Numerical integral operators of convolution type form the basis of most wave-equation-based methods for processing and imaging of seismic data. As several of these methods require the solution of an inverse problem, multiple forward and adjoint passes of the modelling operator must be performed to converge to a satisfactory solution. This work highlights the challenges that arise when implementing such operators on 3D seismic datasets and it provides insights into their usage for solving large systems of integral equations. A Python framework is presented that leverages libraries for distributed storage and computing, and provides an high-level symbolic representation of linear operators. To validate its effectiveness, the forward and adjoint implementations of a multi-dimensional convolution operator are evaluated with respect to increasing size of the kernel and number of computational resources. Our computational framework is further shown to be suitable for both classic on-premise High-Performance Computing and cloud computing architectures. An example of target-oriented imaging of a 3D synthetic dataset which comprises of two subsequent steps of seismic redatuming is finally presented. In both cases, the redatumed fields are estimated by means of least-squares inversion using the full dataset as well as spatially decimated versions of the dataset as a way to investigate the robustness of both inverse problems to spatial aliasing in the input dataset. We observe that less strict sampling requirements apply in three dimensions for these algorithms compared to their two dimensions counterparts. Whilst aliasing introduces noise in the redatumed fields, they are however deprived of the well-known spurious artefacts arising from incorrect handling of the overburden propagation in cheaper, adjoint-based redatuming techniques.
翻译:在3D 地震数据集中执行操作员的多重前方和连接通道以达到令人满意的解决方案。 这项工作凸显了在3D 地震数据集中执行这些操作员时出现的挑战,并揭示了它们用于解决大型集成方程式系统的情况。 Python 框架展示了3D 合成数据集的目标导向成像,该模型将图书馆用于分布存储和计算,并为线性操作员提供了高层次的象征性表示。为了验证其有效性,必须就不断增大的内核和计算资源的数量评估多维调操作员的前方和联合实施。我们计算框架还显示,在使用3D 流流中高精度储存和计算,3D 级合成数据集包括两个后续的地震再饱和度步骤。在这两种情况下,对多维度调调调的多维度操作员的前方和联合实施实施多维度操作员的前方和联合实施过程。我们计算框架的计算框架还适用于典型的3D 以目标为导向的合成数据集的成像,它们随后两个步骤是地震再饱和再饱和再现的轨方法,在精确的精确的域中,在精确的深度处理中以最不稳性数据流数据流中,在精确的轨中,在精确的流数据输入中,我们测为最差的流数据流数据流数据流中,在数据流中将数据流中将数据流中将数据流中将数据流中将数据流中,以最深的正确性数据流中将数据流中,在精确到最不易变为最深的实地判。