Accident case-driven study on the causal modeling and prevention strategies of coal-mine gas-explosion accidents: A systematic analysis of coal-mine accidents in China

Abstract

Post-accident prevention measures in Chinese coal mining enterprises were based on single accident case training, and the learning of accident information is limited to the violation in the accident or incident. The nature of accidents proves that accident prevention encompasses not only visible violations and hazardous materials, but also deep-rooted organizational factors. How to obtain valuable and learnable safety information from limited accident date remains a key issue. This study investigated the use of a modeled accident-prevention framework for combining accident case data to construct an accident causation network for targeted data mining, which in turn could help obtain accident-prevention measures with more control values. The results of the study showed that (1) A framework for categorizing the factors influencing coal mine accidents, including four layers and 42 indicators, was identified through a literature review and factor extraction from 86 coal mine accident reports. (2) A causation network model of gas-explosion accidents comprising 163 accident causation nodes and 681 connected edges was established, and SP02(failure to implement the stop production instruction), SA01 (poor awareness of system compliance), PB03(fluke psychology) and SH01(habitual violation of procedures and rules) were the most influential accidental causes in the accidental network. (3) An F-Score-based attack strategy for accident-causing networks was proposed, which is superior to generalized attack strategies for reducing accident network connectivity and more conducive to rapidly obtaining accident prevention priorities. (4) A gas explosion accident prevention strategy based on the combination of accident network model and dynamic Bayesian network was constructed, and inference simulation was performed for four cases such as IA01 (unchecked gas concentration), verifying that it has the possibility of predicting accident hazards in the workplace. This study provided accident data-driven prevention interventions to reduce future mining accidents.