基于无损约束降噪稀疏自编码的滚动轴承故障诊断技术

在滚动轴承故障诊断过程中，时域振动信号容量大且易受噪声污染，难以建立准确的故障诊断模型。针对上述难题，本文采用无损约束降噪方法对稀疏自编码进行优化，提出了基于无损约束降噪稀疏自编码的滚动轴承故障诊断方法。该方法可直接作用于时域振动信号，消除对人工特征提取的依赖性，无需降噪预处理，降低了故障诊断模型建立的难度。为验证本方法的有效性，利用滚动轴承时域振动信号进行仿真实验，并对诊断过程中学习到的故障特征进行可视化分析。实验结果表明，本方法可以在噪声数据下建立有效的故障诊断模型，且比传统的栈式稀疏自编码诊断算法具有更强的噪声鲁棒性。

Fault Diagnosis Technology of Rolling Bearings Based on Lossless-constraint Denoising Sparse AutoEncoder

In the process of rolling bearing fault diagnosis, the time domain vibration signal has large capacity and easily suffers from noise pollution so that it is difficult to establish an accurate fault diagnosis model. To solve the above problems, this paper uses the Lossless-constraint denoising method to optimize sparse autoencoder. And a fault diagnosis method for rolling bearing based on Lossless-constraint denoising sparse autoencoder is proposed. The method can directly be applied to the time domain vibration signal, thus eliminating the dependence on the artificial feature extraction, and this method does not need the noise reduction preprocessing, which reduces the difficulty of establishing the fault diagnosis model. In order to prove the effectiveness of the method, this paper uses the time domain vibration signal of the rolling bearing to carry out the simulation experiment, and visually analyze the fault characteristics learned during the diagnosis process. The experimental results show that the proposed method can establish an effective fault diagnosis model with noise data and has stronger noise robustness than the traditional stack sparse autoencoder diagnostic algorithm.