Sensor fault diagnosis for uncertain dissimilar redundant actuation system of more electric aircraft via bond graph and improved principal component analysis

This paper develops a sensor condition monitoring method integrating the model-based bond graph (BG) technique and data-driven principal component analysis (PCA) for the dissimilar redundant actuation system of more electric aircraft with uncertain parameters. The uncertain dissimilar redundant actuation system is modeled by BG in linear fractional transformation form. After that, the analytical redundancy relations containing the nominal part and the uncertain part can be derived, based on which the adaptive thresholds and the fault signature matrix (FSM) can be obtained for robust fault detection and fault isolation. To improve the fault isolation performance under the multiple faults condition, a new fault isolation method integrating FSM and improved PCA (IPCA) is developed, where the possible fault set generated from the FSM is further refined by the IPCA module with an improved reconstruction algorithm and cyclic PCA monitoring model to achieve a more efficient fault isolation result. The effectiveness of the proposed approach is validated by simulation investigations.