ArjunAir: updating and parallelizing an existing time domain electromagnetic inversion program

Results from ongoing work to parallelize the existing 2.5D airborne electromagnetic inversion program ArjunAir are presented here. ArjunAir is the only code known to the authors to see extended use in the mineral exploration industry for the rigorous inversion of time domain airborne electromagnetic (EM) data using a two-dimensional (2D) conductivity model. This study sought to increase the efficiency of the code by re-implementing the most computationally expensive calculations with modern high-performance routines, employing parallel algorithms wherever possible. Distributed memory and shared memory versions of the ArjunAir forward solver have been developed. Speedups as high as 23.7 for the distributed memory code and 15 for the shared memory workstation version, relative to the original code, have been achieved. Ongoing work is focused on developing a hybrid MPI/OpenMP forward solver, and on building a minimum structure inversion code using the new implementation of the forward solver. This will replace the existing inversion algorithm, which is based on a non-linear damped least-squares fit to the data.

[1]  Glenn A. Wilson,et al.  2.5D inversion of airborne electromagnetic data , 2006 .

[2]  K. Vozoff,et al.  The joint use of coincident loop transient electromagnetic and Schlumberger sounding to resolve layered structures , 1985 .

[3]  Michael A. Saunders,et al.  LSQR: An Algorithm for Sparse Linear Equations and Sparse Least Squares , 1982, TOMS.

[4]  Michael S. Zhdanov,et al.  3D inversion of airborne electromagnetic data , 2012 .

[5]  Douglas W. Oldenburg,et al.  Calculation of sensitivities for the frequency-domain electromagnetic problem , 1994 .

[6]  Niels B. Christensen,et al.  Optimized fast Hankel transform filters , 1990 .

[7]  Walter L. Anderson,et al.  Algorithm 588: Fast Hankel Transforms Using Related and Lagged Convolutions , 1982, TOMS.

[8]  F. Sugeng,et al.  Comparing the Time-Domain EM Response of 2-D and Elongated 3-D Conductors Excited by a Rectangular Loop Source. , 1993 .

[9]  David L.B. Jupp,et al.  Stable Iterative Methods for the Inversion of Geophysical Data , 2007 .

[10]  Jeffrey S. Ovall,et al.  A parallel goal-oriented adaptive finite element method for 2.5-D electromagnetic modelling , 2011 .

[11]  Kerry Key,et al.  Marine EM Inversion Using Unstructured Grids: A 2D Parallel Adaptive Finite Element Algorithm , 2012 .

[12]  Bruce M. Irons,et al.  A frontal solution program for finite element analysis , 1970 .

[13]  Douglas W. Oldenburg,et al.  Simultaneous 1D inversion of loop loop electromagnetic data for magnetic susceptibility and electrical conductivity , 2003 .

[14]  Olaf Schenk,et al.  Solving unsymmetric sparse systems of linear equations with PARDISO , 2004, Future Gener. Comput. Syst..

[15]  Nicholas I. M. Gould,et al.  A numerical evaluation of sparse direct solvers for the solution of large sparse symmetric linear systems of equations , 2007, TOMS.

[16]  W Hohmann Gerald,et al.  Numerical modeling for electromagnetic methods of geophysics , 1987 .

[17]  Wolfgang Fichtner,et al.  Efficient Sparse LU Factorization with Left-Right Looking Strategy on Shared Memory Multiprocessors , 2000 .

[18]  Patrick Amestoy,et al.  A Fully Asynchronous Multifrontal Solver Using Distributed Dynamic Scheduling , 2001, SIAM J. Matrix Anal. Appl..

[19]  Douglas W. Oldenburg,et al.  Three dimensional inversion of multisource time domain electromagnetic data , 2013 .