摘要
针对目前已有的各种位移反分析方法存在的缺陷,利用神经网络具有的非线性映射能力和遗传算法具有的全局随机搜索能力,提出了一种基于遗传神经网络进行深基坑支护的位移反分析方法。该方法改变了BP算法依赖梯度信息的指导来调整网络权值的方法,而是利用遗传算法全局性搜索的特点,寻找最合适的网络连接权和网络结构等来达到优化的目的。结合地铁深基坑支护位移计算,应用该方法对某一地铁深基坑土体的力学参数进行了反演。结果表明:将位移观测值作为网络输入数据,土体力学参数作为输出数据,在较大的解空间内,该位移反分析方法收敛速度快、解的稳定性好、反演结果精度高,是一种理想的位移反分析方法。最后,采用该软件结合一个工程实例实现了应用遗传神经网络进行的基坑支护位移反分析。
Aiming at subsistefit limitation in diversified displacement back analysis methods, an approach based on neural network and genetic algorithm for displacements back analysis of deep foundation pit for metro is proposed. This approach utilizes nonlinearity of neural network and whole random search capability of genetic algorithm. It can search the best appropriate weight and framework of neural network by using whole searching characteristic of genetic algorithm, which formerly depends on gradient information to adjust weight of network, The proposed approach has been used to carry out inverse calculation for soil mechanical parameters of deep foundation pit for metro. A case is conducted using the software developed in the paper. The result shows： considering measured deformation value as input data and taking soil dynam parameter as output data of neural network, the approach can rapidly get a stable and accurate solution within a relatively large solution space; and the approach is superior to current back analysis approaches.
出处
《岩土力学》
EI
CAS
CSCD
北大核心
2007年第10期2118-2122,共5页
Rock and Soil Mechanics
基金
国家自然科学基金资助项目(No.50308029)
关键词
神经网络
地铁
位移
反分析
遗传算法
neural network
metro
displacement
back analysis
genetic algorithm
