https://orcid.org/0000-0002-1825-6632
Figures
Abstract
Both the number of dust explosion accidents and the resulting number of casualties have increased dramatically in recent years. To reduce this risk of dust explosions, we use the functional resonance analysis method (FRAM) to analyze the cause of the dust explosion accident at the Kunshan factory and propose barrier measures to prevent such accidents. The functional units that changed in the production system during the accident and how these functional units coupled to eventually cause the dust explosion were examined and explained. In addition, barrier measures were developed for functional units that changed during production and emergency systems defined to block the propagation of changes between functions and prevent resonance. Through case study, the identification of key functional parameters in both triggering the initial explosion and in then allowing its spread are key to define barriers to prevent a recurrence of such an event. FRAM uses system function coupling instead of traditional linear causality to explain the accident process, and develops barrier measures for changing function units, providing a novel thinking strategy and method for the analysis of accidents and their prevention.
Citation: Zhang M, Zhang L, Cao X, Li B, Zhou A (2023) FRAM-based causal analysis and barrier measures to mitigate dust explosions: A case study. PLoS ONE 18(6): e0287328. https://doi.org/10.1371/journal.pone.0287328
Editor: Kang Wang, Xi’an University of Science and Technology, CHINA
Received: March 21, 2023; Accepted: June 2, 2023; Published: June 15, 2023
Copyright: © 2023 Zhang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper.
Funding: This work was supported by: 1. Fundamental Research Program of Shanxi Province, China (Grant 202203021222043, 202203021222031). The funder had important role in study design, and preparation of the manuscript; 2. National Natural Science Foundation of China (51774292). The funder had important role in writing – review & editing of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
A broad swath of industries generates a growing mass of explosive dust as a hazardous waste product [1, 2]. This includes the production of food and industrial minerals together with the mining, especially of coal [3–6], that has resulted in repeated dust explosion accidents [7–9]. Moreover, dust explosions typically trigger secondary explosions, expanding the range, reach and intensity of the accident and potentially causing significant loss of life and impact on value of the environment [10–12]. According to incomplete statistics, there were 67 dust explosion accidents in China from 2005 to 2020, with more than 547 deaths and more than 600 injuries [13]. It is known that fuel, oxidant, ignition source, suspension, and confinement are five essential elements for dust explosions [14]. Thus, safety measures eliminating one or more of these five essential factors are effective methods to prevent or diminish the hazard. For example, the use of metal tools, brooms, compressed air, and fans during line cleaning disperses combustible dust in potentially explosive concentrations and also causes it to settle on elevated flat surfaces [15], thus ameliorating the danger. Based on this, the concept of inherent safety is considered a proactive approach to elevate safety where hazards are removed or minimized to reduce risk. In this, four basic principles are used to attain an inherently safe environment, viz. measurement and designed-minimization, substitution, moderation, and simplification [16–18].
In search of an intrinsically safe environment, there have also been many attempts to model various dust explosion scenarios for both risk analysis and safety assessment [19]. These include methodologies for risk analysis of dust explosion based on Bayesian networks [20] where a diagnostic analysis is used and dust particle properties, oxygen concentrations, and safety training of staff are identified as the most critical root-cause events leading to dust explosions. Other analyses have used physics-based modeling of the progression of dust explosions [21] and the analysis of experimental data [22] to develop predictive models to assess the probability of dust explosions and the development of nomographs for assessing risk.
The prediction of both accidents and systems failures can be driven by an appropriate accident causation model [23]. Accident causation models are developed to systematically probe accidents, a posteriori, and are crucial in understanding the underlying reasons accidents occur and how to then improve system safety [24]. The FRAM protocol [25–28] provides a structured method to develop an overall understanding of how a complex sociotechnical system operates. FRAM can be applied by identifying functions with detailed information about how a task is completed, characterizing the variability of the functions, interpreting possible couplings of the variability, and providing suggestions to manage any unexpected variability [29]. Thus, FRAM is potentially an effective and applicable tool for the analysis of dust explosions, but there are few related applications. We use FRAM to analyze the cause of a particularly serious aluminum dust explosion accident that occurred in the Kunshan factory in Jiangsu province on August 2, 2014. We apply FRAM to examine the dynamics of the complex social-technological systems manifest in this factory, together with the nonlinear coupling relationships amongst these elements, and use this analysis to propose barrier measures.
Background
FRAM analysis mechanism
The functional resonance analysis method (FRAM) may be applied to explain underlying mechanisms contributing to accidents. This is based on the principle of functional resonance, which in turn stems from the concept of stochastic resonance [30]. Stochastic resonance (SR) posits that the signal-to-noise ratio (SNR) can be enhanced by adjusting the noise intensity to a certain value under the premise of regular steady-state signal input. Functional resonance refers to the case when a functional performance changes slightly and regularly, then other functions will change through a coupling connection. Other functions in the system, as the operating environment of this function, will show irregular fluctuations due to the superposition of functional changes. If the regular change of function and the irregular fluctuation of operating environment produce a coupling, the system change will exceed positive constant range, eventually leading to system failure [29, 31]. At this time, the regular fluctuation of the function is equivalent to ‘steady-state signal input’, while the irregular change of the operating environment is equivalent to ‘noise’. If the two are coupled, the fluctuation value of the system will exceed the normal limit and cause accidents.
According to the FRAM, any function in the system is not stable and will change due to various factors, but the system can automatically adjust the functional change, so that the final output of the system is normal. However, if the system functional changes and the irregular fluctuation of the operating environment are coupled to produce resonance, the change value at this time has exceeded the range that the system can automatically adjust, the system will deviate from the normal operation direction, and eventually lead to accidents.
FRAM analysis process
The process of analyzing accidents with FRAM is shown in Fig 1. The method relies on progressing through a specified sequence of steps. These are as follows:
- (1) Determine the system boundary and identify the basic functions
The premise of FRAM in either risk analysis or accident case analysis is to establish a system to be analyzed. After that, the system elements should be divided into functions. The identification of basic functions is to identify a series of activities in the system.
- (2) Characterize functions with parameters
Six parameters can be used to describe the basic performance of the function. They are input, output, resource, precondition, control, and time [32] as shown in Fig 2.
The instructions for each parameter are shown in Table 1.
- (3) Build a functional network diagram
The output of one function in the system can be used as other characteristic parameters except not as the output of other functions. In this way, upstream functions can communicate with downstream functions through the same parameter characteristics to establish a functional connection channel. The connection channels between all functions form a system functional network diagram.
- (4) Identify performance variabilities
Functional performances are variable during operation, and may be changed due to the coupling of upstream functional performance or due to their own potential changing performance. So there are two reasons for the change:
- ① Coupling of upstream functions
Coupling links between functions are established through common parameters. If the upstream function changes, it may affect the operation of the downstream function through the coupling effect, so that its function will also change.
- ② Functional changes
The change of a function depends on its potential change performance, which can be studied from the function classification. For the implementation subject of functions, they can be divided into personnel, technology and organizational functions [33]. Different types of functions have different possibilities and reasons for change. Compared with the other two functions, personnel functions are more likely to change. The operation of technical functions is not easily affected by the external environment, and the possibility of change is lower than that of personnel and organizations.
The performance variabilities of system function can be directly reflected as the change of function output, that is, the output of function deviates from the expected goal. Function output changes are divided into accuracy output changes and timing output changes. When using FRAM for accident case analysis, you can specifically analyze which functions have changed in accuracy and timing output according to the actual situation of the accident, so as to determine the functional unit that has changed for subsequent resonance analysis; using FRAM for risk analysis, the potential changing performance of each function should be evaluated to judge the function that may have changed.
- (5) Coupling analysis to determine functional resonance
Based on the identified functional performances that have changed, the failure connection between the functions is determined, and a coupling analysis of the changed functional performances and the failure connection is conducted, so as to determine how the coupling between functions cause the system resonance and eventually lead to accidents.
- (6) Development of barrier measures
The goal of accident analysis is to prevent similar accidents from reoccurring. When using the FRAM to analyze accidents, barrier measures are established for functions with large changes in performance to control changes in function performance and to prevent resonance. Barrier measures are specifically divided into four categories, as shown in Table 2.
Case study
We apply FRAM analysis to a catastrophic dust explosion in China as a case study [34, 35]. At 7:34 on August 2, 2014, a particularly serious explosion of aluminum-alloy dust occurred in the second polishing workshop (namely plant 4) of the Kunshan factory in Jiangsu province. This occurred during manual polishing of the surfaces of aluminum-alloy wheel hubs for the car industry. There were 75 immediate fatalities with another 185 injured. Subsequently, 71 of the most seriously injured also died, increased the total number of fatalities to 146. The direct economic loss was 351 million yuan (~$52M). The timeline of the accident is shown in the sequence diagram of Fig 3.
We use FRAM to study this case, analyze the causes of the accident, and propose barrier measures to avoid similar accidents.
Identify system functions
The main production operation in the workshop where the accident occurred was manual polishing of automobile wheels to specified finish requirements. This confirms that production operation system S1 is included in the system of the accident analysis.
The identification of the basic functions in S1 is to identify a series of production activities carried out during the normal operation in the system. According to ‘Dust Explosion Protection Safety Regulations’, it is determined that the functions included to maintain the normal operation of the system S1 are: F1 assign tasks, F2 pre-job inspection, F3 dust cleaning, F4 equipment maintenance, F5 set-up of de-dusting system, F6 grinding and polishing, F7 environmental parameter monitoring and F8 on-site safety inspection. The functions are described as follows:
Before daily production, the workshop manager arranges the tasks to be completed on that day. Then, the operators check whether the production and de-dusting equipment are in good condition, ensure that there is no dust accumulation at the operation post and in the dust hood, and that the de-dusting system pipes and dust collecting bucket have been cleaned. If the requirements are not met, timely repair is applied to the damaged equipment, the deposited dust cleaned-up and the de-dusting system started after confirming that it is in good condition. The operator must ensure that the de-dusting system has been started normally before grinding and polishing is resumed.
Environmental parameters should be monitored during operation, including temperature and dust concentration. The monitored temperature refers to the temperature in the dust collection bucket and the air inlet of the dust-collector. The dust concentration is monitored through dust concentration sensors installed in the de-dusting system pipes and fan outlet. In addition, it is necessary to arrange safety management personnel or inspectors to conduct safety inspections on the work site, check whether there is dust accumulation in the workplace, whether the staff wear personal protection equipment correctly and according to the regulations, whether the workshop is equipped with dust explosion-proof electrical appliances and whether the emergency evacuation route is unblocked.
Functional changes in the production system can be obtained through environmental parameter monitoring and on-site safety inspections. However, if the monitored functional changes are not controlled and handled in time, causing functional resonance and dust explosion, the emergency system S2 can prevent the accident damage from further expanding. S2 includes functions F9 explosion suppression, explosion venting and arrestment, and F10 emergency evacuation. In the initial stage of the explosion, the explosion flame is extinguished by the explosion suppression system, so that the explosion can be effectively controlled in the initial stage. In addition, explosion venting and arrestment devices can also reduce casualties and equipment damage. At this time when the explosion is controlled, staff can evacuate through the emergency route.
The functions contained in S1 and S2, and their time sequence are shown in Fig 4.
Describe the basic functions
Six characteristic parameters are used to describe the functions in systems S1 and S2, as shown in Table 3.
Build the functional network diagram
According to the construction method, the system functional network diagram is shown in Fig 5.
Analyze performance variabilities
According to the actual situation shown in the accident investigation report, some functions in S1 and S2 deviated from the expected target during in the actual output, causing functional resonance through coupling. The function output changes are shown in Table 4.
In Table 4, except for F1, the output of the other functions in systems S1 and S2 have all changed. Among them, the normal output of function F5 means that the de-dusting system is operating normally, but it is known from the accident report that the de-dusting system exploded during the operation, so function F5 is the causal unit of the accident. The output of function F10 is abnormal, causing many deaths and serious injuries, which is the unit expanding accident loss.
Among the changed functions, the functional units that changed due to their own potential performance are functions F2, F7, and F8. The output changes in these performance functions, being upstream functions, make the output of downstream functions F3, F4, F5, F6, F9 and F10 change through coupling, and finally cause resonance, leading to the dust explosion and many casualties.
Determine functional resonance
Determining functional resonance requires to an estimation of which functional connections failed on the basis of identifying the changed functions in S1 and S2, and performing coupling analysis on the changed function units and failed connections. This is to determine how functional coupling can cause dust explosions during the operation of the de-dusting system.
- (1) Failure connections
The factors influencing resonance are determined by changing functional units in S1 and S2 and the coupling relations between functions established by the same parameters. Then the functional connections leading to system failure are determined, as shown in Table 5.
- (2) Coupling analysis
The output of function F2 is the input of functions F3 and F4, and these changes cause output changes of functions F3 and F4, resulting in the rusty and worn dust collecting bucket not being repaired in time, the deposited dust in the de-dusting equipment not being cleaned in time, causing the oxidation-reduction reaction of deposited aluminum powder that is affected by moisture, and the generation of a large amount of heat. The change of F7 and F8 results in the operation system failing to timely monitor the abnormal state of temperature and dust concentration in the dust collecting bucket, culminating in reaching the ignition temperature of the aluminum powder. The staff continue to work as no abnormality is detected, and an aluminum dust cloud is formed above the dust collecting bucket. This series of functional changes are coupled and force the production system towards a critical state with the possibility of a dust explosion, which ultimately leads to the occurrence of the aluminum dust explosion. The coupling of function changes leading to the accident is shown in Fig 6.
If the emergency system S2 can act quickly after the dust explosion, the loss caused by the accident will not be increased. However, the output of function F9 in S2 has changed, which shows that the automatic fire extinguishing and explosion suppression system do not extinguish the flame in the initial stage of explosion, and that the dust-collector and de-dusting pipeline are not equipped with explosion venting and arrestment devices. This causes the explosion shock wave and combustion products to spread to the workplace along the de-dusting pipeline, further expanding the loss caused by the accident. The output of functions F8 and F9 is the premise of F10, the emergency evacuation, and their output changes affect the output of F10 through coupling, eventually producing resonance, which enlarges the loss caused by the accident.
Develop barrier measures
To prevent accidents, barriers must be established to change functional units. For this accident, it is necessary to establish barrier measures for the changed function units F2, F3, F4, F5, F6, F7, F8, F9, and F10 to monitor and control functional changes and block the propagation path of the changes. The barrier measures developed are shown in Table 6.
Thus, the identification of key functional parameters in both triggering the initial explosion and in then allowing its spread are key to define barriers to prevent a recurrence of such an event.
Conclusions
We use the FRAM protocol to analyze the cause of a dust explosion accident at the Kunshan factory in Jiangsu province on August 2, 2014. The following conclusions are drawn:
- (1) The coupling of a series of function changes in the production system S1, primes the production system for a potential dust explosion that is ultimately manifest. Simultaneously, the function output of the emergency system S2 changes, producing resonance, which further expands the damage from the accident.
- (2) The identification of key functional parameters in both triggering the initial explosion and in then allowing its spread are key to define barriers to prevent a recurrence of such an event. To minimize/prevent dust explosions, it is necessary to strengthen the pre-job inspection to remove dust deposits in the workplace and provide for the timely repair of damaged equipment. Any abnormalities in the system should be detected early through environmental parameter monitoring and on-site safety inspection. In addition, dust explosion-proof appliances should be installed in the plant to ensure that the emergency system can start and function normally when an accident occurs, preventing further expansion of the hazard and resulting damage.
- (3) Causes of the accident and barrier measures recovered from the FRAM analysis are congruent with the conclusions of the investigation report. Both suggest mitigative measures of dust removal, equipment inspection and maintenance, the installation of dust explosion-proof appliances and the adoption of safety production management.
- (4) Case analysis with FRAM identifies that a linear causal relationship between the causative factors in the traditional linear model must be replaced with a mutual coupling between functions based on the normal operation of the system functions. A new accident mechanism is proposed in FRAM which effectively accommodates the dynamics of the complex system and the nonlinear relation of elements.
References
- 1. Bhattacharjee RM, Dash AK, Paul PS. A root cause failure analysis of coal dust explosion disaster—Gaps and lessons learnt. Engineering Failure Analysis. 2020;111. https://doi.org/10.1016/j.engfailanal.2019.104229
- 2. Santamaria-Herrera A, Hoyuelos FJ, Casado-Marcos C. Characterization of the explosiveness of wood dust. Process Saf Environ. 2023;169:252–9. https://doi.org/10.1016/j.psep.2022.10.087
- 3. Reed WR, Joy GJ, Shahan M, Klima S, Ross G. Laboratory results of a 3rd generation roof bolter canopy air curtain for respirable coal mine dust control. Int J Coal Sci Technol. 2019;6(1):15–26. https://doi.org/10.1007/s40789-019-0237-6
- 4. Shahan MR, Reed WR. The design of a laboratory apparatus to simulate the dust generated by longwall shield advances. Int J Coal Sci Technol. 2019;6(4):577–85. pmid:32341808
- 5. Kuracina R, Szabova Z, Skvarka M. Study into parameters of the dust explosion ignited by an improvised explosion device filled with organic peroxide. Process Saf Environ. 2021;155. https://doi.org/10.1016/j.psep.2021.09.011
- 6. Zhou XH, Li XX, Cui ZW, Wu L, Zhou HD, Lu XN. Combustible wood dust explosions and impacts on environments and health-A review. Environmental Research. 2023;216. pmid:36374653
- 7. Blair AS. Dust explosion incidents and regulations in the United States. J Loss Prev Process Ind. 2007;20(4–6):523–9. https://doi.org/10.1016/j.jlp.2007.03.012
- 8. Yan X, Yu J. Dust explosion incidents in China. Process Saf Prog. 2012;31(2):187–9. https://doi.org/10.1002/prs.11482
- 9. Su J, Cheng YF, Song SX, Ma HH, Wang WT, Wang YX, et al. Explosion Characteristics and Influential Factors of Coal dust/sodium Chlorate Mixture on Basis of an Explosion Accident in China. Combustion Science And Technology. 2021;193(8):1313–25. https://doi.org/10.1080/00102202.2019.1689965
- 10. Taveau J. Secondary dust explosions: How to prevent them or mitigate their effects? Process Saf Prog. 2012;31(1):36–50. https://doi.org/10.1002/prs.10478
- 11. Marmo L, Ferri A, Danzi E. Dust explosion hazard in the textile industry. J Loss Prev Process Ind. 2019;62. https://doi.org/10.1016/j.jlp.2019.103935
- 12. Lin S, Liu ZT, Qian JF, Li XL. Comparison on the explosivity of coal dust and of its explosion solid residues to assess the severity of re-explosion. Fuel. 2019;251:438–46. https://doi.org/10.1016/j.fuel.2019.04.080
- 13. Chen G, Zhang X, Xu S, Miao Y, Yang W, Gao J. Statistical analysis on dust explosion accidents in China from 2005 to 2020. China Saf Sci J. 2022;32(8):76–83. https://doi.org/10.16265/j.cnki.issn1003-3033.2022.08.0812
- 14. Abbasi T, Abbasi SA. Dust explosions-cases, causes, consequences, and control. J Hazard Mater. 2007;140(1–2):7–44. Epub 2006/12/30. pmid:17194531
- 15. Amyotte PR, Pegg MJ, Khan FI. Application of inherent safety principles to dust explosion prevention and mitigation. Process Saf Environ. 2009;87(1):35–9. https://doi.org/10.1016/j.psep.2008.06.007
- 16. Amyotte PR, Pegg MJ, Khan FI, Nifuku M, Tan Y. Moderation of dust explosions. J Loss Prev Process Ind. 2007;20(4–6):675–87. https://doi.org/10.1016/j.jlp.2007.05.011
- 17. Amyotte PR, Cloney CT, Khan FI, Ripley RC. Dust explosion risk moderation for flocculent dusts. J Loss Prev Process Ind. 2012;25(5):862–9. https://doi.org/10.1016/j.jlp.2012.05.007
- 18. Salzano E, Di Benedetto A. Inherently Safe indexes for Explosions in the Process Industry. Revista de Ingeniería. 2012;(37):73–8.
- 19. Yuan Z, Khakzad N, Khan F, Amyotte P. Risk-based optimal safety measure allocation for dust explosions. Safety Sci. 2015;74:79–92. https://doi.org/10.1016/j.ssci.2014.12.002
- 20. Yuan Z, Khakzad N, Khan F, Amyotte P. Risk analysis of dust explosion scenarios using Bayesian networks. Risk Anal. 2015;35(2):278–91. Epub 2014/09/30. pmid:25264172
- 21. Cloney CT, Amyotte PR, Khan FI, Ripley RC. Development of an organizational framework for studying dust explosion phenomena. J Loss Prev Process Ind. 2014;30:228–35. https://doi.org/10.1016/j.jlp.2013.09.002
- 22. Hassan J, Khan F, Amyotte P, Ferdous R. A model to assess dust explosion occurrence probability. J Hazard Mater. 2014;268:140–9. Epub 2014/02/04. pmid:24486616
- 23. Fu G, Xie XC, Jia QS, Li ZH, Chen P, Ge Y. The development history of accident causation models in the past 100 years: 24Model, a more modern accident causation model. Process Saf Environ. 2020;134:47–82. https://doi.org/10.1016/j.psep.2019.11.027
- 24. Ge J, Zhang YY, Chen SK, Xu KL, Yao XW, Li JS, et al. Accident causation models developed in China between 1978 and 2018: Review and comparison. Safety Sci. 2022;148. https://doi.org/10.1016/j.ssci.2021.105653
- 25.
Hollnagel E. Barriers and Accident Prevention. United Kingdom: Aldershot; 2004.
- 26. Hollnagel E. FRAM: The Functional Resonance Analysis Method: Modelling Complex Socio-Technical Systems. Surrey, United Kingdom: Ashgate Publishing Limited; 2012.
- 27. Tian WL, Caponecchia C. Using the Functional Resonance Analysis Method (FRAM) in Aviation Safety: A Systematic Review. Journal of Advanced Transportation. 2020;2020. https://doi.org/10.1155/2020/8898903
- 28. Bjornsen K, Jensen A, Aven T. Using qualitative types of risk assessments in conjunction with FRAM to strengthen the resilience of systems. Journal of Risk Research. 2020;23(2):153–66. https://doi.org/10.1080/13669877.2018.1517382
- 29. Salehi V, Veitch B, Smith D. Modeling complex socio-technical systems using the FRAM: A literature review. Human Factors And Ergonomics in Manufacturing & Service Industries. 2021;31(1):118–42. https://doi.org/10.1002/hfm.20874
- 30. Patriarca R, Di Gravio G, Woltjer R, Costantino F, Praetorius G, Ferreira P, et al. Framing the FRAM: A literature review on the functional resonance analysis method. Safety Sci. 2020;129. https://doi.org/10.1016/j.ssci.2020.104827
- 31. Grabbe N, Kellnberger A, Aydin B, Bengler K. Safety of automated driving: The need for a systems approach and application of the Functional Resonance Analysis Method. Safety Sci. 2020;126. https://doi.org/10.1016/j.ssci.2020.104665
- 32. Wu YL, Fu G, Han M, Jia QS, Lyu Q, Wang YX, et al. Comparison of the theoretical elements and application characteristics of STAMP, FRAM, and 24Model: A major hazardous chemical explosion accident. J Loss Prev Process Ind. 2022;80. https://doi.org/10.1016/j.jlp.2022.104880
- 33. Bazaluk O, Koriashkina L, Cheberiachko S, Deryugin O, Odnovol M, Lozynskyi V, et al. Methodology for assessing the risk of incidents during passenger road transportation using the functional resonance analysis method. Heliyon. 2022;8(11). pmid:36468135
- 34. Li XY, Chen HY, Zhang YS, Zhang XX, Li RZ. Research on law and mechanism of dust explosion in bag type dust collector. Advanced Powder Technology. 2022;33(6). https://doi.org/10.1016/j.apt.2022.103619
- 35. Li G, Yang HX, Yuan CM, Eckhoff RK. A catastrophic aluminium-alloy dust explosion in China. J Loss Prev Process Ind. 2016;39:121–30. https://doi.org/10.1016/j.jlp.2015.11.013