摘要
提出了一种基于小波包分解与EMD的故障诊断特征提取方法。对故障振动信号进行小波包分解,并将其高频部分节点信号进行重构,对2个节点的重构信号分别进行EMD分解,得到一系列的IMF分量;提取每个节点的各个IMF分量的能量值并归一化后作为轴承的故障特征量输入神经网络进行诊断。通过实验证明2种方法的结合具有良好的局部分析能力及自适应分解的特点,可以提取更加有效的特征值,因此在进行诊断时,具有更快的速度与更高的准确率。
The method of feature extraction based on wavelet packet decomposition and Empirical Mode Decomposition is proposed in this paper. The vibration signal of fault is decomposed by the method of wavelet packet decomposition. The high frequency part of the node signal is reconstructed. After decomposing the two reconstructed nodes of signals by the method of EMD respectively, a series of the IMF component are acquired. The IMF component of each node's energy is calculated and defined as the input neural network after their normalization. Experiments prove that the combination of the two methods have good capability of performing local analysis and also the characteristics of adaptive decomposition. It can extract more effective characteristic values. Therefore the proposed method has faster speed and higher accuracy for diagnosis.
出处
《煤矿机械》
2015年第2期270-272,共3页
Coal Mine Machinery
基金
国家自然科学基金项目(51075220)
青岛市基础研究计划项目(12-1-4-4-(3)-JCH)
关键词
小波包分解
EMD
故障诊断
wavelet packet decomposition
empirical mode decomposition
fault diagnosis