Research On Multi-Objective Robust Optimization Methods Based On Sequential Kriging And SVM

Design optimization problems of complex engineering products are often multi-objective,multi-constrained,and may have uncertainties in their input variables/parameters.This uncertainty is continuously transmitted and accumulated in the design optimization problem,which causes the objective functions to change and shift,and even makes the feasible optimal solution unfeasible.Therefore,it is important to obtain robust optimal solutions through multi-objective robust design optimization.However,in the process of design optimization of complex engineering products,it is often necessary to obtain optimal solutions by time-consuming simulations,resulting in too high computation cost and long solution cycle.In order to improve the efficiency of computational solution,this paper takes the robust design optimization of complex engineering products as the research object,considers introducing the surrogate model and classification model to replace objective and constraint functions of the design optimization problem,respectively,and studies how to choose the appropriate model and sequential modeling method.The specific research work is as follows:Firstly,this paper proposes a multi-objective robustness design optimization method based on Kriging and SVM.Using the multi-objective robust design optimization based on constraint cuts as the framework,a Kriging approximation model is constructed for each objective function to predict its response value,and the SVM classification model is constructed for all constraint functions,according to whether the design alternative is in the feasible domain(i.e.whether all constraint functions are satisfied)divides it into two categories: feasible and infeasible.The proposed method is compared with typical multi-objective robust design optimization methods by two numerical examples and an engineering case of structure design optimization of micro-aerial vehicle fuselage.The results show that the proposed method can significantly reduce the computational cost and improve the efficiency of robust design optimization while ensuring that the obtained Pareto solution set satisfies the robustness requirements.Secondly,the Kriging model used to approximate the objective function,is sequentially updated by maximizing the expected improvement to iteratively find the updating point.The relationship between local optimization and global optimization can be effectively balanced.Then a multi-objective robust design optimization method based on sequential Kriging and SVM is proposed.The example verification results show that the proposed method can further reduce the computational cost under the premise of guaranteeing the quality of the obtained Pareto solution set.The proposed method is also applied to the robust design optimization of complex engineering products,and the engineering applicability and superiority of the proposed method are verified.