Mitigation of boundary-induced bias in image-based analyses for large soil deformations
Mitigation of boundary-induced bias in image-based analyses for large soil deformations
Image-based measurement of large soil deformations has become increasingly important in geotechnical physical modelling. The hybrid Eulerian-Lagrangian PIV-NP method provides an effective framework for measuring soil motion under large-strain conditions by combining numerical particles with Eulerian velocity fields. However, its reliability can be compromised by systematic biases near material boundaries, particularly when image masking introduces undefined (NaN) nodal information. This paper analyses the formulation-based origins of these border-induced errors and proposes two practical correction strategies: (i) geometric filtering of numerical particles based on mesh element definition, and (ii) a local nodal velocity replacement scheme to handle NaN values prior to interpolation. Validated against synthetic benchmarks with analytical kinematic solutions, the results demonstrate that these strategies effectively suppress artificial strain localization and boundary noise. The proposed corrections significantly enhance the robustness of PIV-NP for modelling flow-like landslides and large-deformation phenomena without altering its core computational framework.
N. M. Pinyol; Gabriela Morales; Mauricio Alvarado
11th International Conference on Physical Modelling in Geotechnics (ICPMG2026)
Special Session 5: Combination of numerical and physical modelling
https://doi.org/10.53243/ICPMG2026-419