Decision making framework in deterministic and stochastic forward/reverse logistics supply chain design
Date
2021-08
Authors
Azizi, Vahid
Major Professor
Advisor
Hu, Guping
Ommen, Danica
Olafsson, Sigurdur
De Brabanter, Kris
MacKenzie, Cameron
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Altmetrics
Abstract
Nowadays an efficient supply chain system plays a crucial role in manufacturing production. Suppliers, manufacturers, and customers are the main stakeholders of a supply chain system. Forward and reverse networks are two main types of supply chain networks. In a forward supply chain network, raw materials are transported from suppliers to manufacturing facilities and are manufactured into final products which are then shipped to the customers. In a reverse supply chain network, the final products which their quality do not meet the minimum standards, or reached to end of their lives are transported to upstream facilities such as manufacturing facilities, recycling centers, or disposal centers. To design an efficient supply chain system, decision makers encounter a wide variety of strategic, tactical, and operational decision making problems in the forward/reverse logistics in deterministic and stochastic environments. In this dissertation, various mathematical methods have been formulated to study a forward supply chain problem in a deterministic environment and two reverse logistics problems under uncertainties.
In the first paper, two integrate mixed integer linear programming models have been proposed for a forward supply chain network. These models aim to find optimal investment on establishing distribution centers to minimize the transportation costs from suppliers to distribution centers and from distribution centers to customers. In the first model, vehicles are allowed to load different types of products to deliver to the customers (multi-product delivery) while in the second model delivery vehicles are allowed to load only one type of product (single-product delivery). Various instances with different sizes were generated to validate the introduced models. Three solution methods including deterministic mode, opportunistic mode, and benders decomposition algorithm in CPLEX have been employed to solve the proposed models. The results show that integrated model reduces total system costs by 24.72\% on average. Also, multi-product delivery approach results in saving rate up to 31.27\% compared to single-delivery approach. Among the solution methods to solve the proposed models, opportunistic mode outperforms other methods on average in terms of objective function value and computational run-time.
In the second paper, a two-stage stochastic programming model has been developed for a reverse logistics network under return and demand quantity uncertainties. This model aims to minimize the total cost of network including establishing costs of sorting centers, warehouses, recycling centers, and disposal centers, and transportation cost. Decision on the number of opened facilities among some candidates are the strategic decisions and tactical decisions include lot sizing, transportation plan, level of inventory, backorder, and shortage over the planning horizon. Probability distributions of return and demand quantities are considered normal. Moment matching method was used to generate discrete sets of scenarios and fast forward selection algorithm was applied as scenario reduction method. A case study was conducted to validate the proposed model. Numerical results indicate that the stochastic model solution outperforms the expected value solution.
In the third paper, a multi-stage programming model has been formulated to address a multi-echelon, multi-period reverse logistics network in which the main uncertain parameters are primary markets' return quantity and quality, and secondary markets' demand quantity. The formulated model minimizes the total cost of establishing facilities, inventory cost, and backorder and shortage cost. Moment matching method for scenario generation and fast forward selection for scenario reduction have been adopted to generate a finite number of discrete scenarios. Extensive form of the problem is used to solve the introduced stochastic programming model. A case study has been conducted to validate and evaluate the proposed model and solution method.
Series Number
Journal Issue
Is Version Of
Versions
Series
Academic or Administrative Unit
Type
dissertation