What question did this study set out to answer?

Investigate transverse shear locking in physics-informed neural networks applied to stiff structural mechanics equations and develop strategies to eliminate training deterioration.

August 14, 2026Computer Methods in Applied Mechanics and EngineeringOpen Access

Locking-free energy-based physics-informed neural networks for shear-deformable beams and plates

View Full Paper

Ask AI

Bookmark

Authors

TWTill WeithoffUniversität der Bundeswehr München LSLukas StrieflerUniversität Hamburg BOBastian OesterleUniversität Hamburg

Discussion

Member takes

Implication

Computational study demonstrates a locking-free neural network framework for beams and plates, enabling fast and slenderness-independent convergence.

Key Points

Investigate transverse shear locking in physics-informed neural networks applied to stiff structural mechanics equations and develop strategies to eliminate training deterioration.
Evaluated energy-based physics-informed neural networks (PINNs) across standard and hierarchic kinematic parametrizations for Timoshenko beams and Reissner-Mindlin plates.
Introduced a mitigation strategy combining direct shear deformation parametrization, decoupled neural networks for primary variables, and slenderness-dependent energy scaling.
Standard PINN formulations displayed severe shear locking, characterized by major training slowdowns and accuracy loss as structural slenderness increased.
The proposed hierarchic multi-network approach achieved robust, slenderness-independent training speed and accurate convergence across all primary kinematic variables.

Cite This Study

Weithoff et al. (2026) studied this question.

synapsesocial.com/papers/6a7ec69fb70b84ec8b912c22 https://doi.org/10.1016/j.cma.2026.119295

View Full Paper

Ask AI

Bookmark

Implication

Computational study demonstrates a locking-free neural network framework for beams and plates, enabling fast and slenderness-independent convergence.

Locking-free energy-based physics-informed neural networks for shear-deformable beams and plates

Discussion

Member takes

Implication

Key Points

Cite This Study

Locking-free energy-based physics-informed neural networks for shear-deformable beams and plates

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider