Thruster Fault Diagnosis Based On Multi-AUV Information Fusion

As Autonomous Underwater Vehicle (AUV) becomes more and more widely available, the demand of its intelligence level has also seen higher and higher, status monitoring and fault diagnosis of AUV and its components play an important role in safeguarding the safe operation of robot. At the same time, along with the complication of submarine work mission, multi-AUV operation arouses people's interest. The condition monitoring and fault diagnosis technology of single AUV based on multi-AUV information is of significant importance in theoretical research and practical application in enriching and perfecting the autonomy of AUV and enhance the intellectualized level of AUV and AUVs.As a main movement executive component of AUV, fault diagnosis of thruster is the fatal part to ensure the security of the underwater vehicle. The thesis proposes a strategy of thruster fault diagnosis based on sensor output information processing according on output difference of sensors pre/post fault and performance analysis for the propeller and sensors of "beaver". First, a forward speed model of underwater vehicle that reflect the time-lapse property constructed based on Elman neural networks. We get the valid fault information through comparing model output and the actual output. Avoiding the disturbance diagnosis information that introduced by speed output at the speed step. For the Multiple Autonomous Underwater Vehicle Systems (MAUVS), a approach of taking the landscape orientation relative distance as the characteristic signal of fault evaluate is proposed based on the difference of relative position between fault robot and natural robot.In order to improve the uncertainty of fault diagnosis based on sensor output, we introduce D-S evidencen theory. The fault probability function of absolute residual value is proposed through residual change of three kind sensors output value and target value.And fault is judged by whether fault probability value achieve or exceed set threshold. The result of fault diagnosis from one sensor is fused with time, and it is fused from different sensor at the same time in space to implement thruster fault diagnosis based on multi-information fusion. According to the characteristic of D-S evidence fusion theory, an improved information fusion method of AUV thruster diagnosis is proposed, realized the thruster fault diagnosis based on multi-AUV information fusion.The fault simulation and pool experiment result shows that the failure diagnosis method is feasibility and validity to detect the single AUV thruster fault in the system of multi-AUV, and provide important technical basis to further fault tolerant control.