Loading...

Hab and Spoke Network Design Considering Location Demand Dependency

Khosravi, Shahrzad | 2018

525 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 51093 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Akbari Jokar, Mohammad Reza
  7. Abstract:
  8. In this thesis, a new hub-and-spoke location problem is investigated in which, the location of spoke nodes should be decided upon, in addition to the location of hub nodes. Three research problems are investigated: In the first problem, the goal is to determine both hub and spoke node locations. In this problem, the objective function considers the following costs, to be minimized: flow transportation costs, lost demand costs, in addition to the hub and spoke location costs. The demand distribution among different nodes follow the gravity rule, and the hub network is considered to be fully connected. Following the gravity rule, the demand among different node is not set a priory and depends on the location of the spoke nodes, which is referred to as “location-dependent demand”. The problem is solved using Gamma solution algorithm. This problem is further developed to form the second problem, in which the demand follows the gravity rule for spoke locations only. Once the demand reaches a spoke location, it is the transportation firm (decision maker) to decide how to transport the demand across the network. The decision variables in this problem is on the location of hub and spoke facilities, as well as the vehicles assigned to each hub. The objective is to minimize the total cost function, including: transportation costs, lost demand costs, and hub and spoke location costs. A two-stage iterative algorithm is proposed to solve the second problem. In the first stage, the location of hub and spoke facilities are determined, and the demand is assigned to each facility. In the second stage, number of vehicles for each hub is determined, as well as the routing for the vehicles. Once the routing problem is solved and the distances are known in the second stage. These results are fed to the first stage to be re-solved. In the third problem, another assumption is relaxed from the first problem, and the hub network is not considered to be fully connected. In this problem, the link establishment cost is also considered, and since the hub network is not fully connected, flow might traverse more than one hub link to reach its destination. The aim is to decide upon the location of hub and spoke facilities, and hub links to be established in order to minimize the total costs of hub and location, link establishment, transportation and lost demand costs. In this model, additional assumptions are included, to further reflect the real-world considerations: a) there should be a balance among the flow congestion of different hub nodes across the network, b) there is flow-based discount on hub links, which is considered in this model. Benders decomposition approach is used to solve the problem, along with two enhancing algorithms to accelerate the Benders decomposition performance. The results of all three problems show that solving the integrated version of the problem(both hub and spoke locations) leads to significantly lower total cost, than solving the hub location only, based on the given location of spoke nodes. It is also observed that the hub locations may change depending on whether the spoke locations are given, or should be decided upon simultaneously. If the spoke locations are not given and treated as decision variable, the model might decide whether to satisfy or lose some demands, which directly affect the location of the spoke facilities
  9. Keywords:
  10. Routing ; Network Design ; Hub Location Problem ; Demand Dependency ; Hub and Spoke Location ; Variable Discount Factor ; Flow Congestion

 Digital Object List

 Bookmark

No TOC