Influence matrix approach to fault diagnosis and controller tuning

Abstract

An experimental verification of the influence matrix approach to fault diagnosis and automatic tuning is proposed. The coefficients of the system transfer function are identified and the Jacobian of the coefficient vector with respect to physical parameters, termed the influence matrix, is computed. The influence matrix is used to detect and isolate faults. The optimal controller parameters are obtained using the influence matrix. It is assumed that the system is linear and time invariant, and the physical parameters enter the coefficient vector multilinearly. The proposed method is verified on a physical robot manipulator arm. >