基于改进EEMD样本熵的高速列车滚子缺陷AE信号提取

摘要

A new method of high speed train roller AE signal extraction was presented.The EMD acted as a dyadic filter bank and noise with proportional amplitude and frequency was added to the tested data.The sifting number was set by frequencies from high to low.Then,IMF's segment sample entropy was calculated along the timeline,which took a larger proportion and was identified as an AE event.The experimental result shows that the Linear-Sinusoidal noise spectrum and sifting number could restrain the mode mixing and the little wave vanishing.The modified EEMD obtained a tangible physical meaning and improved results compared with the original EEMD.Segment sample entropy could capture the AE events in a continuous monitoring data.The AE signal was intuitively reflect in the corresponding Hilbert spectrogram.The roller AE signals were consistent with the theoretical calculation.%提出一种改进EEMD滚子缺陷声发射(AE)信号提取新方法,该方法根据EMD分解的二进滤波器组特性构造幅值与频率成线性-正弦规律变化的噪声添加进测试数据,给出了噪声构建原则,并按频率由高到低限定各阶IMF筛选次数,选取同一时段高频IMF归一化样本熵较小的数据段作为声发射事件参考.对实测数据计算表明特殊构造的噪声和筛选次数能有效抑制中低频段模态混叠和高阶IMF小波消失现象,改进后的EEMD方法分解出的IMF分量物理意义明确,性能优于传统EEMD方法.分段的IMF样本熵变化能在连续监测中捕捉声发射事件,对应的Hilbert谱能直观凸显出滚子缺陷声发射信号,滚动体AE信号事件周期与理论计算相吻合.