Research On The Optimization Design Method Of Complex Equipment Structure Based On The Adaptive Surrogate Model

Since the 20th century,the modernization of industrial development results the complexity and accuracy of industrial products is continuously improving.Therefore,many engineering design optimization methods have been developed,which dedicates to meet the product design requirements under limited costs of the mathematical calculation.When the efficient engineering optimization design method is introduced into the product design process,the product design cycle will be shortened and the product competitiveness will be improved.However,when solving engineering optimization problems,many objectives and constraints cannot be described explicitly in practical engineering and the optimal solution can only be obtained by calling time-consuming simulation models.Therefore,the optimization method based on surrogate model is proposed.By using surrogate model technology,the number of calling the real model could be greatly reduced and the efficiency of solving the problem can be accelerated.However,for the optimization method based on the surrogate model,its accuracy and computational efficiency largely depend on the sample information,moreover,a part of the surrogate model is also affected by its internal kernel function.Therefore,the optimization method based on adaptive surrogate model is studied in this thesis from the following aspects:firstly,an optimal design method based.on adaptive surrogate model is proposed,which adopts the multi-criteria sample updating and management strategy.Secondly,an optimal kernel function prediction method of surrogate model based on random forest is proposed to further improve the accuracy and computational efficiency of the optimal solution.Finally,the structure optimization software platform of bucket wheel machine forearm based on Visual Basic 6.0 is developed,which reduces the cumbersome operation process in the optimization design process and reduces the design burden of designers.The specific research contents of this thesis are as follows:1.An optimal design method based on adaptive surrogate model using multi criteria sample updating and management strategy is proposed.Firstly,the optimal Latin hypercube design(OLHD)is used to obtain the initial sample points,which can ensure the projection uniformity and spatial distribution of the sample points,thus the initial surrogate model could be constructed with fewer sample points and the most basic change trend of the original function can be reflected.Then,the approximate optimization problem is established based on the initial surrogate model.By using the function fluctuation index as the local evaluation index of the surrogate model,the character of the real function in the interested area is fed back to guide the updating direction of samples.Secondly,based on the sample update management strategy,the limited sample resources are concentrated in the interested areas,and the sample structure of the surrogate model is improved to avoid the ill-conditioned matrix.Finally,the effectiveness of this method is proved by two numerical examples and an engineering structure optimization design problem.2.An optimal kernel function prediction method based on random forest is proposed.By the full use of limited sample resources,the optimal kernel function type of surrogate model could be predicted.Firstly,the nonlinear degree of the sample set of the surrogate model is analyzed,thus the key feature parameters of the sample set are obtained by feature extraction which are used to construct the random forest classifier.Then,the accuracy of the random forest classifier is evaluated based on the "k-fold cross validation method",and the classifier is continuously modified for obtaining more accuracy predict result.Secondly,the feature parameters of the original sample set or the updated sample set of the adaptive surrogate model are extracted and input into the random forest classifier,thus the optimal kernel function type of this surrogate model could be accurately predicted.Finally,the effectiveness of this method is proved by two numerical examples and an engineering structure optimization design problem.3.A software platform based on Visual Basic 6.0 is developed for the structure optimization design of the bucket wheel machine forearm,which can not only reduce the tedious operation steps in the optimization design process,but also shorten the optimization cycle.Moreover,it has the characteristics of easy to learn and operate.The software design platform is applied to the optimization design of the boom structure of the bucket wheel machine,and the accuracy of the software platform is verified by the optimization results.