Afet sonrası enkazın kaldırılmasındaki gecikmeyi minimize etmek

After a natural disaster, roads and bridges can be damaged or blocked by debris, causing inaccessibility between critical locations such as hospitals, disaster response centers, shelters and disaster-struck areas. To facilitate immediate response operations, it is essential to clear debris from the roads in the shortest time possible. Moreover, in the longer term, management of waste and restoration of damaged roads are important activities for recovery of the region. In this thesis, we focus on the immediate response phase, such as the first 72 hours after the disaster and aim to optimize debris clearance decisions. First, we study the post-disaster road clearing problem with the aim of providing a fast and effective method to determine the route of a work troop responsible for clearing blocked roads. The problem is to find a route for the troop that starts at the depot and visits all of the critical locations. The objective is to minimize the total latency of critical nodes, where the latency of a node is defined as the travel time from the depot to that node. A mathematical model for this problem has already been developed in the literature. However, for real-life instances with more than seven critical nodes, this exact formulation cannot solve the problem optimally in a 3-hour limit. To find a near-optimal solution in a short running time, we develop a heuristic that solves a mixed integer program on a transformed network and a lower bounding method to evaluate the optimality gaps. Alternatively, we develop a metaheuristic based on a combination of Greedy Randomized Adaptive Search Procedure (GRASP) and Variable Neighborhood Search (VNS). We test both the matheuristic and the metaheuristic on Istanbul data and show that optimal or near-optimal solutions are obtained within seconds. We also compare our algorithms with existing work in the literature. Finally, we conduct an analysis to observe the trade-off between total and maximum latency. Second, we incorporate multiple work troops, multiple depots and visiting a subset of nodes (instead of visiting all nodes) in our problem. To the best of our knowledge, this problem with this generalization (and with the objective function of minimum latency) has not been investigated in the literature. We introduce a novel multi-level network model. We test the mathematical model with two categories of data sets: 1) Kartal data which is based on a real data set, located in Istanbul, Turkey. 2) random data sets in which all parameters are considered randomly to verify the strength of the proposed mathematical model. Our findings testify that the mathematical model is able to solve all instances optimally within reasonable computational times (less than 15 minutes in average). Since these algorithms should be executed in the post-disaster stage, having short computational times is crucial and enables the problem to be resolved in real-time as new information is obtained over the response stage. Note that given the small and reasonable computational times in all proposed algorithms, any organization associated with disaster management can employ these methodologies to mitigate the negative impacts caused by a disaster.