Fault Prognosis and Simulation of Sensor via Hidden Markov Model

Fault prognosis of sensor is vital for measurement system, or even the whole system to work safely, maintenance and repair. We adopted HMM (Hidden Markov Model) to solve the problem of the sensor fault prediction, established the basic structure of sensor fault prognosis system and HMM model, used Bayesian Toolbox in Mat lab for simulation and data sample for training model parameters, obtained the reasoning initial model after modifying, and then got the optimal estimation sequence of states which can predict the current state after Viterbi decoding. The simulation result shows that the optimal estimation sequence meets the process of sensor degradation. Therefore this method is suitable for fault prognosis of sensor.