A Huygens subgridding approach for efficient modelling of Ground Penetrating Radar using the Finite-Difference Time-Domain method

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

Authors

External departments

  • University of Edinburgh

Details

Original languageEnglish
Title of host publication2018 17th International Conference on Ground Penetrating Radar, GPR 2018
PublisherIEEE
ISBN (Electronic)9781538657775
ISBN (Print)978-1-5386-5778-2
DOIs
Publication statusPublished - 23 Aug 2018
Event17th International Conference on Ground Penetrating Radar, GPR 2018 - Rapperswil, Switzerland
Duration: 18 Jun 201821 Jun 2018

Conference

Conference17th International Conference on Ground Penetrating Radar, GPR 2018
CountrySwitzerland
CityRapperswil
Period18/06/1821/06/18
Publication type

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

Abstract

Numerical modelling of Ground Penetrating Radar (GPR) using the Finite-Difference Time-Domain (FDTD) method can be computationally intensive when large volumes must be discretised at fine spatial resolutions. This requirement exists because of the conditionally explicit nature of the FDTD algorithm, and the need to simulate fine geometrical details as commonly found in GPR antennas or in regions of high permittivity dielectric materials. We have developed and implemented a Huygens subgridding (HSG) approach in the open source software gprMax. The HSG can be applied to multiple localised regions of the FDTD grid where it is required, whilst the rest of the grid can be discretised at a more appropriate spatial resolution. An example of applying the HSG to a GPR model that includes a detailed antenna model, demonstrates the accuracy and efficiency of the approach. The relative error compared with a uniform fine grid is generally less than 1.3%. The computational time is reduced by a factor of 16. This level of computational saving is especially important for optimisation and inversion algorithms where many GPR forward models are computed in a single iteration.