Machine Vision to Provide Quantitative Analysis of Meltpool Stability for a Coaxial Wire Directed Energy Deposition Process

Abstract

Wire-based additive manufacturing (AM) is at the forefront of complex metal fabrication because of its scalability for large components, potential for high deposition rates, and ease of use. A common goal of wire directed energy deposition (DED) is preserving a stable process throughout deposition. If too little energy is put into the deposition, the wire will stub into the substrate and begin oscillating, creating turbulence within the meltpool. If too much energy exists, the wire will overheat, causing surface tension to take over and create liquid drips as opposed to a solid bead. This paper proposes a computer vision technique to work as both a state detection and event detection system for wire stability. The model utilizes intensity variations along with frame-to-frame difference calculations to determine process stability. Because the proposed model does not rely on machine learning techniques, it is possible for an individual to interpret and adjust as they see fit. The first part of this paper describes creation and implementation of the model. The model’s capability was then evaluated using a 1D laser power experiment, which generated a wide range of stability states across varying powers. The model’s accuracy was evaluated through 3D geometry data gathered from the experimentally deposited beads. The model proved to be both capable and accurate and has potential to be used as a real-time control system with future work.