Wang, A., Pulatsu, B., Andrews, S. et Malomo, D..
2025.
« Breaking boundaries: Discontinuum failure analysis of dry-joint masonry using physics engine models ».
Engineering Structures, vol. 342.
Prévisualisation |
PDF
Andrews-S-2025-31237.pdf - Version publiée Licence d'utilisation : Creative Commons CC BY-NC-ND. Télécharger (8MB) | Prévisualisation |
Résumé
Discontinuum approaches, including the Distinct Element Method (DEM), are well-established for simulating the failure of dry-joint unreinforced masonry (URM) structures, particularly under in-plane (IP) shear-compression and out-of-plane (OOP) loading. However, they may be computationally intensive, with building-scale DEM analyses requiring up to 3 days for 15 s of seismic loading. This paper breaks boundaries between structural engineering and computer science by presenting the first systematic evaluation of PyBullet, an open-source physics engine based on Bullet Physics – originally conceived for visually plausible virtual animations – for simulating the mechanical response and collapse of dry-joint URM assemblies at different scales. Leveraging PyBullet’s rigid body algorithms, contact models, and efficient constraint solvers, 3D simulations were performed for IP shear-compression walls, settlement-induced damage in interlocking panels, and OOP tilting of URM. Results were benchmarked against experimental data and established discontinuum models – PyBullet predicted peak loads within + 16 % of DEM for IP shear-compression walls. Numerical stability was maintained with time steps in the order of 0.001 s, and full simulations completed within 5 min – up to 6 times faster than DEM. OOP tilting analyses reproduced expected collapse modes (diagonal cracking, overturning) with critical collapse angles underestimated by up to 32 %, largely due to premature block slippage linked to contact stiffness and friction force coupling. Settlement-induced failure in interlocking panels showed good agreement with experimentally observed failure patterns, with ultimate displacements within ±3 % for non-interlocking cases. The study demonstrates that PyBullet offers a computationally efficient alternative for dry-joint URM analysis, providing reduced runtimes and acceptable predictive accuracy, especially for preliminary or large-scale probabilistic assessments. Further refinement of contact stiffness calibration strategies would enhance predictive consistency, supporting the adoption of physics engines as viable alternatives to conventional discontinuum methods for rapid masonry collapse and debris simulations.
| Type de document: | Article publié dans une revue, révisé par les pairs |
|---|---|
| Professeur: | Professeur Andrews, Sheldon |
| Affiliation: | Génie logiciel et des technologies de l'information |
| Date de dépôt: | 30 juill. 2025 13:24 |
| Dernière modification: | 13 août 2025 22:55 |
| URI: | https://espace2.etsmtl.ca/id/eprint/31237 |
Actions (Authentification requise)
 |
Dernière vérification avant le dépôt |