Application of Multi-Way Principal Component Analysis on Batch Data

In this paper, we propose the use of Multi-way Principal Component Analysis (MPCA) to classify batch operation data as normal or abnormal. MPCA also helps in isolating the batches having higher noise levels than the nominal conditions and the variables that deviate significantly from the nominal trajectory. Additionally, this paper proposes a novel approach of generating data for building the robust MPCA model based on limited number of batch run data. The paper successfully demonstrates the application of proposed method using data from a milk pasteurization process.