Fault diagnosis of AUV rudder hydraulic system based on ISSA-SVM

Abstract

As a critical actuator that ensures the navigational safety of autonomous underwater vehicles (AUV), the reliability of the rudder plays a pivotal role in both mission completion and operational security. To address the low efficiency and insufficient accuracy of traditional fault diagnosis methods, this paper proposes a fault diagnosis approach that combines an improved sparrow search algorithm (ISSA) with a support vector machine (SVM). First, a rudder system simulation model is built, and fault data under various operating conditions are obtained through fault injection. Next, by incorporating a salp swarm algorithm and levy flight, the sparrow search algorithm (SSA) is enhanced to improve its search capability. Finally, ISSA is used to optimize the penalty parameter and kernel function of the SVM. A comprehensive set of experiments was conducted on a simulated dataset containing both normal and multiple fault states, comparing ISSA-SVM with nine other methods. The results show that ISSA-SVM achieves an average accuracy of 97.02%, precision of 96.87%, recall of 97.34%, and an F1-Score of 97.01%, with each metric improved by at least 0.43%, 0.37%, 0.46%, and 0.42%, respectively. These findings indicate that ISSA-SVM effectively captures the characteristic patterns of different fault types, providing higher performance in AUV rudder fault diagnosis. Consequently, it ensures strong operational continuity and safety for AUV even in challenging conditions, offering substantial significance for further AUV research and development.