在线极限学习机在岩爆预测中的应用  

Attempt to study the applicability of the online sequential extreme learning machine to the rock burst forecast

作　　者：兰明[1] 刘志祥[1] 冯凡[1] 

机构地区：[1]中南大学资源与安全工程学院,长沙410083

出　　处：《安全与环境学报》2014年第2期90-93,共4页Journal of Safety and Environment

基　　金：国家科技支撑计划项目(2013BAB02B05;2012BAB08B01);中南大学教师基金项目(2013JSJJ029);国家自然科学基金和上海宝钢集团公司联合资助项目(51074177)

摘　　要：为有效预测地下工程岩爆的发生及烈度,结合地下工程岩爆的特点,分析岩爆影响因素及相关判别依据,选取围岩最大切向应力σ与岩石抗压强度σc之比σ/σc、岩石抗压强度σc与岩石抗拉强度σt之比σc/σt以及弹性能量指数Wet为判别因子,引入在线极限学习机理论,建立了岩爆预测的OS-ELM判别模型。以搜集到的国内外15组工程岩爆数据进行训练建模,训练完成后将样本数据做输出预测,得到模型的预测精度达97.98%,并与SVM、BP模型进行对比分析,结果表明,OS-ELM模型精度优于SVM和BP模型。利用该模型对国内两处隧道岩爆情况进行预测,结果与实际情况基本相符。研究表明,OSELM判别模型在岩爆烈度分级上具有良好的适用性和有效性。The present paper takes it as a target to come up with an approach to forecasting the rock burst based on the Online Sequential Extreme Learning Machine （OS-ELM）, a single-hidden layer feed-forward neural network algorithm, a response model between the sampling data and the desired output according to a straightforward nonlinear mapping relation. As is known, rock burst is a kind of primary hazards that are likely to take place in the underground construction, which makes it necessary to forecast the likely frequency and intensity of such bursts efficiently and timely. Our approach lies in the following key points： The first, we have done an analysis of the influential factors and the relevant distinct criteria of such hazards in line with the features of the rock burst in underground construction. In doing so, we have set up the three discriminating factors, including σ θ / σ c , the ratio of the surrounding rock maximum circumferential stress to the rock compressive strength, σ c / σ t , the ratio of the rock compressive strength to the tensile strength and the elastic energy index for classifying the rock burst intensity into three degrees, that is, none, inferior and medium. The second, we are going to introduce the OS-ELM briefly for forecasting the underground geological rock burst. In doing so, we have chosen 15 groups of typical geological rock burst sampling data collected both at home and abroad as the model input with the model output which should be regarded as the rock burst intensity. And, the third, we have established a corresponding OS-ELM discrimination model based on the said input and output data for the targeted rock burst. And, the fourth, when the sampling data was substituted into the model, the result we have gained from the model training would be able to reach the forecast precision, that is, as high as 97.98%. Comparative analysis between the OS-ELM model and the SVM model and that between the OS-ELM model and the BP model in terms of the discriminating precision and trainin

关 键 词：安全工程 地下工程 在线极限学习机 岩爆分级 预测 

分 类 号：X936[环境科学与工程—安全科学]