Process modelling, control and monitoring for robotic wire-arc additive manufacturing

Wire-arc additive manufacturing (WAAM) is an emerging technology in advanced fabrication. In contrast to other additive manufacturing (AM) technologies, WAAM makes use of an electric arc as a heat-source to deposit metal material layer-by-layer to makes up the final part. WAAM has the advantages of high deposition rate and low costs, which makes WAAM more applicable for the production of large-scale complex components. Modern manufacturing states the importance of accuracy, stability, and repeatability of production. This issue also exists in WAAM, since it's an emerging manufacturing technology. To improve these issues, this study utilized advanced modern control algorithms and AI technology to realize process control, monitoring, and optimizing. In this study, a visual sensing system for the melt pool was developed, and the image of melt pool can be collected in real-time. Through using deep learning algorithm (Mask R-CNN), the melt pool in the image can be detected, segmented, and measured automatically. Through stimulating the WAAM process using Pseudo-Random Ternary (PRT) Signals and measuring the real-time width of the melt pool, its process dynamic can be modelled. Based on the dynamic model, a Model Predictive Control (MPC) was designed to implement real-time feedback control. As a repetitive process, previous layers may have effects on the geometry of the newly deposited layer. Therefore, this study also proposed a feedforward control strategy for WAAM, which utilized the information and experience of previous layers to adjust the strategy in the next layer. An up-to-data control algorithm, Model-Free Adaptive Iterative Learning Control (MFAILC), was employed. Experiments were conducted to verify the tracking and robustness performance of the proposed MPC and MFAILC algorithm.