Intelligent modeling for estimating weld bead width and depth of penetration from infra-red thermal images of the weld pool

Abstract

In order to develop remote welding process methodologies, it is first important to develop computational methodologies employing soft computing techniques for predicting weld bead width and depth of penetration using a real time vision sensor during welding. Welding being a thermal processing method, sensing using infra-red (IR) camera is most extensively employed for monitoring and control of welding process. In the present work, attempt has been made to develop predictive methodologies using hybrid soft computing techniques for accurately estimating the weld bead width and depth of penetration from the IR thermal image of the weld pool. IR thermal images have been recorded in real time during A-TIG welding of 6 mm thick type 316 LN stainless steel weld joints with varying current values to produce different depth of penetration. From the acquired IR images, hot spot was identified by image segmentation using the cellular automata image processing algorithm for the first time. The current and the four extracted features from the hot spot of the IR thermal images were used as inputs while the measured bead width and depth of penetration were chosen as the output of the respective adaptive neuro fuzzy inference system and artificial neural network based models. Independent models were developed for estimating weld bead width and depth of penetration respectively. There was good correlation between the measured and estimated values of bead width and depth of penetration using the developed models.