基于遗传算法优化神经网络的产品造型设计评价

Evaluation of product form design based on BP neural network trained by genetic algorithm

为全面科学地评定、筛选最佳设计方案,借助遗传算法的全局寻优能力对BP神经网络进行优化,构建混合GABP算法,并将其应用于笔记本产品造型设计评价中。在建立笔记本造型二级指标评价体系的基础上,利用18款设计方案中的15款对混合GA-BP评价系统进行训练,利用其余3款设计方案对训练后的系统进行验证。验证结果表明,模拟值与实际值的相对误差分别为3.6%、-1.7%和2.8%,显示了较高的精度,反映了混合GA-BP产品造型设计评价系统能对设计方案进行科学的评价。

To carry out all-sided and scientific evaluation and screen the best design scheme,a hybrid GA-BP algorithm was established based on the global optimization of genetic algorithm to optimize the BP neural network,which was applied in evaluating the laptop modeling design.Based on the second indicator evaluation system of laptop,15 out of the 18 design schemes were selected to train the hybrid GA-BP evaluation system,and the residual 3design schemes were used for validating the trained system.The validation results show that the simulated and the actual scores of the 3design schemes were close with a higher accuracy（relative error was 3.6%,-1.7% and 2.8%,respectively）,reflecting the hybrid GA-BP product modeling design evaluation system can scientifically evaluate the design schemes.