kNN Fault Detection Method for Batch Process Based on Principal Sample Modeling Upgraded Online

Batch process has some problems： a high volume of training data,strong repeatability,a lot of noise interference,low data utilization and so on. To solve these problems,we proposed a kNN fault detection method based on principal sample modeling (PSM) upgraded online. The original data space is compressed through the selection of the principal sample based on the analysis of covariance,correlation coefficient,variance, and other statistical characteristics of the original data. When new samples are collected,they are put into the principal sample model to realize the updating of the principal sample space. Then,the fault detection model is built based on the principal sample space using the k-nearest neighbor (kNN) method. The effectiveness of the proposed method is illustrated by applying it to the multiphase semiconductor process. The experimenal results verify the effectiveness of the method.