Optimizing Underground Coal Mine Safety: Leveraging Advanced Computational Algorithms for Roof Fall Rate Prediction and Risk Mitigation

Abstract

The utilization and consumption of coal in various nations have emphasized the pivotal role played by coal mines. However, aside from the substantial contribution of coal mines, miners, engineers, and craftsmen in this industry have long been exposed to numerous risks and financial losses resulting from roof collapses in underground coal mines. Hence, due to the heightened sensitivity surrounding this issue, the accurate and low-error forecasting and assessment of the roof fall rate (RFR) are deemed crucial and of utmost importance. Nonetheless, due to the intricate and uncertain inherent characteristics of the rock formations, assessing the RFR has encountered multiple challenges that cannot be precisely approximated through traditional methods. In this paper, algorithms such as the harmony search algorithm (HS) and the invasive weed Optimization algorithm (IWO) are harnessed to address the aforementioned challenges. To model the RFR, a total of 109 data points were used, incorporating input parameters such as primary roof support (PRSUP), depth of cover (D), coal mine roof rating (CMRR), mine height (MH), and intersection diagonal span (IS). For effective data analysis and model development, the dataset was split into two separate groups: one for training and the other for testing. Specifically, 80% of the data was used to build the model, while the remaining 20% was allocated for model evaluation and validation. Based on the outcomes of three statistical metrics R2, MSE, and RMSE, it is evident that the deployment of HS and IWO algorithms demonstrates high performance, with predicted values closely aligning with actual ones. Consequently, the utilization of intelligent algorithms in the field of rock engineering is positioned as a potent tool for researchers and engineers. In conclusion, a sensitivity analysis is carried out with the help of the @RISK software as a means of ranking the influence that the input parameters have on the output of the model. Its results indicate that among different parameters, the CMRR parameter with a sensitivity degree of 0.11 has the most impact on the model, even with the smallest change in this parameter, a significant change is made in the model output.