Fault-Tolerant Control of Steam Temperature in HRSG Superheater under Actuator Fault Using a Sliding Mode Observer and PINN

TL;DR

The paper tackles robust regulation of HRSG superheater steam temperature under actuator faults modeled as multiplicative losses. It fuses a Sliding Mode Observer for unmeasured states, a Physics-Informed Neural Network for real-time fault estimation, and a one-sided Sliding Mode Controller to adapt control while minimizing actuation. Lyapunov analysis establishes uniform ultimate boundedness of estimation and tracking errors, and validation on real plant data demonstrates accurate fault estimation, state reconstruction, and improved temperature tracking compared with PID. The approach offers a practical, physics-guided framework for enhancing resilience in thermal power plants with multiplicative actuator degradation and is extensible to similar thermal systems.

Abstract

This paper presents a novel fault-tolerant control framework for steam temperature regulation in Heat Recovery Steam Generators (HRSGs) subject to actuator faults. Addressing the critical challenge of valve degradation in superheater spray attemperators, we propose a synergistic architecture comprising three components: (1) a Sliding Mode Observer (SMO) for estimation of unmeasured thermal states, (2) a Physics-Informed Neural Network (PINN) for estimating multiplicative actuator faults using physical laws as constraints, and (3) a one-sided Sliding Mode Controller (SMC) that adapts to the estimated faults while minimizing excessive actuation. The key innovation lies in the framework of closed-loop physics-awareness, where the PINN continuously informs both the observer and controller about fault severity while preserving thermodynamic consistency. Rigorous uniform ultimate boundedness (UUB) is established via Lyapunov analysis under practical assumptions. Validated on real HRSG operational data, the framework demonstrates effective fault adaptation, reduced temperature overshoot, and maintains steam temperature within 1°C of the setpoint under valve effectiveness loss. This work bridges control theory and physics-guided machine learning to deliver a practically deployable solution for power plant resilience, with extensions applicable to thermal systems subject to multiplicative faults.

Figures (11)

• Figure 1: HRSG in Pareh-Sar Powerplant
• Figure 2: A fully functional desuperheater nozzle produces a finely atomized spray into the steam flow (top left). A loss-of-effectiveness fault, caused by blockage or mechanical damage, reduces spray quality (top right) or allows undispersed water to enter the steam pipe (bottom), leading to thermal shock and water hammer damagegibbons_how_2025.
• Figure 3: A loss-of-effectiveness actuator fault in the desuperheater can lead to over-temperature conditions, water hammer, or thermal shock, resulting in severe damage to piping and equipment and potentially causing extended outages for repairgibbons_how_2025mukhopadhyay2011failure.
• Figure 4: A loss-of-effectiveness actuator fault can manifest through various forms of nozzle damage, each degrading spray performance. Examples include plugging (top left), out-of-spec spray patterns (top right), broken nozzles (lower left), and completely detached or blown-off nozzle tips (lower right) gibbons_how_2025.
• Figure 5: Schematic of the proposed SMO–PINN–SMC architecture. The Sliding Mode Observer (SMO) reconstructs unmeasured states; the Physics-Informed Neural Network (PINN) estimates the multiplicative actuator fault; and the Sliding Mode Controller (SMC) ensures fault-tolerant steam temperature regulation. The structure creates a closed-loop adaptive control scheme that responds to loss-of-effectiveness faults in real time.

Theorems & Definitions (7)

• Theorem 1: Uniform Ultimate Boundedness
• proof
• Lemma 1
• proof
• Theorem 2
• proof
• Corollary 1