A hybrid framework for process monitoring: Enhancing data-driven methodologies with state and parameter estimation

Francesco Destro, Pierantonio Facco, Salvador García Muñoz, Fabrizio Bezzo, Massimiliano Barolo
2020 Journal of Process Control  
In this study we bridge traditional standalone data-driven and knowledge-driven process monitoring approaches by proposing a novel hybrid framework that exploits the advantages of both simultaneously. Namely, we design a process monitoring system based on a data-driven model that includes two different data types: i) "actual" data coming from sensor measurements, and ii) "virtual" data coming from a state estimator, based on a first-principles model of the system under investigation. We test
more » ... igation. We test the proposed approach on two simulated case studies: a continuous polycondensation process for the synthesis of poly-ethylene terephthalate, and a fed-batch fermentation process for the manufacturing of penicillin. The hybrid monitoring model shows superior fault detection and diagnosis performances with respect to conventional monitoring techniques, even when the first-principles model is relatively simple and process/model mismatch exists. (M. Barolo). measurements under abnormal process conditions. This issue is particularly relevant if the variables embodying the root-cause of the fault are not measured, and therefore cannot be included in the LVM. To overcome this limitation, monitoring methodologies exploiting first-principles knowledge about the process under investigation may be considered. Process monitoring methodologies based on knowledge-driven (KD) models have been thoroughly reviewed elsewhere [7, 8] . The most popular KD approaches are based on parity relations [9] or on state estimators [10] [11] [12] [13] [14] , possibly implemented for simultaneous state and parameter estimation [15] . Generally speaking, KD models have the advantage of embedding the available understanding on the mechanisms driving the process under investigation. This piece of information can help fault detection and diagnosis, and is missing in DD monitoring approaches. However, KD models are generally more complex to develop than their DD counterparts and, when used for monitoring, the performances can be severely affected by process-model mismatch. In addition, the fault models have typically to be known a priori [13, 14] . Hybrid models [16, 17] combine DD methods with the information available from first-principles knowledge about the process, and are promising techniques for overcoming the limitations of DD and KD monitoring [18, 19] . Hybrid models for process monitoring usually consist of a KD soft-sensing framework in which a DD component is added to make up for missing
doi:10.1016/j.jprocont.2020.06.002 fatcat:27mtpjtlhffmtl5ztb2j44j3h4