Machine learning-based algorithm selection for irregular three-dimensional packing in additive manufacturing - Research

Aug 3, 2026

Additive manufacturing faces the challenge of efficiently arranging arbitrary three-dimensional objects under geometric constraints. This work frames the task as three-dimensional irregular packing (3DIP) and applies machine learning for algorithm selection during the design stage. The study addresses how to match instance characteristics to a portfolio of packing algorithms, formalizing the relationship between problem features and algorithm performance. Extensive experiments evaluate supervised classifiers and parameters, build a labeled dataset from two widely used 3DIP algorithms, and analyze decision-support features. Results show ML-based selection can outperform standalone packing algorithms, improving average build volume utilisation by 1.48% across 2000 instances.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages26

Reading Time1 h 6 min

AccessFree to download

Introduction

Three-dimensional irregular packing in additive manufacturing

Algorithm selection and related work

Research contributions and objectives

Machine learning-based algorithm selection for irregular three-dimensional packing in additive manufacturing Luiz Jonat˜a Pires de Ara´ujoa,∗, Ender ¨Ozcanb , Jason A.D. Atkinb , Martin Baumersc , John H. Draked a The Machine Learning Group, School of Computer Science, University of Lincoln, UK b Computational Optimisation and Learning Lab, School of Computer Science, University of Nottingham, UK c Centre for Additive Manufacturing, Faculty of Engineering, University of Nottingham, UK d School of Computing and Mathematical Sciences, University of Leicester, UK Abstract In additive manufacturing (AM), a common problem is the efficient arrangement of arbitrary threedimensional objects, subject to geometric constraints. This can be mapped to three-dimensional irregular packing (3DIP) problems, which have been systematically addressed by many academicsand practitioners. This study demonstrates the utilisation of machine learning for algorithm selection to find efficient layout configurations during the design stage of the AM process. The choice of the most suitable approach to use, typically a packing algorithm, is not trivial, and depends on the non-obvious relationship between the characteristics of the instance in hand and the portfolio of algorithms available. The matching between problem features and algorithm performance forms the basis of the well-known algorithm selection problem. This study introduces the first empirical investigation of algorithm selection for 3DIP problems, conducting extensive experiments with hundreds of combinations of well-known supervised machine learning classifiers and different parameter settings to identify an initial state-of-the-art for this problem. We generate a comprehensive dataset, labelled with the performance of two of the most popular 3DIP algorithms, and analyse the features which can be used to support decision making when selecting a method to solve a 3DIP instance. Our results show that deploying machine learning-based algorithm selection methods are able to outperform the results obtained by the individual constituent packing algorithms applied independently, with the best algorithm selection method obtaining a 1.48% higher average build volume utilisation over the 2000 problem instances tested. Keywords: Heuristics, Algorithm Selection, Machine Learning, Additive Manufacturing, Packing 1. Introduction The objective of the three-dimensional irregular packing (3DIP) problem is to minimise the space wasted when configuring a set of parts within a constrained volume. 3DIP problems naturally occur in several traditional key industry sectors, such as logistics and transportation, among others ∗ Corresponding author Email addresses: [LAraujo@lincoln.ac.uk](LAraujo@lincoln.ac.uk) (Luiz Jonata˜ Pires de Arau´jo), Ender .Ozcan@nottingham .ac .uk (Ender ¨Ozcan), [Jason.Atkin@nottingham.ac.uk](Jason.Atkin@nottingham.ac.uk) (Jason A.D. Atkin), [Martin.Baumers@nottingham.ac.uk](Martin.Baumers@nottingham.ac.uk) (Martin Baumers), [john.drake@leicester.ac.uk](john.drake@leicester.ac.uk) (John H. Drake) Author’s Accepted Manuscript. Released under the Creative Commons license: Attribution 4.0 International (CC BY 4 .0)[https://creativecommons.org/l](https://creativecommons.org/l)icenses/by/4 .0/ deed.en[https://creativecommons.org/l](https://creativecommons.org/l)icenses/by/4 .0/ (Romanova et al., 2021) . This study focuses on a 3DIP problem in additive manufacturing (AM), an area receiving increased research attention in the contemporary literature (Wang and Cheung, 2022; Ying et al., 2022) . The methods presented here aim to contribute to more efficient AM processes, specifically in the context of laser sintering technology, leading to improved use of productive capacity and the sustainable use of resources. These improvements can be achieved by maximising the volume utilised in each build, reducing each printed item’s cost (Baumers et al., 2017) . Over the pas

FAQs

What problem does the document target in additive manufacturing?

It targets efficient packing of arbitrary 3D objects in a constrained build volume, formulated as the 3D irregular packing (3DIP) problem.

How does the study use machine learning?

It uses supervised machine learning to select among packing algorithms based on features of a specific 3DIP instance.

What evidence of effectiveness is reported?

The best ML-based algorithm selection method achieves 1.48% higher average build volume utilisation than using individual packing algorithms independently across 2000 test instances.

Aug 3, 2026

Uploaded by River Wang