Unlocking Thickness Modeling for Codimensional Contact Simulation

TL;DR

This work addresses the fundamental problem of contact-locking artifacts in codimensional yarn- and shell-model fabrics with thickness as mesh resolution increases. It introduces a barrier-filtering collision-processing approach that pre-labels near-thickness contact stencils, computes a conservative minimum distance $ar{d}_{min}$, and applies a safe-filtered barrier $ ilde{b}(d_{i,j}(x), h)$ to preserve thickness modeling without inducing non-physical forces, all integrated with the IPC barrier and capable of handling biphasic thickness via $etaeta$. The method preserves thickness-consistent contact behavior across rod and shell models, enabling intersection-free, robust simulations across a wide range of real-world yarns, fabrics, and mesh resolutions, including irregular and graded meshes. Practically, this yields accurate, artifact-free equilibrium and dynamic simulations of complex knitted and woven materials without sacrificing resolution or robustness, improving predictive capabilities for real-world textile design and analysis.

Abstract

In this work we analyze and address a fundamental restriction that blocks the reliable application of codimensional yarn-level and shell models with thickness, to simulate real-world woven and knit fabrics. As discretizations refine toward practical and accurate physical modeling, such models can generate non-physical contact forces with stencil-neighboring elements in the simulation mesh, leading to severe locking artifacts. While not well-documented in the literature, this restriction has so far been addressed with two alternatives with undesirable tradeoffs. One option is to restrict the mesh to coarse resolutions, however, this eliminates the possibility of accurate (and consistent) resolution simulations across real-world material variations. A second alternative instead seeks to cull contact pairs that can create such locking forces in the first place. This relaxes resolution restrictions but compromise robustness. Culling can and will generate unacceptable and unpredictable geometric intersections and tunneling that destroys weaving and knitting structures and cause unrecoverable pull-throughs. We address these challenges to simulating real-world materials with a new and practical contact-processing model for thickened codimensional simulation, that removes resolution restrictions, while guaranteeing contact-locking-free, non-intersecting simulations. We demonstrate the application of our model across a wide range of previously unavailable simulation scenarios, with real-world material yarn and fabric parameters and patterns, challenging simulation conditions and mesh resolutions, and both rod and shell models, integrated with the IPC barrier.

Figures (18)

• Figure 1: Contact-based locking occurs for standard codimensional barrier models when local resolution in the reference mesh (a simple polyline midline segment in this example), has distances for contact stencils, here e.g., between vertex $V_0$, and vertices $V_2$ and $V_4$, below the activation threshold, $a$. In turn, this generates nonphysical expansive contact forces, $F_C$ for these stencils, leading to inflated domains (even at rest), artificially stiff material responses, and nonzero rest forces. See Figure \ref{['fig:show_pushing_artifacts_toy_example']}.
• Figure 2: Contact locking in codimensional simulations. Gravityless equilibrium of a 5cm yarn with only contact forces enabled. Left: contact forces at rest when varying resolution (top, fixed thickness) or thickness (bottom, fixed resolution). Note that, at rest, no contact forces should be active. Right: due to nonphysical forces (see Figure \ref{['fig:diagram_barrier_contact_locking']}), the yarn expands beyond its rest length when the resolution is too fine for the chosen thickness. This limits both feasible resolution and achievable accuracy.
• Figure 3: Resolution vs. material thickness. In barrier models, yarn thickness limits the maximum resolution achievable without artifacts. Here, we simulate the DKP pattern with a 150D/72F Polyester yarn Sperl2022 using our filtered barrier (green) and the standard barrier Li2021CIPC (red), from coarse to fine resolution (bottom to top; discretizations at left). Bottom: at coarse resolution both methods agree, but accuracy is low. Middle: the barrier method exhibits locking artifacts (expansions and gaps), while our method remains stable. However, the resolution is too coarse for many applications. Top: at finer resolution, our filtered barrier models the real pattern smoothly, whereas locking completely destroys the standard barrier solution.
• Figure 4: Knot test. A 0.3 mm thick knotted yarn is pulled from both ends. We compare our barrier-filtered simulation (green) with a contact-culled simulation (orange) Kaldor2008, using the smallest culling radius that removes expansive contact forces. Left: initial polyline knot configurations at different resolutions. As the knot is pulled, all culled simulations fail—the knot breaks apart due to intersections and pull-throughs, regardless of resolution or culling radius. In contrast, our barrier-filtering preserves the knot’s tightened structure at all resolutions, free of expansion artifacts.
• Figure 5: Yarn relaxation comparison with real-world captures. We compare the relaxation simulation results of the A1 pattern with real-world images both published in Sperl et al. Sperl2022. Close-up insets highlight differences between the simulated and physical patterns. Our barrier-filtered method accurately reproduces the relaxed yarn configuration without introducing artificial expansion artifacts. In contrast, the barrier method Li2021CIPC introduces spurious contact forces that distort the overall pattern, resulting in noticeable deviations from the real-world reference pattern.