波场分解算法与逆时偏移角道集

Wavefield decomposition and RTM angle gathers

逆时偏移生成的角度域共成像点道集(ADCIGs)有利于提高成像质量,可用于偏移速度分析和AVA/AVO研究等。使用Poynting矢量估计波传播方向可以快速生成角道集,但当模型和波场比较复杂时,不同方向传播的波场分量相互重叠,无法准确计算波的传播方向。波场分解可以提高Poynting矢量的准确性。为此,研究了波场分解算法,在计算Poynting矢量之前先将波场分解成不同方向传播的分量,用于生成角道集。为解决传统的傅里叶变换在频率-波数域进行波场分解时计算量大的问题,对波场分解算法进行了优化,并提出了SSE(单指令多数据流扩展)向量化和多线程并行计算实现算法,从而提高了计算效率。理论和数值计算结果表明:使用波场分解可以显著提高逆时偏移角道集的质量,此优化算法能够得到与传统方法一致的结果并且提高了计算效率。二维Sigsbee2A模型和三维SEAM TTI模型的数值计算也证明了该算法用于生成高质量角道集和逆时偏移图像的有效性和稳定性。优化后的波场分解算法并不局限于上、下行波分解,可容易地拓展到其它方向的波场分解,也可应用于最小二乘偏移和全波形反演,以消除成像噪声,提高成像质量。

Angle domain common image gathers （ADCIGs） from reverse time migration （RTM） can help with AVO/AVA study, improve image quality, and perform velocity analysis.Using the Poynting vectors to estimate wave propagation directions can efficiently generate ADCIGs, but it loses accuracy when the model is complex and different events in the wavefields overlap.Wavefield decomposition can improve the accuracy of Poynting vectors.Therefore, we investigate the wavefield decomposition method, sepa rating the wavefield into different directional components prior to the calculation of Poynting vectors, and apply it to generate AD CIGs.The conventional algorithm is computationally intensive as it is implemented straightforwardly in frequency wavenumber （f-k） domain by Fourier transforms.To solve this problem, we investigate the optimization of the wavefield decomposition algorithm, and propose to implement it with vectorization by Streaming SIMD Extensions （SSE） and parallelization with multiple threads. The efficiency improvement is about five times.Theoretical and numerical examples showed that using wavefield decomposition can largely improve the quality of ADICGs, and the optimized algorithm can provide equivalent results to the traditional algorithm and improve the computational efficiency.The numerical examples from the 2D Sigsbee2A and 3D SEAM TTI models validate the effectiveness and robustness of the proposed algorithm for high quality ADCIGs and RTM images.The proposed algorithm is not limited to up down wavefield decomposition and is easy to extend to wavefield decomposition in other directions.It can be applied to least-squares RTM and FWI to remove artifacts and improve the imaging quality as well.