Autonomous underwater vehicle fault diagnosis model based on a deep belief rule with attribute reliability

• Published in: Ocean engineering Vol. 321; p. 120472
• Main Authors: Mai, Jiahao, Huang, Haolan, Wei, Fanxu, Yang, Cuiping, He, Wei
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 30.03.2025
• Subjects: Attribute reliability; Autonomous underwater vehicle; Belief rule base; Evidential reasoning; Fault diagnosis; Rule explosion; Fault diagnosis; Belief rule base; Rule explosion; Attribute reliability; Autonomous underwater vehicle; Evidential reasoning
• ISSN: 0029-8018
• DOI: 10.1016/j.oceaneng.2025.120472

Autonomous underwater vehicles (AUVs) are sophisticated equipment designed to autonomously navigate and execute missions in complex waters, which makes them susceptible to malfunctions. Therefore, effective fault diagnosis is critical for ensuring the stable and reliable operation of AUVs. Owing to its sophisticated internal structure and the influence of its external environment, the fault diagnosis model faces many uncertainties. The belief rule base (BRB), which is capable of dealing with uncertain information, is an effective solution. However, the data generated by AUV sensors during actual operation are multi-dimensional and subject to disturbances. This can lead to problems such as rule explosion and attribute unreliability. To address these problems, a new deep belief rule base with attribute reliability (DBRB-r) is presented in this study. First, a gradient ascent framework is established to address the rule explosion issue. Second, a statistical approach for calculating attribute credibility is proposed to assess the reliability of attributes. Then, evidence reasoning is used as the model's inference engine. Finally, an interpretable whale optimization algorithm (WOA) is proposed to enhance the model parameters. The results of the verification experiment show that DBRB-r effectively solves the rule explosion problem and also manages unreliable data. This is an effective, accurate and interpretable diagnostic method. •A method for constructing AUV fault diagnosis model based on deep belief rule base.•Evaluate the reliability of attributes to remove the influence of disturbance factors.•Use key attributes gradient modeling to solve rule explosion and improve accuracy.•Optimized interpretability allows scientists to better grasp diagnostic results.•This is an interpretable and reliable fault diagnosis method for AUVs.