声发射机械密封端面摩擦状态识别

摘要

针对现有监测技术对于机械密封端面摩擦状态的识别较难推广到工业现场进行实时监测，提出一种基于声发射的机械密封端面摩擦状态识别的方法。通过建立实验台对机械密封在工作过程中的信号进行采集；利用小波包分析法对信号进行降噪；对预处理后的数据进行特征提取；建立Elman神经网络对机械密封端面的摩擦状态进行识别。实验结果表明：通过建立的Elman神经网络对提取的声发射特征进行识别，能够很好地识别机械密封在工作过程中所处的摩擦状态。因此，提出的方法可以对机械密封端面的摩擦状态进行有效的实时监测。%Since existing monitoring technology to identify the friction state of the mechanical seal face is difficult to extend to industrial field for real-time monitoring, proposed a method which recognized the friction state of mechanical seal face based on acoustic emission. Signal of mechanical seals in working condition was collected by the established bench. Signal noise was reduced by wavelet packet analysis and features were extracted after the data preprocessing. Then, Elman neural network was established to identify the friction state of mechanical seal face. Experimental results showed that: Through the establishment of Elman neural network to identify the feature extraction acoustic emission signal, it could well recognize the friction state of mechanical seal face in working process. Therefore, the proposed method can effectively monitor mechanical seal face friction state in real time.