A Method for Fermentation Process Monitoring via Nonlinear PCA

摘要

A nonlinear PCA method based on an autoasso-ciative network and input training is pesented. The demapping subnet of the autoassociative network is trained by input training method. During training the whole network the weights are keeped and biases of the demapping subnet unadjusted. The results show this non-linear PCA method is a good tool to analyze batch process data.