基于近似技术的序列化多学科可靠性设计优化  

Sequential multidisciplinary reliability design and optimization based on approximate techniques

作　　者：刘成武 钱林方[2] 花海燕 刘继红[3] 

机构地区：[1]福建工程学院机械与汽车工程学院,福建福州350118 [2]南京理工大学机械工程学院,江苏南京210094 [3]北京航空航天大学机械工程及自动化学院,北京100083

出　　处：《计算机集成制造系统》2018年第1期107-116,共10页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(51175019);福建省自然科学基金资助项目(2018J0106)~~

摘　　要：为提高多学科可靠性设计优化的计算效率,集成近似技术与解耦思想,提出一种基于近似并行子空间优化策略(ACSSO)和角度更新功能测度法的序列化多学科可靠性设计优化方法。基于解耦思想将其分解为确定性多学科优化与可靠性分析顺序执行的过程,避免每次优化对整个可靠性模型的反复计算。提出基于ACSSO策略进行确定性多学科设计优化,采用近似灵敏度信息替代实际灵敏度值,仅计算灵敏项构建全局灵敏度方程,避免了每轮都对学科级和系统级灵敏度信息的计算。基于角度更新策略对功能测度法进行改进,以极限状态函数的搜索角度值替代函数值进行可靠性评估,减少了耦合状态变量与多学科分析次数。以汽车侧撞工程设计为实例,验证了所提方法的有效性及工程实用价值。To improve the computational efficiency of Reliability-based Multidisciplinary Design Optimization（RBMDO）,a sequential RBMDO method based on an Approximate Concurrent Subspace Optimization strategy（ACSSO）and an Angle updating-based Performance Measure Approach（APMA）was proposed,which integrated decoupled idea and approximate techniques.This method decomposed RBMDO into a series of Deterministic MDO（DMDO）and multidisciplinary reliability analysis,which avoided the repeatedly calculations of reliability analysis model.DMDO could be implemented by ACSSO,in which the actual sensitivity was replaced by approximate sensitivity information and only calculated the sensitive index to model the global sensitivity analysis.As a result,all the non-sensitivity terms of discipline and system level were avoided.In addition,the efficiency of multidisciplinary reliability analysis was improved by APMA,in which the function value was replaced by angle updating of limit station function searching to reduce the calculations of coupling state variables and multidisciplinary analysis.The efficiency and engineering applications of proposed method was verified by an example of vehicle side impact design.

关 键 词：近似并行子空间优化 功能测度法 角度更新 多学科设计优化 可靠性分析 

分 类 号：TB114.3[理学—概率论与数理统计]