【摘要】 针对目前滚动轴承故障诊断中,模式识别研究主要依靠有监督式机器学习的问题,提出一种基于堆栈稀疏自编码(Stacked Sparse Autoencoder,SSAE)的无监督式深度神经网络的滚动轴承智能故障诊断方法,在滚动轴承故障诊断试验台上提取正常、内圈故障和外圈故障三种状态信号进行验证,试验结果表明,SSAE网络可以有效、准确地识别滚动轴承具体的故障诊断类型,其诊断精度优于反向传播神经网络(Back Propagation,BP)及支持向量机(Support Vector Machine,SVM)。更多还原

【Abstract】 In view of the situation of using supervised learning in pattern recognition in rolling bearing fault diagnosis,an unsupervised deep neural network based on stack sparse autoencoder( SSAE) is proposed,in which the signals of normal condition,inner ring fault and outer ring fault extracted on the rolling bearing test bench are identified. The results show that SSAE network can identify the specific fault diagnosis types of rolling bearing effectively and precisely,and its diagnostic accuracy is superior to back propagation neural network and support vector machine.更多还原