Enhancing wire arc directed energy deposition for challenging printing tasks: A VR-based skill-learning approach

The recent development of wire arc directed energy deposition (WA-DED) has made the additive manufacturing of medium to large-sized metal parts possible with acceptable cost. However, the current WA-DED system commonly uses a fixed set of parameters for each weld bead, which results in a uniform weld bead geometry that fall short of meeting the demands of high-challenging tasks such as fabricating complex parts with intricate geometries. In contrast, experienced human welders can control the weld bead shapes during the deposition process by flexibly adjusting deposition parameters, thus realizing the fabrication of components with more complex geometries. This study aims to bridge this gap by distilling human skills and applying them to WA-DED system for the fabrication of metal parts with complex geometry. To learn the skills of experienced human welders, a skilled welder was asked to execute a series of predefined welding tasks, and a virtual reality (VR) based human motion capture system was developed to capture data on the real-time torch pose and welding parameters. The data collected from the human welder, after being processed through the developed data processing module, was utilized in the proposed PSO-BPNN-based predictive models. The inputs to the developed predictive model include travel speed (TS), welding angles (WAs), and the contact tip to workpiece distance (CTWD), while the outputs are the bead width (BW) and bead height (BH) of the two segments of a geometry-varying weld bead. The model can predict a great number of continuous parameter combinations for complex weld bead deposition with high precision. Then, the accuracies of the forward and backward were validated through experimentation. Finally, a real-world case study demonstrates the effectiveness of the proposed strategy, addressing its potential to broaden the applicability of WA-DED technologies significantly.