基于支持向量机的隧道变形预测模型研究
刘宇
（吉林铁道职业技术学院 铁道工程系，吉林，吉林132001 ）
关键词：隧道，变形预测，支持向量机，遗传算法
中图分类号：TU454 文献标识码：A
摘要:本文采用遗传算法（Genetic Algorithm，GA）对支持向量机(Support VectorMachines, SVM)的核函数参数g和惩罚因子c 进行优化，建立基于参数寻优支持向量机的蠕动型滑坡隧道变形位移预测模型，以与蠕动型滑坡隧道变形密切相关的10个影响因子作为模型输入向量，隧道变形实测数据值作为模型的目标输出。以吉林省蛟河市长珲高速老爷岭隧道实地监测数据为样本对模型进行训练与预测分析，仿真结果表明本文方法训练速度快且预测值与真实值平均相对误差小于2%，具有很强的工程应用价值。
Research on tunnel displacement forecast model basedon support vector machines
LIU Yu
Department of Railway TechnologyEngineering，JilinRailway Vocational and Technology College，Jilin 132001， China
Key words:tunnel，displacement forecast，support vector machines，Genetic Algorithm
Abstract: Geneticalgorithm was used to optimize the penalty factor parameters c and g of supportvector machines. The tunnel displacement forecast model was established basedon the optimized parameters of support vector machines. The model inputswere 10 impact factors closely related to creep landslide deformation of tunnel,andthe value of the measured data of the tunnel deformation were used as thetarget output of the model. Actual data of Laoyeling tunnel in Changchun toHunchun highway of Jilin province were adopted as a sample to train the modeland predict tunnel deformation displacement．The simulation results show that this method has afast training speed, and the average relative error between testing value and truevalue is less than 2%.The method has a strong value for engineeringapplication.