On-line batch process monitoring using modified dynamic batch PCA

In this paper, a new method incorporating batch process dynamics into PCA is proposed by introducing an exponentially weighted forgetting factor. The pro´s and con´s of the new method are evaluated by comparing it with hybrid- wise MPCA and batch dynamic PCA (BDPCA). Two well-known batch processes are used as examples and results suggest that the proposed m-BDPCA and traditional BDPCA method have similar fault detectability but the modified method is less computational demanding.