Prediction of weld bead cross-sectional area in wire arc additive manufacturing using vision system integrated with machine learning approach

Wire and Arc Additive Manufacturing (WAAM) has gained as an efficient technology for the rapid manufacturing of several applications. The cross-sectional area of weld bead geometry is the most influencing parameter that determines the quality of the component produced and its production time. The weld bead cross-sectional area depends on various WAAM operating parameters. In this work, the weld bead cross-sectional area is predicted by integrating a vision system with a machine learning approach. The experiments were conducted using full factorial experiments by varying specific energy, welding travel speed, and arc length at three different levels. The cross-sectional area of the weld bead is determined using the K-means clustering segmentation technique and the process is modelled using Levenberg-Marquardt back propagation algorithm to predict the cross sectional area of the weld bead. The modelling results indicated that the 5-10-1 network precisely predicted the cross-sectional area of the weld bead.