基于EEMD_BP网络的滚珠丝杠副故障模式识别

针对在机械故障诊断领域,对信号的时频域处理分析提取特征值往往不能准确判断机械故障状态的问题。在对数控机床滚珠丝杠副振动信号研究中,提出了利用集合经验模态(EEMD)方法分析受到噪声干扰的3种不同状态的滚珠丝杠副振动信号。利用BP神经网络理论,以振动信号的时频域特征值及EEMD分解得到内禀模态函数(IFM)特征值作为输入,建立BP神经网络模型,并通过实验验证诊断网络模型的可靠性。

Fault Pattern Recognition of Ball Screws Based on EEMD_BP Network

In the field of mechanical fault diagnosis, extracting eigenvalues from time-frequency analysis of signals cannot accurately determine the mechanical fault state. Therefore, in the study of vibration signal from ball screw pair of NC machine tools, three different states of ball screw vibration signals that are affected by noise are analyzed by using EEMD method. Based on the BP neural network theory, the time-frequency eigenvalues of the vibration signal and the IMF from EEMD analysis are used to establish the BP neural network of fault diagnostic recognition in ball screw. And the reliability of network model diagnosis is verified through experiments.