基于Transformer的智能轴承声-振融合故障诊断  

Acoustic-Vibration Fusion Fault Diagnosis for Intelligent Bearing Based on Transformer

作　　者：林昙涛 牛青波[2] 马天旭 王强 朱永生[1] LIN Tantao;NIU Qingbo;MA Tianxu;WANG Qiang;ZHU Yongsheng(Key Laboratory of Education Ministry for Modern Design and Rotor-Bearing System,Xi’an Jiaotong University,Xi’an 710049,China;Luoyang Bearing Research Institute Co.,Ltd.,Luoyang 471039,China)

机构地区：[1]西安交通大学现代设计与转子轴承系统教育部重点实验室,西安710049 [2]洛阳轴承研究所有限公司,河南洛阳471039

出　　处：《轴承》2023年第2期67-73,共7页Bearing

基　　金：国家重点研发计划资助项目(2020YFB2007900)。

摘　　要：针对单一信源(振动或声音)轴承故障诊断方法所蕴含信息不全面的问题,开展了具有多源传感器集成的智能轴承的声-振融合故障诊断研究,引入Transformer架构作为声-振融合诊断模型的基本模式以加强信号的时序特征提取能力,利用交叉自注意力机制使声音信号与振动信号在特征提取过程中交互与融合,从而实现端到端的智能轴承故障诊断。搭建智能轴承试验台采集声音与振动数据进行验证的结果表明,基于Transformer的智能轴承声-振融合故障诊断方法相对于单独使用声音、振动的方法以及基线Transformer方法,诊断性能均有提升。Aimed at incompleteness of information contained from single-source(vibration or acoustic) for bearing fault diagnosis, a study on acoustic-vibration fusion fault diagnosis is conducted for intelligent bearings integrated with multi-source sensor.Transformer architecture is introduced as basic pattern of acoustic-vibration fusion diagnosis model to enhance the temporal feature extraction capability of signals. The cross self-attention mechanism is used to interact and fuse the acoustic and vibration signals during feature extraction process, the end-to-end fault diagnosis is realized for intelligent bearings. A test bench for intelligent bearings is built to collect the acoustic and vibration data for verification. The results show that compared with methods using acoustic and vibration alone, as well as baseline Transformer method, the proposed method has improved diagnostic performance.

关 键 词：滚动轴承 智能轴承 故障诊断 声发射信号 振动 TRANSFORMER 

分 类 号：TH133.33[机械工程—机械制造及自动化] TB53[理学—物理]