Fault detection in nonlinear dynamical systems using multi-sensor data fusion based on Hybrid Extended Information Filter

An effective way to enhance the system reliability is to develop a fault detection algorithm to perform the monitoring task instantly. In a dynamic system, fault is defined as any deviation from a desired operating condition. According to system dimensions, there are different architectures to implement fault detection algorithm including centralized, decentralized and distributed. In this paper, a centralized approach is designed using multi sensor data fusion technique based on Hybrid Extended Information Filter (HEIF). This approach has the advantages of both existing algorithms, the Hybrid Extended Kalman Filter (HEKF) and the Information Filter (IF). Similar to HEKF, it has better performance compared to conventional Kalman filter and as the IF, it can be implemented non-centrally. The proposed centralized algorithm is more efficient for low-order nonlinear dynamic systems. It is also important for the high-order systems because it is the basis for performance comparison of non-central approaches. This approach not only enables us to distribute the algorithm for non-central schemes, but is also superior to the conventional Kalman filter in precision and computational burden with a same convergence speed which helps to move toward a real time implementation. It also acts more timely in fault detection task. In this work, in addition to improved results, we are going to establish a basis for further investigation in large-scale systems.