Turbine Fault Diagnosis Method Based on Improved Convolutional Neural Network

Authors

  • Fenghao Sun

Abstract

An improved convolutional neural network (CLCNN) for turbine mechanic rotor is constructed to address the problem of low accuracy and recognition efficiency of current detection methods for turbine mechanic rotor faults, and an adaptive convolutional neural network fault algorithm is proposed based on the principle that the larger the size of the convolutional kernel is, the larger the perceptual field is. This paper discusses the influence of the size of the convolutional kernel on the convolutional neural network. First, the original vibration signals of various fault states are divided into training samples and test samples by using public fault data; then the training samples are expanded and then the hyper-parameter training for adjusting the neural network is carried out to verify the accuracy of the model; finally, the ZT-3 rotor test is used to simulate the turbine mechanic rotor fault and the obtained test data of the 4 states are used to verify the effectiveness of the optimization model. The experimental results show that the improved convolutional neural network can accurately and efficiently diagnose turbine mechanic rotor faults; which proves the effectiveness and feasibility of this method for turbine mechanic rotor fault diagnosis.

Published

2020-02-28

Issue

Section

Articles