基于判别分析法的岩爆烈度预测研究

Study on rockburst prediction based on discriminant analysis method

岩爆是岩土工程中棘手的地质灾害,工程中以预防为主。现有岩爆分级预测模型大多存在选取样本较少和准确率较低的问题。综合岩爆的参考指标,现选取围岩最大切向应力与岩石单轴抗压强度比σθ/σc(应力系数)、岩石单轴抗压强度与单轴抗拉强度比σc/σt(脆性系数)和弹性能量指数W_(et)作为分级评判指标,广泛收集不同工程的104组岩爆实例,选取其中84组作为样本集进行训练,20组作为测试集进行检验,应用SPSS的判别分析中的Bayes判别和Fisher判别训练及测试,输出结果中,选取了训练效果较好的Bayes判别模型。对95.23%的样本集进行了正确分类,验证集检验准确率为85%,将该模型应用于工程实例中,预测结果与实际结果相符,预测结果表明该模型有较好的应用前景。

Rockburst is a difficult geological hazard in geotechnical engineering,and prevention is the main focus in engineering.Most of the existing rockburst classification prediction models have the problems of small sample selection and low accuracy.Combined with the comprehensive reference indexes of rock burst,the maximum tangential stress of surrounding rock and rock uniaxial compressive strengthσθ/σc(stress coefficient),rock uniaxial compressive strength and uniaxial tensile strengthσc/σt(brittleness coefficient)and the elastic energy index of W_(et) were selected as grading evaluation indexes,different engineering 104 groups of rock burst were widely collected as examples,84 groups were selected as the training sample set,the group of 20 was used for the test in the test set,Bayes discriminant of discriminant analysis of SPSS and Fisher discriminant were applied to train and test,and the better training effect the Bayes discriminant model was selected from the output.95.23%of the sample sets were correctly classified,and the accuracy of verification set was 85%.When the model was applied to an engineering example,the predicted results were consistent with the actual results.The predicted results showed that the model had a good application prospect.