Robust Elasticity and Damping Models for High-Fidelity Textile Simulation

Abstract

Textiles are ubiquitous in our daily lives, being the fashion and textile industry one of the main drivers of the global economy. Despite technological advances, the fashion industry resists the irruption of digital design processes. The reason lies in the fact that existing computational models found in engineering, albeit accurate, do not satisfy the performance requirements compatible with industrial processes. Recently, advances in computer graphics simulation methods have resulted in the development of highperformance solutions for the simulation of fabrics at the yarn level that narrow the accuracy gap with engineering approaches. Although current trends point towards the emergence in the near future of high-fidelity models that will satisfy the requirements for predictive digital prototyping, today there remain open challenges that limit the achievement of this objective. In this thesis, we address some of these challenges and contribute towards the creation of high fidelity simulation models. To this end, we approach two relevant aspects for the characterization and behavior of cloth. First, we address the modeling of dissipative behavior in cloth. We propose a framework for the design of damping forces based on dissipation potentials formulated as functions of strain rate. We study its application to continuum and discrete deformation models, its practical implications, and advantages over commonly adopted dissipation models. We demonstrate its applicability to two highly different deformation models, Saint Venant-Kirchhoff (StVK) hyperlasticity and yarn-level cloth with sliding persistent contacts. Then, we address the challenge of simulating multilayered clothing. We introduce a robust method for the simulation of complex rod assemblies and stacked layers with implicit contact handling. Previous methods fail to robustly handle such complex situations due to ubiquitous and intrinsic degeneracies. Our novel discretization supports accurate and efficient contact while remaining insensitive to such degeneracies. Our solution is simple and elegant, producing robust simulations even on large-scale scenarios with pervasive degeneracies.