改进的自适应EEMD方法及其应用

Improved Adaptive EEMD Method and Its Application

针对目前总体经验模式分解(EEMD)方法中两个参数依靠人工选择难以准确获取的问题,提出了一种可自适应确定二者取值的改进EEMD方法。首先通过提取信号中的高频成分来确定加入白噪声的幅值,再根据减小白噪声影响的统计规律得到总体平均次数。同时,为提高分解效率及降低白噪声影响,在EEMD过程中引入有限带宽高斯白噪声消除模态混叠,实现对信号的快速准确分解。信号仿真试验表明改进EEMD方法可以得到比EMD和原始EEMD更加高效的分解结果。最后将其应用于混合信号输入的模拟电路故障特征提取中,以输出响应EEMD分解得到的IMF能量作为特征进行不同故障的分类,仿真结果表明该方法提取的电路各状态特征可作为故障识别和诊断的依据。

To solve the problem of two parameters obtaining difficultly by artificial selection in ensemble empirical mode decomposition（EEMD）, an improved EEMD method which could obtain their values adaptively was proposed. Firstly, obtaining the amplitude of added noise by extracting the high frequency information, then, according to the statistic law of eliminating noise effect, it was convenient to obtain the number of ensemble members. Meanwhile, in order to improve decomposition efficiency and decrease noise interference, a band-limited noise was used to replace random white noise for decomposing the signal exactly. Signal simulation example indicates that the improved EEMD method can obtain the more efficient result than EMD and original EEMD. Finally, the proposed method was applied to fault features extraction of analog circuit with mixed-signal input, and the IMF energies of output response were extracted as features to class different faults. The results show that the features extracted by improved EEMD can be the basis of faults identification and diagnosis.