(529a) Model-Based Fault Diagnosis and Fault Tolerant Control in Closed-Loop Safety-Critical Chemical Reactors: An Experimental Study

In recent years, the heightened emphasis on process safety has spurred significant advancements in fault detection and isolation methodologies within the chemical industry. Fault diagnosis (FD) techniques generally fall into two categories: model-based and data-driven approaches. For safety-critical chemical reactors, model-based methods using first-principles offer distinct advantages by generating physically interpretable residual signals derived from unclosed material and energy balances. These signals not only facilitate reliable fault diagnosis but also enhance understanding of system behaviour, outperforming purely statistical approaches.

We have experimentally demonstrates fault diagnosis in an open-loop Continuous Stirred Tank Reactor (CSTR) system, specifically applied to the alkylpyridine N-oxidation process^1,2. Model-based functional observers are employed to produce robust fault indicators and estimate fault magnitudes independently of reaction rate disturbances. Accurate and timely detection of cooling system faults is critical to mitigating safety hazards associated with the reaction's exothermic nature and the potential decomposition of hydrogen peroxide. Expanding on the open-loop findings, our research further examines fault detection and isolation within a closed-loop CSTR system³. Various control strategies, including adjustments to coolant flow rate and inlet coolant temperature, are investigated to maintain reactor temperature stability under rapid and frequent setpoint changes and varying fault conditions. We presented an experimentally validated approach demonstrating the effectiveness of closed-loop feedback control systems in detecting and estimating faults through the analysis of residual signals, thereby enhancing operational safety and reliability.

Based on these work, we then advance to achieve fault-tolerant control (FTC). Active FTC integrates fault detection (FD), a reconfiguration scheme, and an adaptable controller⁴. FD is critical to active FTC, enabling adaptive reconfiguration of the controller to accommodate new operating conditions by detecting, isolating, and estimating faults. The reconfiguration scheme in FTC is guided by specific criteria dictating when system adjustments should occur to ensure the reactor remains within safe operating limits. To rigorously evaluate system states and outputs against safety thresholds, the concept of the maximal admissible set (MAS) is employed⁵. This set defines the permissible boundaries for states, guiding reconfiguration decisions by indicating safe operating conditions and necessary intervention points.

In this work, we develop a DSS-based FTC strategy for a CSTR model, addressing the challenge of maintaining safety under both minor and major faults. The results show that the nominal controller effectively manages minor faults, maintaining the system within the DSS without further intervention. However, major faults exceed the capability of the nominal controller alone. To counteract this, we introduce a "Plan B" intervention strategy involving a controlled injection of cold solvent to expand the DSS and restore safety margins. Two practical criteria for activating Plan B, based on fault estimates and DSM thresholds, ensure precise and timely fault responses. Thus, accurate DSS computation facilitates effective FTC actions, significantly enhancing system resilience and safety under demanding operational conditions.

(1) Du, P.; Wilhite, B.; Kravaris, C. Model‐based Fault Diagnosis for Safety‐critical Chemical Reactors: An Experimental Study. AIChE J. 2024, 70 (12), e18565. https://doi.org/10.1002/aic.18565.

(2) Du, P.; Wilhite, B.; Kravaris, C. Model-Based Fault Diagnosis for Safety-Critical Chemical Reactors: An Experimental Study; AIChE, 2023.

(3) NASCRE-5 2025 || Houston, Texas || 16 - 19 February 2025. https://2025.nascre.org/view_paper.php?PaperNum=1047 (accessed 2025-03-31).

(4) Mhaskar, P.; Gani, A.; El‐Farra, N. H.; McFall, C.; Christofides, P. D.; Davis, J. F. Integrated Fault‐detection and Fault‐tolerant Control of Process Systems. AIChE J. 2006, 52 (6), 2129–2148.

(5) Kolmanovsky, I.; Gilbert, E. G. Maximal Output Admissible Sets for Discrete-Time Systems with Disturbance Inputs; IEEE, 1995; Vol. 3.