Fault diagnosis for hydraulic system of naval gun based on BP-Adaboost model

Abstract

There is strong nonlinearity between the fault states and performance parameters of naval gun hydraulic system. The BP neural network can be trained to represent the nonlinear relationship between variables effectively. But it is sensitive to the initial weights of the network, so the training results are relatively unstable. To solve the problem, this paper presents a new approach to the naval gun hydraulic system fault diagnosis based on BP-Adaboost model. Firstly, the BP neural network is used as a weak classifier, which can fit the relationship between the fault states and the parameters. By training the BP neural network repeatedly, several weak classifiers are obtained. Then by using the Adaboost algorithm, a strong classifier is obtained by merging the multiple BP neural network weak classifiers. The strong classifier can finally be used to diagnose the fault of naval gun hydraulic system. The simulation results demonstrate that the fault diagnosis model has a higher convergence speed and diagnosis accuracy, which can meet the requirements of hydraulic system fault diagnosis.