Fast array analysis using parallelised domain decomposition methods

Simple item page
dc.contributor.advisorLudick, Danieen_ZA
dc.contributor.authorEbrahim, Tameezen_ZA
dc.contributor.otherStellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.en_ZA
dc.date.accessioned2020-11-15T07:10:50Z
dc.date.accessioned2021-01-31T19:38:40Z
dc.date.available2020-11-15T07:10:50Z
dc.date.available2021-01-31T19:38:40Z
dc.date.issued2020-12
dc.descriptionThesis (MScEng)--Stellenbosch University, 2020.en_ZA
dc.description.abstractENGLISH ABSTRACT: The requirements for antennae increase daily and as such, antenna arrays are used to meet them. These arrays are often large, consisting of hundreds of antennae, and typically require multiple simulations for optimisation. Due to the computational requirements needed for these runs, it is necessary to apply specialised methods in order to reduce runtime while maintaining accuracy. In this thesis, three domain decomposition techniques based on the Method of Moments (MoM), namely, the Characteristic Basis Function Method (CBFM), the Domain Green's Function Method (DGFM) and the Improved Domain Green's Function Method (i-DGFM) will be investigated and implemented. These techniques will then be parallelised using shared and distributed high performance computing methods such as Open Multi Processing (OpenMP) and Message Passing Interface (MPI). In addition, implementation and parallelisation will also be performed for a Graphical Processing Unit (GPU).en_ZA
dc.description.abstractAFRIKAANSE OPSOMMING: Die vereiste vir groot antenna samestellings neem daagliks toe. Hierdie groot samestelling bestaan uit honderde antennas en benodig tipies verskeie simulasies vir optimering doeleindes. Weens the verwerkingsvereistes wat hiervoor benodig word, is dit nodig om spesialis metodes in te span om sodoende the looptyd te verminder, terwyl die akkuraatheid van die oplossing behoue bly. In hierdie tesis word drie gebied ontbindings metodes ondersoek wat almal onderliggend op die Moment Metode (MoM) baseer is, naamlik die Karakteristieke Basis Funksie Metode (KBFM), the Gebied Green's Funksie Metode (GGFM) en die Verbeterde Gebied Green's Funksie Metode (v-GGFM). Hierdie tegnieke word geparalleliseer deur middel van gedeelde en verspreide hoà spoed verwerking tegnieke soos oop multi ververking (OpenMP) en boodskap deel koppelvlak (MPI). Addisioneel word die parallelisering deur middel van 'n Grafiese Verwerker (GPU) ook ondersoek.af_ZA
dc.description.versionMastersen_ZA
dc.format.extent93 pages : illustrationsen_ZA
dc.identifier.urihttp://hdl.handle.net/10019.1/109181
dc.language.isoen_ZAen_ZA
dc.publisherStellenbosch : Stellenbosch Universityen_ZA
dc.rights.holderStellenbosch Universityen_ZA
dc.subjectParallelised domain decompositionen_ZA
dc.subjectUCTDen_ZA
dc.subjectAntenna arrays -- Analysisen_ZA
dc.subjectMessage passing interfaceen_ZA
dc.subjectGPUs (Graphics processing units)en_ZA
dc.titleFast array analysis using parallelised domain decomposition methodsen_ZA
dc.typeThesisen_ZA
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
ebrahim_fast_2020.pdf
Size:
8.16 MB
Format:
Adobe Portable Document Format
Description:
Download
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Plain Text
Description:
Download
Collections
Masters Degrees (Electrical and Electronic Engineering)