结合非线性频谱与贝叶斯网络的复杂装备传动系统故障诊断  

Fault diagnosis of driving system for complex equipment based on nonlinear spectrum and Bayesian network

作　　者：张家良[1] 曹建福[1,2] 高峰[1] 

机构地区：[1]西安交通大学机械制造系统工程国家重点实验室,陕西西安710049 [2]西安交通大学苏州研究院,江苏苏州215123

出　　处：《电机与控制学报》2014年第3期107-112,共6页Electric Machines and Control

基　　金：国家"863"计划(2009AA01Z126)

摘　　要：针对复杂装备传动系统的故障诊断问题,提出了一种非线性频谱特征与贝叶斯网络相结合的故障诊断方法。为了克服基于Volterra级数的广义频率响应函数所存在的计算量膨胀问题,采用一维的非线性输出频率响应函数获取传递频谱特征。针对非线性频谱故障征兆与多故障模式之间存在不确定性,采用朴素贝叶斯网络进行故障识别。建立了数控装备伺服传动系统模型,重点研究了伺服传动系统永磁同步电动机定子过热与机电过载故障的诊断问题。通过前3阶非线性频谱特征构建了伺服传动系统贝叶斯诊断网络模型,给出了故障检测与识别的具体算法。故障诊断实验结果表明提出方法的故障识别率高、实时性好,平均识别率和识别时间分别为97.5%和0.26 ms。A fault diagnosis approach is proposed based on nonlinear spectrum feature and Bayesian network for complex equipment driving system. In order to overcome the problem of calculated amount expansion of generalized frequency response function based on Volterra series, the single-dimensional non-linear output frequency response function is used to obtain transmission frequency spectrum feature. In view of the uncertainty between fault symptoms of nonlinear spectra and multiple fault modes, naive Bayesian network is used to identify fault. The model of servo drive system of numerical control equipment was constructed and the fault diagnosis of stator overheating and overload of permanent magnet syn- chronous motor were mainly studied. The Bayesian diagnosis network model was constructed based on the features of the first three orders nonlinear frequency spectra, then the algorithm of fault detection and i- dentification was given. The experimental results of fault diagnosis indicate the proposed method has high recognition rate and good timeliness, and the average recognition rate is 97. 5% and the average time is 0. 26 ms.

关 键 词：非线性输出频率响应函数 贝叶斯网络 复杂装备 传动系统 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]