An evidential reasoning-based information fusion method for fault diagnosis of ship rudder

Rudder is an important device for controlling the heading of ships. Due to reasons such as water freezing, foreign matter entanglement, and improper maintenance, the rudder system often encounters different failure states including rudder blade jamming, rudder stock breakage, motor failure, and so on, so as to deteriorate the effectiveness of heading control. Commonly, these faults can be diagnosed by analyzing the monitored vibration signals. Because of the complexity and harshness of the rudder operating conditions, the changing of vibration signals has significant uncertainty. Evidential Reasoning (ER) method can combine the multi-source fault feature information extracted from vibration signals to effectively reduce the uncertainty in diagnosis. Hence, a novel ER-based cascade fusion model in time and space domains (TS-ER for short) is presented for rudder fault diagnosis: (1) time domain diagnostic evidence fusion ER model (T-ER). It can transform the temporal fault feature data into diagnostic evidence and use the current diagnostic evidence to update the historical diagnostic evidence to obtain the fused evidence in time domain; (2) space domain diagnostic evidence fusion ER model (SER). It can combine the local fused evidence given by T-ER in different space positions to obtain the joint diagnostic evidence in the time and space domains. The joint diagnostic evidence can focus the belief degree on the real fault state or fault mode, so as to reduce the uncertainty of diagnostic decision-making. Finally, in the diagnosis experiments of rudder blade jamming fault in a test ship, TS-ER is compared with other some classical data-driven methods to illustrate its accuracy and reliability.