Fault problem in on-line diagnosis for thrusters of autonomous unmanned vehicle (AUV) is studied. The interest to us is not only to detect but also to isolate the fault. This is necessary in order to correct the control program of the vehicle to fulfill pre-assigned task. This problem is difficult because of interdependence among the state vector elements. To overcome this difficulty, so-called test regimes of motion when the stern thrusters work in turn are suggested and the vehicle moves in the horizontal plane or in the vertical one. This allows one to solve the diagnostic task, in particular, to improve the fault isolation property of the diagnostic system.
The autonomous underwater vehicles (AUVs) are complex technical systems which safety and reliable operation deserves particular attention. Navigation-piloting sensors and thrusters are important components of the AUVs which are necessary to control a motion trajectory. This paper is a logical sequel of (Pisarets and Zhirabok, 2002) where the problem of navigation-piloting sensors fault diagnosis in the AUVs was studied. The problem of interest to us is to detect and isolate the faulty thrusters. They can be failed due to faults or another reasons. Since malfunctions and faults occurring in these components can lead to erroneous mission fulfilment or loss of the vehicle, it is necessary to detect and isolate a faulty one early in order to correct the control program of the vehicle to fulfill pre-assigned task. There are some different methods of fault diagnosis: signal-based, analytical model-based, knowledge-based (Frank, 1990; Frank 1996). To diagnosis the AUV components, analytical model-based methods are used that allows one to use the redundancy of the AUV mathematical model. These diagnostic methods based on checking analytical relationships between input and output signals of the AUV are performed.