Detection of model-plant mismatch and model update for reaction systems using concept of extents. (December 2018)

• Record Type: Journal Article
• Authors: Kumar, D.M. Darsha; Narasimhan, Shankar; Bhatt, Nirav
• Is Part Of: Journal of process control. Volume 72(2018)
• Page Start: 17
• Page End: 29
• Publication Date: 2018-12
• Subjects: Fault diagnosis and identification -- Model-plant mismatch -- Reaction systems -- Model identification -- Extent of reaction -- Bank of Kalman filters; Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2018.08.005
• Languages: English
• ISSNs: 0959-1524

Abstract

Highlights: A framework for online model-plant mismatch and model update is developed for reaction systems. Model-plant mismatch due to structural and parametric changes in process models can be handled in the proposed framework. The framework uses the concept of decoupled extents of reaction and flow and incremental identification methods. Abstract: The performance of model-based control and optimization depends on the accuracy of process models. However, changes in physiochemical and operational conditions can result in a mismatch between the process and its model. This model-plant mismatch (MPM) must be detected and rectified quickly to achieve the desired performance. In this work, we consider model-plant mismatch due to structural and parametric changes in the underlying process model of reactor systems. We formulate MPM detection as a fault detection and identification problem. We propose an online model-plant mismatch detection and model re-identification framework using the concept of the reaction extents and incremental model identification for detecting and isolating the faults and appropriately re-identifying the faulty part of the model. The proposed approach is illustrated through simulation studies of acetoacetylation of pyrrole in a batch, semi-batch and continuous stirred tank reactor configuration for different fault scenarios.