|
K3D: 3-D Kirchhoff
Prestack Depth Migration Fusion's 3-D Kirchhoff prestack
depth migration (K3D) package is based on a
state-of-the-art implementation of the algorithm
designed to achieve maximum efficiency on
parallel-vector and cache-based supercomputers. The
algorithm includes several choices of anti-aliasing and
an amplitude equalization operator to preserve the
amplitude information in the final image even when the
spatial sampling of the input data is irregular. Through
scaling of the input and output data K3D can be run on
workstations as well as supercomputers, though the best
efficiency is obtained on larger computing platforms.
Geophysical
Aspects
Kirchhoff migration is recognized as the most flexible
method of imaging prestack 3-D seismic data. 3-D
prestack migration of surface seismic data is the
ultimate tool for imaging the Earth's subsurface in
areas of complicated geological structure and velocity.
Kirchhoff migration is particularly appropriate for 3-D
prestack data, because of its flexibility in imaging
irregularly sampled data and because of its relative
computational efficiency. Kirchhoff migration images
seismic data by approximately solving the wave equation
with a boundary integral method. The acoustic
reflectivity at every point of the Earth's interior is
computed by summing the recorded data on
multidimensional surfaces; the shape of the summation
surfaces and the summation weights are computed from the
Green's functions of the single scattering wave
propagation experiment.
Green's function
computation One of the main elements of 3-D prestack
Kirchhoff depth migration is the traveltime tables used
to parameterize the asymptotic Green's functions.
Fusion's GETT3D software module offers a very fast and
stable solution for generating the 5-D traveltime tables
needed for 3-D Kirchhoff migration. GETT3D is a 3-D
traveltime computation module based on an extremely fast
proprietary finite-difference algorithm (patent
pending). The algorithm is unconditionally stable in
complex velocity media and correct to the order of
accuracy of the finite-difference solution. Green's
function tables can be compressed before storage. K3D
offers both the option of using uncompressed or
compressed traveltime tables.
K3D: an advanced
migration algorithm At the core of the K3D package is
the data summation engine, designed to execute
efficiently on parallel-vector and cache-based
supercomputers. The K3D inner kernel uses very long
vectors to take advantage of vector architectures while
the output image is divided into processing blocks that
are distributed over parallel processors. The problem of
load balancing the computation across processors is
solved by exploiting the locality of the Kirchhoff
migration operator; that is, a seismic data record
collected in a certain area of the survey will mostly
contribute to the part of the seismic image that is
directly below the recording location. Locality is
exploited by designing an optimal partitioning scheme
for both the input and the output data.
The user can
choose from several anti-aliasing methods that enhance
the imaging and resolution of steep faults and
salt-sediment interfaces.
Scalable software
implementation 3-D prestack data processing requires
very powerful computers for efficiently handling the
large number of computations necessary to accurately
image the data. While more efficient on large platforms,
Fusion's K3D software can be run on a variety of computer
sizes and architectures. Through scaling of the input
and output data, the design of the K3D ensures that the
same problem can be solved on workstations or
supercomputers.
Easy Integration
The K3D package is modular in design and can be
customized and integrated into a client's specific
environment. K3D is available on Unix workstations,
parallel-vector and cache-based supercomputers, and
other architectures. If desired, Fusion can integrate the
package into a client's processing environment or add it
as a Promax, Focus or Views module. Of course it is
fully integrated into Fusion's
GeoPRO™ platform.

Back |