De-noising Fault Diagnosis Of Rolling Bearing Based On Deep Learning

As one of the vulnerable parts of mechanical equipment,rolling bearing fault has the characteristics of uncertainty,complexity and variability.Therefore,it is of great significance to study the fault diagnosis technology of rolling bearing.However,due to the signal characteristics of transient,nonlinear and non-gaussian,it needs a large amount of analysis and comparisons to get the effective fault features,which is a time-consuming and labor-intensive process.In order to extract fault features automatically,this paper devotes to introducing deep learning to solve the problem.Deep learning methods have been widely used in the field of intelligent fault diagnosis due to their powerful feature learning and classification capabilities.However,it is easy to overfit depth models because of the large number of parameters brought by the multilayer-structure.As a result,the methods with excellent performance under experimental conditions may severely degrade under noisy environment conditions.In this paper,a de-noising algorithm based on de-noising convolutional autoencoder and convolutional neural network is proposed to address this problem,whereby the former is used for de-noising of raw vibration signals and the latter for fault diagnosis using the de-noised signals.Using onedimensional convolution operation to adapt to the vibration signal,so as to realize the end-toend diagnosis of the rolling bearing fault.Considering the complex and changeable characteristics of the actual noise environment,gaussian noise with different signal-to-noise ratio is added to the original sample to simulate the actual noise environment.In order to improve the de-noising ability and strengthen the robustness to noise of the model.The joint diagnosis is optimized and improved respectively.The main contents are as follows:Aiming at the problem that the bearing vibration signal is easy to be polluted by noise and the effective features are difficult to be extracted,this paper proposes a one-dimensional denoising convolutional autoencoder based on full convolutional network to de-noising of the original signal.Compared with the traditional de-nosing autoencoder based on the full connection structure,improved model can effectively solve the problem of too many parameters in the full connection layer by convolution operation,and can improve the ability of feature extraction and reconstruction by increasing the depth of the network.On this basis,1d-dfcae abandons the pooling layer of the traditional convolutional neural network.Traditional CNN uses full connection layer as classifier,resulting in too many parameters,and the model is easy to fall into the risk of over fitting.This paper proposes an improved CNN to solve these problems,a global average pooling layer,instead of fully-connected layers,is applied as a classifier to reduce the number of parameters and the risk of overfitting.In addition,manifold learning is introduced to visualize features.By comparing with many classical feature classification algorithms,the superiority of this algorithm is verified.