Show simple item record

dc.contributor.advisorTurner, Cameron J.
dc.contributor.authorAnderson, Daniel W.
dc.date.accessioned2007-01-03T07:13:29Z
dc.date.accessioned2022-02-03T12:52:11Z
dc.date.available2007-01-03T07:13:29Z
dc.date.available2022-02-03T12:52:11Z
dc.date.issued2015
dc.identifierT 7749
dc.identifier.urihttps://hdl.handle.net/11124/17107
dc.description2015 Spring.
dc.descriptionIncludes illustrations (some color).
dc.descriptionIncludes bibliographical references (pages 57-64).
dc.description.abstractCompetitive manufacturing companies require systems that can respond quickly to disruptions. In the event of a disruption, system performance information is provided to managers, schedulers and operators, to provide assistance in making informed decisions. Often, past data from a system or simulation is used to predict future performance, so that decisions can be made proactively, rather than reactively. Before calling in a worker for overtime to meet a quota, or taking a machine offline for maintenance, a manager might examine a variety of scenarios to provide a comprehensive perspective on the effects of any decision that is made. Such examinations are a part of what is called manufacturing forecasting. Many tools have been developed to forecast performance in a manufacturing environment, including regression analysis, artificial neural networks and simulation. The use of simulation is particularly common; however, realistic and accurate simulations are computationally expensive. In a scenario study, thousands or millions of simulations may be required. To save on computational expense, a manager can make use of a computationally efficient surrogate model that approximates the response of the simulation. Surrogate models imitate the response or behavior of a system; they generally sacrifice some accuracy to gain computational efficiency. With a good surrogate model, an accurate scenario study can be completed much more quickly than with a simulation. One method of surrogate modeling makes use of Non-Uniform Rational B-splines (NURBs), which can be used to precisely represent curves and surfaces of any dimension. NURBs have been used extensively in geometric modeling, but have seen only limited use as the basis for surrogate models. NURBs curves and surfaces can model any arbitrary topology, which makes them ideal for systems where little is known about the topology a priori, especially those which may be non-linear. This work implements a novel method of approximating throughput of a simulated flexible job-shop using NURBs as the basis for surrogate models. To the author's knowledge, it is the first instance of NURBs-based surrogate models used in manufacturing simulation surrogate modeling. This research does not compare the NURBs-based method directly with state-of-the-art forecasting techniques; rather, it is intended to determine the feasibility modeling the behavior of a manufacturing simulation with NURBs-based surrogate models. Three scenario studies are conducted that investigate the response of the simulation to variable resource availability and changing production quotas. Results from the scenario studies show that the NURBs-based surrogates accurately approximate the output of the flexible job-shop simulation. Additionally, using surrogate models demonstrate a clear computational advantage over simulation when evaluating many simulations. These findings indicate that NURBs-based surrogates are a promising method of manufacturing forecasting. Given the nature of NURBs-based models (they can model any arbitrary topology, are invariant under affine transformation, and are computationally efficient), they ought to be more thoroughly examined as a general tool for manufacturing forecasting and benchmarked against state-of-the-art methods.
dc.format.mediumborn digital
dc.format.mediummasters theses
dc.languageEnglish
dc.language.isoeng
dc.publisherColorado School of Mines. Arthur Lakes Library
dc.relation.ispartof2015 - Mines Theses & Dissertations
dc.rightsCopyright of the original work is retained by the author.
dc.subjectsurrogate Modeling
dc.subjectsimulation
dc.subjectmanufacturing scheduling
dc.subjectNURBs
dc.subjectnon-uniform rational B-splines
dc.subject.lcshManufacturing industries -- Forecasting
dc.subject.lcshSimulation methods
dc.subject.lcshComputer-aided design
dc.subject.lcshComputer-aided engineering
dc.titleFeasibility of NURBs-based surrogate models for manufacturing systems forecasting
dc.typeText
dc.contributor.committeememberBlacklock, Jenifer
dc.contributor.committeememberNewman, Alexandra M.
dc.contributor.committeememberSteele, John P. H.
thesis.degree.nameMaster of Science (M.S.)
thesis.degree.levelMasters
thesis.degree.disciplineMechanical Engineering
thesis.degree.grantorColorado School of Mines


Files in this item

Thumbnail
Name:
Anderson_mines_0052N_10698.pdf
Size:
10.74Mb
Format:
PDF
Description:
Feasibility of NURBs-based ...

This item appears in the following Collection(s)

Show simple item record