Research And Application Of Multi-objective Global Optimization Algorithm Based On Design Space Reduction

In recent years,environmental degradation and energy scarcity problems have followed and become increasingly prominent with the rapid development of the automobile industry.Based on the development needs of a conservation-oriented society,the lightweighting of automobiles is an important way to save energy and reduce emissions,improve the overall performance of automobiles,and promote the sustainable development of the automobile industry in the new era,both for traditional fuel vehicles and new energy electric vehicles.In the face of a large number of complex black-box design optimization problems with high-dimensional,highly nonlinear and large design spaces in the field of automotive lightweighting engineering,although the simple use of an optimization algorithm based on an agent model can effectively reduce the number of calls to the finite element simulation model,the modeling accuracy,optimization efficiency and computational cost still need to be reflected in the breakthrough.The advent of big data and information era also provides a brand new idea and channel to solve this bottleneck.Considering that the modeling accuracy and computational efficiency of the agent model are directly related to the design space,it becomes an active research branch of great significance to develop a design space reduction strategy based on data mining theory to further improve the comprehensive performance of the combined optimization algorithms.First,considering that the design space reduction strategy depends on the performance of the underlying optimization algorithm,the multi-objective EGO algorithm based on the Eulerian distance EIM criterion is improved with parallel strategies and nested genetic algorithms to improve its comprehensive performance.Inspired by data mining technology,a multi-objective global optimization method(FCMEIMOEGO)based on FCM fuzzy clustering design space reduction strategy is proposed,and the comprehensive performance of FCM-EIMOEGO is improved by testing and analyzing numerical cases such as ZDT series.Meanwhile,a multi-objective global optimization method(DT-EIMOEGO)based on the design space reduction strategy of C4.5 decision tree technology is proposed,and the improved algorithm is found to perform better in the two-objective optimization problem through numerical algorithmrelated tests,so it is applied to the T300/EXPOY composite single-layer plate parameter inverse,and it is found that the simulation response of the optimal parameter design solution obtained by optimization is approximately equal to the experimental response,and then it is verified that DT-EIMOEGO can efficiently identify the parameters of automotive lightweight materials,which has strong practical guidance significance for engineering.Finally,a multi-objective global optimization method(DTFCM-EIMOEGO)based on C4.5 decision tree technology and FCM fuzzy clustering with integrated design space reduction strategy is proposed to integrate the two design space reduction ideas into each other and to realize the optimization algorithm to explore the optimal search in the design space combined with dynamic and static reduction until convergence.The comprehensive performance of DTFCM-EIMOEGO in the high-dimensional optimization problem has significant advantages through numerical case testing,and it is also applied to the structural dimensional optimization design of the TRB B-pillar of the body under side impact conditions,with mass,intrusion and energy absorption as the objective functions,and finally the lightweight optimization design scheme of 10.94%mass reduction,13.25% intrusion reduction and 4.91% energy absorption increase is obtained efficiently to verify its engineering practical application value.