Модел система за аутоматско препознавање технолошких форми на призматичним деловима

Abstract

U ovom radu, predstavljen je model jednog sistema za automatsko prepoznavanje tehnoloških formi na prizmatičnim delovima. Ulaz u sistem predstavlja STEP fajl, eksportovan iz CAD modela realizovanog u Pro/E WF2. Specijalizovani program pretražuje dobijeni .stp fajl, formira B-rep dela, a zatim taj prikaz konvertuje u tzv. vektore zbira bodova strana tog dela, koji predstavlja ulaz u veštačku neuronsku mrežu, sa "back-propagation" obučavanjem. Zadatak ove neuronske mreže, koja se projektuje i obučava korišćenjem MatLab ANN Toolbox-a, jeste da izvrši prepoznavanje tehnoloških formi na prizmatičnom delu. Razvijeni delovi sistema su detaljno opisani u radu, a date su i osnovne smernice za dalja opsežna istraživanja u ovoj oblasti.

In this paper, a model of a system for automatic recognition of machining forms on prismatic parts is presented. The input to the system is a STEP file, exported from a CAD model realized in Pro/E WF2. A specialized programme searches the resulting .stp file, forms B-rep parts, and then converts that view into the so-called. vectors of the sum of the points of the side of that work.piece, which represents the input to the artificial neural network, with "back-propagation" training. The task of this neural network, which is designed and trained using the MatLab ANN Toolbox, is to recognize machining forms - features on the prismatic work.piece. The developed parts of the system are described in detail in the paper, and basic guidelines for further extensive research in this area are given.