基于AL-Kriging模型及PSO算法的桥机主梁优化
Main Beam Optimization of Overhead Crane based on AL-Kriging Surrogate Model and PSO Algorithm
肖辉 1范小宁1
作者信息
- 1. 太原科技大学 机械工程学院,山西 太原 030024
- 折叠
摘要
为解决有限元模型和群智能算法相结合的起重机结构优化计算的计算成本昂贵的问题,该文基于AL-Kriging代理模型和粒子群智能优化算法构建了起重机主梁优化方法,在该方法中通过EFF学习函数选择优化所需的有效样本点,从而用较少的高保真计算样本构建出满足精度要求的代理模型,再通过PSO算法以及所构建好的代理模型完成结构优化.通过工程案例验证证明,在取得同样计算结果的情况下,与基于静态Kriging模型相比,该方法的优化时间节省了 70%,调用样本数仅为静态代理模型的 27%,证明了所构建的优化方法是可行和有效的.
Abstract
In order to solve the expensive calculation cost of crane structure optimization combining finite element model with swarm intelligence algorithm,a crane girder optimization method is constructed in this paper based on active learning Kriging surrogate model and particle swarm intelligence optimization algorithm.In this method,the effective sample points required for optimization is selected by the EFF active learning function,so as to construct a surrogate model that meets the accuracy re-quirements with as fewer high-fidelity samples as possible.Finally,the structural optimization is completed based on the con-structed surrogate model through the particle swarm optimization algorithm.Through the engineering case,under the condition of obtaining the same optimization results,the optimization cost time could be saved by about 70%compared with the static Kriging surrogate model,the number of call samples is only 27%of the static surrogate model,which verifies the feasibility and effectiveness of the established optimization method.
关键词
起重机主梁/Kriging代理模型/EFF学习函数/智能优化Key words
crane girder/Kriging surrogate model/EFF active learning/intelligent optimization引用本文复制引用
出版年
2024
NETL
NSTL
万方数据