基于PSO-SVM算法的层状软岩隧道 大变形预测方法
基于PSO-SVM算法的层状软岩隧道
大变形预测方法
杨文波 1 ,王宗学 1 ,田浩晟 1 ,吴枋胤 1 ,杨自成 2
(1.西南交通大学土木工程学院,四川 成都 610031；2.四川绵九高速公路有限责任公司,四川 绵阳 621700)
为提高复杂地质条件下层状软岩隧道大变形预测的可靠性，提出基于粒子群优化（particle swarm optimization,PSO）-支持向量机（support vector machine,SVM）算法的隧道大变形预测方法，解决大变形预测中多项评价指标权重计算复杂及界限值多样等问题。为充分考量层状软弱围岩强度、围岩结构类型、地应力及地下水对隧道大变形的影响，选取岩体抗压强度、层理倾角、初始地应力状态、埋深、岩体修正质量指标[BQ]、地下水发育情况6项亚级指标对大变形等级进行预测。根据大变形等级划分标准，构建以地应力反演、现场大变形监测信息为基础的大变形预测模型，并采用粒子群算法优化惩罚参数 C 与核函数参数Gamma，以提高模型的准确性。研究结果表明：采用粒子群优化-支持向量机(particle swarm optimization,PSO-SVM)算法可以避免传统预测方法如地质综合判断法和强度应力比法由于单一指标和主观原因引起的误差，预测精度高；该方法利用隧道已发生的大变形信息，构建出符合目标隧道现场实际规律的PSO-SVM大变形预测模型；PSO-SVM模型对样本测试集预测的准确度达86.36％，优于SVM和GS-SVM模型；以九绵高速典型层状软岩隧道白马隧道为研究对象，应用提出的PSO-SVM模型进行大变形预测，通过与现场实测对比发现，预测精度达80％，验证了该方法的可行性。
关键词： 层状软岩隧道;大变形预测;支持向量机;粒子群优化
杨文波（1985—），男,四川成都人，博士，教授，博士生导师，主要研究方向为大规模隧道群结构安全标识系统，车致振动荷载作用下隧道结构动力响应特性。 E-mail:yangwenbo1179@hotmail.com
引用格式： 杨文波,王宗学,田浩晟,等.基于PSO-SVM算法的层状软岩隧道大变形预测方法[J].隧道与地下工程灾害防治,2022,4(1):29-37.YANG Wenbo,WANG Zongxue,TIAN Haosheng,et al.Large deformation prediction method of layered soft rock tunnel based on PSO-SVM algorithm[J].Hazard Control in Tunnelling and Underground Engineering,2022,4(1):29-37.
Large deformation prediction method of layered soft rock tunnel based on PSO-SVM algorithm
YANG Wenbo 1 ,WANG Zongxue 1 ,TIAN Haosheng 1 ,WU Fangyin 1 ,YANG Zicheng 2
(1.School of Civil Engineering,Southwest Jiaotong University,Chengdu 610031,Sichuan,China; 2.Sichuan Mianjiu Expressway Co.,Ltd.,Mianyang 621700,Sichuan,China)
In order to improve the reliability of prediction of large deformation level of layered soft rock tunnels under complex geological conditions,a particle swarm optimization (PSO) based support vector machine (SVM) was proposed.The large deformation prediction method of tunnel solved the problems of complex calculation of multiple evaluation index weights and diverse boundary values in large deformation prediction.In order to fully consider the influence of layered soft rock surrounding rock strength,surrounding rock structure type,in-situ stress and groundwater on the large deformation of the tunnel,the 6 sub-indices of value:uniaxial compressive strength,rock inclination angle,initial in-situ stress state,burial depth,rock mass correction quality correction index [BQ] and groundwater development,were used to predict the large deformation level.This method constructed a large deformation prediction model based on in-situ stress inversion and on-site large deformation monitoring information according to the classification standard of large deformation grades,and used particle swarm optimization algorithm to adjust the penalty parameter C and the kernel function parameter Gamma to improve the model′s performance accuracy.The research results showed that the use of particle swarm optimization-support vector machine (PSO-SVM) algorithm could avoid errors caused by traditional prediction methods such as geological comprehensive judgment method and strength-stress ratio method due to a single index and subjective reasons,and the prediction accuracy was high;this method used the large deformation information of tunnels that had occurred,a PSO-SVM large deformation prediction model conforming to the actual law of the target tunnel site was constructed;the PSO-SVM model had an accuracy of 86.36％ in the prediction of the sample test set,which was better than the SVM and GS-SVM models.Taking the Baima Tunnel,a typical layered soft rock tunnel of the Jiu-mian Expressway as the research object,the proposed PSO-SVM model was used to conduct large-scale research.The deformation prediction was compared with the field measurement,and it was found that the prediction accuracy reached 80%,which verified the feasibility of the method.
Keywords
