M.S. Thesis Presentation: Leveraging Large-Scale Data For Supply Chain Network Design: A Location-Allocation Model for Rwanda by Zeynep Göze Gürkan
Thesis Advisor: Assoc. Prof. Ayşe Selin Kocaman
Thesis Co-Advisor: Research Asst. Prof. Pablo Duenas Martinez
Date: August 26 Monday 2024
Time: 10:00
Place: Zoom
This is an online seminar. To obtain event details please send a message to department.
Abstract:
Clean cooking strategies are significant contributors to the enhancement of development and sustainability. Regarding the lower emission levels compared to biomass usage, we consider Liquefied Petroleum Gas (LPG) a clean cooking strategy to promote, especially in developing countries. Hence, we design large-scale supply chain operations for the LPG distribution in Rwanda. This involves addressing the location-allocation problem of facilities by utilizing a large dataset on the location and LPG demand of each rooftop by formulating a Mixed-Integer Linear Programming (MILP) model. In order to decrease the size of the problem, we propose three methods. First of all, we design the system on a monthly basis rather than annual basis by adhering to some conditions. Next, we use the agglomerative hierarchical clustering-based heuristic approach to cluster the rooftops and locate retailers on the distance-constrained geomedian point of each cluster. Finally, we propose to decompose the formulated MILP model to get adequate solutions in less time. For computational analysis, we compare the system configurations with different retailer locations obtained by the village centroid approach and agglomerative hierarchical clustering based-heuristic approach. In addition, we investigate whether the existing system configuration can be extended when the projected increase in yearly LPG demand is introduced. Moreover, we conduct a sensitivity analysis to show the trade-off between the infrastructure and transportation costs due to the volatility in diesel fuel prices. Finally, we compare the results and performances of the main model and the decomposed model.