PSO-BP神经网络在隧道围岩变形预测中的应用  

Application of PSO-BP Neural Network in Prediction of Tunnel Rock Deformation

作　　者：李海斌 翟秋柱 张优 王亚暐 张霄 

机构地区：[1]中铁十七局集团第四工程有限公司,重庆401121 [2]中国中铁西南科学研究院,成都610000 [3]西南交通大学地球科学与环境工程学院,成都610000

出　　处：《路基工程》2017年第5期164-169,共6页Subgrade Engineering

摘　　要：针对滇西复杂地质条件下隧道围岩变形预测问题,以BP神经网络为基础,引入了改进后的粒子群算法,通过调试和改进建立了PSO-BP神经网络。该神经网络结合了粒子群算法的全局搜索能力和BP神经网络的局部搜索能力,非线性映射能力强,泛化能力强,具有一定的容错能力。计算结果表明:PSO-BP神经网络预测精度高,平均绝对误差为2.4 mm,平均相对误差为2.7%,满足隧道围岩变形预测精度的需要。In relation to the prediction of rock deformation surrounding tunnel under complicated geological conditions in western Yunnan, this paper on the basis of BP neural network introduced an improved particle swarm optimization （PSO） to find out PSO - BP neural network by debugging and improving. This neural network , which combines the PSO global searching ability with the BP local searching ability, has strong capability of nonlinear mapping and generalization with certain capability of fault tolerance. The result of calculation shows that PSO - BP neural network is of high accuracy with the average absolute error of 2.4 mm and the average relative error of 2. 7 %, and meets the requirement of rock deformation prediction.

关 键 词：隧道工程 围岩变形预测 粒子群算法 PSO-BP神经网络 

分 类 号：U456.31[建筑科学—桥梁与隧道工程]