Forecasting the stress concentration coefficient around the mined panel using soft computing methodology

Stress analysis around the mined panel is one of the most critical topics in longwall mining for stability analysis of surrounding pillars and access tunnels. Actually, stress redistribution due to roadway extraction along with the longwall mining-induced stress must be considered in designing the panel surrounding structures. In this research, three new methods, i.e., radial basis function neural network (RBFNN), fuzzy inference system (FIS) and statistical analysis (SA) models have been developed to predict the stress concentration coefficient (SCC) around a mined panel. The transferred stress due to longwall mining has been also considered in predicted SCC from these models. Proposed models have been constructed based on the sufficient datasets gathered from the literatures. For SCC prediction, the height of destressed zone above the mined panel, rock mass unit weight, overburden depth and horizontal distance from the panel edge have been regarded as input variables. Based on the actual testing datasets, performance of the suggested models has been evaluated using the statistical indices. Accordingly, it was proved that RBFNN and FIS models have better capability compared to the SA model and their results are in a great agreement with the real values. Moreover, proposed models were practically applied in Tabas longwall coal mine of Iran and verified by comparing their results with the results of existing models. Finally, conducted sensitivity analyses of the proposed models indicate that height of destressed zone and unit weight are the most and least influencing variables on the SCC in all models.