Journal of Japan Society of Civil Engineers, Ser. A2 (Applied Mechanics (AM))
Online ISSN : 2185-4661
ISSN-L : 2185-4661
Journal of Applied Mechanics Vol.15 (Special Feature)
Acceleration of Image-based EFIT using GPU Computing and Validation by Ultrasonic Measurement
Kazuyuki NAKAHATATakayuki TANAKA
Author information
JOURNAL FREE ACCESS

2012 Volume 68 Issue 2 Pages I_227-I_236

Details
Abstract
For nondestructive testing, we developed a simulation tool of elastic wave propagation in a material with complex outer surfaces or various inclusions. The tool was based on the elastodynamic finite integration technique (EFIT) with an image-based modeling. In this study, the image-based EFIT code is accelerated using graphics processing units (GPUs) with the CUDA Fortran programming language. The methodology on the speed-up using the CUDA Fortran is described, and checked by numerical experiments with multiple GPU boards. The calculation speed with a GPU can be dramatically improved in comparison with the speed obtained by running the same simulation on a classical CPU. In addition, the validation of the image-based EFIT is performed by ultrasonic measurements of SH Lamb wave. The results of the simulation show good agreements with the measurement.
Content from these authors
© 2012 by Japan Society of Civil Engineers
Previous article Next article
feedback
Top