Session: AM3D1.1 - Current and Emerging Trends in AM

Paper Number: 158782

158782 - Enabling Dimensional Accuracy in Wire Arc Additive Manufacturing: A Machine Learning Based Approach 

Wire arc additive manufacturing (WAAM) is a fast-growing metal 3D printing process where a metallic wire feedstock is melted and deposited using a welding electric arc heat source to manufacture components. WAAM is best known for its high deposition rates, enabling large structural components to be built faster and cheaper than any other metal 3D printing process. In addition, WAAM is lean and clean in terms of materials and energy consumption compared to conventional manufacturing processes. However, some key issues, such as heat accumulation, severely affect the quality of the components made with WAAM. Such challenges can be overcome by controlling the key WAAM deposition parameters (welding current, voltage, and travel speed) in real time. This research focuses on enabling dimensional accuracy in WAAM by controlling travel speed with a machine learning (ML) approach. This work encompasses three sub-systems to create an integrated intelligent manufacturing system: a vision system, a machine learning model, and a relay network. The real-time vision system consisting of a high-speed high dynamic range welding camera captures the WAAM bead widths. Multiple beads are deposited with varying travel speeds to capture the change in bead widths with travel speed. A relay network is integrated with the welding robot to adjust the travel speed in real time. Then, a machine learning model is developed to predict the relationship between the bead width and the travel speed. An integrated manufacturing system is developed using the combination of the real-time vision system, the ML model, and the relays, where the desired bead widths can be achieved in real time. A predefined shape geometry with varying bead widths is designed, and the required bead widths are used as input to the ML model. The model generates the required travel speeds to be continuously fed to the relay system, and the robot's travel speeds are updated accordingly. Simultaneously, the real-time vision system measures the bead width, continuously monitoring any fluctuations in bead width. Experimental results showcase the effectiveness of the integrated system in achieving dimensional accuracy for WAAM components.

Presenting Author: Ajay V Indian Institute of Technology Bombay

Presenting Author Biography: Ajay V is a PhD scholar at the Mechanical Engineeringng Department at the Indian Institute of Technology Bombay, Mumbai, India. Ajay's research area includes process monitoring of wire arc additive manufacturing.