Design of Robust Approach for Failure Detection in Dynamic Control Systems

Introduction

The application of modern control theories plays an important role for the improvement of the dynamic performance and safety of nuclear reactors (Subramaniam & Rajakumar, 1995). The model equations of the nuclear reactor are nonlinear. Conventional PI controllers may require several tuning adjustments to get satisfactory performance. Nonlinear control strategy is a must for the control of nuclear reactors. Input-output linearizing controller for a nuclear reactor is designed which shows that the performance of the closed loop system is good. However, it has not evaluated the robustness of both system parameters perturbations and component failure detection (Guimara & Lapa, 2004, 2005). The operation of any industrial plant is based on the readings of a set of sensors. The ability to identify the state of operation, or the events that are occurring, from the time evolution of these readings is essential for the satisfactory execution of the appropriate control actions. In supervisory control, detection and diagnosis of faults, adaptive control, process quality control, recovery from operational deviations, and determining the correct mapping from process trends to overcome some problems of operational conditions are the pivotal task of (Roverso, 2000).

Control systems are becoming more and more powerful and sophisticated. Reliability, availability, and safety are primary goals in the operation of the process systems (Odgaard & Thøgersen, 2010). The aim is to develop a fast and reliable control system that could detect undesirable changes in the process (referred to as “faults”) and isolate the impact of faults has been attracting much attention of researchers. Various methods for fault detection and control of process systems have been studied and developed over recent years (Staroswiecki, 2005; Zhang & Jiang, 2003; Ducard & Geering, 2008; Li & Parker, 2007) but there are relatively few successful developments of controller systems that can deal with faults in stochastic hybrid sense where faults are modeled as multiple-model set with variable structure and use of a stochastic model predictive control algorithm. Faults are difficult to foresee and prevent. Traditionally, faults were handled by describing the result behavior of the system and were grouped into a hierarchical structure of fault model (Li & Parker, 2007). This approach is still used for some fields in practice. When a failure occurs, the system behavior changes and should be described by a different mode from the one that corresponds to the normal mode. A more appropriate mathematical model for such a system is the so-called stochastic hybrid approach. It differs from the conventional hierarchical structure in that its state may jump as well as it may vary continuously. Apart from the applications to problems involving failures, hybrid systems have found some success in such areas as target tracking and control that involve possible structure changes (Zhang & Campillo, 2005). For detecting the faults, a critical assumption, in the simulation test runs, which is the availability of the system component models under different input excitations. Such models may not be available for the necessary input conditions in an operating system (Sarkar & Yasar, 2008; Gupta & Ray, 2008).