Research Of Multidisciplinary Collaborative Optimization And Its Application To Product Design

Currently,engineering product design system involves multiple objectives and multiple disciplines which always act and couple with each other.To obtain the design scheme of the best product overall performance,multidisciplinary design optimization(MDO)arises and becomes a hot research topic at home and abroad.While,how to'efficiently coordinate the coupling effect among disciplines and reduce the huge computational cost are the two key factor for the application to the engineering practice.For the theoretical research of MDO method,this paper focuses on the study of the Collaborative Optimization(CO)method to improve its performance.Combined with the Kriging approximation model,the computational complexity of the collaborative optimization is effectively reduced.Two optimization examples are tested for these methods.And the results prove that the methods proposed in this paper are feasible and effective.Firstly,as the collaborative optimization method using quadratic structure of the consistency constraints,it is difficult to get the computational convergence or easy to fall into local convergence.This paper proposed Modified Collaborative Optimization(MCO)method.The design variables is decomposed into system-level design variables and subsystem-level local design variables.The objective function of the system level is expanded as the first order Taylor series.An expression is deduced to represent the overall influence of subsystem design variables on the system-level objective function,replacing the original consistency constraints as the objective function of subsystem level,which avoids the problem of the convergence difficulties.A classic coupling function is tested as an example to compare MCO method with the standard CO method.The optimization results show that the MCO method is stable and efficient.Secondly,although MCO method improved the convergence performance of the collaborative optimization method,it is inevitable that MCO method has its limits:due to the complex objective function of engineering design problems may not conductivity,multimodal or discreteness,the objective function of the derivative information relating to the design variables is difficult to obtain or does not exist.In order to solve this problem,this paper presented a new collaborative optimization method fused global approximation with local accuracy(FCO).The method has double-phase characteristics which contains the global approximation optimization phase and local accurate optimization phase.Through geometric analysis method,this paper discusses the features of the dynamic slack variable method and fixed slack variable method respectively.Dynamic slack variable method was applied to the global approximation optimization phase using the particle swarm optimization algorithm(PSO),which ensures the adequate feasible solution in the system optimization process,solves the problem that the CO method optimization result is sensitive to the initial point.Local accurate optimization phase using smaller fixed relaxation factor,sequential quadratic programming method(NLPQL)as the optimization algorithm,quickly and accurately converge to global optimal solution.Through the gear reducer test case,it is verified that the FCO is stable and accurate.Thirdly,Kriging approximation model based on Latin hypercube experimental design optimization method is applied to the collaborative optimization method fused global approximation with local accuracy,which simplifies its system analysis model,improves the efficiency of optimization,reduces the computing cost.Equally,the example of gear reducer is tested to compare FCO method with FCO method based on Kriging approximation model.The optimization efficiency of the FCO method has been improved significantly.Finally,a parameter optimization design of ship is tested as an example of the MCO method,and the FCO method based on Kriging approximate model is employed in the optimization design of disc brake.Results indicate that the two proposed methods can coordinate the coupling effect among disciplines efficiently,reduce the computational cost of MDO method considerably and be applied to the practice engineering product design successfully.