Hybrid Grey Wolf Optimization Algorithm–Based Support Vector Machine for Groutability Prediction of Fractured Rock Mass

Abstract

Groutability determination is a very important task in grouting quality control. There is little research on the groutability of cement-based grout in a fractured rock mass, and the prediction is hampered by a small number of samples along with multidimensional and nonlinear problems. This study proposes an intelligent predictive model that integrates hybrid grey wolf optimization (HGWO) and a support vector machine (SVM) to predict the groutability. The model was built in three steps: HGWO was embedded in a SVM to search for the best hyperparameters (C, g); crossvalidation and error analysis were introduced into the HGWO-SVM model to ensure the generalization performance and prediction accuracy; and the classification and regression prediction of groutability with cement-based grout in a fractured rock mass were predicted by the established HGWO-SVM intelligent prediction method. Taking a curtain grouting project as a case, the applicability of the method was verified. The performance of the proposed prediction model is improved compared with other methods, and the prediction accuracy meets engineering needs. The results show that this method can accurately and conveniently predict the groutability of cement-based grout in a fractured rock mass and provide practical assistance to field projects.