Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Combining Heuristics and Machine Learning for Hybrid Flow Shop Scheduling Problems
This dissertation presents a new generic scheduling approach to makespan and flowtime minimization in hybrid flow shops with unrelated machines. In addition, possibilities of utilizing machine learning tools in job scheduling are explored.The proposed scheduling approach is twofold. First, a new heuristic based on the divide et impera strategy is introduced, that can be optionally enhanced by different local search improvement schemes. Moreover, several parameters are provided that enable a trade-off between solution quality and computation time.Then, a new hybrid variant of the heuristic is presented, that drastically reduces computation times through the application of machine learning techniques. Initially, the formal problem definition is given in form of a mixed integer linear programming formulation, and, based on an extensive literature review, the research gap is outlined. To evaluate the performance of the new approach, a testbed representing various production settings is introduced. Due to the 102, NP-hardness of the problem, optimal solutions cannot be determined for all instances. Therefore, several lower bounds are proposed for makespan minimization. Additionally, the well-known heuristic of Nawaz et al. (1983) is considered as benchmark for makespan and flowtime, as well as solutions of the mathematical optimization model determined in a specified time limit. A t-test for paired, dependent samples is conducted to statistically validate the results of the new approach. Furthermore, stability aspects of the results are evaluated and a study on computation times is included. Lastly, possible scope for future work is highlighted. ; Die vorgelegte Dissertation stellt einen neuen generischen Planungsansatz zur Minimierung von Makespan und Flowtime in hybriden Flowshops mit nichtidentischen Maschinen vor. Darüber hinaus werden Möglichkeiten des Einsatzes von Methoden des maschinellen Lernens in der Maschinenbelegungsplanung untersucht. Der vorgeschlagene Planungsansatz ist zweigeteilt. Zunächst wird eine ...
Combining Heuristics and Machine Learning for Hybrid Flow Shop Scheduling Problems
This dissertation presents a new generic scheduling approach to makespan and flowtime minimization in hybrid flow shops with unrelated machines. In addition, possibilities of utilizing machine learning tools in job scheduling are explored.The proposed scheduling approach is twofold. First, a new heuristic based on the divide et impera strategy is introduced, that can be optionally enhanced by different local search improvement schemes. Moreover, several parameters are provided that enable a trade-off between solution quality and computation time.Then, a new hybrid variant of the heuristic is presented, that drastically reduces computation times through the application of machine learning techniques. Initially, the formal problem definition is given in form of a mixed integer linear programming formulation, and, based on an extensive literature review, the research gap is outlined. To evaluate the performance of the new approach, a testbed representing various production settings is introduced. Due to the 102, NP-hardness of the problem, optimal solutions cannot be determined for all instances. Therefore, several lower bounds are proposed for makespan minimization. Additionally, the well-known heuristic of Nawaz et al. (1983) is considered as benchmark for makespan and flowtime, as well as solutions of the mathematical optimization model determined in a specified time limit. A t-test for paired, dependent samples is conducted to statistically validate the results of the new approach. Furthermore, stability aspects of the results are evaluated and a study on computation times is included. Lastly, possible scope for future work is highlighted. ; Die vorgelegte Dissertation stellt einen neuen generischen Planungsansatz zur Minimierung von Makespan und Flowtime in hybriden Flowshops mit nichtidentischen Maschinen vor. Darüber hinaus werden Möglichkeiten des Einsatzes von Methoden des maschinellen Lernens in der Maschinenbelegungsplanung untersucht. Der vorgeschlagene Planungsansatz ist zweigeteilt. Zunächst wird eine ...
Combining Heuristics and Machine Learning for Hybrid Flow Shop Scheduling Problems
Zacharias, Miriam (Autor:in) / Gottschling, Johannes
02.09.2020
Hochschulschrift
Elektronische Ressource
Englisch
New formulation for scheduling flexible flow shop problems
| Springer Verlag | 2023
A Hybrid Genetic Algorithm for Hybrid Flow Shop Scheduling with Load Balancing
| British Library Online Contents | 2009
Exploring the scheduling problems in a flow shop environment via simulation
| British Library Online Contents | 1996
Exploring the scheduling problems in a flow shop environment via simulation
| British Library Online Contents | 1996