in chemical plants Although intended as a study to supplement the insufficiency of HAZOP MacGregor 2017 proposed a unique method merging HAZOP with failure mode and ef fects analysis FMEA As studies on combinations of risk assessment in addition to Guo and Kang 2015 studies have combined quantitative risk assessment FTA and severity analysis eg Abuswer et al 2013 Moreover AlSharrah et al 2007 proposed a risk index for the chem ical processing industry consisting of four terms accident frequency hazardous effects of chemicals inventory of chemicals released and plant size Despite progress from the above studies serious accidents still occur at petrochemical plants some of which should not have happened based on the results of these studies Furthermore analysis of annual numbers of accidents in Japan shows that accident frequency is not decreasing at a sufficient rate These facts appear to suggest the need to introduce new measures in addition to conventional measures to further reduce acci dents and failures Many companies in Japan have identified issues that are considered factors leading to accidents such as deteriorating industrial safety technology capabilities aging of facilities reduced capital investment and maintenance spending due to cost reduction requests and decreased numbers of personnel with industrial safety skills Any measure to address these issues will require additional costs in personnel time or money Accident reductionrelated investments should be made for economically effective projects and risk reduction effects should be evaluated based on comprehensive and ongoing risk assessment as well as from an appropriate management resource allocation perspective In this study we focus on acrylonitrile AN plants and suggest a method for selecting economically feasible risk reduction method in vestments We chose AN plants because the AN production process in volves a gas phase partial oxidation reaction which poses high risks Additionally AN has been suspected to be a carcinogen and have reproductive as well as general toxicity therefore risk management involves physical environmental and health risks Few studies have been conducted on methods for estimating risks associated with AN plant industrial accidents meaning that the findings of this study will be crucial for preventing accidents at AN plants In this study to reduce fireexplosion risks in AN plants we aimed 1 to identify highrisk parts of the AN production process 2 assess risks in those parts 3 establish a new technical system and protective measures to prevent fireexplosion accidents and 4 propose a new method for selecting safetyrelated investment priorities based on a costtobenefit analysis 2 An production process AN is produced from propylene and ammonia and the main by products are acetonitrile and hydrogen cyanide see reaction formulae Van der Bann 1980 developed an ammoxidation reaction to convert propylene to AN and showed that it is a firstorder reaction propylene concentration reaction The AN production process flow is shown in Fig 1 CH2 ¼ CHCH3 þ NH3 þ 32 O2 CH2 ¼ CHCN þ 3H2O Acrylonitrile CH2 ¼ CHCH3 þ 32 NH3þ 32 O2 32 CH3CN þ 3H2O Acetonitrile CH2 ¼ CHCH3 þ 3NH3 þ 3 O2 3 HCN þ 6H2O Hydrogen cyanide A catalyst with both suitable chemical activity and physical prop erties eg abrasion resistance particle size distribution specific sur face area apparent density catalyst shape etc is employed in a fluidized bed reactor Several papers have reported on catalyst compo sitions and production methods from with respect to improving AN production yield Centi and Perathoner 1998 Grasselli 1986 1999 ThanhBinh et al 2016 Additionally many patents related to catalyst reactions have also been found From a viewpoint of improving equip ment Dutta and Gualy 1999 focused on remodeling the reactor and reported on attaching a dispersion plate and improving flow charac teristics to prevent gas drift in the catalyst layer Maccallion 1996 investigated cyclone types to improve catalyst recovery rates Reaction conditions vary depending on the catalyst used and reactor structure and are summarized in Table 1 Ammoxidation is an exothermic reaction and cooling coils in the reactor recover the reaction heat as steam which is effectively used in the plant with the excess exported outside From the perspectives of controlling the reaction and improving operational efficiency research exists that focuses on enabling timely reaction condition adjustment through online analysis of the reaction product gas without requiring manual analysis Goodrich 1972 Hawkins and Wood 1999 reported that fluidized bed reactors can be controlled using multivariate statis tical process analysis by predicting changes in output due to input Fig 1 An process flow K Sano et al K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 Table 1 revalidation relative to HAZOP Zhao et al 2009 developed a new Condition for acrylonitrile AN synthesis learning HAZOP expert system based on integrating casebased Feed gas molar ratio Unit C3H6 NH3 Air 1 1012 912 reasoning and ontology to improve HAZOP expert system learning Resctiontempeate GéSOOCNSNSNOO They used an example of the AN production process with an startup Reaction pressure MPaG 005015 furnace ignition failure Shah et al 2005 presented a new method Linear velocity ms 0410 using an automated software tool in which a hierarchical approach re Contact time s 210 veals the degree of nonideality of chemical processes with regard to safety health and environment aspects at different process layers They variable variations provided an example of acrylic fiber production in which they showed Product gases from the reactor are washed and cooled using an the safety assessment of an tank As studies on reaction risks Cozzani aqueous sulfuric acid solution in the quench column After this sub et al 1 998 studied unwanted reactions caused by accidental contact of stances with high boiling points and unreacted ammonia in the product reactive substances and Miyake ct al 2005 Proposed an ev aluation gas are then separated and removed Ammonium sulfate generated from flow of hazards of mixing organic peroxides with other chemicals such unreacted ammonia and sulfuric acid is then recovered as either as AN As described above there have been many risk assessment ammonium sulfate or sulfuric acid studies but there have not been enough studies on the costeffectiveness Products such as AN hydrogen cyanide and acetonitrile in outlet gas of risk reduction for AN plants thus there is room for improvement at the quench column are absorbed using water in the absorber to form an aqueous solution which is then withdrawn from the bottom of the 3 Consideration for physical risk assessment absorber In contrast noncondensable gases which are water insoluble are discharged from the top of the column and incinerated 31 Risk assessment targets in a waste gas incinerator Because the boiling points of AN 773 C and acetonitrile 813 C After quantitatively deriving each AN plant units risk level It Is are not significantly different AN is difficult to separate completely crucial to grasp relative risk levels between all units and those with high using a common distillation operation Therefore an extractive distil tisk levels In this study we conducted a safety assessment of an plant lation process using water as an extractant is conducted in the recovery using Dows Fire Explosion Index Hazard Classification Guide AIChE column to increase the relative volatility of both substances Separated 1994 hereinafter referred to as FREI acetonitrile is transported to an acetonitrile recovery plant and purified This method classifies the plant into different units based on char and unrecovered acetonitrile is incinerated in the waste gas incinerator acteristics and or physical layout Characteristic classification isa Water generated from the reaction is recycled as an absorbent for the concept similar to unit operation used in chemical engineering Processes absorber and a solvent for the recovery column and surplus water is in which ordinary plants are divided around the main equipment In this discharged from the bottom of the recovery column as wastewater that is study the goal was to find hazards of the process Therefore an FEL treated and detoxified in an activated sludge facility before being unit was not selected as a section but rather each main equipment discharged component constituted a section Because evaluation focused on pro Crude AN containing hydrogen cyanide from which acetonitrile was cesses tanks that did not involve chemical engineering unit operations separated in the recovery column is then sent to the heads column were not evaluated We employed the following assumptions where hydrogen cyanide low boiling point substances and water are an separated Separated hydrogen cyanide is then sent to a hydrogen cya a AN plant operation consists of reaction recovery and purification nide recovery plant or is incinerated in the waste gas incinerator if not POCESSEs Because these definitions are somewhat coarse in this recovered Lastly high boiling point substances are separated in the case each main piece of equipment composing an plant was taken as product column to obtain AN an FEI unit Those units are the reactor the quench column the In addition although not shown in Fig 1 there is a wastewater absorber the recovery column the heads column and the product treatment process a waste gas treatment process and a tank process In column the AN production process the reactor quench column absorber re b Plant size ie production capacity should be cost competitive covery column heads column and product column are unique compo therefore we targeted large plants worldwide 200 kty nents and are the main equipment used c The plant follows the world standard and has two reactors two Studies related to AN recovery and purification processes have quench columns one absorber one recovery column one heads focused on rationalization and improvement of processes from the column and one product column perspective of reducing manufacturing cost For example Pujado 1977 d Material coefficients Ge material factors were determined by studied improvement of AN recoverypurification processes and Gu and weighted averaging from the mixture composition The mass of Chen 1991 studied the use of xylene instead of water during extraction materials in each piece of equipment was determined by calculating to separate AN and acetonitrile Additionally AN plants also generate retention amounts and equipment size large quantities of wastewater and waste gas therefore there have also been studies from a waste reduction perspective Hopper et al 1993 The method for calculating the FEI value is described using a and Sanghavi 1998 studied waste reduction and Shelly 1995 studied reactor as an example The FEI value is calculated using two compo catalytic oxidation treatment of waste gases nents Material Factor MF and Process Unit Hazards Factor F3 which Many previous studies have reported on the AN process environ consists of General Process Hazards F1 and Special Process Hazards mental and health risks USEPA 1983 WHO 1983 TERA 1997 F2 The FEI value is determined by Eqs 31 and 32 AIChE USEPA 1998 IARC 1999 The following studies reported on fire and 1994 explosion in AN production processes AlSharrah et al 2007 proposed FEI MF x F3 31 a safety risk index to estimate the maximum number of people affected if an accident occurred that caused the release of a plants entire inventory F3F1 x F2 32 of a chemical For the risk analysis First 2010 presented a simplified MF is th f ial f hemical ial chemical process risk analysis method that is effective at providing a oo 1s the rate of potential energy release Tom emuca smateria Ss semiquantitative measure of consequence and mentioned that this indicated bya value of 140 and can be determined from chemical method minimizes overall time required for scenario development and materials reactivity and flammability The MF for main substances is given in the table of guidelines AIChE 1994 The MF value in the 3 K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 reactor was calculated to be 36 Table 3 Fl determines a potential incidents magnitude and covers Fire Explosion Index FEI evaluation result exothermic chemical reactions endothermic processes material Unit FEI handling and transfer enclosed or indoor process units access and ReactortCsSstCSsSstS 8 drainage and spill control The score standard is described in the AIChE Quench Column 134 1994 guidelines The reactor corresponds to an exothermic reaction and Absorber 132 F1 was calculated to be 150 Recovery Column 150 F2 determines a potential incidents probability and consists of 12 Heads Column 136 items toxic material subatmospheric pressure operation in or near Product Column flammable range dust explosion relief pressure low temperature F EI Fire and Explosion Index quantity of flammableunstable material corrosion and erosion leakagejoints and packing use of fire equipment hot oil heat exchange necessarily sufficient with respect to countermeasures after the earth system and rotating equipment From the score given in the AIChE quake Yu et al 2017 Research on Natech is progressing with case 1994 guidelines F2 in the reactor was calculated to be 378 studies the advancement of risk assessment methods and the organi Therefore from Eq 32 F3 150 x 378 567 and from Eq zation of matters necessary for management Although understanding of 31 FEI 36 x 567 204 Natech risk has deepened there are several features compared to The degree of hazard based on F EI values is classified as shown in fireexplosion risk etc that need further study and proposals for better Table 2 AIChE 1994 and the resulting values shown in Table 3 clearly management methods are expected in the future indicate that all degrees of hazard of AN plant units were classified as To concretely implement physical risk reduction measures it is Heavy and Severe The reactor has a maximum FEI value of 204 necessary to break down scenarios leading to accidents or disasters We whereas the product column had the next highest value at 185 Here selected FTA as the analysis method to obtain the following results by after we focus on the reactor which has high risks associated with fires risk assessment and explosions as this studys subject Clarify serious event scenarios Y Extract and clarify fatal events in the system and lead to system 32 Risk assessment method improvement Y Evaluate the importance of each basic event To enhance risk identification completeness in this study we con Y Compare the occurrence probability of a serious event top event ducted risk identification using the following considerations from those of basic events a Partial oxidation reaction process characteristics obtained from To enhance risk identification completeness in the FTA we incor actual plant operation porated all reasonably foreseeable events related to matters a through b AN fluidized bed reactor characteristics obtained from actual plant d in Section 32 We will thus conduct a costbenefit analysis based on operation FTA results c Natural disasters d Cyberattacks and terrorism 33 FTA events By using HAZOP andor FMEA risk identification completeness for We examined events related to physical risks in the reaction section a and b can be improved based on the perspectives described in Section 32 As a result top events Natural disasters have rarely been considered in past risk assess were derived from events caused by characteristics of partial oxidation ments however natural disasters can trigger industrial accidents at reaction process obtained from actual plant operation Top events with chemical plants In recent years industrial accidents caused by natural a high priority in handling due to serious accidents were identified as disasters have attracted attention and are referred to hereinafter as fireexplosion outside the reactor fireexplosion inside the reactor Natech For example two earthquakes in northwest Turkey caused and reactor destruction Fig 2 It can be said that these are accidents heavy damage to many industrial facilities Korkmaz et al 2011 and that would be troublesome if they were to happen hurricanes and earthquakes have caused hazardous material releases We obtained intermediate events constructing accident scenarios from chemical plants Cruz and Krausmann 2009 Krausmann et al and basic events from characteristics of equipment including fluidized 2010 One such example is Hurricane Harvey which hit the state of bed reactors In other words these events were single failures of Texas in the United States of America in 2017 The hurricane seriously equipment and parts failures caused by excessive stress on equipment damaged public infrastructure and plant equipment forcing many of the and parts and failures due to operationhandling errors chemical plants in the area into a force majeure declaration and caused As for the events that could be caused by natural disasters and cyber them to shut down for several months In Japan damages from typhoon attacks or terrorism loss of utility loss of operation control equipment winds and flooding and due to earthquakes or tsunami cannot be destruction etc were considered These were events that affected all ignored Regarding Japanese chemical plants it has been revealed that events and were defined as common failures correspondence behaviors at the time of the 2011 Great East Japan The above descriptions are summarized in Fig 3 Earthquake also known as the Tohoku Earthquake were not Common failures are often related to top events as well as many intermediatebasic events so putting common failures in an FTA makes Table 2 the FT diagram complicated Consequently common failures were Fire Explosion Index FET risk rank separated from FTA because targets for reducing risks that affect AN FEI Range Degree of Hazards reaction process physical hazards targeted in this study are ambiguous 160 Light 6196 Moderate 34 Accident case in the reaction process 9727 Intermediate 128158 Heavy To verify the identified risk we collected 102 cases of AN plant re 159 Severe action process accidents from AN manufacturing companies around the F El Fire and Explosion Index world and classified and consolidated AN fluidized bed reactor accident 4 Journal of Loss Prevention in the Process Industries 63 2020 104015 5 cases worldwide Table 4 shows the aggregated results obtained by this work Table 4 refers to accident cases but we found that abnormalities can be classified as either accidentsdisasters or incidents Accidents and disasters include events such as explosions fires and pressure abnor malities Notification of accidents or disasters is given both inside and outside the company with public reports being made depending on the scale of the accidentdisaster Incidents include events such as reaction temperature abnormalities hot spots leakage and cyclone abnormal ities These events were determined by postinspection and treated using unsteady operation though in some cases urgent action was taken and incidents were calmed down before an accidentdisaster occurred We defined 30 cases14 explosions one fire and 15 pressure abnormalitiesas accidentsdisasters accounting for 29 of all cases and the remaining 71 were defined as incidents Among these 51 cases of high temperature low temperature hotspots leakage and cyclone abnormalities accounted for 50 of the total The remaining 21 in cidents 21 were due to blackouts machine issues machine or instrumentation failures and natural disasters typhoons Although these incidents did not lead to accidents or disasters they caused un intended emergency reactor shutdowns Table 4 shows that accidentsdisasters have occurred not only during unsteady operation such as plant startup and shutdown but also during steady operation Thus it is shown that measures must be taken regardless of operating conditions From the cases described in Table 4 it is possible to understand the types of physical risk that occur during the AN reaction process Table 5 Because incidents can lead to accidentsdisasters it is possible to connect accidentsdisasters with related incidents Likewise relevant incidents are also connected with other incidents For example in the pressure abnormality case measures were taken before the reactor was destroyed but it was not distinguished from a breakdown Therefore the pressure abnormality case was not classified as an incident but rather was treated as an accidentdisaster Fig 4 was created based on the above idea Explosions and fires were mainly classified in the block with chemical reactions wheras destruc tion was mainly classified in the block with physical phenomenon As shown in Fig 4 accidents and disasters are consistent with top events identified using the method described in Sections 32 and 33 suggesting the validity of those methods Furthermore events identified using the method described in Section 32 and 33 covers accident disaster cases In other words it was not necessary to supplement the identified events with accidentdisaster cases If it is possible to detect the occurrence of incidents that can lead to explosion fire or destruction thereby suppressing their expansion it can be said that accidentdisaster risks can be reduced Thus it is important to systematically extract incidents leading to accidents or disasters from the perspective of preventing them and taking countermeasures 35 Evaluation of severity of accidents The Center for Chemical Process Safety CCPS proposed an index known as the CCPS Process Safety Metrics for determining the degree of event severity CCPS Process Safety Leading and Lagging Metrics 2011 This index is a common metric for conducting selfevaluations for petrochemical and fine chemical industries The CCPS index aims at preventing process accidents and disasters and quantitatively evaluates four levels of four items human health fireexplosion potential impact of leakage and environmental impact The final score is the total of points for each of these items and shows accident severity The Japan Petrochemical Industry Association added a minor accident level to the Fig 2 Top events Events related to physical risk in an acrylonitrile plant reaction section The top events identified are fireexplosion outside the reactor fire explosion inside the reactor and reactor destruction Fig 3 Identified events Top events are derived from events caused by char acteristics of partial oxidation reaction process obtained from actual plant operation We obtained intermediate events for constructing accident scenarios and information to be considered as basic events from characteristics of equipment including fluidized bed reactors Events that could be caused by natural disasters cyberattacks or terrorism are positioned as causes of com mon failure and thus would be related to all top events K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 6 CCPS index and set each item to five levels as shown in Table 6 Using the degree of severity from this index as a single indicator allows us to directly compare the sizes of each accident Fig 5 shows the relationship between total loss amounts and corre sponding CCPS scores of accident cases for AN fluidizedbed reactors shown in Table 4 As seen in Fig 5 the coefficient of determination is low R2 ¼ 043 indicating that the CCPS index cannot estimate damage amounts incurred by a company and society This is probably because the CCPS evaluation index only covers direct costs To meet the needs of company with accountability to stakeholders it is desirable to employ an evaluation index that can express substantial damage the company experiences Direct costs are expenses such as repair exchange cleaning Table 4 Examples of accidents in acrylonitrile fluidized bed reactors Type of event Abnormal Form Number Contents Cause Accident Disaster Explosion 14 SD Explosion occurred in the reactor quench column or absorber when C3H6 and NH3 were stopped at the same time Under air shutoff delay andor air purge continuation the explosive mixture gas was formed Accident Disaster Fire 1 SU During SU after emergency stop gas blew out of the bottom of the reactor and ignited When the emptying stopped Mo pieces dropped and accumulated inhibiting catalyst fluidization and abnormal high temperature was formed due to insufficient heat removal Incident Reaction Temperature High 3 SU The temperature of the upper part of the reactor sharply increased at the stage of temperature rise or temperature decrease due to NH3 combustion Dense phase temperature dropped due to excessive use of cooling coils and a large amount of NH3 burned in a dilute phase Incident Reaction Temperature Low 2 Nor Ope Mist was entrained from the raw material gas vaporizer and the reactor temperature sharply dropped This led to increased unreacted and induced temperature runaway in the waste gas incinerator The level meter of the raw material gas vaporizer failed Accident Disaster Pressure High 15 Nor Ope The flow of reaction gas was inhibited and reactor pressure rose Clogging of reactor effluent cooler clogging of quench column packed bed closing of quench column outlet CV Incident Hot spot 3 SU NH3 abnormally burned and damaged on the upper tube sheet of the reactor effluent cooler The catalyst accumulated on the tube sheet and the temperature rose due to NH3 combustion 11 Nor Ope Sparger nozzles flanges and support lugs were melted Because internal cleaning was not done when the reactor was stopped Mo pieces accumulated and deteriorated catalyst fluidization Inadequate distance and structure between sparger and air distributor Incident Leakage 9 Nor Ope The cooling coil was broken causing catalyst cracking and catalyst scattering Cracking occurred due to stress concentration in the welded parts of the cooling coils and BFW blew out Reactor shell thickness was reduced due to erosion 13 Nor Ope Sparger was broken unreacted raw material increased and waste gas incinerator temperature increased Breakage of sparger due to material degradation Flange fastening problem Incident Cyclonesabnormality 10 any time Cyclone was blocked by catalyst and a large amount of catalyst scattered Dipleg obstruction Trickle valves malfunction Nor Ope A hole opened in the cyclone and a large amount of catalyst scattered Dust balls were broken by erosion Incident External Factor 21 Nor Ope Blackout machineinstrumentationelectrical equipment failure emergency stop due to natural disaster SU Start up SD Shut down NorOpe Normal operation Sparger Disperser nozzles that discharge raw material gas Air distributor Plate with nozzles to discharge air Cyclones Apparatus for recovering catalyst from gas flow Cooling coils Cooling device for removing heat of reaction BFW Boiler feed water CV Control valve Table 5 Types of physical risks in acrylonitrile fluidized bed reactors Fire Gas leakage leads to jet fire Explosion Gas explosion Explosion due to pressure rise abrupt pressure change expansion of liquid Leakage Leakage of raw material gas from gas supply system Leakage of reaction gas from reactor leads to fireexplosion Destruction Destruction due to pressure rise due to explosion inside the reactor Due to the reactor internal pressure being increased due to reactor outlet closing and gas supply continuation Fig 4 System diagram of the accidents Incidents can be accidents or disasters or can cause them and it is possible to connect accidents and disasters with related incidents Likewise relevant incidents are also frequently connected with other incidents Explosion and fire are mainly classified in a block with chemical reactions and destruction is mainly classified in a block with phys ical phenomena K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 7 disposal environmental restoration and emergency response Gener ally direct costs do not include lost opportunity business interruption loss of raw materials and products lost profits due to equipment shut down costs of procuring and operating temporary equipment or costs of procuring alternative products to meet customer requests These ex penses are thus known as indirect costs However when expressing damage incurred by a company the weight of indirect costs in the total loss amount cannot be ignored therefore it is crucial to also account for indirect costs Several items in this study are indirect costs however we solely adopted opportunity loss due to the size of its effect on indirect cost and to simplify evaluation That is to measure the degree of severity we proposed a revised CCPS hereinafter referred to as RCCPS evaluation index that adds opportunity loss to the conventional CCPS evaluation index Table 7 We set opportunity loss severity to the same level as existing evaluation items For this purpose we first set length of stoppage to roughly the same level by referring to the explanation given for each level described under CommunityEnvironment Impact in the conventional CCPS evaluation index Next we confirmed that there was no significant difference in the amount of opportunity loss calculated from length of stoppage loss production volume and marginal profit compared with existing items Specifically we set the AN plant pro duction amount to the world standard size of 200 kty 600 tD and set the marginal profit to 30000 t We set length of stoppage levels as follows 2 months or more was set to level 1 1 week2 months was set to level 2 1 day1 week was set to level 3 and 3 hours1 day was set to level 4 Additionally we corrected historical loss amount data consid ering current price levels and used an exchange rate of 100 JPY ¼ 1 US Table 4 was rearranged following the RCCPS evaluation index and thus we drew Fig 6 which depicts the relationship between total loss amounts and corresponding RCCPS points and indicates a good fit of the model R2 ¼ 071 The RCCPS evaluation index can use degree of severity as a single indicator and thus compare the scale of accidents For past accidents scores are 45 or less In other words no accident cases exceeded a score of 45 4 Risk assessment by FTA and risk reduction 41 Severity of top events The analysis in the previous section identified three top events which are addressed with high priority to reduce risk We expressed the severity for each of these three events using RCCPS scores Severity level has thresholds for considering human health fireexplosion leakage environmental costs and opportunity loss We determined the average severity level to be an intermediate value between the maximum and minimum severity levels In the absence of an intermediate value we selected a point based on a judgment of whether there was a shift toward the maximum side or the minimum side due to influence of the occur rence of the target top event We then calculated the loss amount using the correlation formula shown in Fig 6 From Fig 6 RCCPS scores were applied up to 50 to apply the correlation equation When RCCPS scores were 50 it is possible that catastrophic damage to some equipment and significant increases in opportunity loss occurred due to the acci dent as this would be an accident with an unprecedented score The loss amount when a RCCPS score is 50 is calculated to be 984 M about 1000 M based on the correlation formula Therefore considering the possibility that extrapolating the loss amount seriously affected esti mation accuracy we did not use the correlation equation to calculate the loss amount and when RCCPS score is 50 it was set to 1000 M Fig 7 We found fire and explosion outside the reactor to be high thus we discuss fire and explosion outside the reactor hereafter We explain creating the FT diagram estimating the probability of occurrence identifying portions for risk reduction and selecting capital investment Table 6 Center for Chemical Process Safety CCPS points Severity Level points Human Health Fire or Explosion Potential Chemical Impact Community Environment Impact 1 27 Multiple fatalities Direct Cost 1 Byen Chemical release with potential for significant on site or offsite injuries or fatalities National media coverage over multiple days OR Environmental remediation required and cost in excess of 250Myen Federal government investigation and oversight of process OR other significant community impact 2 9 1 fatality 100 Myen 1 Byen Chemical release with potential for injury off site Shelterinplace or community Evacuation OR Environmental remediation required and cost in between 100Myen250Myen State government investigation and oversight of process OR Regional media coverage or brief national media coverage 3 3 Lost work accident 10 Myen 100 Myen Chemical release outside of containment but retained on company property Minor offsite impact with precautionary shelterinplace OR Environmental remediation required with cost less than 100Myen No other regulatory oversight required OR Local media coverage 4 1 First aid 25 Myen 10 Myen Chemical released within secondary containment or contained within the unit Shortterm remediation to address acute environmental impact No long term cost or company oversight Examples would include spill cleanup soil and vegetation removal 5 03 Does not meet or exceed Level 4 threshold 25 Myen Does not meet or exceed Level 4 threshold Does not meet or exceed Level 4 threshold 100 yen ¼ 1 Fig 5 Relationship between CCPS points and loss amounts This figure shows the relationship between CCPS points and total loss amounts for the AN fluid ized bed reactor accident case shown in Table 4 K Sano et al K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 Table 7 Revised Center for Chemical Process Safety CCPS points Severity Human Health Fire or Potential Chemical Impact CommunityEnvironment Impact Opportunity Level Explosion Loss points 1 27 Multiple fatalities Direct Cost Chemical release with potential National media coverage over multiple days OR Environmental 1 Byen 1 Byen for significant onsite or offsite remediation required and cost in excess of 250Myen Federal injuries or fatalities government investigation and oversight of process OR other significant community impact 2 9 1 fatality 100 Myen1 Chemical release with potential Shelterinplace or community Evacuation OR Environmental 100 Myen 1 Byen for injury off site remediation required and cost in between 100Myen250Myen Byen State government investigation and oversight of process OR Regional media coverage or brief national media coverage 3 3 Lost work accident 10 Myen Chemical release outside of Minor offsite impact with precautionary shelterinplace OR 10 Myen 100 100 Myen containment but retained on Environmental remediation required with cost less than 100Myen Myen company property No other regulatory oversight required OR Local media coverage 41 First aid 25 Myen Chemical released within Shortterm remediation to address acute environmental impact 25 Myen 10 10 Myen secondary containment or No long term cost or company oversight Myen contained within the unit Examples would include spill cleanup soil and vegetation removal 5 03 Does not meet or 25 Myen Does not meet or exceed Level 4 Does not meet or exceed Level 4 threshold 25 Myen exceed Level 4 threshold threshold 100 yen 18 following formula holds 10000 dN yn 42 1000 dt z 100 gp y19678x where N is the number of pieces of equipment in operation The number 8 10 é R 07142 of specimens in the initial state is No and both sides are divided by No If P NNo dP 10 20 30 40 50 JP 43 3 0 dt a 8 When t 0 as P 1 9 pj fj yd Pe 44 Revise CCPS point P is known as availability and 1P is unavailability F That is Fl1e 45 Fig 6 Relationship between revised CCPS points and loss amounts This figure Generally in quantitative risk assessment analysis because 1 At shows the relationship between the revised CCPS points and total loss amounts holds the second term on the right side of Eq 46 can be approximated by the expanded formula aimed at risk reduction For loss amount we used the average value 2 3 Patex11 4 449 at 46 42 FTA and top event occurrence probability 2 3 that is unavailability also known as occurrence probability can be 421 Creating the FT diagram calculated by multiplying failure rate 4 and time t We created an FT diagram Fig 8 that used fireexplosion outside the reactor as the top event and considered the intermediate and basic 423 Failure rate events identified in Sections 32 and 33 Failure rate data can be obtained from present databases Abundant The FT diagram made the accident scenario clear From an FT dia data exists on nuclear power plants such as the US Nuclear Regulatory gram it is possible to identify incidents that do not lead to major acci Commission Component Reliability Data Sheet 2010 USNRC dents like the top event Also when new risk identification is performed September 2012 etc As for chemical plant data the CCPS Guideline for and risks are added in the future it will be possible to improve Process Equipment Reliability Data with Data Tables 2010 can be completeness by adding to this diagram used In contrast failure rates of specific equipment used in the AN re action process have not been disclosed until now In this study data on 422 Unavailability of system the number of failures and total operating time observed from AN plants If many installed components m are present and a long time period were obtained and we used observed failure numbertotal operating T elapses the failure rate 2 can be given using the following equation time 1h as the failure rate Total operation time exceeds 100000 h n which is sufficient to observe the number of failures mT 41 There were some AN reaction process failures that have no history of occurring Although the occurrence probabilities of some failures are where n is the number of observed failures of independent components extremely small they cannot be said to be zero The total operation time during the operation time T for cases with no records of failure exceeded 100000 h which is Here we assume that the proportion of failures over time is propor considered a satisfactory time period and the probability of failure tional to the equipment in operation at that time and the proportional occurrence is considered to have reached its minimum value Quoting constant is 4 In this case an ordinary differential equation of the 8 Journal of Loss Prevention in the Process Industries 63 2020 104015 9 Fig 7 Revised CCPS points and loss amounts for three top events The severity for three events is expressed using RCCPS points In the pentagon graphs the thin thick and doted lines represent the maximum average and minimum points respectively The loss amounts were calculated using the correlation formula in Fig 6 K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 10 Fig 8 Fault tree diagram fireexplosion outside reactor This is the fault tree diagram for the case of fireexplosion outside reactor Hatching value indicates the occurrence probability of the basic event Black hatching indicates events related to the equipment or phenomena of acrylonitrile reaction section For interpretation of the references to colour in this figure legend the reader is referred to the Web version of this article K Sano et al K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 the allowable value of locationspecific individual risk LSIR ISOTS probability is d and an improved loss amount is calculated as d x total 16901 2015 the probability of the failure event occurring was 10 x loss amount 10 on the safety side That is in this study the failure occurrence Table 10 shows case study results when multiple modifications are probability was considered the minimum value in the AN plant performed simultaneously In Case 1 modification with the relatively As for unavailability of human error HE in the AN reaction process short investment recovery year in Table 9 is implemented and Cases 2 to we selected work described in the literature like that of the AN reaction 4 add further modifications In Table 10 indicates implementation process with the assumption that humans always have the same Case 3 implementation reduces top event occurrence probability by the reliability order of 10 Also it can be implemented with a relatively short in Table 8 summarizes failure rates and occurrence probabilities of vestment recovery year Case 4 implementation is not a feasible safety basic events used to estimate occurrence probabilities of top events by investment and we found that implementations up to Case 3 are the FT diagram sufficient Cases 3 and 4 differ in the absence or presence of sparger modifi 424 Probability of top event occurrence cation From the perspective of improving loss amount the investment First each basic events occurrence probability was calculated as effect of sparger modification cannot be seen The sparger affects gas follows and substituted into the green hatched portion of the FT diagram distribution in the fluidized bed and is thought to govern reaction effi Fig 8 We set the operation period to 2 years 16000 h as regular ciency in a fluidized bed reactor Sparger modification is not an in repairs are often performed every 2 years and calculated each basic vestment focusing on design optimization for safety purposes but is events occurrence probability from Eq 46 using the failure rate 1h rather an optimum design meant to improve reaction results listed in Table 8 In Table 8 basic events with an occurrence probability of 12 years were used as listed 5 Discussion The and gate and or gate in the FT diagram Fig 8 were calculated as follows This study first revealed that in the AN process the reactor has the And gate highest risk level In general the fluidized bed reaction has the advan Ft FltF2tFnt tages of uniform progress of the catalyst phase reaction improved I Fit 47 control of reaction temperature due to easy removal of heat easy il handling of gases with explosion range composition and easy catalyst Or gate loading and unloading Fluidized bed reaction systems are used in AN synthesis reactions and other commercialscale chemical plants for these Ft 1 i 1 F1t1 F21 Faa 48 reasons Many studies have focused on improving reaction yields by 1 II 1 1 Fit optimally controlling the partial oxidation reaction and improving the catalyst and equipment to improve economic efficiency as noted in where Ft is system unavailability and Fi is unavailability for event i Section 2 Additionally it is typically difficult to grasp the possibility of Fig 8 shows that the occurrence probability per two years of top risk until an accident occurs in a fluidized bed reaction system as it is event was calculated as 73 x 10 Because the loss amount is 846 M difficult to understand the accident occurrence mechanism Because risk average loss amount in Fig 7 the expected value of the loss amount arises from nonlinear interactions between failures andor normal per two years becomes 62 M based on the product of the occurrence operating fluctuations it is difficult to think of ways to reduce risk probability and the loss amount Solving these issues requires comprehensive identification of such risks In this study we investigated and evaluated accident cases reac 43 Priority of investment for risk reduction tion processes plant components natural disasters and terrorist attacks that had not yet been considered for AN plants Because the influences of It is assumed that generalpurpose single equipment machinery natural disasters and terrorist attacks on chemical plants are not negli instrumentation and electrical equipment commonly used in chemical gible we regarded these influences as common faults plants such as pumps control valves thermometers flow meters and From the evaluation of investment cost and benefits the priority of cables that make up a chemical plant have ordinary reliability for use in single investment was shown initially Table 9 Furthermore to reduce chemical plants Thus we excluded improving reliability of general fireexplosion risks several investments were selected in order of high purpose equipment from this study priority as single investments Table 10 Because investment amounts Although it can be easily inferred that internal and external factors and construction periods are limited in actual plants it is believed that can change HE we set it to a constant value in this study From the selecting investment destinations based on the method described in this perspective of understanding the influence of HE on top event occur study is sufficiently effective rence probability we doubled each HEs probability of occurrence The Our study clearly demonstrates that risks related to fluidized bed top events occurrence probability reached a maximum of 92 x 107 reaction systems can be traced to specific equipment subsystems and compared to the original maximum of 73 x 1077 suggesting that the operations Thus this study also demonstrated an effective method for influence is small in this case reducing AN process risks by preventing certain operations or protecting Events in this study were related to AN reaction process equipment against such risk events and finding safety investment priorities How and are indicated by black hatching in the FT diagram Fig 8 Equip ever this study has the following limitations ment improvement andor introduction of safety protection measures were carried out for target events with the failure rate being reduced 51 Estimating occurrence probability using an FT diagram from the perspective of reducing top event occurrence probability The target event investment priority for reducing failure rate is In this study we calculated failure rates of specific components in the indicated by the investment recovery year shown in Table 9 Table 9 AN reaction process using data obtained from actual AN plants to details failure rates after equipment improvement or introducing safety determine top event occurrence probabilities As noted above failure protection measures Data are from AN plant observations and data with rates of specific components in the AN reaction process are not available no failure occurrence after remodeling are represented by an occurrence in the public literatures Because failure rates can generally be obtained probability of 10 x 107 same as Table 8 In Table 9 top event from statistical data failure rate uncertainties are discussed here In a occurrence probability is the value when each modification is done discussion of blower failure rates the Centralized Reliability Data Or alone Reduced displacement from the original top event occurrence ganization established by the US Department of Energy in 1985 found 11 Journal of Loss Prevention in the Process Industries 63 2020 104015 12 Table 8 Failure rate and occurrence probability Failure rate unit is 1hour Occurrence probability describes the probability of occurring within two years Basic Event No Facilities Operation Failure Value Unit Cited Document Remarks 1 Flow control device Air operated valve Operation failure 86E 08 1h 1 Consider failure of air operated valve constituting FCV flow measurement unit RO and transmitter respectively or Air operated valve Accidently open or close 18E 08 1h 1 Air operated valve Clogging 20E 08 1h 1 Orifice Internal damage 13E 08 1h 1 Orifice Clogging 20E 08 1h 1 Flow transmitter Inoperative 31E 08 1h 1 Flow transmitter High outputlow output 72E 08 1h 1 2 Operation by operators DCS operation mistake 42E 02 1 2years 2 3 Pressure control device Air operated valve Operation failure 86E 08 1h 1 Consider failure of air operated valve constituting PCV and transmitter respectively or Air operated valve Accidently open or closed 18E 08 1h 1 Air operated valve Clogging 20E 08 1h 1 Pressure transmitter Inoperative 13E 08 1h 1 Pressure transmitter High outputlow output 42E 08 1h 1 4 Flow control device Same as 1 1h 1 Consider C3H6 and NH3 5 Operation by operators DCS operation mistake 42E 02 1 2years 2 6 Sparger Crack header 66E 07 1h AN data Crack nozzles 34E 07 1h AN data 7 Reactor Undetected temperature abnormality 30E 05 1h AN data 8 Reactor Leakage due to corrosion of condensed acidic material 34E 07 1h AN data 9 Cooling Coils Crack welding part 10E 05 1 2years AN data Consider crack form or Crack piping 10E 05 1 2years AN data Crack reactor wall weld 73E 07 1h AN data 10 Manual valves Failure of opening or closing 11E 08 1h 1 Clogging 11E 08 1h 1 11 Operation by operators Valve operation 18E 03 1 2years 3 12 Electric pumps water Continuous operation failure 81E 07 1h 1 13 Pumps Cavitation 10E 05 1 2years No data To be sufficiently low 14 Operation by operators Pump operation 10E 03 1 2years 3 15 Cooling coils Dirt on the coil surface difficulty in recovering heat transfer capability 15E 06 1h AN data 16 Operation by operators Valve operation 18E 03 1 2years 3 17 Reactor Obstacles to fluidization 10 Eþ00 1 2years AN data 18 Reactor Formation of Mo pieces size that exacerbates fluidization 36E 05 1h AN data 19 Reactor Accumulation of Mo pieces 97E 05 1h AN data 20 Reactor Generation of reaction heat 10 Eþ00 1 2years AN data 21 Reactor Undetected hot spot 12E 05 1h AN data 22 Operation by operators Missed anomalies insufficient operation monitoring 10E 01 1 2years No data Estimated from actual results to be less than 005 Reference 3 is 015 Adopt simple average of both 23 Sustainable gas Always present continued on next page K Sano et al K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 Table 8 continued Basic Facilities Failure Value Unit Cited Remarks Event No Operation Document 10 1 E00 2years 24 Ignition source 10 1 Always present E00 2years 25 Reactorpiping Fastening problem 10E 1 AN data 05 2years AN Acrylonitrile FCV Flow control valve PCV Pressure control valve DCS Distributed control system 1 JANSICFR02 2016 Estimation of domestic general equipment failure rate considering uncertainty of failure number 19822010 29 years 56 plants data June 2016 wwwgenanshinjparchivefailureratedataJANSICFR02pdf 2 Gertman DI Blackman HS 1994 Human Reliability Safety Analysis Data Handbook John Wiley Sons 3 Williams JC 1999 Human reliability data the state of the art the possibility in ProcReliability 89 Vol1 UK June 1416 Table 9 Investment selection 1 Reduced displacement from the original occurrence probability of the top event is d and d x total loss amount is calculated as the loss amount to be improved Improvement The investment priority is indicated by Year to recovery investment Basic FacilitiesOperation Failure Failure Rate or Unit Top Event Investment Improvement Years to event Occurrence Occurrence Myen Myen Recover No Probability Probability Investment 6 Sparger Crack header 66E07 1h Crack nozzles 34E07 1h Modification Crack header 17E07 1h 68E02 280 43 65 1 Modification Crack nozzles 10E07 1h 2 7 Reactor Undetected temperature 30E05 1h abnormality Modification Undetected temperature 11E05 1h 45E02 35 24 1 1 abnormality Modification 10E05 1 28E02 70 38 2 2 2years 8 Reactor Leakage due to corrosion 34E07 1h of condensed acidic material Modification Leakage due to corrosion 10E05 1 68E02 100 43 23 of condensed acidic 2years material 9 Cooling coils Crack welding part 10E05 1 2years Crack piping 10E05 1 2years Crack Reactor wall 73E07 1h weld Modification Crack Reactor wall 13E07 1h 64E02 172 77 22 1 weld Modification 10E05 1 63E02 264 93 29 2 2years 17 Reactor Obstacles to fluidization 10 E00 1 2years Modification Obstacles to fluidization 10E05 1 The cost and effect of fluidizing obstacle removal are described in 19 2years 19 Reactor Accumulation of Mo 97E05 1h pieces Modification Accumulation of Mo 29E06 1h 63E02 40 9 4 pieces 21 Reactor Undetected hot spot 12E05 1h Modification Undetected hot spot 10E05 1 63E02 56 9 6 2years 100 yen 18 four distributions normal distribution exponential distribution Wei improvement are expected to be different in each plant This stands in bull distribution and lognormal distribution and concluded that fail contrast with nuclear power plants which have relatively unified ure rates successfully fitted a lognormal distribution Cadwallader and equipment design Taylor 1993 Since then the lognormal distribution has been The FT diagram also considered HEs which at chemical plants frequently applied to equipment failures Because this study employed typically include operating a valve in the wrong direction or overlooking data from one AN plant statistical fluctuation cannot be considered signals It was the case that reliability depended on random variables Additionally even if it were possible to collect data from several AN other than time As noted in Section 42 we employed data on similar plants it would be difficult to evaluate uncertainties in failure rates of human error probabilities HEP obtained from the references Gertman AN reaction process components This is because the design specifica and Blackman 1994 Williams 1989 However as expected HEPs tions of fluidized bed reactors which have undergone refinement and generally depend on plant operation thus further research using HEP 13 Journal of Loss Prevention in the Process Industries 63 2020 104015 14 data on AN plants is required 52 Safety investment selection Because selecting combinations of multiple investments depends on top event occurrence probability there might be priorities depending on the optimal combination that are not in accordance with the single in vestment rank order However in this study we did not achieve this level of analysis To reduce risks associated with accidental fires and explosions at AN plants this study clearly suggested a method for judging the importance of technical tasks This method would allow engineers to notice the priority of technologies and provide a clear goal 6 Conclusions In this study we evaluated a method for understanding safety in vestment priorities using risks in AN plants as an example This study demonstrated that fires and explosions posed the greatest risks in the reaction process As for the severity of the risk standard applied to AN plants we proposed a modified index that considered both direct and indirect costs By considering process characteristics and natural di sasters we added new accident scenarios and set up a policy to derive accident occurrence probabilities develop accident prevention tech nical systems and determine investment costs from FTAs Our results highlighted safety protection equipment and investment costs As a measure of the investment effect the safety investment priority was indicated by using loss amount differences from before and after investment Declaration of competing interest The authors declare that there are no conflicts of interest Acknowledgments This study received funds from the Faculty of Environment and In formation Sciences Yokohama National University Japan Appendix A Supplementary data Supplementary data to this article can be found online at httpsdoi org101016jjlp2019104015 References Abuswer M Amyotte P Khan F 2013 A quantitative risk management framework for dust and hybrid mixture explosions J Loss Prev Proc 26 283289 AlSharrah GK Edwards D Hankinson G 2007 A new safety risk index for use in petrochemical planning Process Saf Environ 85 533540 American Institute of Chemical Engineers AIChE 1994 Dows Fire and Explosion Index Hazard Classification Guide seventh ed AIChE New York Baybutt P 2018 Guidelines for designing risk matrices Process Saf Prog 37 4955 Cadwallader LC Taylor GL 1993 Failure rate data for glovebox components and cleanup systems at the Tritium Systems Test assembly J Fusion Energy 12 12 CCPS Process Safety Leading and Lagging Metrics 2011 You dont improve what you dont measure Retrieved January 2011 httpswwwaicheorgsitesdefaultfiles docspagesCCPSProcessSafetyLagging2011224pdf Accessed 13 May 2019 Centi G Perathoner S 1998 Unraveling a catalytic pathway acrylonitrile from propane Chem Tech 28 1318 Cox LA 2008 Whats wrong with risk matrices Risk Anal 28 497512 Cozzani V Smeder M Zanelli S 1998 Formation of hazardous compounds by unwanted reactions in industrial accidents J Hazard Mater 63 131142 Cruz AM Krausmann E 2009 Hazardousmaterials releases from offshore oil and gas facilities and emergency response following Hurricanes Katrina and Rita J Loss Prev Proc 22 5965 Curcurù G Galante GM La Fata CM 2013 An imprecise fault tree analysis for the estimation of the rate of occurrence of failure ROCOF J Loss Prev Proc 26 12851292 Dutta S Gualy R 1999 Overhaul process reactors Hydrocarb Process 78 4350 EPA US 1983 Health Assessment Document for Acrylonitrile Final report NoEPA 600882007F EPA US 1998 Ambient Water Quality Criteria for the Protection of Human Health Acrylonitrile EPA822R98006 Ferdous R Khan F Veitch B Amyotte PR 2009 Methodology for computer aided fuzzy fault tree analysis Process Saf Environ 87 217226 First K 2010 Scenario identification and evaluation for layers of protection analysis J Loss Prev Proc 23 705718 Gertman DI Blackman HS 1994 Human Reliability Safety Analysis Data Handbook John Wiley Sons Inc New York Goodrich Chemical Company BF 1972 Acrylonitrile process analyzer system Adv Instrum 27 15 Grasselli RK 1986 Selective oxidation and ammoxidation of olefins by heterogeneous catalysis J Chem Educ 63 216229 Grasselli RK 1999 Advances and future trends in selective oxidation and ammoxidation catalysis Catal Today 49 141153 Gu Y Chen W 1991 Separation of acrylonitrile and acetonitrile by xyrenwater dual solvent extraction Proc Int Conf Pet Refin Petrochem Process 2 552558 Guo L Kang J 2015 An extended HAZOP analysis approach with dynamic fault tree J Loss Prev Proc 38 224232 Hawkins C Wood M 1999 Slashing waste through tighter control Chem Eng 106 107 Hopper JR Yaws CL Ho TC Vichailak M 1993 Waste minimization by process modification Waste Manag 13 314 IARC 1999 Monographs on the Evaluation of Carcinogenic Risks to Humans 71 Re evaluation of Some Organic Chemicals Hydrazine and Hydrogen Peroxide pp 43108 Part I ISOTS 16901 2015 Guidance of Performing Risk Assessment in the Design of Onshore LNG Installations Including the Shipshore Interface JANSICFR02 2016 Estimation of domestic general equipment failure rate considering uncertainty of failure number 19822010 29 Years 56 plants data June 2016 wwwgenanshinjparchivefailureratedataJANSICFR02pdf Accessed 13 May 2019 Kang J Guo L 2016 HAZOP analysis based on sensitivity evaluation Saf Sci 88 2632 Korkmaz KA Sari A Carhoglu AI 2011 Seismic risk assessment of storage tanks in Turkish industrial facilities J Loss Prev Proc 24 314320 Krausmann E Cruz AM Affeltranger B 2010 The impact of the 12 May 2008 Wenchuan earthquake on industrial facilities J Loss Prev Proc 23 242248 Maccallion J 1996 New separation approach saves catalyst and energy Chem Process 59 7374 MacGregor RJ 2017 Results matter three case studies comparing and contrasting PFFM and HazOp PHA reviews J Loss Prev Proc 49 266279 Miyake A Yamada N Ogawa T 2005 Mixing hazard evaluation of organic peroxides with other chemicals J Loss Prev Proc 18 380383 Pujado PR 1977 Newest acrylonitrile process Hydrocarb Process 56 169172 Sanghavi VD 1998 Pollution control and environmental protection in chemical industry Chem Eng World 33 6162 Shah S Fischer U Hungerbühler K 2005 Assessment of chemical process hazards in early design stages J Loss Prev Proc 18 335352 Shelly S 1995 Catalytic oxidation unit meets stringent emission limits Chem Eng 102 135 TERA 1997 Acrylonitrile Inhalation Cancer Risk Assessment ThanhBinh N Dubois JL Kaliaguine S 2016 Ammoxidation of acrolein to acrylonitrile over bismuth molybdate catalysts Appl Catal A 520 712 Table 10 Investment selection 2 Case study results when multiple modifications are performed at the same time indicates implementation Case Investment Top Event Occurrence Probability Investment Improvement Years to Recover Investment 6 7 8 9 17 19 21 12 years Myen Myen Years 0 73E02 1 17E02 166 48 3 2 12E02 266 52 5 3 31E05 530 62 9 4 31E05 810 62 13 100 yen ¼ 1 K Sano et al Journal of Loss Prevention in the Process Industries 63 2020 104015 15 Van der Bann HS 1980 The acrylonitrile process NATO Adv Study Inst Ser E 39 522533 WHO 1983 International Programme on Chemical Safety vol 28 Environmental Health Criteria Acrylonitrile Williams JC 1989 Human reliability data the state of the art the possibility In Proc Reliab vol 89 UK June 1416 1 Yu J Cruz AM Piatyszek E Lesbats M Tardy A Hokugo A Tatano H 2017 A survey of impact on industrial parks caused by the 2011 Great East Japan earthquake and tsunami J Loss Prev Proc 50 317324 Zhao J Cui L Zhao L Qiu T Chen B 2009 Learning HAZOP expert system by casebased reasoning and ontology Comput Chem Eng 33 371378 K Sano et al