基于SOM-PCA-RBF的过程质量预测与整体调优

为实现过程输出质量的预防性改进，提出一种将SOM(self-organized map)神经网络、PCA(principal component analysis)法和RBF(radial basis function)神经网络集成的过程质量预测和多因素整体调优方法。采用SOM神经网络对过程数据进行分类；采用主成分分析法对分类后的过程输出进行评价，建立过程因素备选库；采用RBF神经网络建立过程预测模型，通过预测判断过程输出质量的符合性，给出过程多因素整体调优的方案。实例分析结果表明，该方法能够有效实现预防性质量改进。

Process Quality Prediction and Overall Tuning Method Based on SOM-PCA-RBF

In order to improve process output quality, a method for process quality prediction and multi-factor integrated optimizing is proposed. This method combines self-organized map neural network (SOMNN), principal component analysis (PCA) and radial basis function neural network (RBFNN). SOMNN is used to classify the process data. PCA is used to evaluate the classified data and establish the process factors repository. RBFNN is used to establish the process prediction model, determine the conformity of the process output quality by predicting, and propose the scheme for multi-factor integrated optimizing. The case analysis result shows that the method is effective and can achieve preventive quality improvement.