基于EMD的声波全波列几种时频分析方法对比

时频分析方法是处理阵列声波测井数据的有效手段,但不同时频分析方法特性不同,处理效果也存在差异。为此,利用经验模态分解(EMD)将单点的实测声波全波列分解为有限个固有模态函数(IMF)分量,再采用短时傅里叶变换、Choi-Williams分布、平滑伪Wigner-Ville分布及重排谱图处理原始的声波全波列和前3个IMF分量,并从时间分辨率、频率分辨率和交叉项干扰3方面对比这4种方法的优缺点。结果表明:前3个IMF分量表征了声波全波列中不同的模式波,而4种方法均无交叉项干扰影响识别模式波的问题,其中重排谱图的时间分辨率、频率分辨率最高。引入时频分布的边缘特性,利用基于EMD的4种方法处理干层、水层、油水同层、油层井段的声波全波列,结果表明重排谱图使能量聚焦明显,能较清晰地显示能量很低的纵波,波峰时间、主频的定位更精准,对不同流体性质储集层的区分效果最好,进一步验证了重排谱图的优势。

Comparison of time-frequency analysis methods for acoustic full waveform based on empirical mode decomposition

Time-frequency analysis is an effective way to process non-linear and non-stationary signals such as the array acoustic logging data.The array acoustic logging signal is decomposed into finite intrinsic mode functions(IMFs)by empirical mode decomposition(EMD).Four time-frequency analysis methods,namely the short-time Fourier transform,Choi-Williams distribution,smoothed pseudo Wigner-Ville distribution and reassigned spectrum,were used to process the original array acoustic logging data,as well as IMF1,IMF2,and IMF3.These four methods were compared in terms of time resolution,frequency resolution,and cross term interference.The results showed that IMF1,IMF2,and IMF3 represent different components of the original array acoustic logging data.In fact,the four methods did not exhibit significant cross-term interference.Among them,the time and frequency resolution obtained by the reassigned spectrum method were the highest.By introducing the edge characteristics of time-frequency distribution,the four aforementioned methods based on EMD were also used to process the array acoustic logging data of dry layers,water layers,oil-water layers,and oil layers.The results showed that the reassigned spectrum method produced a clearer energy focus and the compressional waves with low energy were also displayed more clearly.Furthermore,the location of the peak time and main frequency was more accurate,and the performance in distinguishing the dry layer,water layer,oil-water layer,and oil layer was the best,thereby verifying the superior performance of the reassigned spectrum method.