相对小波包能量谱的电机振声信号故障分类方法研究

Audio Fault Classification Research Base on Relative Wavelet Packet Energy Spectrum

文章研究利用相对小波能量谱作为特征的电机振声信号故障分类方法。对振声信号采用小波包分解,利用重构系数计算各个频段内的相对小波能量谱值,各频带的能量变化反映电机是否有故障;再根据标准样本故障结果建立"特征-故障"对应关系的信息表;以容差范围作为误差判别标准,力使故障诊断误差最小化,从而确定待测信号为何种故障;最后通过比较各频带相对小波能量谱大小判断故障所处的频带位置。电机振声信号诊断实例验证了该方法的有效性。

In this paper we reserch audio fault classification which use relative wavelet packet energy spectrum as the characteristics.It makes wavelet packet decomposition,uses decomposition coefficients to calculate the relative wavelet packet energy Spectrum,the energy changes in each frequency determine whether the motor is faulty.Establish ＂characteristics-fault＂ information table according standard faults.Tolerance range was introduced as an error standard to trive for diagnose error minization,which determine what kind of test fault.At last,comparing the relative wavelet energy spectrum in each frequency to determine the fault location.The example of audio fault detection has verified the effectiveness of the method.