Efficient Sparse Representation Response Surface Modelling and Simulation Method for Complex Products

2018 3rd International Conference on Advanced Materials Research and Manufacturing Technologies

材料科学;机械制造

Li, Pu^1 ; Huang, Changzheng^1 ; Hu, Songxi^1 ; Yang, Senquan^1 ; Wen, Haoxiang^1 ; Zhou, Lin^2

School of Physics and Mechanical and Electrical Engineering, Shaoguan University, Shaoguan, China^1

School of Education, Shaoguan University, Shaoguan, China^2

关键词: Electro-mechanical products;    Engineering practices;    Response surface approximation;    Response surface modeling;    Response surface modelling;    Response surface models;    Simulation efficiency;    Sparse representation;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/436/1/012005/pdf
DOI  :  10.1088/1757-899X/436/1/012005

学科分类：材料科学（综合）

来源: IOP

【 摘 要 】

With the composition and control of modern electromechanical products more and more complex, the time of solving simulation model is longer and longer. The response surface simulation optimization method is an effective method to reduce the simulation time, however, the number of design sampling points of response surface approximation is still large. In this paper, we propose a sparse representation response surface model to accurately reconstruct the source model with small amount sampling points. By means of the sparse representation of the source model on a specific basis, most of the coefficients are zero which can be solved by the equations constructed from small number of sampling. Sparse representation response surface models include sparse response surface and quasi-sparse response surface which are respectively applied to the case where the number of sampling points is greater than and less than the degree of sparse. Sparse response surface runs quicker and quasi-sparse response surface has higher accuracy. Two test functions and one engineering practice problem are employed to compare the performance between sparse representation response surface model and other common response surface models. The results show that the sparse representation response surface model has better performance in approximate accuracy and simulation efficiency.