Enhancing decision support for solutions of packing problem in additive manufacturing: features, datasets and experimental studies
Publication Date
July 24, 2019
Creator
Abstract
Additive manufacturing (AM) encompasses a set of technological advancements that enable objects to be produced in an incremental layer-by-layer material deposition process. The advantages of such techniques include a more flexible production chain and the capacity to manufacture highly customised products. The manufacturing process takes place within an enclosed build container, referred to as a `build volume', which should be fully utilised to achieve more efficient production times and reduce costs. This requirement is at the core of cutting and packing problems, which are well-known combinatorial problems that have been algorithmically addressed by the operations research community. This study devotes particular attention to the understanding of three-dimensional irregular packing (3DIP) problems, i.e., the task of arranging arbitrary three-dimensional geometries. It is motivated by the necessity for more precise and well-informed terminology and categorisation criteria in this problem domain. The thesis also investigates the properties of existing 3DIP algorithms and the performance patterns with respect to build volume utilisation and the feature space. These topics have been scarcely addressed in the literature due to the amount of available data and relevant features on this problem domain. The primary objective of this work is to contribute to more efficient AM processes by assessing how volume utilisation can be maximised within the machine at every build. First, the research assists in the characterisation of 3DIP problems by introducing new measurements for assessing part complexity. Experiment results demonstrate that such metrics are suitable for describing entrant geometric features in non-convex three-dimensional objects. Second, this study extends the existing taxonomy for cutting and packing and provides the most significant benchmark for 3DIP in the literature, which is aligned with the challenging requirements observed in the AM environment. Third, it evaluates some of the most commonly used packing approaches based on the deepest bottom left with fill heuristic. Lastly, this thesis presents one of the first reported applications of algorithm selection to 3DIP problems, mapping the problem instance features, including the newly proposed ones, to the best packing algorithm. The results confirm the potential of the algorithm selection approach to deliver increased build volume utilisation in AM processes.
Item Type
ethesis
Thesis Type
PhD
Supervisors
Subjects (LC)
Associated Schools / Departments
School of Computer Science (UK)
eprints ID
56995
UoN Repository URI
Except where otherwise noted, this item's license is described as
File(s)![Thumbnail Image]()
Name
THESIS_LUIZ_ARAUJO.pdf
Type
Full-text
Description
Examined. Correct version
Size
4.45 MB
Format
Adobe PDF
Checksum (MD5)
1bed87c60ee2d4328976fa065e4b1644