Taguchi与RBF神经网络在注塑翘曲参数优化中的应用

Application of Taguchi and RBF Artificial Neural Network to the Optimization of Warp Injection Parameters

本文在CAE仿真的基础上,采用田口试验设计(Taguchi)和径向基神经网络(RBF)技术对引起翘曲的塑参数进行了优化。结果表明:Taguchi技术可以在较少的试验次数的情况下,确定各因素对翘曲的影响程度,获取各因素最佳的水平组合;运用RBF建立翘曲产生的神经网络模型,可以预测各因素在不同水平组合下的翘曲变形程度,达到离线预测的目的。

In this paper, Taguchi and Radial Basis Function Artificial Neural Network (RBF) are applied for the optimization of injection molding parameters in sense of product warp. By Taguchi technique the best level combination of different factors can be obtained with fewer experiments. Then, the results obtained from Taguchi technique are used to train an RBF model. With the RBF trained, it can predict the optimal injection parameters off-line.