摘  要：為抑制經驗模態分解（empirical mode decomposition，EMD）處理過程中的端點效應，在整理和研究現有方法的基礎上，提出一種鏡像延拓和極值平移相結合的端點處理方法，在最大程度地融合兩種傳統方法優點的同時盡可能地還原信號邊界特征。該方法通過構造特征平行四邊形使延拓極值處于理想區域，從而避免三次樣條差值過程中包絡線與信號交叉的產生，并引入Blackman窗函數對延拓信號進行邊界處理，進而有效地控制延拓誤差影響。經過仿真信號驗證與實測信號分析，對比鏡像延拓、極值平移與加窗邊界處理方法的端點抑制效果，證明該改進方法能有效地抑制分解過程中出現的端點效應，并能在高頻噪聲干擾下較完整地提取低頻信息，為燃氣輪機工頻特征的獲取提供可靠的保證。
關鍵詞：經驗模態分解；端點效應；鏡像延拓；極值平移；Blackman窗函數
文獻標志碼：A       文章編號：1674-5124（2016）01-0107-07
Improved EMD and its applications in gas turbine power frequency extraction
CUI Xinhan1，2， MA Liyuan1， WEI Zhonglin1， LI Shilong1， WANG Tianhui1
（1. The 4th Department，Ordnance Engineering College，Shijiazhuang 050003，China；
2. Baicheng Ordnance Test Center，Baicheng 137001，China）
Abstract: In order to dispose the end effect during empirical mode decomposition （EMD）， there comes the improved EMD. The treatment of end effect decides the extraction of signal time-frequency characteristics. based on the study of current approaches， an effective method has been proposed in combination with mirror extension and extreme point shifting. The new method comprises the merits of the two traditional approaches and serves as the maximum restoration of signal boundary characteristics. A characteristics parallelogram has been constructed to circumscribe the prolongation extremum within an ideal region， thus preventing the envelope curve from intersecting with the signal during the process of cubic spline interpolation. Blackman window function has been introduced to dispose the extended signal boundary so as to control the influence of prolongation errors. The simulation test and actual measurement show that the method can effectively restrain end diffusion and completely extract low frequency information under the interference of high-frequency noise， compared to mirror extension， extreme point shifting and windowing boundary treatment.
Keywords: EMD; end effect； mirror extension； extreme point shifting； Blackman window function