Prediction of rockburst classification using Support Vector Machine

Abstract

With the deepening of underground excavation, rockburst has become a serious geological disaster, which will always cause casualties, machine damage and delay of construction schedule. Therefore, many scholars at home and abroad have carried out researches on rockburst. Because the mechanism of rockburst is complex, and there is no unified understanding of its generation mechanism at present, thus it is hard to predict the rockburst happen or not and the intensity to guide for the underground engineering construction. Support Vector Machine (SVM) was used to classify the rockburst. Some main factors of rockburst, such as the maximum tangential stress σθ, the compressive rock strength σc, the tensile strength σt, the stress coefficient Ts, the brittleness coefficient of rock B, and the elastic energy index Wet were selected in the analysis. The factors were divided into two combinations: index I and II. SVM model and criterion were acquired through 36 training samples. Another 10 testing samples were used to evaluate the model. As a consequence, the evaluated results agree well with the measured record. No matter the training samples or the testing samples, the misjudgement ratio using the combination index I is smaller than that using the combination index II. It is suggested that using SVM model and the index I can classify the rockburst grade very well.