The online fault diagnosis for batch processes has received increasing attention.The most popular method is multivariate statistical methods (MSM)

such as multiway principal component analysis (MPCA) and multiway partial least square (MPLS).Abnormal events can be detected by calculating the T2 and SPE values of the sensor data without the need to build first principal based mechanism models while fault diagnosis is realized by employing contribution plot.In industrial applications to batch processes

especially to multiphase batch processes，however

contribution plot analysis is not always accurate and reliable

and the rate of wrong diagnosis is unfavorably high.To solve the aforementioned problem

a novel fault diagnosis strategy was proposed by combining dynamic locus analysis (DLA) and online dynamic time warping (DTW).The efficiency and accuracy of fault diagnosis were improved.The utility of the new fault diagnosis strategy was demonstrated through the penicillin fermentation case study.