Hybrid fault diagnosis method for underwater thrusters based on the common features of multi-source signals

Integrating the motor and driver into the confined space of an underwater thruster's sealed shell can lead to current sensor failures, primarily due to high temperatures, pressures, and electromagnetic interference. Despite progress in distinguishing sensor malfunctions from propeller issues, a significant gap exists in understanding simultaneous sensor and propeller failures. This study addresses this by thoroughly analyzing the propeller's condition during sensor failure. We introduce two virtual sensors, ingeniously derived from the motor's voltage and mechanical model, to estimate the thruster's current from different perspectives and effectively separate sensor and propeller faults. Recognizing the potential discrepancies between the virtual and real sensors, we developed a multi-source signal common features extractor inspired by transfer learning. This extractor obtains common features from measured and estimated currents, leveraging these variations to detect and assess faults accurately. The effectiveness of this approach has been corroborated through simulation and experiment, demonstrating the ability to distinguish between sensor and propeller faults and accurately evaluate the system's status.