摘要
针对深度学习在轴承故障诊断中出现的多分类时测试准确率低、数据集不足的情况,提出基于格拉姆角场(GAF)法和卷积神经网络(CNN)的轴承故障诊断模型以及采用重叠采样20%的方法扩充数据集。通过对轴承振动信号采用格拉姆角场法变换构建数据集,导入到搭建的六层卷积神经网络中实现故障分类。在搭建的CNN中测试了不同的轴承数据集以及不同数据长度下的测试准确率和抗噪性能。结果表明,在不同数据集的测试中,所搭建的模型最高测试准确率可达100%,搭建的CNN有着良好的性能,在多分类问题上具有较高的精度;扩充数据集的方法具有一定的可行性,可以有效提升模型的测试性能。
In view of the low test accuracy and insufficient data sets in the multi-classification of deep learning in bearing fault diagnosis,a bearing fault diagnosis model based on Gramian Angular field(GAF)method and convolutional neural network(CNN)was proposed,and the overlapping 20%sampling method was used to expand the dataset.The data set was constructed by using Gramian Angular field method to transform the bearing vibration signals and imported into the constructed six-layer convolutional neural network to realize the fault classification.The test accuracy and anti-noise performance under different bearing datasets and different data lengths were tested in the built CNN.The results showed that in the testing of different datasets,the highest test accuracy of the built model was able to reach 100%,and the built CNN had good performance and high accuracy in multi-classification problems.The method of data set expansion has a certain feasibility,able to improve the test performance of the model effectively.
作者
周孟然
宋乾坤
ZHOU Mengran;SONG Qiankun(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan Anhui 232001,China;School of Artificial Intelligence,Anhui University of Science and Technology,Huainan Anhui 232001,China)
出处
《安徽理工大学学报(自然科学版)》
CAS
2023年第1期8-14,共7页
Journal of Anhui University of Science and Technology:Natural Science
关键词
格拉姆角场
卷积神经网络
轴承故障诊断
抗噪性能
Gramian Angular field
Convolutional neural network
fault diagnosis
noise immunity