基于MATLAB神经网络工具箱的岩爆预测模型  

A model for predicting rock burst by MATLAB neural network toolbox

作　　者：孟陆波[1] 李天斌[1] 王震宇[1] 

机构地区：[1]成都理工大学工程地质研究所,四川成都610059

出　　处：《中国地质灾害与防治学报》2003年第4期81-85,共5页The Chinese Journal of Geological Hazard and Control

基　　金：四川省杰出青年学科带头人培养计划资助项目(03ZQ026 045)

摘　　要：文章介绍了BP人工神经网络的基本原理,针对其收敛差的缺点,发挥MATLAB神经网络工具箱的优势,分别采用VLBP和LMBP算法建立了改进后的BP神经网络。对于影响岩爆发生的关键因素,总结了专家经验,选取地下硐室围岩最大切向应力与岩石单轴抗压强度比值、岩石单轴抗压强度和抗拉强度比值和岩石冲击性倾向指数作为岩爆预测的评判指标,建立了岩爆预测的神经网络模型,并利用国内外一些岩石地下工程实例进行分析计算校验,计算结果表明,用该模型进行岩爆预测是可行有效的。This paper introduces the basic principle of the BP neural network. Aiming as its disadvantage of constringency, this paper exerts the superiority of MATLAB neural network toolbox, and builds the improved BP neural network through using VLBP and LMBP algorithm. To the key factors of affecting the rock burst′s occurrence, we summarize the experiences of former experts, select the ratio of wall rock′s maximal tangential stress and rock′s single axle pressive strength, the ratio of rock′s single axle pressive and rock′s single axle tensile strength, and impact proneness index as the judging indexes of rock burst, and built neural network model for the prediction of rock burst. Through real rock underground project examples of home and abroad, we carry through analysis and computing, the results show that the model is feasible and valid to predict rock burst.

关 键 词：MATLAB 神经网络 岩爆 预测模型 冲击性倾向指数 改进BP算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]