University of Twente Student Theses
Preference-based multi-objective optimization: a comparative case study in the Dutch steel manufacturing industry
Wal, E.W. van der (2024) Preference-based multi-objective optimization: a comparative case study in the Dutch steel manufacturing industry.
PDF
1MB |
Abstract: | Economic and environmental factors are putting pressure on the business model of the steel manufacturing industry. Companies are exploring ways to improve their processes and use of materials to remain competitive. One approach focuses on optimizing which parts are cut from which stock material, known as the \textit{Nesting Problem}, a sub-problem of the family of \textit{Cutting Stock Problems}. However, current algorithms that focus on a single objective do not meet the demand of the industry. In this context, we investigate a multi-objective optimization approach to the nesting problem. Our approach focuses on incorporating the decision-maker's knowledge and preferences in the optimization. We first present a review of the state-of-the-art literature and an investigation into the factors that affect how a decision-maker evaluates solutions to the nesting problem. Then, we select two evolutionary algorithms based on this review and compare these on the ZDT and DTLZ test sets. From there, we apply one of these algorithms to a case study using cases from Dutch steel construction companies. We show that the multi-objective optimization outperforms the single-objective optimization, and gives the decision-makers more control over which solution they want to accept. We also show that, in the context of the nesting problem, the local optimization and repair steps in the evolutionary algorithm are profoundly impactful on which solutions the algorithm finds. Based on these observations, we postulate that some techniques that work on theoretical test sets may not have a significant impact when applied to real-world nesting cases, and propose various research directions based on this notion. |
Item Type: | Essay (Master) |
Clients: | Voortman Steel Group, Rijssen, The Netherlands |
Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
Subject: | 54 computer science |
Programme: | Business Information Technology MSc (60025) |
Link to this item: | https://purl.utwente.nl/essays/101555 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page