面向齿轮箱故障诊断的序贯概率比检验理论和方法

Theories and methods of gearbox fault diagnosis oriented sequential probability ratio test

提出了一种基于序贯概率比检验的齿轮裂纹故障诊断方法,并选用了无裂纹和有裂纹的齿轮模拟故障模式.实验中提取的振动信号夹杂着噪声等干扰,运用具有良好去噪效果的小波包方法对齿轮箱振动信号进行预处理.采用时域分析法提取预处理后信号的特征值,提取对冲击性振动非常敏感的峭度值作为特征值.将序贯概率比检验算法应用于齿轮箱故障模式的检验和识别.为了验证所提出方法的诊断能力,本文选用均方根误差的方法来计算同种故障之间,以及不同种类故障之间的识别误差,结果表明了所提出的方法是有效且强大的.

A novel method of fault diagnosis of gearbox based on sequential probability ratio test（SPRT） was proposed in this paper. The gear with the crack is used to simulate the gear fault mode. The wavelet packet transform was used to preprocess the signals. The time domain analysis was applied to extract the fault feature values of the vibration signals after preprocessing. The kurtosis value is used as the feature parameter because it is sensitive to impact vibration. The SPRT algorithm was developed to detect the gearbox fault mode. In order to test the diagnostic capability of SPRT,the root mean square error （RMSE） was used to calculate the errors between the same faults and between the different faults. The results show that the presented method was effective and robust.