A Real-to-Sim Framework for Evaluating Fragmented Material Dynamics from 3D Observations

Principal Investigator(s)

Emmett Ientilucci

Research Team Members

Fei Zhang

Michael Gartley

Project Description

We developed a real-to-sim framework for building dynamic digital twins of rock-fragment movement following blasting. Rather than representing the site as a static 3D model, it captures how the 3D scene evolves over time—effectively creating a four-dimensional digital twin (3D + time) of the blast. PhysTerra derives simulation-ready fragments and scene boundaries directly from reconstructed 3D meshes, reducing reliance on idealized shapes and manually defined geometry, while combining physics simulation with machine learning-based closed-loop calibration and quantitative evaluation. The core pipeline has been implemented and tested on three real-world blast scenes, with further validation and manuscript preparation underway.

Figures and Images

Post-Blast Muckpile

From observation to simulation: a post-blast muckpile captured by UAV (left) and its physics-based dynamic digital twin (right).