dc.contributor.advisor Erera, Alan dc.contributor.author Kues, Brian Joseph dc.date.accessioned 2016-08-22T12:22:11Z dc.date.available 2016-08-22T12:22:11Z dc.date.created 2016-08 dc.date.issued 2016-05-13 dc.date.submitted August 2016 dc.identifier.uri http://hdl.handle.net/1853/55583 dc.description.abstract This thesis investigates how a service provider responsible for completing repairs of durable goods deployed throughout a region can improve profitability by reconfiguring its existing supply chain. This supply chain stocks spare parts inventory both in kits that technicians take to the repair sites, and at forward stocking locations (FSLs) distributed throughout the region. The investigation was conducted in three parts. The first part examines the decision of which spare parts to stock in the kit for a single technician who must complete multiple repairs within a single period. The technician has access at a single FSL to all parts not stocked in the kit but at a time penalty which decreases the likelihood of completing a satisfactory number of repairs by the end of the period. This decision is modeled with a binary optimization problem which has a single probabilistic constraint, namely, that the allocation of spare parts between the kit and the FSL allows the technician to complete the required number of repairs in expectation. Formulated as such, the decision problem is $\mathcal{NP}$-hard. Six heuristics are proposed to quickly find spare parts allocations that have low cost. A method to find a lower bound on the optimal value is developed and then used to demonstrate that many of the heuristics can generate solutions whose costs are within ten percent of optimal. The heuristics are also compared against one another on a test suite of problem instances with important parameters varied over ranges of values. Finally, the first part concludes with a case study in which the heuristics are analyzed for effectiveness using instances drawn from real-world data from an industry collaborator. The second part of the thesis broadens the scope of the first part to include the decision of how to sequence the repairs for the technician when the geographic locations of the customers are explicitly incorporated into the model. The total travel time, which depends on the sequence in which customers are visited, affects the expected number of completed repairs, which again is constrained to be above a certain required level in expectation. Travel time, however, does not directly impact the objective to minimize total inventory cost. Given a method for sequencing customers, the decision problem is identical to that in part one. The fact that the customers change from one period to the next is accounted for by evaluating the expected fill rate for a repair kit on average across multiple customer instances. Heuristics for sequencing customers are proposed, and heuristics for determining inventory are reused from the first part of the thesis. Computational results show that random routing leads to inventory costs twenty percent higher than those for all other proposed methods of routing. Furthermore, it is shown that almost all benefits of smarter routing can be generated even when using a simple greedy heuristic for routing decisions. The third and final part of the thesis again broadens the scope of the first part, but in a different direction to include the decisions of how many technicians to employ and how many FSLs to operate. Each technician must complete the repairs assigned to him or her, the number of which depends on the total number of technicians employed, and shares access to inventory at all of the FSLs, the number of which influences the time delay needed to retrieve a part not in the repair kit. The objective of the decision problem is to minimize the average total cost of repair kit inventory, FSL inventory, and technician labor per customer repair job in a single period. A straightforward algorithm to find a good solution, completely specified by a kit-or-FSL decision for all part types, a number of technicians, and a number of FSLs, is motivated by the fact that the inventory-setting algorithms developed in the first part of the thesis can be reused for a given number of technicians and number of FSLs. Computational tests reveal that the best solutions have a single FSL and either fewer but busy technicians with close-to-full kits or more but less busy technicians with empty kits. When inventory is four times more costly than labor, the solutions fall somewhere between these two extremes. Such an arrangement offers savings of $10\%$ to $30\%$ over a supply chain without inventory recourse at an FSL. % a single FSL that stocks all part types and many technicians with empty kits is best, and that this configuration can save fifty percent of the total costs for a spare parts supply chain without FSL recourse. dc.format.mimetype application/pdf dc.language.iso en_US dc.publisher Georgia Institute of Technology dc.subject Spare parts dc.subject Supply chain optimization dc.subject Logistics dc.subject Inventory optimization dc.subject Heuristics dc.title On spare parts supply chains with forward stocking location recourse dc.type Dissertation dc.description.degree Ph.D. dc.contributor.department Industrial and Systems Engineering thesis.degree.level Doctoral dc.contributor.committeeMember Atasu, Atalay dc.contributor.committeeMember Savelsbergh, Martin dc.contributor.committeeMember Swann, Julie dc.contributor.committeeMember White, Chelsea dc.date.updated 2016-08-22T12:22:11Z
﻿