Prediction Model for Tensile Shear Strength of Gas Metal Arc Weld using a Laser Vision Sensor

Abstract

This study proposes a method to predict tensile shear strength of the overlap welded joint of aluminum alloy (Al5083-O, Al6061-T5) plates applied to the cowl-cross part of a vehicle. The profile of a weld bead was measured using a laser vision sensor, and a technology that can predict tensile shear strength of a welded joint was developed and evaluated. Welded joints were fabricated by using AC pulse GMAW to overlap the configuration of the aluminum alloy plates. The data required for training the prediction model were obtained by measuring the profiles of the welded joints using a laser vision sensor and conducting a tensile test. A CNN-based regression model was thus developed to predict tensile shear strength of welded joints. The model uses a weld?bead profile and material information as input and estimates tensile shear strength of a welded joint as output. The average prediction error of the proposed model was calculated to be approximately 3%.