基于改进1D-CNN的轴承故障实时诊断方法

Real-time Diagnosis Method of Bearing Fault Based on Improved 1D-CNN

电机作为生产生活中重要的动力设备,如果出现故障将会造成巨大的损失,所以对电机故障进行实时检测具有重要意义。为此提出一种基于改进1D-CNN的电机轴承故障智能诊断模型。该方法将一维电机振动信号直接作为1D-CNN输入而不进行数据重构等过程,大大提高了模型的诊断效率。同时,在1D-CNN中引入残差结构来提升模型的学习能力,从而弥补1D-CNN特征提取方面的缺陷,实现实时性和高准确率的统一。使用凯斯西储数据库设计仿真实验,对电机轴承的10种工作状态进行识别,取得99.3%的准确率,高于许多基于2D-CNN搭建的模型,且诊断时间也要明显少于2D-CNN。实验结果表明,所提模型在高效和高精度方面具有优异的性能。

As an important power equipment in production and life,the motor will cause huge losses if it fails.So the real-time detection of motor fault is of great significance.This paper proposed an intelligent diagnosis model of motor bearing fault based on the improved 1D-CNN.In this method,one-dimensional motor vibration signals are directly used as the input of the 1D-CNN without data reconstruction or other pro⁃cesses,which greatly improves the diagnostic efficiency of the model.At the same time,the residual structure is introduced into the 1D-CNN to improve the learning ability of the model,so as to make up for the defects in feature extraction of the 1D-CNN,which achieves the unity of real-time and high accuracy.This paper uses Case Western Reserve database to design simulation experiment.10 working states of motor bear⁃ings are identified,which achieved a 99.3%accuracy,higher than many models based on 2D-CNN,but the diagnostic time is significantly less than that based on 2D-CNN.The experimental results show that the proposed model has excellent performance in real-time and precision.