Potential hazard analysis of accidents in Indian underground mines using Bayesian network model

The underground mining industry is recognized as one of the most hazardous industries in the world since it is characterized by a high rate of fatal and non-fatal accidents. In most cases, mine accidents result in the death of miners, and damage to machines which directly affect the production and safety. Thus, to ensure mine safety, it is necessary to predict the accidents, determine their causes, and take preventive measures within a predetermined timeframe. In this study, we examined 224 accidents which occurred in Indian underground mines from 2010 to 2023. We categorize the collected data into four layers: root causes of potential hazards, major accidents, effects of accidents, and basic information of accidents. The Bayesian Network (BN) model is used to make a probabilistic relationship between four layers containing 29 variables and establish the conditional probability table (CPT). The real-time evidence of the accident is fed into the BN model and updates the CPT. The prior, likelihood, and posterior probability used to develop the reasoning to categorize the accident into three groups namely; catastrophic, degraded and cascaded accident. The conflict analysis is used to measure the conflict between real-time sets of evidence as well as validate the model. The five objective-based sensitivity analyses are used to identify the best combination of three elements; evidence, interesting, and investigating parameter that can used to make the preventive measure and avoid the effect of accidents. The proposed research work helps to develop the modern mining intelligence system and improve the safety governance system for the mining industry.