Optimization And Application Based On Ensemble Of Surrogate Models

In the engineering field,it is often faced with the need to conduct a large number of engineering experiments or physical simulation experiments to model or optimize the design of real engineering problems.Especially in the optimization design,a large number of iterative calculations are required.The introduction of surrogate model greatly reduces the amount of calculations and improves engineering design efficiency and optimization efficiency,but there are still problems such as large number of samples,low robustness,and sometimes difficult to find the real optimal solution when combined with intelligent optimization algorithms,and low engineering applicability.This article starts research from the following aspects,the main research contents are:(1)At present,there are few infill strategies that are applicable to different types of surrogate models.In view of the common characteristics of different types of surrogate models,an infill strategy based on local error expectation points is proposed,and the effectiveness and versatility of the strategy is proved by classic test examples.In addition,because there are few researches on combining the ensemble of surrogate models with the infill strategy,based on the proposed infill strategy,an adaptive ensemble of surrogate models construction method based on local error expectation addition is proposed.Then it is compared with the three most widely used strategies for constructing the ensemble of surrogate models through the classic test functions.The results show that the method has higher prediction accuracy and robustness,and the palletizing robot driving arm is used as the engineering case study object,which also shows that the effectiveness of the proposed method further proves the engineering applicability of the method.(2)Nowadays,the most widely used particle swarm optimization algorithm with linear decreasing inertia weight has problems such as easy to fall into local optimum and premature.Therefore,two major improvements about the initialization of the group and the change of inertia weight have been made to it,and then an improved particle swarm optimization algorithm has been designed.Through the comparison experiment between the classic test functions and the existing intelligent algorithm,the result proves the superiority of the proposed improved particle swarm optimization algorithm.Then based on the excellent performance of the proposed improved particle swarm algorithm,on the basis of the proposed algorithm,an improved particle swarm optimization algorithm based on the ensemble of surrogate models is proposed.It is compared with the existing currently commonly used optimization algorithm based on the surrogate model through the classical test functions.The results prove that the proposed algorithm requires fewer sample points and has higher prediction accuracy,which improves the optimization efficiency to a certain extent.(3)In order to solve the problems in the cooling and heat dissipation of automotive lithium battery packs,the optimization design goal is to minimize the maximum temperature difference of the lithium battery pack.First,from the aspects of structural rationality,fluid mechanics and heat transfer,two types of liquid cooling channel structures are designed for automotive lithium battery pack cooling and heat dissipation systems.The thermal management system selects the serpentine cooling channel as the design structure through FLUENT thermal simulation contrastive analysis,and then optimizes the selected serpentine cooling channel.The optimization algorithm proposed in the article is used to optimize the structure.The results show that the maximum temperature difference has been reduced to a certain extent,which makes the temperature distribution of the automobile lithium battery pack more uniform,and further illustrates the engineering applicability of the proposed improved particle swarm algorithm based on the adaptive ensemble of surrogate models.