Development of a model to predict weld bead geometry during the GMAW process using high speed camera

摘要

Among the various sensors used in welding processes, there are no sensors capable to directly sense some of the weld characteristics as depth and heat affected zone (HAZ) during the process. This is a limiting factor in the process control due to the fact that the determination of the weld final characteristics can only be done at the end of the process, through desctructive or non destructive tests, when any control law cannot be taken. In this work it is shown the study of the relation among some weld quality parameters and observable variables, as voltage, current, welding speed and weld pool geometry, focusing on the later one, in order to determine if the information about welding pool geometry is relevant on the determination of the weld bead geometry or if the information about voltage, current and welding speed is sufficient. Based on the experimental results, a mathematical model is developed relating welding parameters and measurable variables with non-measurable ones during the GMAW process. This model makes possible the inference of some weld quality parameters as depth, height, width and HAZ by measuring some variables as voltage, current, welding speed and weld pool width. Therefore, in order to obtain the desired weld bead geometry, the model allows a real time control system to change the welding parameters. It was generated some welding parameter sets and, during the welding process using each set, the weld pool was monitored by a high-speed camera. So as to automatically detect the weld pool geometry, computer vision techniques were used. The measures bead geometry was obtained with metallographic analysis. The obtained model shows that the observation of the welding parameters and the pool geometry makes possible the inference of the weld geometry, better than just observing the welding parameters.