基于声信号的轴承故障检测算法研究与实现  

Research and Implementation of Bearing Fault Detection Algorithm Based on Acoustic Signal

作　　者：张哲 刘智丰 肖仲喆 黄敏 周佳 郁彬 ZHANG Zhe;LIU Zhifeng;XIAO Zhongzhe;HUANG Min;ZHOU Jia;YU Bin(School of Optoelectronic Science and Engineering,Soochow University,Suzhou Jiangsu 215006,China;Kunshan AutoRoad Automation Technology Co.,Ltd.,Suzhou Jiangsu 215311,China)

机构地区：[1]苏州大学光电科学与工程学院,江苏苏州215006 [2]昆山奥德鲁自动化技术有限公司,江苏苏州215311

出　　处：《电子器件》2024年第3期656-660,共5页Chinese Journal of Electron Devices

摘　　要：滚动轴承是机械设备中重要的零部件,与其他机械零部件相比,滚动轴承的一大特点是其寿命离散性很大。通过研究轴承的故障检测和诊断方法,设计了基于声信号的轴承故障检测算法。首先提出了故障检测系统的流程,包括特征提取、特征处理与选择、模型训练算法和优化方法。之后使用处理后的特征,采用神经网络训练出一个可以区分内圈轴承故障和正常轴承且兼容附带油脂声音的二分类模型。该模型测试集准确率达到了97.8%。使用模型对剩余数据进行测试,内圈故障轴承的准确率为97.5%,正常轴承的准确率为97.3%,附带油脂的正常轴承准确率为98%。表明该算法在对轴承故障的检测上有着优良的性能。Rolling bearing is an important part in mechanical equipment,compared with other parts of machinery,its life span is very discrete.The fault detection and diagnosis of bearing is studied,and acoustic signal-based bearing fault detection algorithm is designed.Firstly,the process of the fault detection system is proposed,including feature extraction,feature processing and selection,model training algorithms and optimization methods.Then,by using the processed features,a binary classification model,which is capable of distinguishing inner ring bearing faults from normal bearings along with the normal bearing with grease sound,is trained by means of neural networks.The accuracy of the model on the test set reaches 97.8%.When testing the remaining data,the accuracy is 97.5%for the inner ring faulty bearing,97.3% for the normal bearing,and 98% for the normal bearing with grease sound.It shows that the algorithm has excellent performance in bearing fault detection.

关 键 词：轴承故障检测 声信号算法 神经网络 

分 类 号：TH133.33[机械工程—机械制造及自动化]