Estimation of Yarn-Level Simulation Models for Production Fabrics

Georg Sperl, Rosa M. Sánchez-Banderas, Manwen Li, Chris Wojtan and Miguel A. Otaduy
ACM Transactions on Graphics (TOG), Proc. of SIGGRAPH, 2022





Abstract

This paper introduces a methodology for inverse-modeling of yarn-level mechanics of cloth, based on the mechanical response of fabrics in the real world. We compiled a database from physical tests of several different knitted fabrics used in the textile industry. These data span different types of complex knit patterns, yarn compositions, and fabric finishes, and the results demonstrate diverse physical properties like stiffness, nonlinearity, and anisotropy.

We then develop a system for approximating these mechanical responses with yarn-level cloth simulation. To do so, we introduce an efficient pipeline for converting between fabric-level data and yarn-level simulation, including a novel swatch-level approximation for speeding up computation, and some small-but-necessary extensions to yarn-level models used in computer graphics.



Citation

@article{sperl2022eylsmpf,
	author    = {Sperl, Georg and Sánchez-Banderas,  Rosa M. and Li, Manwen and Wojtan, Chris and Otaduy, Miguel A.},
	title     = {Estimation of Yarn-Level Simulation Models for Production Fabrics},
	journal   = {ACM Transactions on Graphics (TOG)},
	number    = {4},
	volume    = {41},
	year      = {2022},
	publisher = {ACM}
}

Description

We introduce the first technique for the modeling and estimation of yarn-level fabric mechanics that succeeds to capture the macroscopic (swatch-level) response of textile-production knitted fabrics. We achieve this through three main contributions:

Data set of real-world fabrics. We compiled a data set of physical test data from 33 different knitted fabrics used by industry professionals in the production of casual and sports garments. The fabrics span different knit patterns (e.g., multiple layers), yarn compositions (e.g., plated yarns), and yarn finishes. On a macroscopic level, they show diverse stiffness, nonlinearity, and anisotropy. The data set consists of physical information about each yarn type, manually registered yarn geometry, high-resolution photographic scans, and physical measurements from experiments for each fabric type, and it will be made available to the research community.

Efficient fitting procedure. To estimate yarn-level parameters from swatch-level physical tests, we have designed a two-step procedure that circumvents the computational cost of simulating full fabric swatches at yarn level: we first fit a thin-shell model to swatch-level data, then we generate analytical stress-strain data using the thin-shell model, and we finally fit a periodic version of the yarn-level model.

Practical and versatile simulation models. Basic models for yarns and thin shells cannot capture the diversity of behaviors of real-world knitted fabrics, while complex models with a large number of parameters are vulnerable to overfitting. After experimenting with many of these models and observing how well they fit real-world data, we propose a few minimal extensions to typical models used in computer graphics to help strike the balance between simplicity and expressive power: an anisotropic area-preserving thin shell model, and a yarn model with two-phase stretching and contact energies.

In this work, we focus on the major aspects of macroscopic mechanical response, including nonlinearity and anisotropy of stretch, shear, and bending deformation. We leave for future work more complex aspects such as extreme nonlinearity, hysteresis, or curling. Under these limitations, we maximize parallelism between the data, parameterization, and estimation processes of thin-shell and yarn-level fitting; we do this to minimize the error introduced by using the thin-shell model as an intermediate representation, while circumventing the challenge of simulating full-swatch non-uniform deformations at yarn level.



  • Contact

    Georg Sperl – georg.sperl@ist.ac.at
    Rosa M. Sánchez-Banderas – rosa.sanchez@seddi.com
    Manwen Li – manwenli9302@gmail.com
    Chris Wojtan – wojtan@ist.ac.at
    Miguel A. Otaduy – miguel.otaduy@urjc.es