基于ShuffleNet-DELM的轴承故障诊断研究
Research on bearing fault diagnosis based on ShuffleNet-DELM
李睿智 1杨芳华 1张伟 1周旗开1
作者信息
- 1. 军事科学院系统工程研究院,北京 100141
- 折叠
摘要
滚动轴承信号是一种典型的非平稳、非线性数据,深度学习模型能够有效提取此类数据特征.为获得更高的精度,深度学习模型不断增加计算量和参数规模,而工程实际中计算机硬件能力和可供训练的数据有限,更注重较快的响应速度和泛化能力.为解决此类矛盾,提出一种基于ShuffleNet-DELM的轴承故障诊断方法.首先将一维的时序信号变换为二维频域张量,再使用改进的ShuffleNetV2模型提取特征,最后经由深度极限学习机(deep extreme learning machine,DELM)方法进行分类,在不同工况的滚动轴承数据集合上取得 95.47%的平均准确率.结果表明:该方法响应速度快,能够进一步提高ShuffleNetV2模型对轴承故障的分类精度和泛化能力,有较好的实用价值.
Abstract
Rolling bearing signal is a typical non-stable,non-linear data,and deep learning models can effectively extract such data features.For higher accuracy,the deep learning model continues to increase the amount and parameter scale of the computing and parameters,while the computer hardware capacity and the data available for training are limited in the actual project.A bearing fault diagnosis method based on ShuffleNet-DELM is proposed.First,the one-dimensional time-series signals are transformed into two-dimensional frequency-domain tensors.Then,an improved ShuffleNetV2 model is employed to extract features,followed by classification using the deep extreme learning machine(DELM)method.This approach achieves an average accuracy of 95.47%on a dataset comprising bearing vibration signals under various operating conditions.The results show that the method has a fast response,which can further improve the classification accuracy and generalization of ShuffleNetV2 model for bearing faults,and has greater practical value.
关键词
深度学习/ShuffleNet/深度极限学习机/轴承Key words
deep learning/ShuffleNet/deep extreme learning machine/bearing引用本文复制引用
出版年
2024
NETL
NSTL
万方数据