Critical evaluation of PINN for FWD inverse analysis and differentiable FEM as an alternative
Yongjin Choi, Hyeonbin Moon, Seunghwa Ryu · 2026
Read Critical evaluation of PINN for FWD inverse analysis and differentiable FEM as an alternative by Yongjin Choi, Hyeonbin Moon, Seunghwa Ryu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-03210.txt.
Description
Choi, Moon, and Ryu test physics-informed neural networks on a pavement inverse problem, and find them wanting. The task is falling weight deflectometer backcalculation: infer layer stiffness from surface deflections. Methods based on automatic differentiation promise sharp gradients and fast convergence.
A standard PINN failed to recover the layer moduli. Sharp jumps between layers in the system break its soft physics fit. An extended PINN with domain decomposition did better, but stayed touchy about loss weights and network shape. Noise in the data hurt it more. A differentiable finite element method, DiffFEM, did better on all counts: accuracy, stability, and cost. It enforces the physics as a hard constraint, not as a loss term. The authors advise care in choosing between the two for inverse work.
Catalog record
| CREATOR | Yongjin Choi, Hyeonbin Moon, Seunghwa Ryu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.03210