Abstract: A new method of multi-mode partition in the batch processes was presented on the basis of comparing the similarity of principal angles， which effectively overcomes the defects of noise or redundant data impaction on the phase partition. The main idea is to set up the principal component models of the batch data along time axis by employing the principal component analysis (PCA) approach， and then uses the principal angle method to compare the similarity of the principal component models， finally identify the steady phases and the transition phases effectively. Furthermore， an improved strategy is proposed to complete the partition criterion of the principal angle similarity. The simulation results demonstrate the effectiveness of the proposed method.

Key words: multi-mode partition, batch process, principal component analysis (PCA);principal angle, subspace similarity