Abstract

Closed loop supply chains comprise, in addition to the conventional forward flows from suppliers to end-users, a reverse flow of products, components, and materials from end-users to the manufacturers and secondary markets. Designing a closed-loop supply chain is a strategic level planning which considerably impacts on tactical and operational performance of the supply chain. It refers to the decisions taken on the location of facilities involved in the supply chain network along with the management of the physical flows associated with forward and product recovery channels. Our problem of interest is mainly motivated by the case of durable products including but not limited to large household appliances, computers, photocopying equipment, and aircraft engines. Such category of products has a modular structure, composed of independent components. As opposed to simple structured products, e.g., printer cartridges, that can only be recycled, each of the components in the reverse bill of materials of durable products can be recovered by a particular type of recovery process. Besides, durable products share a long life cycle characteristic which indeed makes designing their CLSC networks more complicated.

In this thesis, in keeping with the abovementioned motivation, we focus on designing closed-loop and reverse supply chains in the context of durable products that are of various quality conditions. The recovery decisions for product return include remanufacturing, part harvesting, bulk recycling, material recycling, and landfilling/incineration. Moreover, we take into account environmental concerns regarding the harmful impacts of used products in the closed-loop supply chain planning. As the closed-loop supply chains typically encounter uncertainty in quality and quantity of the profitable return stream, we further aim to consider the impact of uncertainty in designing the recovery network. For such purposes, in the first phase, we address a closed-loop supply chain planning problem in the context of durable products with generic modular structures. The problem is formulated as a mixed-integer programming model which is then solved by an accelerated Benders decomposition-based algorithm. The performance of the proposed decomposition approach is enhanced through incorporating algorithmic features including valid inequalities, non-dominated optimality cuts, and local branching strategies.

Next, in the second phase, we propose a precise approach to model the uncertain quality status of returns, in which the availability of each component in the reverse bill of materials is modeled as discrete scenarios. We propose a two-stage stochastic programming model to address this problem setting. Then, since the cardinality of the scenario set grows exponentially with the number of involved components, we detail on a scenario reduction scheme to alleviate the computational burden of the proposed model. The stochastic problem is solved by a L-shaped algorithm enhanced through valid inequalities and Pareto-optimal cuts.

Finally, we investigate designing a dynamic reverse supply chain where the quantity of the return flows is uncertain. We introduce a multi-stage stochastic programming model and develop a heuristic inspired by scenario clustering decomposition scheme as the solution method. It revolves around decomposing the scenario tree into smaller sub-trees which consequently yields a number of sub-models in accordance with sub-trees. The resulting sub-models are then coordinated by Lagrangian penalty terms. On account of the fact that each sub-model per se is a hard to solve problem, a Benders decomposition-based algorithm is proposed to solve sub-models.