摘要
探讨了计算速度和可靠性这两个在位移反分析工作中非常重要的问题。一方面用 BP网络代替有限元计算提高了计算效率 ,另一方面用遗传算法代替常规的优化算法 ,使反分析结果与初值无关。三介质边坡算例验证了上述解决方案的可行性。
Two important aspects in displacement back analysis, calculation efficiency and reliability, are discussed. Firstly the error back-propagation neural network (BP network) is used to substitute the time-consuming finite element analysis, and secondly genetic algorithm is used for the optimization of objective function, making back-analyzed solutions irrelevant to initial values. The feasibility is verified by back analysis of a three-media slope.
出处
《岩石力学与工程学报》
EI
CAS
CSCD
北大核心
2001年第1期1-5,共5页
Chinese Journal of Rock Mechanics and Engineering
基金
国家"攀登计划"B40项资助项目
关键词
BP网络
遗传算法
岩石
边坡
位移反分析
Backpropagation
Finite element method
Genetic algorithms
Neural networks
