Font Size: a A A

Reserch On United Optimization Strategy Based On SiPESC Platform And Tensorflow Frame

Posted on:2022-09-17Degree:MasterType:Thesis
Country:ChinaCandidate:D Y JiaFull Text:PDF
GTID:2492306509979129Subject:Engineering Mechanics
Abstract/Summary:PDF Full Text Request
With the development of computer science,the application of computer-aided engineering is more and more extensive,and the use of optimization design becomes more and more important.The purpose of structural optimization design is to get a reliable optimal solution that has a more economical,safe,and stable structure.Therefore,the research of united optimization process based on machine learning method is constructed to explore the applicability of the strategy.Firstly,based on the existing experimental design model and optimization model in SiPESC.OPT,a united optimization strategy that introduces open-source machine learning algorithm libraries is established in this paper.By the employment of SiPESC.OPT and Python script language,a friendly optimization environment is created to provide a more convenient and efficient possibility for structure and multidisciplinary coupling optimization.And then,the correctness of the regression analysis and the operation mechanism of the optimization process are demonstrated by numerical examples and typical finite element examples.Based on SiPESC.OPT and tensorflow framework,a joint optimization strategy is proposed.The accuracy of the prediction model for nonlinear problems is verified by the regression calculation of Three-Hump-Camel,Griewank and Branin functions.The strategy is applied to the optimization of simple built-up beam structure,which proves the feasibility of the method in engineering practice.Further more,the application of the united optimization strategy of SiPESC.OPT is verified by solving the finite element optimization of complex structure.And through these engineering examples,the efficiency and practicability of the strategy are demonstrated.For bracket structure,Using neural network training to obtain the corresponding relationship between structural frequency,maximum stress,mass,and opening location information,and the shape optimization design is completed.According to the location and size of stiffeners in complex modeling environment,the optimal frequency optimization results can be obtained quickly in the proposed optimization strategy.As for the optimization of power station maintenance platform,the laying position and section size of primary and secondary beams are flexibly analyzed to achieve the goal of lightweight.Finally,the strategy is extended to the optimal solution of the trajectory planning problem,and the optimal initial state of the five-bar mechanism and the initial position of the slider-crank mechanism are calculated respectively.The solution is based on the standard experimental design table.After exploring the appropriate network structure,it interacts with the optimization calculation module,which further proves the convenience and portability of different problems.
Keywords/Search Tags:SiPESC, Machine Learning, Optimization Design, Python
PDF Full Text Request
Related items