基于动态贝叶斯网络与DS证据理论的轧机颤振监测方法  

Chatter Monitoring Method of Rolling Mill Based on Dynamic Bayesian Network and Dempster/Shafer Evidence Theory

作　　者：时培明[1] 刘奥运 张逸伦 高浩 SHI Pei-ming;LIU Ao-yun;ZHANG Yi-lun;GAO Hao(Key Laboratory of Measurement Technology and Instrument of Hebei Province,Yanshan University,Qinhuangdao,Hebei 066004,China)

机构地区：[1]燕山大学河北省测试计量技术及仪器重点实验室,河北秦皇岛066004

出　　处：《计量学报》2022年第9期1178-1185,共8页Acta Metrologica Sinica

基　　金：国家自然科学基金(61973262);河北省自然科学基金(E2019203146)。

摘　　要：提出了基于动态贝叶斯网络和DS(Dempster/Shafer)证据理论的轧机颤振实时监测方法,该方法预选多个时域和频域的特征参数表征轧机不同工况下振动信号的不同特征,利用稳定判别率方法筛选敏感度高的特征参数;使用动态贝叶斯网络与DS证据理论实时监测模型建立轧机颤振状态实时监测系统,构建连续的速度载荷时间片,将3个连续的速度载荷时间片作为DS证据理论的证据体,给出了优化基本概率分配的信任度方法,解决了DS证据理论的证据体间冲突问题;最后在轧机实验平台进行实验,诊断结果表明:该方法对轧机颤振不同状态的识别率达到99.05%。A real-time monitoring method of rolling mill chatter based on dynamic bayesian network and DS(Dempster/Shafer)evidence theory is proposed.In the mentioned method,multiple characteristic parameters in time domain and frequency domain are preselected to represent the different characteristics of rolling mill vibration signals under different working conditions,and the symptom parameters with high sensitivity are selected by using the stability discrimination rate method.Then,the dynamic bayesian network and DS evidence theory real-time monitoring model is used to establish the real-time monitoring system of rolling mill chatter state,and the continuous speed load time slices are constructed.Three continuous speed load time slices are taken as the evidence body of DS evidence theory.At the same time,the trust method of optimizing the basic probability distribution is proposed to solve the conflict between the evidence bodies of DS evidence theory.Finally,the experiment is carried out on the rolling mill experimental platform,and the diagnosis results show that the recognition rate of different states of rolling mill chatter can reach 99.05%.

关 键 词：计量学 颤振监测 动态贝叶斯网络 Dempster/Shafer证据理论 轧机 

分 类 号：TB936[一般工业技术—计量学]