SECA: Strict-Elastic Compact Activation for Differentiable Elastoplastic Deformation
arXiv cs.GR details SECA, a Strict-Elastic Compact Activation method for differentiable elastoplastic simulation. The core idea is to preserve an exact elastic domain below yield while smoothing the onset of plastic flow over a finite overstress interval, which is a useful middle ground for inverse problems and differentiable simulation pipelines.
For developers, the practical appeal is in the update behavior: SECA is built into a Perzyna-type flow law and returns exactly zero plastic increment for sub-yield trials, avoiding the sub-yield tail that can complicate calibration and interpretation. The paper derives an exact discrepancy decomposition for scalar and proportional isotropic J2 updates, plus an algebraic budget for the transition width, giving implementers a clearer handle on accuracy versus smoothness.
The reported results are encouraging on the solver side. In a coarse cube-indentation run where the transition is crossed without intermediate increments, SECA reduced whole-run Newton iterations by 47% and backtracking halvings by 93% compared with return mapping. The method was also embedded in finite-deformation loading histories with contact and complete unloading, then differentiated to released equilibria.
Beyond the benchmark, the authors show controlled indentation staying closely matched, a four-method comparison over loading partitions, and a contact scene where the path derivative reaches a target in three iterations. That makes this especially relevant for graphics programmers and simulation-heavy teams looking for more stable differentiable material...
“SECA prescribes a bounded C2 activation over a finite positive-overstress interval.”
- what
- SECA introduces a Strict-Elastic Compact Activation method for differentiable elastoplastic simulation.
- who
- Authors: Yu Ren, Shuangjiu Xiao, and Deli Dong; source: arXiv cs.GR.
- when
- Submitted Jul. 27, 2026; revised Oct. 7, 2026.
- impact
- On coarse cube indentation, SECA cut Newton iterations by 47% and backtracking halvings by 93% versus return mapping.
Promising solver gains with stricter elastic behavior
Follow graphics updates
See relevant stories in your personalized news feed.
Discussion