Quantitative comparison of different ultrasound simulation approaches for civil engineering applications

Interpreting ultrasonic waveforms is often challenging, especially in the presence of tilted boundaries that introduce multiple reflections, mode conversions, and echoes. To analyze such complex signal patterns, wavefield simulations provide a valuable tool. To ensure confidence in the simulation results, the accuracy of the numerical method is crucial. Another consideration when using simulations is their computational expense. Especially in civil engineering applications, large structures need to be investigated, which may not be feasible given the available resources. In this study, different numerical methods and software implementations commonly used for simulations in civil engineering are compared, including the Elastodynamic Finite Integration Technique (EFIT), the Finite Element Method (FEM), and the Spectral Element Method (SEM). The first simulation algorithm was implemented in Fortran by the first co-author; the second and third by utilizing COMSOL and Salvus, respectively. This comparison is conducted through three case studies. In the first case study, the simulation results are compared to analytically determined reflection and transmission coefficients for a two-layered material. Here, COMSOL and Salvus show lower errors than EFIT, but all simulation algorithms achieve relative errors < 8 % . In the second case study, simulated waveforms of reflections from a circular void (2D) are compared to an analytical solution. Simulations were performed for differently fine grids/meshes using identical hardware. The results demonstrate that both EFIT and COMSOL require longer run times to reach the same level of accuracy in the simulated waveforms as Salvus. The third case study shows that an ultrasonic echo array measurement in a PMMA block containing a void can be accurately simulated. However, due to hardware limitations, COMSOL was only able to perform a 2.5D simulation, and not a full 3D simulation. Salvus and EFIT reached similar accuracies in this case.