BATCH PROCESS MONITORING USING MULTIBLOCK MULTIWAY PRINCIPAL COMPONENT ANALYSIS

Batch process monitoring to detect the existence and magnitude of changes that cause a deviation from the normal operation has gained considerable attention in the last decade. There are some batch processes that occur as a single step, whereas many others include multiple phases due to operational or phenomenological regimes or multiple stages where different processing units are employed. Having a single model for all different phases/stages with different covariance structures may not give a sufficient explanation of the system behavior and fault detection and diagnosis can be more challenging with increasing model size. Multiblock methods have been recently proposed to improve the capabilities of the existing statistical monitoring models. In this study, a multiblock algorithm based on concensus principal component analysis is applied to the benchmark fed-batch penicillin fermentation simulator data. The results of a static multiblock model and a sliding window multiblock model are compared. The need for data synchronization, and the effect of block size are discussed. Multiblock multiway principal component analysis methods are found to be effective in fault detection and localization.