Optimizing coupled fault diagnosis in liquid rocket engines via distributed robust multi-model Kalman filter

This paper concerns a novel fault diagnosis method for decoupling and diagnosing liquid rocket engine (LRE) system and sensor failures. LRE faults significantly affect the launch reliability. Many traditional methods applied in system-level fault diagnosis of LRE struggled with robustness when dealing with sensor and system coupling faults. The distributed robust multi-model Kalman filter (DRMM-KF) based fault diagnosis method is proposed for the LRE sensor-system concurrent fault diagnosis problem to solve the aforementioned issues. Firstly, improvements are made to the multi-model Kalman filter (MM-KF) method to enhance the robustness of the hypothetical model conditional probability calculation to obtain the robust MM-KF (RMM-KF). Then, based on the RMM-KF and local sensor fault hypothesis models, the local robust MM-KFs (LRMM-KFs) are established to describe the local sensor faults of the LRE subsystems, and then, the DRMM-KFs are established by integrating each LRMM-KF to realize the adaptive description of the LRE sensor faults. Finally, by fusing the results of the LRMM-KFs, the index fusion weighted sum-squared residual (FWSSR) is established for LRE system and sensor faults decoupling and diagnosis. After that, three hot-test cases are given to illustrate the effectiveness of the proposed method in practice. The contribution of this study is to propose the novel DRMM-KF algorithm to achieve real-time LRE system and sensor faults decoupling fault diagnosis, showing the capacity of distinguishing between system and sensor faults with alarm timely.