Evaluating the Thickness of Broken Rock Zone for Deep Roadways using Nonlinear SVMs and Multiple Linear Regression Model

AbstractSince the traditional methods to estimation of the thickness of broken rock zone (BRZ) are usually difficult, expensive and not feasible in many cases, the development of some predictive models for the thickness of broken rock zone (BRZ) for deep roadways will be useful. To describe the complex relationship between geological factors and BRZ, a nonlinear model-based support vector machines (SVMs) regression analysis was applied on the data pertaining to China mine to develop some predictive models for the thickness of BRZ for deep roadways from the indirect methods in this study. The type of kernel function was Radial basis function (RBF). 132 samples were trained by proposed models; the other 10 samples that were not used for training were used to validate the trained models. The correlation coefficients of SVMs model for predicting the thickness of BRZ is more than 0.90. For the same two similarity groups, the developed SVMs model was also compared with the multiple linear regression analysis (MLRA) model and measured data. As a result of SVMs analysis, a very good model was derived for BRZ estimation. It was shown that SVMs models were more reliable and precise than the regression models. Concluding remark is that the thickness of BRZ values of deep roadways can reliably be estimated from the indirect methods using SVMs analysis