多特征信息融合贝叶斯网络故障诊断方法研究76256.pdfVIP

中国机械工程第 2 1 卷第 8 期 20 10 年 4 月下半月 多特征信息融合的贝叶斯 网络故障诊断方法研究 姜万录 　刘思远 燕山大学 ,秦皇岛 ,066004 摘要 :针对轴向柱塞泵故障特征的模糊性和不完备性特点 ,提出一种多特征信息融合与贝叶斯网络 相结合的故障诊断方法 。该方法从柱塞泵采集的振动信号中提取出频域和幅域的多个故障特征 ,并将 这些特征当作来自多个不同传感器的多源信息 。利用贝叶斯参数估计算法进行多特征信息融合 。通过 构造贝叶斯网络并建立贝叶斯分类器来简化融合后的结果 ,通过最大后验概率估计值的计算进行故障 识别 。经过轴向柱塞泵多故障模式的诊断实验 ,验证了该方法能够有效地实现柱塞泵柱塞松靴和脱靴 故障的诊断 。 关键词 :柱塞泵 ;故障诊断 ;多特征信息融合 ; 贝叶斯网络 ( ) 中图分类号 : TP206 . 3 　　　文章编号 :1004 —132X 20 10 08 —0940 —06 Fault Diagnosis Approach Study of Bayesian Net works Based on Multi - characteristic Inf ormation Fusion J iang Wanlu 　Liu Siyuan Yan shan U niver sit y ,Qinhuangdao , Hebei ,066004 Abstract :Aiming at t he f uzzy and inco mp let e nat ure of f ault charact eri stics of axial pi ston p ump ,a met ho d of Baye sian net wor k s and multi - charact eri stic infor mation f u sion wa s p rop o sed . Fir st ly ,multi - fault charact eri stic s in t he f requency do main an d amp lit ude do main were ext ract ed f ro m vibration signal s and regar ded ,a s multi - so urce infor matio n s co ming f ro m diff erent sen sor s. Then ,Baye sian p a ramet er e stimatio n algorit hm wa s app lied to f u se multi - charact eri stic infor mation . N ext ,t he f u sion re sult wa s simp lified by co n st r ucting a Baye sian net wor k and e st abli shing Baye sian cla ssifier . Finally , t hrough calculatin g t he maximum po st erior p robabilit y e stimation an d t he fault p at t er n s were identi fied . The validit y of t hi s met ho d wa s verified t hrough exp eriment s of multi - f ault p at t er n s on an axial pi ston p ump .