Thesis (MScEng)--Stellenbosch University, 2011. / ENGLISH ABSTRACT: Although a numerical model of the mirror support truss of the Southern African Large Telescope (SALT) has already been developed during the design thereof, this thesis focuses on the development of the methods and techniques that would result in a more accurate numerical model of the actual structure that could be used as a basis for a numerical control system. This control system will compensate for de ections in the structure by adjusting the positioning of the individual mirror segments of the primary mirror. The two main components from which the support truss is constructed are the steel nodes, and the struts that connect to them. For this project a smaller, simpler laboratory model was designed and built to have geometrical properties similar to that of the support truss. The methods and techniques that were investigated were carried out on this model. By using numerical design optimisation techniques, improved numerical models of the different strut types were obtained. This was done by performing tests on the struts so that the actual responses of the struts could be obtained. Numerical models of the struts were then created and set up so that they could be optimised using structural optimisation software. Once accurate strut models had been obtained, these strut models were used to construct a numerical model of the assembled structure. No additional optimisation was performed on the assembled structure and tests were done on the physical structure to obtain its responses. These served as validation criteria for the numerical models of the struts. Because of unforeseen deformations of the structure, not all of the measured structural responses could be used. The remaining results showed, however, that the predictive accuracy of the top node displacement of the assembled structure improved to below 1.5%, from over 60%. From these results it was concluded that the accuracy of the entire structure's numerical model could be signi ficantly improved by optimising the individual strut types. / AFRIKAANSE OPSOMMING: Alhoewel daar reeds 'n numeriese model van die spieëlondersteuningsraamwerk van SALT ontwikkel is gedurende die ontwerp daarvan, fokus hierdie tesis op die ontwikkeling van metodes en tegnieke om 'n numeriese model van steeds hoër gehalte van hierdie spesi eke struktuur te verkry wat kan gebruik word as 'n basis vir 'n numeriese beheerstelsel. Hierdie beheerstelsel sal kan kompenseer vir die ondersteuningsraamwerk se vervormings deur om die individuele spieëlsegmente van die primêre spieël se posisionering te verstel. Hierdie stuktuur bestaan uit hoofsaaklik twee komponente, naamlik staalnodusse en die stutte wat aan hulle koppel. Vir hierdie projek is 'n kleiner, eenvoudiger laboratorium-model ontwerp en gebou om geometriese eienskappe soortgelyk aan die van die ondersteuningstruktuur te hê. Die metodes en tegnieke wat ondersoek is, is op hierdie model uitgevoer. Verbeterde numeriese modelle van die verskillende stut-tipes is ontwikkel deur middel van numerieseoptimeringstegnieke. Dit is gedoen deur toetse op die stutte uit te voer sodat hul werklike gedrag bepaal kon word. Numeriese modelle van die stutte is toe geskep en opgestel sodat hulle geoptimeer kon word om dieselfde gedrag as wat gemeet is, te toon. Hierdie geoptimeerde modelle is toe gebruik om numeriese modelle van die toets-struktuur te skep. Geen verdere optimering is op die numeriese model uitgevoer nie en toetse is op die struktuur gedoen om sy werklike gedrag te meet. Data wat deur die toetse verkry is het as validasie kriteria gedien om die akkuraatheid van die numeriese modelle van die stut-tipes te bepaal. Weens die struktuur se onvoorsiene vervorming kon alle gemete struktuurdata nie gebruik word nie. Die oorblywende data het egter getoon dat die akkuraatheid van die finale numeriese modelle van die struktuur verbeter het en dat dit die translasie van die top-node met 'n speling van 1.5% akkuraatheid kon voorspel, teenoor die oorsponlike speling van meer as 60%. Daar is bevind dat die akkuraatheid van die numeriese model van die hele struktuur noemenswaardig verbeter kan word deur die numeriese modelle van die stut-tipes te optimeer.