基于二维卷积神经网络的轴承故障智能诊断

Intelligent Diagnosis of Bearing Faults Based on Two-Dimensional Convolutional Neural Network

摘要: 机械故障诊断已成为保证设备安全稳定运行的重要手段和关键技术之一。机械故障诊断对于保障设备安全运行意义重大，然而传统诊断方法太过依赖专家经验和先验知识，一直是该领域亟需解决的问题。随着机器学习的诞生和发展，基于深度学习的智能诊断方法应运而生。本文提出一种以短时傅里叶变换时频图为输入的二维卷积神经网络架构，在结构并不复杂的前提下可以实现轴承故障“端到端”的智能诊断。使用凯斯西储大学轴承故障数据集进行验证，结果表明所提方法具备有效性，且优于对比方法。总结整体方法流程并给出结论。

Abstract: Mechanical fault diagnosis has become one of the important means and key technologies to ensure the safe and stable operation of equipment. Mechanical fault diagnosis is of great significance to ensure the safe operation of equipment, however, traditional diagnostic methods rely too much on experts’ experience and a priori knowledge, which has been an urgent problem in this field. With the birth and development of machine learning, intelligent diagnosis methods based on deep learning have emerged. In this paper, we propose a two-dimensional convolutional neural network architecture with short-time Fourier transform time-frequency images as inputs, which can realize “end-to-end” intelligent diagnosis of bearing faults under the premise of uncomplicated structure. Validation using the Case Western Reserve University bearing failure dataset shows that the pro-posed method is effective and outperforms the comparison method. Summarize the overall method process and provide conclusions.