Multivariate statistical monitoring of multiphase batch processes with uneven operation durations

In multivariate statistical monitoring, batch process models should well reflect process characteristics in order to achieve satisfactory fault detection results. In manufacturing systems, many batch processes are inherently multiphase. Usually, process features are different from one phase to another, and gradual transitions are often observed between phases. Another important characteristic of batch processes is uneven operation durations. In multiphase batch processes, not only the entire batch durations but also the phase durations may be unequal from batch to batch. In this paper, the Gaussian mixture model (GMM) method is adopted to solve both the multiphase and the uneven-duration problems simultaneously. A benchmark penicillin fermentation process is utilized to verify the phase division, transition identification and process monitoring results based on the proposed method.