Failure Modes Detection of Nuclear Systems Using Machine Learning

Abstract

A control design scheme is formulated for a pressurized water type nuclear power plant by integrating the optimal linear quadratic Gaussian (LQG) control with the robust integral sliding mode (ISM) technique. A novel robust-optimal hybrid control scheme is further proposed by integrating the LQG-ISM technique with the loop transfer recovery approach to enhance the effectiveness and robustness capability. The control architecture offers robust performance with minimum control efforts and tracks the reference set-point effectively in the presence of disturbances. The multi-input-multi-output nuclear power plant model adopted in this work is characterized by 40 state variables. The nonlinear plant model is linearised around steady-state operating conditions to obtain a linear model for controller design. The efficacy of proposed controllers is demonstrated by simulations on different subsystems of the nuclear power plant. The control performance of the proposed technique is compared with other classical control design schemes. Statistical measures are employed to quantitatively analyse and compare the performance of the different controllers that are studied in the work.