Enhanced control in time-delay processes: Diagnostic, monitoring, and self-tuning strategies for the filtered smith predictor in response to model-plant mismatch and abrupt load disturbances

Abstract

The Filtered Smith Predictor (FSP) is a practical structure in industrial control systems, especially those with dominant time delays. Despite its importance, a notable gap exists in developing tailored algorithms for the FSP that can effectively diagnose model degradation and the presence of abrupt disturbances. This research addresses the challenge of distinguishing between model-plant mismatch (MPM) and unmeasured disturbances (UD), which is crucial to maintaining control system performance and stability. We propose a novel algorithm that diagnoses, monitors, and self-tunes the FSP structure. By leveraging the sensitivity transfer function, our method ascertains nominal closed-loop performance without MPM or UD and meticulously analyzes the real effects of these factors. A dynamic time window and an operating range, used as tuning parameters, enable a precise assessment of the model’s predictive capacity and the detection of UD. Additionally, the algorithm incorporates a self-tuning mechanism for the FSP’s robustness filter, responding adaptively to identified discrepancies and ensuring enhanced system stability. Demonstrated through detailed simulated case studies and a real-life temperature control experiment, our methodology significantly improves the diagnosis and self-tuning capabilities of FSP structures. The results highlight a marked enhancement in system stability and performance, providing our approach’s practical value and effectiveness in real-world industrial applications.