Research On Fault Diagnosis Method Of Steam Turbine Bearing In Thermal Power Plant Based On Deep Learning

With the development of economy,the demand of electricity consumption in production and daily life is increasing,and the installed capacity of steam turbine is also increasing.As an important equipment in thermal power plants,it is very important to ensure the steam turbine running in a normal and stable state.As one of the most critical components of steam turbine,rolling bearing status affects the operation of the whole unit.The effective diagnosis of rolling bearing fault has always been one of the research hotspots in the field of fault diagnosis.With the development of turbines and other mechanical equipment towards complexity and large-scale,it is becoming more and more complicated to realize effective monitoring of their operation status and accurate fault diagnosis.Conventional fault diagnosis methods are gradually difficult to meet the requirements.Fortunately,it is possible to build a more intelligent state monitoring and fault diagnosis method based on large data,with the continuous development of computer technology,multi-sensor technology and deep learning technology.Therefore,this paper constructs a CNN fault diagnosis model based on deep learning technology and multi-sensor information fusion.The model achieves 100%fault recognition rate through a test on common data sets.The generalization ability and robustness of this CNN model was also verified by an experiment.In addition,the ability of fault diagnosis model based on convolutional neural network to minie the implicit correlation information and potential fault characteristics of the data collected by multi-sensor are analyzed and discussed.Finally,the CNN fault diagnosis model constructed in this paper is compared with the fault diagnosis model based on single sensor.The experimental results show that the CNN fault diagnosis model based onmulti-sensor information fusion has better anti-noise ability than that based on single sensor.A C-LSTM fault diagnosis model based on convolution neural network and long-term and short-term memory network is proposed for bearing signals with time series attributes.The fault diagnosis model also achieves 100% fault recognition rate on public data sets.Finally,the fault diagnosis model is compared with the CNN fault diagnosis model constructed in this paper.In noise interference test,the fault recognition rate of C-LSTM fault diagnosis model is higher than that of CNN diagnostic model by 1.24%,which proves that the C-LSTM network can mine more temporal features in data and improve the fault recognition rate under noise interference.