Prediction of tailgate stability in mechanized longwall mines using an improved support vector regression model

Abstract

Tailgate stability in a mechanized longwall mine is serious to mine productivity and personnel’s safety. Although the stability of both roadways is critical in longwall mining, the tailgate is subjected to higher stresses and deformations than the other one. Therefore, the prediction of tailgate stability is a distinctive challenge in mechanized coal mining to ensure its functionality during mining operations. In this regard, an Improved Support Vector Regression (ISVR) model was developed to sustain and secure longwall mine design through stability prediction of the tailgate roadway based on the roof displacements and geomechanical data. For model development, the geomechanical information gathered through laboratory tests and site investigations in Tabas coal mine, Iran, was introduced to the ISVR model as independent variables. The roof displacements values, which were monitored in a 1.2 km long tailgate, were also used as the dependent variable. According to the results, the proposed ISVR model could predict roof displacements in a reasonable accordance with measured ones. The squared correlation coefficient (R2) between ISVR predicted and measured roof displacements showed a high conformity with R2 = 0.91. The results of the ISVR model were compared with those of Artificial Neural Networks (ANNs) and Multivariable Linear Regression (MLR), which respectively yielded R2 = 0.87 and R2 = 0.81. In conclusion, the ISVR model appears to be a proper measure for indicating unstable zones ahead of time in mechanized and high-speed longwall mining.