基于SincNet网络结合L2正则化的故障诊断  

Based on SincNet Network Combined with L2 Regularization Fault Diagnosis

作　　者：魏永合[1] 陈懿翀 谷晓娇 WEI Yonghe;CHEN Yichong;GU Xiaojiao(School of Mechanical Engineering,Shenyang Ligong University,Shenyang 110159,China)

机构地区：[1]沈阳理工大学机械工程学院,沈阳110159

出　　处：《组合机床与自动化加工技术》2024年第8期158-162,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金资助项目(51875368)。

摘　　要：故障诊断对于保持设备和系统的正常运行至关重要,它可以帮助提高效率、减少成本、增强安全性、改善用户满意度,并为决策和优化提供支持。通过及时进行故障诊断和解决,可以提高生产效率、降低风险,并提供更好的产品和服务。针对基于物理信息模型和基于数据驱动模型等传统故障诊断方式的可解释性不强、故障诊断正确率较低的缺点,提出一种基于卷积神经网络(convolution neural network, CNN)、SincNet和L2正则化相结合的故障诊断方法。通过以轴承为例,进行实验验证,并于传统CNN进行对比,准确率达到99.5%,也更具有可解释性。相较于传统的CNN,该模型的可解释性更强、故障诊断准确率更高。Troubleshooting is critical to keeping equipment and systems up and running.It can help increase efficiency,reduce costs,enhance security,improve user satisfaction,and support decision-making and optimization.Through timely fault diagnosis and resolution,productivity can be improved,risks can be reduced,and better products and services can be provided.Aiming at the shortcomings of traditional fault diagnosis methods based on physical information model and data-driven model,which are not strong in interpretability and low in fault diagnosis accuracy,this paper proposes a fault diagnosis method based on convolution neural network,SincNet and L2 regularization.By taking the bearing as an example,experimental verification is carried out and compared with traditional CNN,the accuracy rate reaches 99.5%,which is also more interpretable.Compared with traditional CNN,the model has stronger interpretability and higher fault diagnosis accuracy.

关 键 词：故障诊断 SincNet CNN L2正则化 

分 类 号：TH164[机械工程—机械制造及自动化] TG66[金属学及工艺—金属切削加工及机床]