Modelo preditivo de regressão logística para análise de perfuração em soldagem MIG/MAG de tubos de veículo do tipo baja

Abstract

The Baja SAE is an engineering competition that aims to build a Baja's stiles off-road vehicle. The structural design and performance are good for the vehicle's performance, with the car's weight being an essential variable for the vehicle's good performance. For this reason, the use of tubular metallic structures for their construction becomes advantageous. However, it is frequently the occurrence of burn-through during the welding. The purpose of this study is to obtain a prediction regression model that provides the probability of burn-through on the frame. 72 tests were carried out using MIG/MAG welding for data collection, controlling the parameters: shielding gas, wire diameter, voltage and wire speed nominal values. The R language was used to obtain five regression models. One of these models obtained the best accuracy (90.3%) and sensitivity (87.5%) results for the training data. Therefore, was obtaining a model capable of predicting the occurrence of burn-through with some effectiveness. The results allowed us to conclude that for the conditions of the study the type of gas was not significant and that for lower chances of perforation the most indicated is the wire diameter of 0.8 mm.