一种基于深度融合模型的滚动轴承故障诊断方法

针对噪声环境下滚动轴承故障难以诊断的问题，提出一种基于深度学习融合网络的轴承故障识别新方法。该方法首先对轴承振动信号进行一定程度的随机损坏，并将加噪后的数据输入卷积降噪自编码器(convolutional denoising autoencoder，CDAE)中对网络进行训练，目的是降低信号中的噪声干扰并提取浅层特征；然后，利用深度信念网络(deep belief network，DBN)学习深层特征并建立轴承状态识别模型，输出故障识别结果。在融合模型中，将卷积降噪自编码器作为网络的第一层以增强网络的抗干扰能力，提高故障的识别精度。使用凯斯西储大学(CWRU)滚动轴承数据对所提模型进行验证，结果表明提出的融合模型在噪声环境下能够较好地实现轴承的故障状态识别。

A rolling bearing fault diagnosis method based on the deep fusion model

Bearings are a key component in mechanical equipment. Since rolling bearing faults are difficult to diagnose in the presence of noise, a new fault diagnosis method based on a deep learning fusion network is proposed in this work. In the first step, a convolutional denoising autoencoder (CDAE) is employed to process the noisy vibration signal in order to reduce the noise interference and extract the low-level features. A deep belief network (DBN) is then used to learn the deep features and construct a bearing state identification model, and output the fault diagnosis results. In the fusion model, the CDAE is utilized as the first layer to enhance the anti-noise ability of the network and improve the fault recognition accuracy. The proposed method has been verified using the rolling bearing dataset from Case Western Reserve University (CWRU). The results show that the proposed fusion model can accurately identify the bearing fault status in a very noisy environment.