Afet öncesi hazırlık ve yardım malzemsi dağıtım problemine çok aşamalı risk ayarlı SDDP yaklaşımı

Disaster management subject is highly studied by academicians, researchers and municipals especially in the most recent years due to the increase of their tremendous and destructive effects to the human lives.The increase of these events requires urgent and effective responses with well defined methodologies. However, because of the random occurrence of these events, it is hard to deal with these problems. This study aims to apply an effective methodology to disaster management problem where we cover a three layer supply chain with both the pre-disaster and post-disaster phase of a disaster management problem. The pre-disaster decisions are the inventory decisions at the main depots (suppliers) and the location decisions of the distribution centers. These decisions are made before observing the random data, which are jointly distributed demands and road capacities. After observing the randomness, at the second and later stages, the demands have to be satisfied using only the available inventories at the main depots and passing through the previously selected distribution centers. Mathematically, the first-stage problem is a mixed-integer linear optimization problem, whereas the problems at the later stages are simply linear programming problems. Assuming that the demands and road capacities are jointly and continuously distributed, the deterministic equivalent is obtained through the sample average approximation method. In addition, we built risk-adjusted multistage formulation using conditional value at risk and the problem is solved through the stochastic dual dynamic programming (SDDP) method under fairly general assumptions. The proposed algorithm is applied on a real life data that is obtained for Asian side of Istanbul.