Sequence Iterative Optimization Based On Adaptive Surrogate Model And Its Application

With the development of modern construction machinery becoming more and more complex and comprehensive,traditional optimization design methods based on computer simulation models are gradually unable to meet the needs of engineering practice due to defects such as long design cycles.In this context,the surrogate model is used as a substitute for simulation model in engineering optimization design problems due to its advantages of small computation.Existing studies have shown that the optimization design method based on proxy model can significantly reduce the number of calls to time-consuming simulation model in the optimization process,thus significantly shortening the time spent in the optimization process.Therefore,it is considered to be one of the best methods to solve complex engineering optimization problem.However,in practical engineering application,surrogate model technology also reveals many problems to be solved.In this paper,an improved agent model optimization method is proposed based on the study of the agent model technology.The feasibility and effectiveness of the proposed method are verified by numerical tests commonly used in the field of agent model research and its application in the engineering example of piston optimization design.The main work of this paper can be summarized into the following three aspects:(1)This paper proposes an adaptive agent model selection method to solve the shortcoming that the existing surrogate model construction method is not universal and can not be applied to the optimization design problems with different characteristics.This method can adaptively select an agent model with the best performance for the optimization problem according to the complexity and sample size of the optimization problem.The results show that the final output model of the proposed adaptive agent model has better performance compared with the existing agent model construction methods,and has good applicability to the optimization design problems with different characteristics.(2)Combining the adaptive agent model method with the sequential iterative optimization of the agent model,a sequential iterative optimization method based on the adaptive surrogate model is proposed.In the process of optimization of this method,the optimization algorithm of each iteration after screening potential optimal solution is added to the sample space,at the same time the adaptive agent model method based on the updated sample space choice out of the current conditions of the optimal agent model is used for the next iteration calculation,combination makes the agent model with optimization of process and constantly improve the growing performance,on the premise of guarantee optimization results accuracy reduce optimization process takes time.The test results show that compared with the sequential iterative optimization based on the fixed agent model,the proposed sequential iterative optimization method based on the adaptive agent model has higher solution accuracy and improved optimization efficiency to a certain extent.(3)The method proposed in this paper is applied to the piston optimization design problem,and the feasibility and effectiveness of the method in practical engineering application are verified.The section size of the cooling oil chamber of the parameterized piston model was optimized by this method,and the weight of the piston was reduced under the condition of satisfying the constraints.The optimization results show that,although the solution precision of the proposed method is not as good as that of the traditional finite element model-based direct optimization,the time consumed by the proposed method is shorter and the optimization efficiency is significantly improved.