An Online Gas Path Fault Diagnosis for Aircraft Engine Transient Behavior Using iB-EKF Algorithm

• Published in: IEEE transactions on aerospace and electronic systems Vol. 61; no. 2; pp. 2659 - 2678
• Main Authors: Zou, Zelong, Zhou, Xin, Huang, Jinquan, Jiang, Wenlong, Lu, Feng
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.04.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Aerospace engines; Aircraft; Aircraft engine dynamic behavior; Aircraft engines; Aircraft propulsion; Algorithms; Damping; Engines; Equilibrium equations; Estimation; Extended Kalman filter; Fault diagnosis; Flight envelopes; gas path fault diagnosis; Hidden Markov models; incomplete sensor measurements; Iterative methods; Jacobian matrices; Mathematical models; Noise measurement; Real time; selftuning damping coefficients; State estimation; Vectors
• ISSN: 0018-9251; 1557-9603
• DOI: 10.1109/TAES.2024.3478190

Online state estimation of aeroengine dynamic performance plays an important role in gas path fault diagnosis. However, it is hard for nonlinear state estimation algorithms to be applied online due to the large consuming time, and the linear algorithms are poor of the diagnosis accuracy as the plant feature is strongly dynamic and nonlinear. Hence, an online gas path fault diagnosis method (iB-EKF) is proposed, and it involves a component level model (CLM) without iterative computation and a devised online nonlinear state estimator with the combined structure. The extended Kalman filter is applied to diagnose gas path faults in combination with the single rank inverse Broyden (Rank-1 iB) algorithm, which is used to generate Jacobi matrices to avoid multiple calls to CLM. To reduce the calculation time of CLM, the iteration calculations in CLM are replaced by the EKF outside CLM. Therefore, pressure ratios and the residuals of equilibrium equations are separately set as the outputs and inputs of EKF. Besides, a selftuning ramping coefficient is developed to improve the stability of the fault diagnosis method. Three primary achievements are outlined as follows: 1) The Jacobi matrices are all obtained by the Rank-1 iB algorithm rather than multiple calls to CLM, which greatly reduces the consuming time. 2) The devised iB-EKF algorithm owns a high real-time diagnosis accuracy not only in the steady but also transient process. 3) The satisfactory robustness of fault diagnosis in the wide flight envelope could be drawn out by the designed selftuning damping coefficients. The comparison simulations are carried out on various gas path fault modes and different sensor combinations of aeroengines in the flight envelope. Simulations demonstrate the iB-EKF algorithm takes much less time than the EKF-based method, and its estimation error and required memory are less than those of the LKF-based method. The superiority both in time-consuming and accuracy demonstrates that the proposed approach lays the better candidate to diagnose gas path faults for aircraft engines during the transient behavior.