摘要
针对工况条件下轴承故障振动信号的非平稳特性,分析时变自回归与隐马尔科夫模型的特点,提出了一种基于时变自回归和隐马尔科夫模型的滚动轴承故障诊断方法.振动信号经时变自回归建模后,得到时频分辨率较高、无交叉干扰项的时频谱,基于能量法对时频谱进行特征提取,然后利用隐马尔科夫模型对故障特征统计分类,实现对轴承故障的诊断.轴承信号分析表明,TVAR建模可以有效地提取信号中的故障特征,结合隐马尔科夫模型的动态统计特性可智能识别轴承故障类型,得到良好的诊断效果.
Considering the non-stationary nature of the vibration signals of bearing fault under working conditions,a rolling bearing fault diagnosis method based on time-varying autoregressive(TVAR)model and hidden Markov model(HMM) has been proposed.After the vibration signals was modeled with TVAR,time-frequency spectrum with high time-frequency resolution and no cross term was obtained.Features of the time-frequency spectrum were extracted by energy means and then HMM was adopted to carry out statistic classification of the fault features and realize diagnosis of bearing fault.Analysis of the bearing signals shows that modeling with TVAR can facilitate effective extraction of the fault features from signals,and HMM has dynamic statistic characteristics,the integration of which can achieve intelligent identification of the fault type of bearing with satisfactory diagnosis result.
出处
《天津大学学报》
EI
CAS
CSCD
北大核心
2010年第2期168-173,共6页
Journal of Tianjin University(Science and Technology)
基金
国家自然科学基金资助项目(50805100
50705066)
国家科技支撑计划资助项目(2008BAF32B11)
关键词
时变自回归
隐马尔科夫模型
谱估计
故障诊断
time-varying autoregressive
hidden Markov model
spectrum estimation
fault diagnosis