论文详情
时频分析是地震数据处理和解释过程中重要的数学工具,其精度和分辨率决定了后续处理和解释成果的质量。提出了一种结合贝叶斯学习方法(sparse bayesian learning,SBL)和魏格纳威利分布(wigner-ville distribution,WVD)的两步高分辨率时频分析方法。第一步基于构建的雷克子波库和贝叶斯学习方法将地震数据分解为子波的线性组合;第二步通过求取子波的魏格纳威利分布获得地震数据的时频分布。其中,贝叶斯最大后验概率和第二型最大似然概率通过迭代求解。贝叶斯学习方法可以用最少数量的、具有不同主频和相位的雷克子波重构地震数据,并同时有效压制随机噪声。求取、分解子波的魏格纳威利分布可有效避免交叉项干扰,分辨率高。模拟数据和实际数据实验结果均验证了方法的正确性和有效性。与常规基于Gabor变换和匹配追踪算法的时频分析方法相比,该方法具有更高的精度和分辨率,有利于后续处理和解释研究。
The precision and resolution of time-frequency analysis determines the quality of processing and interpretation.Here,a two-step high-resolution time-frequency analysis method is presented,which combines sparse Bayesian learning (SBL) and Wigner-Ville distribution (WVD).First,on the basis of a built Ricker wavelet library and SBL,the seismic record is decomposed as a linear combination of wavelets;secondly,the WVD of the wavelets are calculated to obtain the time-frequency distribution of the seismic record.By iteratively solving a Bayesian maximum posterior and a type-II maximum likelihood problem,the Bayesian learning method can find the minimum number of Ricker wavelets with different frequency and phase that are needed to reconstruct the seismic record.By doing so,the method can effectively suppress random noise.Furthermore,the WVD of the decomposed wavelets can effectively eliminate cross-term interference,thereby achieving high-resolution imaging.An application to both synthetic and actual data samples showed higher precision and resolution of the proposed method in comparison with that of conventional time-frequency analysis methods based on Gabor transformation and the MP algorithm,which facilitates subsequent processing and interpretation.
国家科技重大专项“不同缝洞储集体地震识别与预测技术”(2016ZX05014-001)、国家科技重大专项示范工程“缝洞系统结构刻画及描述技术完善”(2016ZX05053-001)、中国博士后科学基金资助项目(2019M662005)和江苏省333工程科研项目(BRA2018308)共同资助。