|dc.description.abstract||Waste reduction is critical in today's manufacturing climate. Producers are being compelled to incorporate reverse logistics into their supply chain by government legislation, the potential for recovering economic value, and consumer demand for ``green'' practices. Complicating network design for reverse logistics is the need to incorporate existing supply chain networks and higher levels of uncertainty in quantity, frequency and quality of return product.
Even though facility location models have been developed for reverse logistics network design, the producer is confronted with a number of high-level decisions before detailed decisions can be made. For instance, how will the product be collected? Will testing be done centrally or near the collection site? Where will processing be performed? In order to answer these questions, decision makers need a means to quantify and analyze tradeoffs inherent in the network design.
This work presents a flexible, generalized decision model that integrates high-level and detailed design decisions and that incorporates uncertainty. The first part of the work is a conceptual framework for the high-level decisions identifying eight possible network configurations based on more than thirty-five case studies. The second part extends the framework into a multicriteria decision making model using Analytical Hierarchy Process (AHP), which quantifies tradeoffs and provides insights through sensitivity analysis. The third part of the work presents a suite of mixed-integer linear programming (MILP) models that integrates the high-level and detailed decisions, and that addresses uncertainty using three methods: chance-constrained programming, stochastic programming, and robust optimization. The deterministic and chance-constrained models provide a comparatively inexpensive way to determine the optimal network configuration, while the stochastic programming and robust optimization models require more computation but better address sensitivity to detailed site locations through recourse variables.
The AHP decision making model is demonstrated on three case studies with different characteristics. The findings show that the AHP preference ranking of network configurations is sensitive to the producer's goals and values, and sensitivity analysis explores the impact of the relationships of those goals and values. For instance, the collection decision is sensitive to the preference for business relations vs.\ cost savings, the sort-test decision is sensitive to potential cost savings from reducing testing costs and identifying scrap early, and the processing decision is sensitive to the need to protect proprietary knowledge and the availability of original facility processing capacity.
The suite of MILPs with uncertainty is demonstrated on a numerical study from the literature and a fourth case study involving consumer electronics recycling. The research found that the choice of network configuration is relatively insensitive to uncertainty, but that there is sensitivity to site location decisions. The findings indicate that the combination of the AHP decision making model and the detailed MILP models provides a strategic approach for decision makers facing the challenge of designing a reverse logistics network into their supply chain.||en_US