摘要
粒子群优化(PSO)算法是近年来发展迅速,并得到广泛应用的一种仿生全局最优化算法。与遗传算法相比,该算法具有操作简单、易于编程的优点。结合铜黄高速公路汤屯段大田连拱隧道施工,采用PSO算法对BP神经网络的权值进行自动优化,获得训练效果最好的BP网络模型参数以提高网络的泛化能力,建立起基于PSO-BP算法的大田隧道施工位移非线性智能分析模型,并采用此模型对后续施工隧道变形进行了预测分析。与实测位移对比表明,本文建立的PSO-BP模型平均预测相对误差仅为3.1%,可很好地作为隧道信息化施工的一种辅助方法,并为其他类似岩土工程提供借鉴。
With the rapid development recently, the particle swarm optimization （ PSO）has been widely used as a bionic global optimization algorithm. Compared with the genetic algorithm, it embodies the characteristics of easy programming and less parameters. Combined with the construction of Datian double - arch tunnel in Tonghuang highway, a novel BP neural network based on PSO algorithm which had been adopted to optimize the weight value of the network is introduced into analyzing monitoring data in this paper. The optimal BP model, with improvement of the generalization ability and the nonlinear mapping relationship between time and displacement is established applied into fitting and predicting tunnel monitoring data. The mean prediction relative error of crown subsidence compared with measured displacement is only 3.1% based on the PSO - BP algorithm, so it can serve as an assistant tool in information construction of tunnel and similar geotechnical work.
出处
《地下空间与工程学报》
CSCD
北大核心
2009年第2期250-253,共4页
Chinese Journal of Underground Space and Engineering
基金
863计划资助项目(2007AA118109)
北京交通大学科技基金项目(2006XM025)
