Adaptive monitoring based on independent component analysis for multiphase batch processes with limited modeling data

An adaptive monitoring method based on multiphase independent component analysis (ICA), termed AMPICA, is proposed for batch processes of long duration. Starting from limited modeling batches, the local process correlation structures are explored and thus multiple phase-specific models are developed, where each phase pattern can be faithfully approximated by different sub-ICA models. Then an adaptive updating strategy is adopted to accommodate more underlying process information and normal batch-to-batch slow-varying behaviors with the accumulation of new batch data. The idea and algorithm are illustrated with respect to the typical data collected from a benchmark simulation of fed-batch penicillin fermentation production. The simulation results show that the proposed method provides a new feasible statistical analysis solution for modeling and monitoring problems with limited data in long-cycle batch processes.