论文详情
当存在噪声时,常规弹性阻抗分解算法不稳定,借助贝叶斯理论引入模型参数的先验分布作为弹性参数提取过程的正则化项,可以有效降低其不适定性。联合一阶差分矩阵和三变量柯西分布构建更为合理的稀疏约束项,使得三参数反射率“尖脉冲化”,提高了层分界面的识别能力;与此同时,引入低频软约束项,可以有效降低弹性参数剖面的“门帘”现象,提高了其横向连续性和光滑度。模型试算和实际数据测试验证了基于双项约束的弹性阻抗分解方法具有较好的稳定性和精度。
In the presence of the noise,the conventional method for decomposition of elastic impedance is not stable.In Bayesian framework,we can improve the stability of elastic impedance decomposition by introducing the regularization term.By combining the first-order differential matrix and trivariate Cauchy distribution,we can build a more reasonable sparse constraint item and obtain a high resolution reflectivity characterized by the three-parameter,thus improve the recognition of layer interfaces.At the same time,introduced low frequency soft constraint can reduce the “curtain” effect,which make the extracted elastic parameter sections smooth in the horizontal direction.The test results of synthetic and real data show that the method for decomposition of elastic impedance with two-term constraint is provided with high stability and accuracy.
国家自然科学基金项目(41502148)资助。