Research On Bearing Fault Diagnosis Based On Improved Convolutional Neural Network

Bearings are the core components that are widely and widely used in production equipment.It is important to carry out relevant research on intelligent fault diagnosis of rolling bearings based on vibration signal analysis and identify faults in a timely and accurate manner,which is very important to ensure the efficient operation of machinery and reduce the probability of safety accidents.The combination of bearing fault diagnosis technology and bearing remanufacturing technology can repair the faulty bearing in a targeted manner,which can increase the life of the bearing,while protecting the environment and improving economic benefits.Most of the traditional fault diagnosis methods are based on the knowledge and experience of experts to extract features artificially from the signal,but the collected signals are often affected by the environment and processing conditions,the signal appears messy and disorderly,and the fault features cannot be extracted quickly and effectively.For such signals,experts also need to analyze a large number of samples for a long time,which is inefficient.Therefore,in order to solve the common problems such as time-consuming,laborious and strong noise in the fault diagnosis of rolling bearings,this paper combines singular value decomposition with convolutional neural network to propose a new intelligent fault diagnosis method for rolling bearings.The specific research contents are as follows:(1)Aiming at the problem of low pattern recognition accuracy under strong noise,a new intelligent rolling bearing fault diagnosis method combining segmented singular value decomposition and one-dimensional convolution network is proposed.Divide the long-sequence original vibration signal of the rolling bearing into equal-length samples,use the segmented singular value decomposition method and the adaptive eigenvalue selection method to decompose and reorganize the original signal of the rolling bearing.Then rely on the advantages of deep learning model pattern recognition to identify the fault signal and prove the effectiveness of the method through experiments.(2)Aiming at the problem that the model costs a lot of time to train a large number of samples,a combination of piecewise singular value decomposition and deep separable interval convolutional network is proposed.The long-sequence original vibration signals of rolling bearings are divided into equal-length samples.Use piecewise singular value decomposition and adaptive eigenvalue selection method.Decompose and reorganize the original signal to be tested.Then rely on the advantages of the pattern recognition of the deep learning model to identify the failure of the signal and prove the effectiveness of the method through experiments.(3)According to actual needs,a set of intelligent identification and automatic classification system for rolling bearing faults was developed based on mixed programming of Lab VIEW and Python.The above algorithm is embedded in the rolling bearing fault diagnosis system.And the research results are put into practical engineering applications,which can be simulated according to the actual working conditions.