Full text loading...
-
Parallel MR-RCGA Based FWI for Building Velocity Model
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 74th EAGE Conference and Exhibition incorporating EUROPEC 2012, Jun 2012, cp-293-00034
- ISBN: 978-90-73834-27-9
Abstract
We present a parallel algorithm for distributed-memory platform to perform Full-Waveform Inversion (FWI) of pre-stack seismic data in time domain that can be used for building velocity model. Our code is written in C and uses Message Passing Interface (MPI) and OpenMP for parallelization. This non-linear and non-unique inverse problem is solved by using Multiple Realization of Real Coded Genetic Algorithm (MR-RCGA); a stochastic search approach. For validation of algorithm, we applied it on realistic synthetic test consisting of an intersection point of dip and strike lines of the SEG/EAGE Overthrust data. Inverting two differently recorded data on the intersection point of two lines should yield the similar velocity model: it is done to demonstrate that the algorithm works with a very little start-up information and can give good estimate of compressional wave velocity (Vp) at each inverted CDP point. It is proposed to run this method on several CDP locations and interpolated in between, to get a good reliable starting velocity model. This method is developed and demonstrated on parallel clusters of PARAM Yuva. The results and benchmarking performance of the algorithm are presented and discussed.