A Fault Diagnosis Method for Autonomous Underwater Vehicles Based on Online Extended Belief Rule Base With Dynamic Rule Reduction

Abstract

Autonomous underwater vehicles (AUVs), sophisticated devices that operate independently in complex underwater environments, are prone to malfunctions. Therefore, fault diagnosis is crucial for maintaining stable operations. The massive amounts of complex multidimensional data generated during AUV operations make it challenging to construct rule bases efficiently. The extended belief rule base (EBRB) offers a solution by employing data-driven approaches. Furthermore, as the data are sequentially generated, online training is more aligned with the practical realities of AUV operations. When processing large amounts of data, the EBRB system generates a substantial number of rules. This process can be very time-consuming. It may trigger the activation of many irrelevant rules, potentially leading to erroneous decisions. In response, a fault diagnosis method for autonomous underwater vehicles based on an online extended belief rule base with dynamic rule reduction (OEBRB-R) is introduced in this paper. This approach involves first establishing an OEBRB model that continually processes new incoming sensor data in real time. Subsequently, it employs a Mahalanobis distance-based k-means clustering technique to dynamically reduce the number of active rules. Finally, evidence reasoning (ER) algorithms are utilized to infer the type of fault present. Through case studies, the proposed method has been empirically validated for maintaining strong accuracy in real-time diagnostics while effectively reducing the model’s spatial complexity.

Keywords

• Fault diagnosis
• extended belief rule base
• rule reduction
• autonomous underwater vehicle
• online model updating