| Rockburst is a typical engineering geological disaster,which is suddenness,difficulty to control and large damage range.However,there are many factors affecting the occurrence of rockburst,and the genetic mechanism is complex.It is too one-sided to predict the tendency of rockburst only from a single index.Therefore,the characteristics of rockburst precursors are analyzed comprehensively with multi-source information to improve the accuracy of rockburst prediction.Firstly,according to the characteristics of various dynamic indicators of AE,MS and electromagnetic radiation,and according to their evolution law,the change law of the precursory characteristics of mine rockburst is analyzed,which provides a theoretical basis for rockburst prediction and prevention and control.Secondly,the rockburst prediction models are constructed from two perspectives.1.From the perspective of rockburst index weight,the combined weighting-improved TOPSIS prediction model is constructed.The subjective and objective weights are obtained by the improved G1 and the anti-entropy method,and combined to solve the problem of bias caused by a single weight.2.From the perspective of rockburst engineering data training,the Ada Boost-BAS-SVM model is constructed.The parameters of SVM are optimized by BAS,which avoids the influence of inaccurate manual settings.At the same time,combined with Ada Boost ensemble algorithm,weak classifiers are combined into strong classifiers,the feasibility and accuracy of the model are proved.Afterwards,the dynamic index verification of the two prediction models is carried out through engineering examples,and the improved PCA method is used to reduce the dimension of the data,which greatly improves the accuracy of the Ada Boost-BAS-SVM prediction model.The two models are compared from many aspects,and the results show that the constructed Ada Boost-BAS-SVM prediction model is more excellent.Figure 36;Table 32;Reference 53... |
PDF Full Text Request