Graph neural network comparison for 2D nesting efficiency estimation

Corentin Lallier (), Guillaume Blin, Bruno Pinaud and Laurent Vézard
Additional contact information
Corentin Lallier: Univ. Bordeaux, CNRS, Bordeaux INP, LaBRI, UMR 5800
Guillaume Blin: Univ. Bordeaux, CNRS, Bordeaux INP, LaBRI, UMR 5800
Bruno Pinaud: Univ. Bordeaux, CNRS, Bordeaux INP, LaBRI, UMR 5800
Laurent Vézard: Lectra

Journal of Intelligent Manufacturing, 2024, vol. 35, issue 2, No 22, 859-873

Abstract: Abstract Minimizing the level of material consumption in textile production is a major concern. The cornerstone of this optimization task is the nesting problem, whose goal is to lay a set of irregular 2D parts out onto a rectangular surface, called the nesting zone, while respecting a set of constraints. Knowing the efficiency—ratio of usable to used up material enables the optimization of several textile production problems. Unfortunately, knowing the efficiency requires the nesting problem to be solved, which is computationally intensive and has been proven to be NP-hard. This paper introduces a regression approach to estimate efficiency without solving the nesting problem. Our approach models the 2D nesting problem as a graph where the nodes are images derived from parts and the edges hold the constraints. The method then consists of combining convolutional neural networks for addressing the image-based aspects and graph neural networks (GNNs) for the constraint aspects. We evaluate several neural message passing approaches on our dataset and obtain results that are sufficiently accurate for enabling several business use cases, where our model best solves this task with a mean absolute error of 1.65. We provide open access to our dataset, whose properties differ from those of other graph datasets found in the literature. This dataset is constructed on 100,000 real customers’ nesting data. Along the way, we compare the performance and generalization capabilities of four GNN architectures obtained from the literature on this dataset.

Keywords: Fashion manufacturing; 2D bin packing; Machine learning; Graph neural networks (search for similar items in EconPapers)
Date: 2024
References: View complete reference list from CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s10845-023-02084-6 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:joinma:v:35:y:2024:i:2:d:10.1007_s10845-023-02084-6

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/10845

DOI: 10.1007/s10845-023-02084-6

Access Statistics for this article

Journal of Intelligent Manufacturing is currently edited by Andrew Kusiak

More articles in Journal of Intelligent Manufacturing from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().