奇异值分解和稀疏自编码器的轴承故障诊断

针对滚动轴承故障特征提取和分类需要进行有监督训练才能实现等问题，提出了一种基于奇异值分解（SVD）和时域统计特征分析并结合堆栈稀疏自编码器（SAE）以及Softmax分类器实现滚动轴承故障诊断方法。该方法利用Hankle矩阵对原始数据进行矩阵重构，利用奇异值分解和时域分析对重构后的故障信号进行特征预提取，融合两种特征并输入到堆栈稀疏自编码器中进行特征优化，将优化后的特征输入到Softmax分类器中进行分类识别。实验结果表明，3种工况下10类故障数据的识别准确率均在96%左右，且高于文中其他方法，因此该方法能有效地进行滚动轴承复杂信号的特征预处理以及分类。

Singular Value Decomposition and Sparse Automatic Encoder for Bearing Fault Diagnosis

In order to solve the problem that the feature information of rolling bearing fault is difficult to be extracted under high-dimensional data and the signal fault classification needs supervised training to achieve classification, a method is proposed based on Singular Value Decomposition（SVD）and time domain feature analysis and Stacked Sparse AutoEncoder（SAE） and Softmax classifier for classification of rolling bearing faults. This method uses Hankle matrix to reconstruct the original data, then singular value decomposition and time domain analysis is used to conduct feature preprocessing. The integrated features are used as the input of the SAE for feature optimization. The optimized features are entered into the Softmax classifier for classification recognition. The experimental results show that the recognition rate of ten types of data under three kinds of working conditions is above 96%. Compared with other methods, the recognition rate is improved. Therefore, this method can effectively perform feature preprocessing and classification of complex signals such as rolling bearings.