Thruster fault diagnosis in autonomous underwater vehicle based on grey qualitative simulation

The paper investigates thruster fault diagnosis for autonomous underwater vehicle (AUV). Since it is difficult to eliminate the spurious behaviors based on the conventional qualitative simulation in AUV qualitative modeling, this paper suggests a grey qualitative constraint filtering method based on higher-order derivative, probability grey number and persistence time. In the process of qualitative modeling, it selects variables whose persistence time is least to derive their possible successors; and then, it filters the set of possible successor states by the developed higher order derivative constraint table. After obtaining the predicted state sequence, taking external disturbance effect into consideration, grey relational analysis based on weighted average is developed to detect and isolate thruster fault by introducing a weighted coefficient about the number of the transition sequence. Furthermore, with respect to fault identification, considering the identification result is an interval based on decision tree technique, the paper proposes a three-dimension identification model based on fractal box dimension and three-dimension surface fitting. Finally, experiments are conducted on Beaver AUV to acquire experiment data, and the comparative results validate the effectiveness of the proposed method. Crown Copyright (C) 2015 Published by Elsevier Ltd. All rights reserved.