| Popis: |
V sklopu magistrske naloge je prikazan postopek obdelave podatkov za namene uporabe v sistemu umetne inteligence. Slednji se uporablja za napovedovanje korekcij na varilnih orodjih znotraj procesa izdelave obese kolesa. Na začetku so predstavljena teoretična izhodišča, potrebna za pravilno razumevanje problema. Tukaj so na kratko predstavljeni načini za aproksimacijo krivulj in površin. Predstavljen je tudi Akaikov informacijski kriterij, ki je v nadaljevanju uporabljen za izbiro stopenj polinomov. V srednjem delu naloge so opisani dejavniki, ki vplivajo na korekcije varilnih orodij oziroma na končno geometrijo izdelane obese. Ugotovljeno je bilo, da ima na izvajanje korekcij največ vpliva geometrija odpreškov. V nadaljevanju so bili predstavljeni načini, s katerimi je bil izveden opis geometrije odpreškov. Kot najbolj ustrezen se je izkazal način, pri katerem so s pomočjo polinomov aproksimirani odstopki v normalni smeri na površino. Na koncu je predstavljen koncept sistema za obdelavo podatkov ter izvajanje korekcij na varilnih orodjih. The process of data processing for the purpose of use in an artificial intelligence system is shown as part of the master's thesis. The latter is used for prediction of welding fixtures adjustments within the trailing arm manufacturing process. The theoretical starting points necessary for a correct understanding of the problem are presented at the beginning. Methods for approximating curves and surfaces are briefly presented here. Akaike information criterion which is used in the following for the selection of polynomial degrees is also presented. The middle part of the thesis describes the factors that affect the welding fixtures adjustments or rather the final geometry of the manufactured trailing arm. It was found that the geometry of the stamping parts has the greatest influence on the implementation of adjustments. In the following, the methods used to describe the geometry of stamping parts were presented. The method in which deviations in the normal direction to the surface are approximated with the help of polynomials proved to be the most appropriate. At the end, the concept of a system for processing data and performing welding fixtures adjustments is presented. |
