Bayesian network-based model for assessing the intensity of outbursts of coal and gas

Fully understanding the major parameters contributing to an outburst is significant in the evaluation of coal and gas outbursts. However, these parameters can also affect each other which significantly increase the difficulty of evaluation. Considering related uncertainties, a Bayesian network (BN) which reflects major variable parameters affects the object without knowing the clear relations between the parameters. In this study, a conceptual BN model is developed to estimate the probability of outburst intensities in the laboratory. A total of 2,000 groups of random samples based on physical experiments are produced by trace linear or 3D interpolation for these factors and outbursts, respectively. A total of 1,800 groups of random samples are used to construct the BNs while the other 200 random samples are used for validation purposes. The proposed BN model estimates the intensity of the outburst with a root-mean-square deviation (RMSD) (3.8) and a linear correlation (0.94) compared with validation values of 200 random samples.