From correlation to causality: Path analysis of accident-causing factors in coal mines from the perspective of human, machinery, environment and management

• Published in: Resources policy Vol. 73; p. 102157
• Main Authors: Fa, Ziwei, Li, Xinchun, Qiu, Zunxiang, Liu, Quanlong, Zhai, Zhengyuan
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.10.2021; Elsevier Science Ltd
• Subjects: Accidents; Algorithms; Association rules; Behavior; Built environment; Causality; Classification; Coal; Coal industry; Coal mine; Coal mines; Coal mining; Data mining; HFACS framework; High risk; Human factors; Incentives; Machinery; Mines; Path analysis; Structural hierarchy; Supervision; Text mining; Unsafe; Text mining; Path analysis; Association rules; Coal mine; HFACS framework
• ISSN: 0301-4207; 1873-7641
• DOI: 10.1016/j.resourpol.2021.102157

Coal mine industry is one of the typical high-risk industries with frequent accidents and numerous incentives. In order to expose the accident-causing factors in the process of coal mine production and explore their internal causal relationship, this paper took 883 coal mine accident reports in China from 2011 to 2020 as the original data. Firstly, driven by data, with the help of text mining technique and Apriori algorithm, a modified Human Factors Analysis and Classification System (HFACS) for coal mines was established and the strong association rules among the contributing factors were extracted. Then, the related hypotheses were put forward according to the frequent patterns within the elements. Finally, under the guidance of the theory-driven approach, hierarchical structure relationships in HFACS-CM framework were identified and analyzed from the perspective of human, machinery, environment and management. The results indicated that machinery and equipment factors, physical environment factors, and unsafe preconditions could directly affect employees' unsafe behaviors, while outside influences, organizational influences and unsafe supervision and could only exert influences on unsafe acts through other intermediary variables. Moreover, the unsafe preconditions had the greatest direct effect on unsafe acts; as for indirect effect and overall effect, the unsafe supervision was the most impactful factor. •A combinational research paradigm of data-driven and theory-driven is proposed.•Text segmentation and Apriori algorithm are applied as data-driven methods to extract strong association rules.•The hypotheses put forward based on the association rules are verified through path analysis as a theory-driven method.•The mediating paths and effects between unsafe acts and associated influencing factors are further identified.