An intelligent fault diagnosis method for fan drive system based on DCNN model

Abstract

The invention discloses a fault intelligent diagnosis method of fan drive system based on Deep Convolution Neural Network (DCNN) model.Firstly, the vibration signals of normal and faulty driving system of wind turbine are collected, and the vibration signals are analyzed by Winger-Ville spectrum analysis algorithm processes the collected vibration data and extracts the time-frequency curves underdifferent operating conditions. Frequency characteristic map as sample set; establishing the DCNN model, selecting a certain number of feature maps as training samples to train the DCNN model; finally, the remaining samples in the feature map sample set are used as test samples to test the trained DCNN model, optimize the structural parameters and training parameters of the model, and select the best structural parameters and training parameters with the best recognition performance and stability. This method can realize the intelligent diagnosis and identification of the fault of the fan drive system, and improve the accuracy and efficiency of the fault identification.

Description

technical field [0001] The invention belongs to the field of intelligent diagnosis, and relates to an intelligent fault diagnosis method for a wind turbine transmission system based on a DCNN (Deep Convolutional Neural Network, deep convolutional neural network) model, in particular to a vibration fault diagnosis for a transmission chain system of a wind turbine. Background technique [0002] As an emerging industry, my country started late in the research on fault diagnosis of wind turbines. Compared with the rapidly increasing installed capacity of wind power, the related fault diagnosis technology is significantly behind. Many aspects such as research, intelligent diagnosis and trend forecasting technology research are relatively weak, and it is still far from the urgent demand for fault diagnosis technology in the wind power industry. [0003] The fault diagnosis of foreign wind turbines mainly focuses on the transmission system and electrical system. At present, there a...

AI Technical Summary

Problems solved by technology

my country's research on fault diagnosis of wind turbines started relatively late, and now the focus of research is mainly on the transmission chain, and the analysis method used is mainly vibration analysis. The eigenvectors are all artificially constructed on the basis of time-frequency diagrams, and there are few studies that directly use time-frequency diagrams as the basis for identification, and the number of hidden layers of shallow networks such as BP neural networks and SVMs used is limited, and the ability of feature learning and expression is limited. Limited, training tends to fall into local extremum

Images