基于进化神经网络的某层状岩体隧道塌方预测  

Study on Collapse Prediction of Layered Rock Tunnel Based on Evolutionary Neural Network

作　　者：杨建成 陈东方 王越 毕钛俊 Yang Jiancheng;Chen Dongfang;Wang Yue;Bi Taijun(YCIC Lijiang Investment Development Co.,Ltd;School of Resources and Environmental Engineering,Wuhan University of Technology;Wuhan Institute of Rock and Soil Mechanics,Chinese Academy of Sciences)

机构地区：[1]云南交投丽江投资开发有限公司,云南丽江674100 [2]武汉理工大学资源与环境工程学院,武汉市430000 [3]中国科学院武汉岩土力学研究所,武汉市430000

出　　处：《勘察科学技术》2023年第3期32-36,共5页Site Investigation Science and Technology

摘　　要：层状岩体隧道塌方是多种因素导致的结果,单一判据方法难以保证预测结果的准确性。该文提出基于进化神经网络的层状岩体隧道塌方预测方法,综合地质与施工两方面因素,以埋深、主应力比、岩体倾角、洞轴线与岩层层面夹角、地质强度指标和支护强度作为输入参数,以塌方深度为输出参数,选用两个隐含层结构,采用遗传算法优化了神经网络模型的结构参数和初始权值。通过训练,建立了满足塌方深度智能预测精度的进化神经网络模型。经实例检验,基于进化神经网络模型的塌方深度预测结果与现场实际塌方深度较为吻合,表明了基于进化神经网络的层状岩体隧道塌方预测的可靠性。Collapse of layered rock tunnel is the result of many factors,and a single judgment method is difficult to ensure the accuracy of the prediction results.This paper proposes a collapse prediction method of layered rock mass tunnel based on evolutionary neural network,which integrates two factors of geology and construction.Taking buried depth,principal stress ratio,rock mass dip angle,angle between tunnel axis and rock layer,geological strength index and the supporting strength as input parameter,and collapse depth as output parameter,two hidden layers are selected and the structural parameters and the initial weights of the neural network model are optimized by the genetic algorithm.Through training,an evolutionary neural network model is established to meet the accuracy of collapse depth intelligent prediction.Tested by a typical layered rock tunnel,the prediction result of collapse depth based on evolutionary neural network model is relatively consistent with the actual collapse depth on site,which indicates the reliability of the collapse prediction of layered rock mass tunnels based on evolutionary neural network.

关 键 词：层状岩体 塌方 神经网络 遗传算法 风险预测 

分 类 号：U458.3[建筑科学—桥梁与隧道工程]