Dual-free reduced-order modeling for parametrized variational inequalities with application to contact mechanics

Read full text Summarise with VIRAT
Authors: Sagbo Jean-Claude Mitchozounou, Amina Benaceur, Mohammed-Khalil Ferradi
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

We develop an energy-based primal-only reduced-order formulation for parametrized variational inequalities arising in contact mechanics. Standard reduced-basis methods for variational inequalities generally rely on mixed primal--dual formulations, for which the reduction of the dual variables is challenging because their non-negativity must be preserved. Building on the convex-hull principle for admissible displacement snapshots previously introduced in a monolithic primal--dual dictionary setting, the proposed formulation circumvents dual cone compression by minimizing an energy over a greedily compressed feasible primal set. Unlike the monolithic primal--dual approach, the online reduced problem requires neither a reduced contact-pressure basis nor a dual dictionary. Under the stated assumptions, the simplex-constrained representation guarantees primal feasibility with respect to the non-penetration constraints.When required, contact forces are reconstructed a posteriori from the reduced displacement using the optimality conditions of the original constrained problem, thereby preserving the complementarity conditions. The proposed reduction strategy is investigated for elastic frictionless contact under small-deformation and matching-mesh assumptions. Numerical experiments involving two- and three-dimensional configurations demonstrate the accuracy, robustness, and substantial online computational savings of the resulting reduced model.

Fields of Study & Subjects

Model Reduction and Neural Networks Contact Mechanics and Variational Inequalities Advanced Numerical Methods in Computational Mathematics

Publication details

Journal / source
HAL
Publication date
24 March 2026
Language
EN
Document type
Preprint