Fault detection for multiphase batch process

In this paper, a fault detection method based on intermediate phase dependency analysis for multiphase batch process is proposed. The contributions of the proposed method are as follows: 1) multi-way kernel independent component analysis (MKICA) are used to extract the principal information; 2) the subspace separation is completed by establishing the relationship of intermediate phase using MKICA; after that, the relative changes (increased part, decreased part and unchanged part) from previous phase to latter phase are obtained; 3) the detection model establishing by the proposed approach is more accurately comparing with the conventional method. The advantage and effectiveness of the proposed method are illustrated with penicillin fermentation process.