Research And Application Of Sequence Iterative Optimization Method Based On Surrogate Model

The method of using surrogate models to replace time-consuming simulation software to predict the objective function has obvious advantages in actual engineering applications,that is,the number of simulations is greatly reduced,the optimization efficiency is improved,and the optimization cost is reduced.However,the accuracy of the surrogate model is the key to the optimization method based on the surrogate model.It is one aspect to use as few sample points as possible to improve the accuracy of the agent model;one aspect is to combine the intelligent algorithm optimization and the process of adaptive point addition of the agent model to simplify the optimization process.Based on the above two aspects,the main research contents of this article include:(1)A global and local search balance strategy is proposed.This strategy is based on the adaptive scaling of the distance constraint,combining the balance constant and the distance constraint,so that the balance constant and the distance constraint are dynamically adjusted according to the needs of the algorithm in different search stages.Variety.In the early stage of the search,the algorithm is biased towards global search;in the later stage of the search,the algorithm is biased towards local search.On this basis,the proposed CGU global optimization algorithm is given,and the algorithm combines the point-adding process of the proxy model and the algorithm optimization process,which simplifies the entire calculation process and avoids the overlapping of errors caused by repeated intelligent algorithm optimization.Finally,the CGU algorithm is compared with the two global optimization algorithms of MSG algorithm and EGO algorithm through the test functions of six different dimensions.The comparison result shows that the CGU algorithm is obviously better than the other two algorithms in terms of final accuracy.(2)The previously mentioned distance-constrained adaptive scaling method is expanded,and on this basis,the single-objective plus point criterion of the CGU algorithm is expanded to the multi-objective plus point criterion.At the same time,the traditional agent model optimization algorithm process based on the multi-objective plus point criterion is improved,and finally the multi-objective optimization algorithm MCGU is proposed.The MCGU algorithm can optimize the Pareto frontier of the objective function group while improving the accuracy of the surrogate model,simplifying the entire calculation process and avoiding repeated intelligent algorithm optimization and error stacking.Finally,a series of test functions verify the effectiveness of the algorithm.(3)Multi-objective structural optimization design of the flywheel.First,the flywheel structure and its main structure parameters are analyzed,and the flywheel structure parameters that can be used as design variables are obtained.After that,the flywheel is analyzed by finite element analysis to analyze the structural parameters of the flywheel that have a greater impact on the two goals of flywheel mass and maximum flywheel stress,and six flywheel unassembled dimensions are determined as design variables.Finally,the multi-objective optimization of the flywheel structure is carried out with the MCGU algorithm proposed in the previous article.The results show that the mass and maximum stress of the flywheel are reduced,which also verifies the effectiveness of the MCGU algorithm mentioned in the previous article.