发动机故障分形诊断中噪声信号的分析和处理  

Analysis and processing of noise in engine fault diagnosis by fractal method

作　　者：张梅军[1] 陈江海[1] 侯宝科[1] 

机构地区：[1]解放军理工大学工程兵工程学院,江苏南京210007

出　　处：《解放军理工大学学报（自然科学版）》2006年第4期380-384,共5页Journal of PLA University of Science and Technology(Natural Science Edition)

摘　　要：为分析信号中噪声成分对信号关联维数计算结果的影响,提高利用信号关联维数诊断机械故障可靠性,模拟产生不同信噪比的含噪混沌信号,分析其关联维数随噪声水平的变化情况。证明噪声成分会极大影响到信号关联维的计算效果,利用含有噪声信号的关联维数刻画系统的特征行为是不可靠的,必须对获得的原始信号进行滤波消噪处理。通过对气阀正常、轻微漏气、较重漏气和严重漏气4种情况下,发动机启动电压信号降噪前后的关联维数对比研究,进一步证明,利用小波降噪对发动机启动电压信号进行预处理,可使得信号关联维数较为接近于系统的真值,极大地提高利用启动电压信号的关联维数判断发动机运行状态的准确性与可靠性。To analyze the influence of noise on the correlation dimension of mechanical fault signals and improve the reliability of the mechanical fault diagnosis using correlation dimension of signals ,the changes of correlation dimensions of simulated chaos signals with different noise level were studied firstly. It is langely unreliable to reflect the system＇ s characteristic with the correlation dimension of noised signal, which would be greatly influenced by noise. So the original signal must be de-noised prior to the correlation dimension calculation. The comparison between the correlation dimensions of the noised and de-noised startup voltage signals of the engine at different gas-leaking state that are normal, light, obvious and serious states also shows that the wavelet de-noising preprocessing on signals could make the correlation dimension quite close to the truth value and could greatly enhance the accuracy and reliability of the fault diagnosis by the correlation dimension of the engine startup signals.

关 键 词：发动机 故障诊断 小波消噪 分形 关联维数 

分 类 号：E92[军事—军事装备学] TK428[兵器科学与技术—武器系统与运用工程]