A real-time fault monitoring and diagnosis for batch process based on dynamic principal component analysis

Batch process monitoring methods based on multivariate statistics are mainly multiway principal component analysis (PCA), its problems are that monitoring process needs predicted data, unequal length process must be aligned on data processing and small batches of data can not modeled and so on. Therefore, this article proposes dynamic PCA modeling methods for batch process based on dynamic characteristics of the batch. The method uses time-lagged technology to regroup for each batch data of the model after obtaining procedure dynamic lag time constant, then all batches combination data make a whole, based on which the PCA monitoring is established. This article gives fusion algorithm for delay data diagnosing information redundancy problems. Ultimately it realizes real- time online fault monitoring and diagnosis. The simulation result shows that the proposed method is effective.