Abstract

Every year, various human actions (e.g., terrorist attacks, strikes, etc.) and natural disasters (e.g., earthquakes, hurricanes, and etc.) cause disruptions in supply networks, and as the result, huge financial and humanitarian loss. Not only they brought loss of services to the system, they, depending on the type, partial or complete, may result in facility failures, roads failures or both, simultaneously. Therefore, having reliable systems are essential in order to reduce risks as well as cost in case of failures. Motivated by the importance of considering the failure in design level, we, in this thesis, focused on problem of locating facilities, allocating demand points to the facilities, and defining the rout among them while considering the complete failure in the elements of the network. The Reliable Location/ Allocation/ Routing Problem (RLARP) formulation which is Mixed Integer Programming model is proposed, taking into account failures in facilities and routs in different scenarios as failure sets. Along with bringing in trustworthy systems, we also contribute an exact decomposition methodology and propose a Column Generation model to tackle the complexity. The idea is to define a supply chain network at the design level to be robust against worst case failures and disruptions scenarios. To the best of author’s knowledge, the Column Generation technique has not been applied previously to solve RLARP problems in the literature. In addition, we consider the facility and transportation method failures in our model, despite the fact that mostly either facility failures or transportation failures are taken into account in the literature. Various data sets designated for validating Column Generation and RLARP formulation proposed in this thesis. Eventually, we compare the performance of CG and RLARP models over a range of instances. Results suggests that CG technique performs significantly better than solving the RLARP model with a general optimization solver (CPLEX) in terms of computational time and the size of instances that can be solved.