Multidisciplinary Reliability Optimization Design Based On Convex Model Considering Tolerances

Multidisciplinary Design Optimization(MDO)is an effective method to solve the design problems of large-scale and complex coupled engineering systems.There are often various uncertainties in complex engineering problems.Therefore,it is necessary to consider these uncertain factors in the process of engineering optimization design.Based on reliability,multidisciplinary design optimization that takes uncertainty into consideration has received a lot of attention.Traditionally,probabilistic methods can deal with reliability problems,but it is sometimes difficult to obtain enough samples in engineering practice and these sample information also contains some subjective factors.Convex model is a non-probabilistic uncertainty theory,and it only needs the change scope of uncertain information.Convex model does not need to consider its distribution law,so it has strong engineering applicability in multidisciplinary reliability optimization design.The current reliability-optimized design based on the convex model does not consider the uncertainty caused by the design variable tolerance.In reliabilityoptimized designs which treat geometric dimensions as design variables,tolerances are usually considered after finding the optimal solution.The optimal solution obtained by the optimization design often falls on the constraint boundary.When the design variables fluctuate due to the uncertainty caused by the tolerance,the optimal solution will change near the constraint boundary.The optimal solution may fall into the failure area.Therefore,when designing the reliability optimization,it is significant to consider the tolerance of the design variables.In view of the above problems,on the base of the reliability optimization design based on the convex model,this paper considers the design variable tolerance besides the research on the reliability optimization design based on the convex model is further studied and discussed.The main contents are:1)Non-probabilistic convex model is used to describe uncertainty and a reliabilityoptimized design model and method based on convex model considering tolerances is proposed.Optimization and reliability assessment are separated by a framework of reliability assessment and sequence optimization.Deterministic optimization and reliability analysis are done by sequential loops.In order to obtain accurate tolerances,the branch-and-bound method in discrete-variable optimization methods is used in deterministic optimization.The performance measurement method is used in reliability analysis and the engineering example of pressure vessel is used to verify the effectiveness of the method.2)The multidisciplinary reliability optimization design model and method based on convex model considering tolerances is presented.Reliability analysis is separated from multidisciplinary optimization by SORA framework.The reliability analysis and multidisciplinary optimization are transformed into deterministic multidisciplinary optimization.Multidisciplinary approach(MDF)is used to do the reliability analysis,and collaborative optimization(CO)is used to do the multidisciplinary optimization.The feasibility of the method is verified by an engineering example of a compression cylinder coil spring.The results show that the optimization results obtained in this paper are still in a feasible domain with the influence of uncertainties.The optimization result meets the reliability index requirements and has a certain engineering application value.