基于改进堆叠式循环神经网络的轴承故障诊断

Bearing Fault Diagnosis Based on Improved Stacked Recurrent Neural Network

提出基于改进堆叠式循环神经网络的轴承故障诊断模型.利用深层网络极强的非线性拟合能力以及循环神经网络特有的沿时间通道传播的特点,通过门控循环单元解决堆叠式循环神经网络梯度消失的问题,实现对轴承健康状况的分类识别.利用美国凯斯西储大学轴承数据集进行了轴承故障诊断试验,同时将支持向量机、粒子群优化的支持向量机、人工神经网络、卷积神经网络AlexNet以及循环神经网络作为对比以检验所提模型的分类性能.结果表明,提出的模型能够对轴承故障进行有效诊断,并且具有一定的可靠性与泛化能力.

A bearing fault diagnosis model based on improved stacked recurrent neural network was proposed, which takes advantage of great nonlinear fitting capability and the characteristics of propagation though time. Gated recurrent unit was used to deal with the vanishing gradient problem, which contributes to classify the bearing health condition. The data set from Bearing Data Center of Case Western Reserve University was used to carry out the bearing fault diagnosis test. Support vector machine, particle swarm optimization-support vector machine, back-propagation network, AlexNet, and recurrent neural network were tested as well for comparison. The results show that the proposed model has exceptional reliability and generalization.