SLIDING MODE OBSERVERS FOR OBSERVING THE DYNAMICS OF NUCLEAR REACTOR SYSTEMS

Abstract

ABSTRACT The behaviour of the control system under parameter uncertainties is of utmost importance in industrial applications. For designing a robust control system, a necessary effort has to be carried out for achieving desirable control characteristics irrespective of the operating conditions. With the shortcomings in the design of a conventional controller, a theory with proper mathematical background is of imminent need for a control designer. The application of a high speed switching with a suitable switching logic provides attractive feature of possessing a new property which is not present in the substructures. This fundamental idea developed into Variable Structure theory. Having derived from variable structure, sliding mode control is the operational tool of variable structure control. The design of sliding surface is important for control of the system during sliding mode. The distinguishing feature of sliding mode is the independence of the system against model dynamics which testifies the robustness with respect to parametric uncertainties, bounded disturbances and noise. Numerous applications of sliding mode in controller design is already proposed. The requirement of all states of the system for effective control was an inherent drawback with respect to practical system. Estimation of non-measurable states using the input and information of any measurable state is the fundamental idea of an observer. Luenberger observer for estimation of states is not robust to handle parametric variations and nonlinearities present in the system. Hence, the fundamental idea of sliding mode for observer design is used. This thesis describes the general framework for the design of sliding mode observer with focus on nuclear reactor system. Point Reactor Kinetics (PRK) model which is derived from the neutron diffusion equation is the model to be considered. With only nuclear reactor power as a measurable quantity, the delayed neutron precursor concentration, reactivity, external neutron source, xenon and iodine concentration is estimated. Out of large number of applications of sliding mode, fault detection in a nuclear and non-nuclear component with actuator and sensor faults is carried out. Residual evaluation is carried out using sliding mode observer while validating the residual signal at different instances where fault occurs. The inherent drawback of chattering due to formulation of first order sliding mode is minimized by using smooth functions. This method would forgo the fundamental idea of robustness of sliding mode. An attempt is made to present the application of higher order sliding mode which minimizes or sometimes eliminates chattering without compromising the robustness. The state estimation of a nuclear reactor system using Super Twisting algorithm is in good agreement with simulated state from PRK model. The independence of states with respect to initial conditions along with faster convergence time is achieved by the application of uniform second order sliding mode algorithm validated using the PRK model.