Inter-Turn Short-Circuit Fault-Based Thruster Remaining Useful Life Prediction

Inter-turn short circuit (ITSC) faults are one of the most common failure types in Autonomous Underwater Vehicle (AUV) thrusters. Predicting the remaining useful life (RUL) of thrusters post-ITSC occurrence provides critical data support for AUV autonomous decision-making systems to determine whether remaining tasks can be completed. This paper establishes a dataset by collecting lifespan data from multiple inter-turn short-circuited motors and constructs a VMD-TCN-MHSA prediction model based on Variational Mode Decomposition (VMD), Temporal Convolutional Network (TCN), and Multi-Head Self-Attention (MHSA) mechanisms. Features extracted from each decomposed component of the original signals (via VMD) are fed into the model for prediction. Through comparative experiments with other prediction models using datasets under diverse operating conditions, the feasibility, effectiveness, and superior accuracy of the proposed model are validated.