基于SVR消除EMD端点效应的研究及其在汽轮机油膜涡动故障中的应用

Research on Eliminating End Effect of EMD Based on SVR and Application in Detecting Turbine Fault

目前,经验模态分解(EMD)广泛应用在信号处理中,但应用过程中不可避免会有端点效应,如果处理不好会"污染"整个数据序列而使所得结果严重失真。对于边界问题的处理采用波形延拓是比较理想的一种方法,利用支持向量回归机(SVR)对原始信号两端进行波形延拓来处理端点效应并应用于汽轮机油膜故障分析中。该方法首先利用SVR分别对波形两端进行延拓,然后对延拓后的信号进行EMD分解,得到结果的中间部分即为原信号的EMD分解结果。实验结果表明,该方法能有效抑制EMD方法的端点效应,得到准确的分析结果。

Presently, empirical mode decomposion （EMD） is widely used in the field of signal processing, but inevitably there will be end effects when use it. If this problem is not treated properly, the whole date sequence will be contaminated and caused the wrong result. Waveform extension is an ideal method for end effect. Support vector regression machine （SVR） is used to extend both ends of the original signal waveform for end effects and this method is applied to the diagnosis of turbine oil film. First, SVR is used to extend both ends of the original signal waveform, then carry on the EMD decomposition. The middle part of the final wave shall be the EMD decomposition of the original signal. The results of experiment shows that the end effect can be controlled effectively and has received the accurate results.