基于动态Kriging模型的起重机金属结构可靠性解耦优化方法  

Reliability Decoupling Optimization Method for Crane Metal Structure Based on Dynamic Kriging Models

作　　者：王凯[1] 范小宁[1] 余畅 吕召国 WANG Kai;FAN Xiaoning;YU Chang;LÜ Zhaoguo(Department of Mechanical Engineering,Taiyuan University of Science and Technology,Taiyuan 030024)

机构地区：[1]太原科技大学机械工程学院,太原030024

出　　处：《机械工程学报》2025年第2期384-394,共11页Journal of Mechanical Engineering

基　　金：山西省基础研究计划资助项目(20210302123212)。

摘　　要：针对基于静态代理模型的起重机金属结构可靠性优化设计双循环法的复杂性和耗时问题,提出基于多点加点策略动态Kriging代理模型的起重机金属结构可靠性解耦优化方法。该方法采用性能度量法中的改进均值法进行逆可靠性分析,通过计算偏移向量完成可靠性解耦,从而将复杂的可靠性优化问题转化为简单的确定性优化问题;同时在优化迭代过程中构建了U学习函数与最优点加点的多点加点策略和约束边界与最优点加点的多点加点策略,抽取更有效的高保真样本点来动态更新Kriging代理模型,使优化建立在不断更新的Kriging代理模型上,优化结束的同时模型精度也达到了要求。最后通过实例验证表明,所提出的方法简化了可靠性优化过程,提高了优化效率,为基于代理模型的起重机金属结构可靠性优化设计提供了一个新的有效方法。In view of the complexity and the time-consuming problem of the double-loop method of the reliability-based design optimization(RBDO)for crane metal structures(CMSs)based on static surrogate model,a reliability decoupling design optimization method for CMSs based on dynamic Kriging agent model with multi-point addition strategy is proposed.In this work,the improved mean value method of the performance measurement method is adopted for inverse reliability analyses,and the reliability decoupling is performed by calculating the offset vector,which render the complex RBDO problem transform into a deterministic optimization one.At the same time,during the optimization iteration process,the multi-point addition strategy of U learning function and optimal point addition and the multi-point addition strategy of constraint boundary and optimal point addition are constructed to extract more effective high-fidelity sample points to dynamically update the Kriging agent model,so that the optimization is built on the continuously updated Kriging agent model,and the optimization ends while the model accuracy meets the requirements.Finally,verification by example shows that the proposed method simplifies the procedure of RBDO,improves the optimization efficiency and provides a new effective method for the RBDO for CMSs based on the agent models.

关 键 词：起重机金属结构 偏移向量 可靠性解耦 动态Kriging代理模型 多点加点策略 

分 类 号：TH215[机械工程—机械制造及自动化]