Fault Diagnosis of Autonomous Underwater Vehicle Using Neural Network

Fault Diagnosis in Autonomous Underwater Vehicles (AUVs) has become more important since the rapid use of these vehicles in more complex missions which may be done in unstructured environments with unpredictable conditions; Due to plant's features like nonlinearity or time variance and also unpredictable external disturbance generated by the sea current fluctuations, implementing a fault diagnosis system to prevent diverse damages to the vehicle, researchers have been occupied in the field. In this paper, an adaptive model is used to implement desired system for the purpose of fault diagnosis. In this model, two kinds of Neural Networks are proposed to design the adaptive filter which estimates the state of our plant in each sample time. They are Multi-Layer Perceptron (MLP) and Adaline. In this approach, the adaptive model is tested by using both neural networks for both limited and persistent faults. The obtained results show that both neural networks are able to diagnose fault occurrence through a proposed algorithm based on the obtained coefficients which describe the system state in each sample time.