摘要
由于很大比例的故障来源于汽车后桥,因此在后桥出厂前对其装配建立一个质量检测是很有必要的。提出用小波分解对后桥装配进行在线检测的算法。首先在现有频率范围内比较正常和异常后桥振动信号FFT的不同。然后这2种信号用Daubechies母波分解成5个标准。基于逼近系数和详细系数,不同频率范围的信号得到重建。通过比较重建的正常和异常信号,发现D5最能反映故障特征。因此,D5被选为特征提取的主要标准。通过计算和比较一组采样信号的D5的统计变量(平均值、方差、边峰值和峰值),发现方差能有效地从正常信号中把异常信号区分出来,而且峰度能进一步地决定故障类型。本故障诊断的核心就是妥善选择正常和异常信号之间的边界方差。
Due to the large ratio of faults originated in automobile rear drive, it is necessary to establish a quality test for Rear Drive Assembly （RDA） before leaving factory. Ppresents an on-line fault diagnosis algorithm for RDA using wavelet decomposition. First the FFT of normal and abnormal RDA vibration signal are compared to determine their difference in existing frequency range. Then both signals are decomposed into five levels by Daubechies mother wavelet. Based on the approximation coefficients and detail coefficients, signals with different frequency ranges are reconstructed. By comparing the reconstructed of normal and abnormal signal, level D5 is found to most reflect the fault characteristic. Therefore, D5 is selected as the main level for feature selection. By calculating and comparing the statistic variabes （including mean, variance, skewness and kurtosis） of D5 for a group of sample signals, variance is found to be effective in distinguishing the abnormal signals from the normal ones, while kurtosis can be used to furthur determine the fault types. The fault diagnosis kernel is established by properly choose a boundary variance value between normal and abnormal signals.
出处
《煤矿机械》
北大核心
2010年第2期217-219,共3页
Coal Mine Machinery
关键词
故障诊断
汽车后桥装配
小波分解
特征提取
fault diagnosis
rear drive assembly
wavelet decomposition
feature extraction
