A Diagnosis Approach for Parameter Deviations in Linear System Using Artificial Neural Networks

A novel approach towards diagnosis of parameter deviations in linear system using artificial neural network based identification algorithm is proposed in this brief. The fault diagnosis is done by identifying the parameters of the system from the measurement of state variables of the system. The input excitation for the system is unit step and the artificial neural network which identifies the parameters of the system is trained using Widrow-Hoff training rule. The system is identified in discrete domain by using the state variables of the system at to consecutive sampling instants in a recursive procedure. The faults are introduced arbitrarily and the neural network identifies the parameter continuously from the measurement of state variables. The proposed method is feasible for on line fault diagnosis through digital implementation. The mathematical formulation of the technique and the simulation results are presented to validate the feasibility of the proposed approach.