基于深度卷积网络的多传感器信号故障诊断方法研究

The study of multi-sensor fault diagnose method based on convolutional neural networks

摘要:

针对传统故障诊断方法中多传感器数据融合技术难度大、特征提取困难等问题,提出了一种基于深度卷积网络的多传感器信号故障诊断方法,通过构建测量数据帧进行卷积计算实现多通道数据的自然融合,利用深度网络结构实现高层特征的自动提取和分类,从而高效地实现了故障分类诊断；经分别采用小规模数据集REF和大规模故障数据集BI02进行实验验证,均取得了较高的故障识别准确率,具有很强的工程应用价值。

Abstract:

This paper presents a multi-sensor fault diagnose method based on convolutional neural networks, which utilizes the convolutional core to fuse the different type of measurement data via constructing the measurement data frame. Meanwhile, the high-level features are abstracted automatically from original signal data, and then fault type can be specified according to the output of classifier. As result, the fault recognition achieves high accuracy in treating both a small-scale dataset REF and a large-scale dataset BI02, which shows a significant effect and strong application value.