REAL-TIME PERFORMANCE MONITORING BASED ON ARMA MODEL ESTIMATION

Abstract A real-time performance monitoring scheme to display the status of a system based on estimating an ARMA model of the measured signal is proposed. The signal model is estimated using the Linear Predictive Coding Algorithm (LPCA) based on the numerically robust singular value decomposition (SVD) technique. The information is displayed in decreasing order of importance but increasing amounts of computational burden such that the essentials are known in the shortest possible time with the complete picture available after some time. The information obtained is stored in a knowledge-base which is used for incipient fault detection and detection of overloads on the system. The proposed monitor is evaluated on an actual position control system that consists of a D.C. armature controlled motor controlled by a ‘PID’ controller.