GPU-Accelerated Transient Electro-Magnetic Modelling

Yikuo Liu, Anton Ziolkowski, Paul Stoffa

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract / Description of output

We present a GPU-accelerated transient electromagnetic modelling method implemented in CUDA C where the
numerical efficiency can be improved by a factor of over 100 compared with the serial C code. The method uses
the Fourier pseudo-spectral method to solve the spatial derivatives, and uses the rapid expansion method (REM)
to solve the temporal evolution of the field by a summation of Chebyshev polynomials. The results are free of
numerical dispersion and accurate to the Nyquist in both space and time. We show that if the model consists only
of 1-D or 2-D structures, the computational effort to solve a 3-D field can be reduced by up to two orders of
magnitude without loss of accuracy. We demonstrate the accuracy of the code by a 2.5-D modelling example in a
VTI anisotropic half space. The significant improvement in numerical efficiency generalizes the use of REM in largescale,
time-domain EM modelling or inversion problems.
Original languageEnglish
Title of host publicationEAGE Annual 81st Conference and Exhibition
Publication statusPublished - 4 Jun 2019

Fingerprint

Dive into the research topics of 'GPU-Accelerated Transient Electro-Magnetic Modelling'. Together they form a unique fingerprint.

Cite this