Sensor-Based In-situ Process Control of Robotic Wire Arc Additive Manufacturing Integrated with Machine Learning

Abstract

Wire Arc Additive Manufacturing (WAAM) is a manufacturing technology that has the capability to fabricate a large-scale metallic part in a layer-by-layer fashion. It is receiving significant attention from many industries as a viable method of manufacturing as it has a high deposition rate, production rate, and cost efficiency. However, numerous challenges still need to be addressed and overcome to ensure the geometrical accuracy of the manufactured goods produced. As the number of deposited layers increases, geometrical errors increase, and the accumulated heat becomes significant, leading to the undesirable slumping of the beads. The quality of the part can be enhanced through in-situ real-time feedback control. However, as WAAM is a time-variant process that is highly non-linear and multi-dimensional, it is difficult to model the relation between the process parameters and the final quality of the produced part. To address this challenge, a sensor-based in-situ data-driven process control framework integrated with machine learning (ML) is proposed to iteratively learn from the feedback, the impacts of various process parameters to ultimately control the geometry of a single-bead multi-layer part to conform to desired geometrical specifications. The proposed control framework is then implemented and validated on a custom robotic large-scale WAAM system. The experiment result showed that the beads printed with the proposed control framework had a noticeable improvement in both consistency and following the user-specified bead’s geometry, in comparison to traditional printing beads.