铁道标准设计2025，Vol.69Issue(3)：122-129,8.DOI:10.13238/j.issn.1004-2954.202407250004
基于"函数变量分离+移动最小二乘拟合"桩板墙高效优化算法探索
Exploration of Efficient Optimization Algorithm for Pile-Slab Wall Based on"Separation of Function Variables+Moving Least Squares Fitting"
摘要
Abstract
The parameter optimization design of pile-slab wall requires the use of geotechnical finite element calculation programs to ensure that the design parameters meet safety constraints,such as stress and deformation,while achieving the optimal cost goal.It is a multi-parameter,multi-constraint,and non-linear optimization problem.In view of the complexity of the calculation process of the geotechnical finite element calculation program for pile-slab wall and the difficulty of quickly obtaining the cost function solution through analytical methods,a parameter optimization algorithm combining separation of function variables and moving least squares fitting was proposed.Firstly,a cost sensitivity analysis of the design parameters was conducted to obtain the reduced-dimensional analytical expression for the separation of pile length variables in the cost function.Next,the moving least squares fitting was performed on the coefficient function of the reduced-dimensional expression to obtain the coefficient function values and select the optimal combination of pile thickness and pile spacing parameters corresponding to the best coefficients.Finally,the geotechnical finite element calculation program was used to evaluate the optimal parameter combinations and the separated pile length parameters.The optimal pile length,pile thickness,and pile spacing parameters that achieved the optimal cost and met the safety indicators constraints were obtained.The algorithm was verified by practical engineering cases The results showed that the optimal cost obtained by this algorithm was the same as that obtained by the traversal parameter method,but the data volume used was 17%of that required by the traversal method.Moreover,the algorithm exhibited more stable convergence compared to the genetic algorithm,with the runtime reduced from 7.33 s to 0.17 s.The optimization effect was improved,and the search time was significantly shortened.关键词
桩板墙/参数优化/敏感性分析/函数变量分离/移动最小二乘法/安全指标Key words
pile-slab wall/parameter optimization/sensitivity analysis/separation of function variables/moving least squares/safety indicators分类
交通工程引用本文复制引用
梅龙月,杨艳,姚洪锡,毛国成,陈罗平,马淑霞..基于"函数变量分离+移动最小二乘拟合"桩板墙高效优化算法探索[J].铁道标准设计,2025,69(3):122-129,8.基金项目
国家重点研发计划项目(2021YFB2600404) （2021YFB2600404）
中国铁建股份有限公司科技研究计划项目(2022-A02) （2022-A02）
