Search for: All records

Gas Metal Arc Welding (GMAW) is a critical industrial technique known for its high productivity, flexibility, and adaptability to automation. Despite the significant advancements in robotic welding, challenges remain in fully automating the arc welding process, particularly due to the complex dynamics of the weld pool associated with GMAW. A human-robot collaborative (HRC) system where humans operate robots may conveniently provide the needed adaptive control to the complex GMAW. While in conventional HRC systems humans receive process feedback to make adaptive adjustments, we propose provide humans with predictive future feedback to further ease the human decision and reduce the needed skills/trainings. To this end, this study explores the integration of deep learning models, specifically Generative Adversarial Networks (GANs) combined with Gated Recurrent Units (GRUs), to model and predict the dynamic behavior of the weld pool during GMAW. By leveraging time-series data of torch movement and corresponding weld pool images, the proposed GRU-GAN model generates high-fidelity weld pool images, capturing the intricate relationship between speed variations and weld pool morphology. Through extensive experimentation, including the design of an acceptable Encoder-Decoder structure for the GAN, we demonstrate that incorporating both temporal and speed sequence information significantly enhances the model's predictive capabilities. The findings validate the hypothesis that dynamic torch speed adjustments, akin to those performed by skilled human welders, can be effectively modeled to improve the quality of automated welding processes. Future work will be devoted to human-based model predictive control (MPC) in an HRC environment. 

This study aims to extract critical scenes/continents in the weld pool region during gas metal arc welding (GMAW). The scenes considered include the wire, arc, and weld pool, while other secondary ones such as oxides are temporarily excluded. They are critical to understanding, analyzing, monitoring and controlling the welding process, in particular the critical correlation how the welding parameter, arc and weld pool are dynamically correlated. Unfortunately, such fundamental correlation has not been studied and lack of effective ways to simultaneously monitor/extract these scenes is responsible. With the development of optoelectronic devices, weld pool regions can be better imaged. However, because of the nature of the scenes in particular the arc which is formed by ionized gas without a clear boundary and highly dynamic, detecting them using computer vision is challenging. Deep learning is an effective method, but model training usually needs a large number of labels. As manually labeling is expensive, we propose an approach to address this challenge that can train a model from a small, inaccurately labeled dataset. This approach is designed, per the characteristics of the scenes and their dynamics All-in-One Network (AOD-Net) was deployed first for defogging, and then the YOLOX network was utilized to identify regions of interest to reduce the impact of molten metal splashes on image sharpness. Subsequently, we developed a timed segmentation network incorporating the Long Short-Term Memory (LSTM) mechanism into U-Net, which can be used to extract more accurate information about the weld pool by combining the temporal and spatial information in the continuous process of welding at a low cost because our scene of interest is in a continuous and dynamic evolutionary process. After defogging and removing the effects of molten metal spatter, we can obtain information on the dynamics of the weld pool and the weld arc at the same time. Experimental results verified that the trained network could extract the critical boundaries accurately under various welding conditions despite the highly dynamic changes and fuzziness of the views.