Investigate Contribution Of Multi-Microseismic Data To Rockburst Risk Prediction Based On Bagging-SVM

Rockburst disasters bring a serious threat to mine safety.Carrying out rockburst risk prediction can reduce rockburst disasters and maintain the safety of personnel and property of mine.The formation mechanism of rockburst is very complicated.The existing methods mainly focus on theoretical prediction research,which has the problems of low prediction efficiency,low prediction accuracy and poor adaptability.At the same time,existing studies often only use microseismic energy data,ignoring the microseismic raw wave data,resulting in incomplete analysis of microseismic data,and the microseismic data has not been fully and effectively used.To solve the above problems,this paper investigate contribution of multi-microseismic data to rockburst risk prediction based on Bagging-SVM integrated classifier.The main work is as follows:(1)Aiming at the problem of ignoring the microseismic raw wave data and the incomplete analysis and utilization of microseismic data in existing studies,from the perspective of fully and effectively using the rich information contained in multi-microseismic data,this paper investigate contribution of multi-microseismic data consisting of microseismic raw wave data and microseismic energy data to rockburst risk prediction.Firstly,a large amount of microseismic raw wave data was collected from the microseismic monitoring system by on-site monitoring,combined with the microseismic energy data into multi-microseismic data.Then,fully mine the rich information contained in the multi-microseismic data by extracting 132 and 24 features from microseismic raw wave data and microseismic energy data in the multi-feature domains such as frequency domain,entropy and time-frequency domain,respectively,and construct multi-dimensional combined feature matrix based on multi-microseismic data.Finally,multi-microseismic data is used to predict rockburst risk,and compared with the results of using microseismic raw wave data or microseismic energy data alone based on the same experimental conditions to analyze the importance of multi-microseismic data in the prediction of rockburst risk.Experiments show that the microseismic raw wave data is an important part of multi-microseismic data and should not be ignored in the prediction of rockburst risk.(2)Aiming at the problems of low prediction efficiency,low prediction accuracy and poor adaptability in existing research,this paper proposes a prediction model of rockburst risk based on Bagging-SVM integrated classifier.Firstly,SVM classifier suitable for high-dimensional small sample problems is reasonably selected as the base learner of the integrated classifier by analyzing the characteristics of multi-microseismic data.Secondly,the RBF kernel function suitable for nonlinear problems is selected as the kernel function of the SVM base learner through comparative research.Then,the Bagging-SVM integrated classifier prediction model is established using the voting mechanism of the Bagging ensemble learning method which can improve the prediction model,and to study the improvement effect of the model compared with the SVM classifier prediction model.Finally,feature selection and hyperparameter optimization of Bagging-SVM integrated classifier are carried out by innovatively combined with genetic algorithm to avoid the empirical dependence of manual parameter selection.Experiments show that this method can effectively improve the accuracy and stability of the rockburst risk predictionMultiple sets of comparative experiments prove that multi-microseismic data plays an important role in the rockburst risk prediction,and rockburst risk prediction can achieve good results based on Bagging-SVM integrated classifier by using multi-microseismic data.