Self-updating Fault Diagnosis Method of Dynamic Positioning Ship Sensor Based on Wavelet Neural Network

Abstract

In order to solve the problem that the fault diagnosis method of intelligent dynamic positioning ship sensors has poor accuracy and the difficulty in obtaining a large number of fault data of each specific ship, this paper proposes a self-updating fault diagnosis method based on wavelet packet decomposition and neural network, combined with semi-supervised learning method. The method uses labeled data sets to train the initial model and unlabeled data sets to improve the accuracy of model. Finally, based on a specific dynamic positioning ship model, the simulation results show that the proposed method can effectively improve the overall accuracy of the fault diagnosis model and decrease the rate of false positives, misstatement of the fault diagnosis model by using the sensor data acquired during ship operation, overcome the lack of historical data of neural network, and improve the performance of fault diagnosis on a specific ship.