Research On Fault Diagnosis Wind Turbine Rolling Bearing Based On Convolutional Neural Networks

In recent years,in order to alleviate the environmental pollution caused by industrialization,vigorously developing wind power has become an important choice.The installed capacity of wind power has increased year by year,and wind power has become the third main power source in China.However,the frequent faults due to severe weather conditions and complex operating conditions have a bad effect on the stable and safe operation of wind turbines.The faults of the transmission system,especially the rolling bearing,often requires a long downtime for carrying out repairs,which seriously damages economic benefits.Therefore,it is of great significance to carry out real-time condition monitoring and early fault identification for the rolling bearing of wind turbine.In recent years,with the rapid development of sensor monitoring technology,data access technology and big data analysis technology,bearing fault diagnosis technology has been developed and improved accordingly.Machine learning methods are more and more applied to the field of fault diagnosis.To reduce the dependence on prior knowledge and expert knowledge has become a common demand in fault diagnosis industry.In addition,many complex problems which are difficult to solve in the past have been paid more attention,such as rolling bearing fault diagnosis under small sample conditions or variable working conditions has become a hot topic in the research.Based on this,this paper proposes a fault diagnosis method of wind power rolling bearing based on convolution neural network,the main contents are as follows:(1)The components of the wind power transmission chain simulation test bench and the corresponding signal acquisition system are introduced,some time-domain,frequency-domain and time-frequency domain features that often need to be extracted in the field of fault diagnosis are listed.(2)Aiming at the problem of lack of selectivity for a large number of extracted features and poor generalization ability of deep learning model,a rolling bearing fault recognition method based on one-dimensional convolution neural network(CNN)and convolution block attention module(CBAM)is proposed.Firstly,one-dimensional convolutional neural network is used to adaptively extract the multi angle and multi-level features of bearing vibration signal,and then the features are weighted by CBAM to increase the weight of important features to play a greater role,and finally the global pooling layer is used to replace the full connection layer,which makes the model selective for features,higher diagnostic accuracy in the case of strong noise,and fasterconvergence speed.(3)Due to the lack of sample size,machine learning and deep learning model is training difficultly,and the problem of over fitting is of common occurrence.A rolling bearing fault diagnosis method under variable conditions based on variational mode decomposition(VMD)and multi-scale convolutional neural network(MSCNN)is proposed.Firstly,VMD is used to decompose the original vibration signal into several intrinsic modal components,and then the envelope order spectrum is obtained by energy operator demodulating,angle domain resampling and Fourier Transform,which is input into MSCNN for rolling bearing fault diagnosis.Experiments prove that this method has high recognition accuracy and good effect in small samples and variable conditions.(4)In order to reduce the dependence of fault diagnosis on expert knowledge under variable condition,a fault diagnosis method of rolling bearing based on convolutional auto encoders(CAE)and CNN is proposed.Firstly,the data under a certain working condition is taken as the source domain,and the data under other working condition is taken as the target domain.The feature extraction and the neighborhood self-adaptation between the source domain and the target domain are realized by CAE.Then the feature vector obtained is input into the convolution neural network for rolling bearing fault diagnosis.