基于改进一维卷积神经网络的滚动轴承故障识别

滚动轴承的故障识别对于防止旋转机械系统故障恶化并保证其安全运行具有重要意义。针对现有智能诊断模型参数多、识别效率低的问题,提出一种基于改进一维卷积神经网络的滚动轴承故障识别(FRICNN–1D)方法。通过引入1×1卷积核增强一维卷积神经网络模型的非线性表达能力;并用全局平局池化层代替传统卷积神经(CNN)网络中的全连接层,以降低模型参数和计算量,且防止过拟合现象。试验结果表明,该方法可以准确识别滚动轴承不同故障状态,具有一定的工程实际应用潜力。

Fault recognition of rolling bearing based on improved 1D convolutional neural network

Fault recognition of rolling bearing is of great significance to prevent fault deterioration of rotating machinery system and ensure its safe operation.Here,aiming at problems of many parameters and low recognition efficiency of existing intelligent diagnosis models,a rolling bearing fault recognition method based on improved 1D convolutional neural network was proposed.By introducing 1×1 convolution kernel,the nonlinear expressing ability of 1D convolution neural network model was enhanced.The global draw pool layer was used to replace the full connection layer in the traditional convolutional neural network,reduce model parameters and calculation amount,and prevent the phenomenon of over-fitting.The test results showed that the proposed method can accurately recognize different fault states of rolling bearing,and have a certain engineering application potential.