基于双域深度学习与特征提取的机床故障定位与诊断

打开文本图片集

中图分类号：TP391 文献标志码：A

Machine tool fault localization and diagnosis based on dual-domain deep learning and feature extraction

WANG Jing，CHEN Pin， YANG Ya

（School of Intelligent Manufacturing，Wuhu Vocational Technical University，Wuhu 241O06，China）

Abstract：Limitations of single-domain （time or frequency）feature extraction can cause weak fault signaturesin machine tool operational data to beobscuredunder complex working conditions，directly compromising feature extraction quality and indirectly impairing fault localization and diagnostic performance.To addressthis，a method for machine tool fault localization and diagnosis based on dual-domain deep learning and feature extractionis proposed.Multi-source operational data （vibration，temperature，currnt， etc.）are synchronously colected viasensors.A dual-domain deep learning network is constructed.The initially collected data are input to extract time-domain and frequency-domain operational features separately. Through dualdomain feature fusion，the final operational featureset is obtained.Diagnostic criteria areestablished based on the mechanisms and manifestations of machine tool faults.The degree offeature matching between the extracted features and the diagnostic criteria is calculated. By comparing this matching degree against apredefined threshold，the fault diagnosis result is determined，subsequently enabling fault point localization.Performance tests demonstrate that，compared to traditional methods，the optimized method greatly improves machine tool fault localization accuracy，significantly lowers the misdiagnosis rate，and markedly increases the area under the curve（AUC）of the receiver operating characteristic（ROC）curve metric.These results prove the distinct advantage of the proposed method in both localization and diagnosis.

Key Words：dual-domain deep learning algorithm； feature extraction；machine tool fault； fault localiz-ation； fault diagnosis

机床故障是指机床在运行过程中，因机械、电气、液压、气动或控制系统等方面的异常，导致其无法按照设计要求正常工作或性能下降的现象，可能会引发加工精度降低、生产效率下降、设备停机甚至安全事故[1。（剩余6588字）