Incipient fault prognosis for hybrid systems without mode change and degradation behavior information

This work concerns incipient fault prognosis for hybrid systems with unknown mode change and degradation behavior. This method utilizes the newly developed concept of Augmented Global Analytical Redundancy Relations (AGARRs) to consider incipient fault prognosis of parametric and nonparametric nature. The incipient fault could develop at all modes (detectable mode/non-detectable mode). Before a fault is detected, the mode information is provided by a mode tracker, which is based on Mode-change Signature Matrix (MCSM) and is only activated when an inconsistency between monitored system and its nominal model is detected. Once a fault is detected, model based mode tracker is not useful anymore. A technique to parameterize the mode change is utilized. The degradation behavior of incipient fault is unknown in advanced, and a degradation model selection mechanism is discussed. The fault hypothesis set, including suspected faults and suspected mode change, is established after fault isolation. A mixed differential evolution (MDE) algorithm is proposed for fault prognosis, which is able to simultaneously handle real and binary unknown variables. Simulation results of different fault scenarios show the effectiveness of the proposed method.