Sensor fault detection and estimation for heat exchanger using Unscented Kalman Filter

The main objective of this paper is to describe the application of Unscented Kalman Filter (UKF) estimation algorithm to the problem of fault detection and estimation of the sensor failures in the heat exchanger systems. The proposed fault detection is based on innovation sequence implementing UKF which is used to generate residuals when a sensor fault is presented. By evaluating the generated residual, it is possible to see if any problems of sensor have occurred. And the estimated state and decision function are built; the indication of sensor malfunction is achieved by UKF, so makes it possible for us to see where the fault is. Experiments in a heat exchanger have demonstrated the power of this method. The proposed algorithm has evident advantages such as less computation, easy programming, easy application and generalization and so on.