Kolmogorov-arnold networks for automated packaging design
Use Kolmogorov-Arnold Networks (KANs) to automate packaging design by learning from preform design data and predicting container shapes and characteristics from design parameters. The approach targets over 95% prediction accuracy and converts complex relationships into human-interpretable functions, enabling faster, lower-cost exploration of packaging designs, particularly PET containers.
Overview
The Kolmogorov-Arnold Networks (KANs)-driven data mining solution provides an approach to automated packaging design. Using a database of preform designs and their associated container outcomes, the solution applies KANs to learn the relationship between design parameters and resulting container shapes and characteristics. The goal is to predict new container designs with high accuracy, reducing time and cost compared with traditional design methods.
Technical specifications
Features:
- Utilizes Kolmogorov-Arnold Networks for data mining
- Predicts container shapes and characteristics with over 95% accuracy
- Transforms complex data relationships into human-interpretable functions
- Applicable to a wide range of packaging designs, particularly PET containers
Process:
- Stage 1: Train and test the model on existing data to achieve high prediction accuracy
- Stage 2: Validate predictions on new design parameters through practical experiments
Technology readiness level
Currently at Technology Readiness Level 2, this solution is in the concept validation phase. Initial tests will be conducted using existing data, followed by practical experiments. The development is structured over a two-year period, focusing first on model training and testing, then on validation with new designs.
About Victoria University of Wellington
Te Herenga Waka—Victoria University of Wellington is a comprehensive public research university serving New Zealand’s capital, with campuses embedded across the city and a strong civic orientation. Industry collaboration is anchored by research institutes based in the Gracefield Innovation Quarter alongside national innovation infrastructure, providing access to advanced laboratories and workshops for joint development. The Pipitea campus hosts the Taiawa Wellington Tech Hub, while proximity to Parliament and government agencies supports regulatory engagement and public‑sector partnerships. Research is supported by competitive national funding and international grants. Wellington UniVentures, the University’s commercialization company, manages IP, licensing, and spin‑outs to accelerate market translation.
delivered to your inbox.