基于ROC-SVD的轴承故障检测  被引量:1

Fault detection of bearings based on ROC-SVD

在线阅读下载全文

作  者:于洪兵[1] 张亚岐[2] 周副权[2] 陈重均 YU Hongbing;ZHANG Yaqi;ZHOU Fuquan;CHEN Zhongjun(Baotou Vocational and Technical College, Baotou 014030, CHN;Technology Centre, Dongfeng Motor Corporation, Wuhan 430058, CHN)

机构地区:[1]包头职业技术学院,内蒙古包头014030 [2]东风汽车公司技术中心,湖北武汉430058

出  处:《制造技术与机床》2018年第2期100-105,共6页Manufacturing Technology & Machine Tool

基  金:国家自然科学基金项目(61374196;51178053);教育部长江学者和创新团队发展计划项目(IRT1286)

摘  要:在利用奇异值分解检测低信噪比的轴承故障信号时,奇异值能量阈值选择的随意性对检测结果有重要影响。为此提出利用受试者工作特征曲线(ROC曲线)对奇异值能量阈值进行优化,以降低阈值选择对检测结果影响。首先对当前滑动时窗内带噪信号进行降噪处理,通过奇异值分解初选奇异值阈值,通过Hilbert变换,得到当前时窗下故障频率;再对下一时窗信号进行检测时,将上一时窗奇异值阈值作为初值,利用ROC曲线对阈值进行修正,使得相邻两时窗故障频率误差在5%以内;并通过试验对所提及的检测算法进行验证,结果表明:ROC-SVD算法检测结果的一致性较好,跟实际的检测结果较为接近。通过该方法计算得到的轴承外圈故障频率为110 Hz,实际为109.35 Hz。Because arbitrariness selection of singular value energy threshold has an important influence on the detection results when the low signal-to-noise ratio of the bearing fault signal was detected by using singular value decomposition. This paper proposed to optimize the singular values of energy threshold through receiver-operating characteristic curve (ROC curve) in order to reduce the influence of the threshold selection on detection results. Firstly, the signal with noise in the current sliding time window was filtered. According to singular value decomposition, the primary singular value threshold can be determined and the fault frequency in current time window can be achieved by using Hilbert transform. The previous singular value threshold would be as initial value when the fault frequency in next time window would be detected. The threshold would be modified through ROC curve. It can ensure the error between the adjacent two window fault frequency was within 5%. In addition, a test would be adopted to verify this mentioned detection algorithm and he results show that ROC-SVD method has good consistency of detection results, and is relatively close to the actual test results. The fault frequency in bearing outer ring calculated by this method is 110 Hz and the actual value is 109.35 Hz.

关 键 词:低信噪比 故障 奇异值 ROC曲线 优化 

分 类 号:TH17[机械工程—机械制造及自动化]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象