Özyeğin Üniversitesi, Çekmeköy Kampüsü Nişantepe Mahallesi Orman Sokak 34794 Çekmeköy İstanbul

Telefon : +90 (216) 564 90 00

Fax : +90 (216) 564 99 99

info@ozyegin.edu.tr

Oca 14, 2021 - Oca 15, 2021

Dissertation Defense - Tonguç Yavuz (PHDIE)

 

Tonguç Yavuz - Ph.D. Industrial Engineering

Assoc.Prof. Ö. Erhun Kundakcıoğlu– Advisor

Asst.Prof. İhsan Yanıkoğlu – Co Advisor

 

Date: 15.01.2021

Time: 19:00

Location: This meeting will be held ONLINE. Please send an e-mail to gizem.bakir@ozyegin.edu.tr in order to participate in this defense.

 

Testing and Scheduling Applications Under Uncertainty

Thesis Committee:

Assoc.Prof. Ö. Erhun Kundakcıoğlu, Özyeğin University

Asst.Prof. İhsan Yanıkoğlu, Özyeğin University

Assoc.Prof. Okan Örsan Özener, Özyeğin University

Asst.Prof. Mehmet Önal, Özyeğin University

Prof. Tonguç Ünlüyurt, Sabancı University

Asst.Prof. Serhat Gül, TED University

 

 

 

 

Abstract:

In this thesis, we consider two different optimization problems under uncertainty. The first problem is identifying defective components in failed k-out-of-n systems. In the second part, we study the parallel machine scheduling problem under uncertainty.

In the first part of the thesis, we propose four exact and two heuristic solution methods for identifying defective parts in failed k-out-of-n systems with the minimum expected cost. We observe that Markov decision-based approaches perform better among the exact solutions where we consider the deterioration rates of all parts in the system based on historical knowledge. Additionally, we determine that dynamic programming for the proposed Markov Decision Process model performs better than linear programming. Since the number of states in the MDP increases exponentially due to the problem's nature, we show that this approach cannot be used on large scale problems. Thus, we propose two heuristic methods and two lower-bound approaches to determine the solution quality of these methods.

In the second part, we study parallel machine scheduling on plastic injection machines at Vestel factory, mimicking a real-life manufacturing problem. We propose two separate robust optimization reformulations and a branch-and-price algorithm that solve real-life instances in a reasonable time, which cannot be achieved with a commercial solver. Additionally, we demonstrate that the scheduling problems may have alternative optimal solutions for a worst-case (nominal) tardiness objective, whose performance under nominal (worst-case) processing times are remarkably different. Therefore, we propose Pareto efficient extensions to consider alternative solutions.

 

Bio: 

Tonguç Yavuz received his B.S. degree and M.Sc degree in Industrial Engineering from FMV Işık University. Then Mr. Yavuz started his Ph.D. Program in Industrial Engineering Department at Özyeğin University in 2016. His contribution to the department has also been recognized by rewarding him with the Best Teaching Assistant Award in the 2017-2018 academic year. He spent one year of his Ph.D. studies at the University of Florida with TÜBİTAK BİDEB 2214-A Fellowship Project under the supervision of Dr. Panos Pardolos. His research interests include decomposition algorithms for large-scale optimization and decision making under uncertainty.