Ernesto Bonomi, director, energy and environment programme at CRS4
C
ommon Refl ection Surface (CRS) theory was developed more than 10 years ago by the German school
of Karlsruhe to approximate, without the knowledge of an earth model, the time-of-fl ight of seismic impulses travelling along trajectories or rays, carrying at the surface the echo of the buried geological strata. At CRS4, the theory has become a mature processing technology to form a high-resolution (non-migrated) image of the subsurface layering, starting from the complete collection of recorded seismic traces. Since then, CRS has become an important tool. CRS theory models a
trajectory with eight unknown parameters (each one with a specifi c geophysical meaning) that provide a local description of the medium associated to each pixel of the image volume (a few Gbytes of output). The CRS application collects the seismic echo carried by independent CRS trajectories used as predictors. The adequacy, or coherence, to seismic data associated to each 8-tuple is measured by the semblance, the highly nonlinear multimodal objective function that must be maximised by the CRS application. Since the local description of the medium, not provided by the user, is inferred from a very large number of subsets of seismic traces, the resolution method is data-driven. Researchers in Imaging and
Numerical Geophysics at CRS4 and our industrial partner have pioneered the industrial use of CRS processing technology for oil and gas prospecting, developing (approximately 10 years ago) a 3D parallel application that concurrently synthesizes the collection of CRS traces forming (in time domain) the fi nal image. Since the simultaneous search of all CRS parameters may become a very intensive computational
www.scientific-computing.com
task, the problem, initially, was heuristically solved running a nested sequence of partial optimisations. In addition, it was not possible to consider a global search due to the small amount of available memory in 32-bit machines. With the acquisition volumes
getting increasingly larger, four years ago the CRS application was redesigned. Starting from the mathematical formulation, the application was implemented on a peer-to-peer datafl ow framework, capable of exploiting the reusability of the data across a large number of computing nodes, each one running asynchronously (a solution developed in collaboration with Nice, an Italian software house). The critical part remained the
design of a more accurate parallel optimizer. A couple of years ago, CRS4 developers introduced an optimisation strategy to effi ciently compute the parameters together. The algorithm starts, fi rst of all, from a cheap suboptimal solution based on a rough method, followed by an accurate search in the whole parameter space along a sequence of conjugate directions. Its implementation has then
demonstrated that it was possible to signifi cantly improve the result accuracy with a moderate increase of the computational cost. The computation of each trace can be treated as an independent unit of processing that takes in input about 3GB of seismic traces (complex numbers) and that, on a conventional X86 core, requires about 20 hours to output the result with a computing effort of thousands of cores for an industrial dataset. Since the coarse parallelism of the CRS application was already extremely optimised, to boost the performance one had to operate on the level of a single computing node. A fi ner level of parallel processing seemed the solution since the search of the best eight CRS parameters can be independently run for each pixel along a single output trace of the image volume.
CHRONOS ANALYSIS SOFTWARE
INTRODUCING:
FINITE DIFFERENCE TIME DOMAIN
LIMITED-TIME OFFER
CONTACT US FOR AN EVALUATION
Analyze and Model:
• Planar microwave and antenna structures
• Wire antennas • UWB antennas
• Microwave circuits, waveguides, and coaxial structures
• Near field and far field applications
Fast. Accurate. Easy-to-use. • Very short learning curve; no scripting required
• Highest level of support in the industry PUT OUR SOFTWARE TO THE TEST
CALL FOR AN EVALUATION AND START IMPROVING PRODUCTIVITY TODAY. A live demo is also available.
Call +1 204.632.5636 25
optimizing designs
email
info@integratedsoft.com or visit
www.integratedsoft.com/products/chronos
choronos_89x261.indd 1 JUNE/JULY 2011 4711-02-28 10:59 AM
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52