波叠加法在机械噪声故障特征提取中的应用研究

声学故障诊断中,测量到的噪声信号是现场所有声信号的混合.为提取待诊设备噪声故障特征,建立机器系统的多声源宽带相关混合声场模型,使用波叠加法重建源表面为任意形状的空间声压场分布,计算出未知声源的数目与位置.提出的算法具有计算速度快、重建精度高,能够消除其他噪声源信号的干扰,从较小的信噪比的观测信号中分离待监测源信号的功率谱,有效提取机械噪声故障特征.实验结果验证模型与算法的可行性.

MECHANICAL NOISE FEATURE EXTRACTION BASED ON WAVE SUPERPOSITION METHOD

Mechanical noise carries affluent information about the working condition of machinery. So, the acoustic signal can be employed to diagnose the mechanical fault. As a diagnosis method, it provides the advantage of easier measurement and so on. While there are always several machines running simultaneously, the signal carrying fault feature is sank in the mixture signals. Theoretical model is established to extract mechanical fault features of monitored sources, and the wave superposition method is proposed to compute the acoustic fields generated by arbitrary-shaped radiators in order to determine the number of sources and sources location. This method is reconstructed precisely and calculated quickly to remove the interference of uncorrelated sources extraction of mechanical noise signal and obtain power spectrum of monitored source from very low signal-to-noise ratio observed signals. At last, the experiment results made in semi-anechoic chamber demonstrate that the wave superposition method is available in mixture signal separation.