Aristotle University Of Thessaloniki (AUTH)
Optimization techniques for production scheduling in flexible industrial processes
Abstract
dc:descriptionIn today’s era of rapid technological advancement, the industrial landscape is characterized by high uncertainty due to increased competition, limited profit margins and continuous market changes. At the same time, two of the main challenges in dynamic industrial environments include the efficient response to unforeseen production disruptions and the absence of specific production data. Consequently, effective production scheduling and rescheduling have emerged as key success factors. The above challenges of modern dynamic environments cannot be addressed using traditional methods whose required computational time increases significantly for larger problems. This thesis aims to address the existing research gap by developing practical solutions that enhance offline and real-time decision-making in complex, large-scale industrial environments. It leverages Artificial Intelligence (AI) techniques alongside mathematical optimization for rapid problem-solving. First, the lot-sizing and production scheduling problem of a real-life beverage industry is addressed. The plant produces numerous intermediate and final products through multiple production stages. The production process involves unique aspects, including the potential temporary storage of intermediate products and the employment of overtime shifts. A multi-bucket immediate precedence Mixed-Integer Linear Programming (MILP) framework is developed to address the problem while minimizing the plant’s operating time. Additionally, the mathematical model is extended to accommodate a new production layout including intermediate storage vessels. A comparative analysis between the two layouts is conducted, revealing significant productivity improvements which can be achieved through simple retrofit design options. Next, a novel solution framework is developed integrating Machine Learning (ML) techniques with MILP to address the optimization of production scheduling in multi-product manufacturing industries with parallel production lines. The main goal is the rapid schedule generation to address unexpected events and frequent changeover operations. The proposed framework compiles a representation space in which distances between products correspond to changeover times. Then, constrained clustering is employed to group products according to the available packing lines, minimizing changeover times within each cluster and the derived allocation is inserted into an MILP-based scheduling model to reduce its computational complexity. Furthermore, a ML model is trained to predict unknown product changeover times. An evaluation study based on a construction materials plant is conducted and it is shown that the proposed approach achieves accurate solutions rapidly, reduces downtime and facilitates the integration of new products into the production process. Finally, a hybrid framework combining Neural Networks (NNs) and MILP is presented to enhance production rescheduling in single-stage, multi-line dynamic manufacturing facilities. The integration of NNs is examined as an alternative solution to further speed up the decision-making processes. Specifically, the embedding and forecasting steps are realized through two NNs, whose outputs are used to improve the computational performance of the MILP scheduling model. The proposed approaches are evaluated against other standard methods and several cases concerning a real-life plant are considered, including a weekly production horizon with daily disruptions. Results show that the proposed method achieves better production schedules in significantly faster solution times. The outcomes highlight the flexibility and robustness of the proposed approaches, making them promising tools for decision-making in complex production environments.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Samouilidou, Maria-Eleni
- Σαμουηλίδου, Μαρία-Ελένη
Subjects
dc:subject × 17- Χρονοπρογραμματισμός παραγωγής
- Βελτιστοποίηση
- Μικτός ακέραιος γραμμικός προγραμματισμός
- Μηχανική μάθηση
- Production scheduling
- Optimization
- MILP
- Machine learning
- Επιστήμη Χημικού Μηχανικού
- Επιστήμες Μηχανικού και Τεχνολογία
- Χημική μηχανική
- Άλλες Επιστήμες Μηχανικού και Τεχνολογίες
- Οργάνωση παραγωγής και Μηχανική των κατεργασιών
- Chemical Engineering
- Engineering and Technology
- Other Engineering and Technologies
- Industrial and Manufacturing Engineering
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/60347