Determinação de geometrias substitutas utilizando redes neurais artificiais para análise de medidas obtidas em máquina de medição por coordenadas

Abstract

Measurements are carried out very naturally in almost all branches of human activity and due to the existence of a great diversity of measures to be measured, it was necessary to separate Metrology in several different branches. Within the Geometric Metrology there is the Coordinate Metrology. The coordinate measuring machine (CMM) has as main objective the analysis of the geometric properties of objects through Coordinate Metrology. Significant features of this equipment are: high inspection speed, accurate results and flexibility of measurement. However, the process of acquisition of coordinates through a CMM isn’t totally free from errors, which includes errors from computational systems responsible for determining the substitute geometries. In this paper are studied the basic elements that constitute the CMMs and their main sources of errors. As a proposal a study is made on Artificial Neural Networks (ANNs) and their application for measurements made in CMM in order to evaluate the ability to generate substitute geometries. The developed RNAs, when fed with coordinate points of circumferences and spheres, are able to determine their centers. With this determined center from the geometric entity it is possible to calculate its radius by the distance between this center and the points that were used as entrance in the neural network. It is noted that errors in the measurement process, such as the impossibility of an operator to capture points equidistantly from these shapes, has a drastic influence on the results obtained.