Open AccessProblem Solving Strategy for Product Variant Design Combined with Improved Neural Network Method
Author(s) -
Lin Yang,
Xiaoqing Gu,
Qiuming Yin
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/825/1/012008
Subject(s) - artificial neural network , computer science , sample (material) , product (mathematics) , parametric statistics , artificial intelligence , machine learning , mathematics , chemistry , statistics , geometry , chromatography
Aiming at the fast solution of product variant design problem, a solving strategy combined with improved neural network method was proposed. Past similar cases were extracted as training samples through case-based reasoning (CBR) technology, empirical parameters were predicted based on neural network; computational parameters were solved by the existing calculation templates. The design results visualization was realized through parametric design. To reduce the influence of artificial sample construction on neural network training effect, the sample construction strategy of the multiple input/single output (MI/SO) structure was proposed to improve the parameter prediction quality. Finally, the feasibility and effectiveness of the proposed method was demonstrated by taking the design of a single-cylinder recuperator for example.