Abstract

The field of fault detection and isolation (FDI) has attracted much attention in control theory during the last three decades which has resulted in development of sophisticated FDI algorithms. However, increasing the complexity of FDI algorithms is not necessarily feasible. Particularly for on-line FDI, the FDI unit must have the minimum possible computation cost to prevent any long delays in fault detection.
In this research, we try to address the FDI problem of a single spool jet engine by using a modified linear multiple model (MM). We first develop a novel symbolic computation-based method for linearization purposes such that the obtained linear models are subjected to the symbolic fault variables. By substituting certain values for these symbolic variables, one can obtain different linear models, which describe mathematically the healthy and faulty models. In order to select the operating point, we use this fact that for a given constant fuel flow (W_f), the system reaches a steady state, that is varying for different values of W_f. Therefore, the operating points for linearization can be determined by the level of the Power Level Angel (PLA) (different values of W_f). These operating points are selected such that an observer, which is designed as a candidate for the healthy mode, can accurately estimates the states of the system in healthy scenario and the number of false alarm then would be kept to minimum. If the system works at different operating points one can then discretize the W_f into different intervals such that in each interval a linear model represents the behavior of the original system. By using the obtained models for different operating points, one designs the corresponding FDI units.
Second, we provide a modified multiple model (MM) approach to investigate the FDI problem of a single spool jet engine. The main advantage of this method lies in the fact that the proposed MM consists of a certain set of linear Kalman filter banks rather than using nonlinear Kalman filters such as the Extended Kalman Filter which requires more computational cost. Moreover, a hierarchical structural multiple model is used to detect and isolate multiple faults. The simulation results show the capability of the proposed method when it is applied to a single spool jet engine model.