Mechanical Engineering Theses and Dissertations
Permanent URI for this collectionhttp://hdl.handle.net/1903/2795
Browse
4 results
Search Results
Item Solving the Inventory Slack Routing Problem for Emergency Medication Distribution(2010) Montjoy, Adam Wesley; Herrmann, Jeffrey W; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)A bioterrorist attack, or natural disaster, would prompt an immediate government response in order to efficiently address the possible health effects of the population. Such a scenario would create a logistics problem of delivering medication (or other supplies) to makeshift dispensing centers in a short period of time and in high quantities while operating. These makeshift centers, or Points of Dispensing, require schedules of delivery that are robust against uncertainty. This inventory slack routing problem is a novel vehicle routing problem. The objective function is to maximize the slack in the schedule. This thesis presents heuristic approaches that separate the problem into routing and scheduling. The routing problem is solved using a route first-cluster second method. The scheduling problem is solved using a heuristic and an improvement approach. This thesis also presents a search approach that uses heuristics to search various neighborhoods in the solution space. These heuristics are chosen randomly based on probabilities that adapt during the search according to their performance. The inventory slack routing problem is also formulated as a mixed-integer program and solved using a column generation procedure that utilizes simulated annealing to generate new vehicle schedules. This thesis presents the results of testing these three approaches on a set of 432 instances that were generated from real-world data to evaluate solution quality and computational effort. The search approach outperformed the heuristic approach with a reasonable amount of computational effort. The column generation approach did not generate desirable vehicle schedules and therefore was not productive in solving the problem.Item Solving continuous replenishment inventory routing problems(2008-08-22) Fomundam, Samuel; Herrmann, Jeffrey; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This research investigates the problem of resupplying points of dispensing (PODs), which will dispense medications to millions of people in case of a bioterrorist attack such as anthrax. After receiving an initial but limited supply of medication, the PODs will operate continuously. Vehicles will resupply the PODs continuously from a central depot that has a stockpile of medication. Each vehicle will repeatedly follow the same route and will deliver at each POD enough medication to replace what was consumed since the last visit. Because the number of drivers and trucks may be limited during an emergency, we wish to minimize the number of vehicles used to resupply the PODs. This thesis presents heuristics and a branch-and-bound approach for solving this NP-hard problem and evaluates their performance. We also analyze a special case in which all of the PODs have the same demand.Item Simulation and Optimization of Production Control for Lean Manufacturing Transition(2008-08-06) Gahagan, Sean Michael; Herrmann, Jeffrey W; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Lean manufacturing is an operations management philosophy that advocates eliminating waste, including work-in-process (WIP) inventory. A common mechanism for controlling WIP is "pull" production control, which limits the amount of WIP at each stage. The process of transforming a system from push production control to pull is not well understood or studied. This dissertation explores the events of a production control transition, quantifies its costs and develops techniques to minimize them. Simulation models of systems undergoing transition from push to pull are used to study this transient behavior. The transition of a single stage system is modeled. An objective function is introduced that defines transition cost in terms of the holding cost of orders in backlog and material in inventory. It incorporates two techniques for mitigating cost: temporarily deferring orders and adding extra capacity. It is shown that, except when backlog costs are high, it is better to transform the system quickly. It is also demonstrated that simulation based optimization is a viable tool to find the optimal transition strategy. Transition of a two-stage system is also modeled. The performance of two simple multi-stage transition strategies is measured. In the first, all of the stages are transformed at the same time. In the second, they are transformed one at a time. It is shown that the latter strategy is superior. Other strategies are also discussed. A new modeling formalism, the Production Control Framework (PCF), is introduced to facilitate automated searches for transition strategies in more complex systems. It is a hierarchical description of a manufacturing system built on a novel extension of the classic queue server model, which can express production control policy parametrically. The PCF is implemented in the form of a software template and its utility is shown as it is used to model and then find the optimal production control policy for a five stage system. This work provides the first practical guidance and insight into the behavior and cost of Lean production control transition, and it lays the groundwork for the development of optimal transition strategies for even the most complex manufacturing systems.Item Development and Evaluation of Algorithms for Scheduling Two Unrelated Parallel Processors(2007-08-09) Leber, Dennis D; Herrmann, Jeffrey W; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Given a group of tasks and two non-identical processors with the ability to complete each task, how should the tasks be assigned to complete the group of tasks as quickly as possible? This thesis considers this unrelated parallel machine scheduling problem with the objective of minimizing the completion time of a group of tasks (the makespan) from the perspective of a local printed circuit board manufacturer. An analytical model representing the job dependent processing time for each manufacturing line is developed and actual job data supplied by the manufacturer is used for analysis. Two versions of a complete enumeration algorithm which identify the optimal assignment schedule are presented. Several classic assignment heuristics are considered with several additional heuristics developed as part of this work. The algorithms are evaluated and their performance compared for jobs built at the local manufacturing site. Finally, a cost-benefit tradeoff for the algorithms considered is presented.