Robert G. Landers

Balark Tiwari, Nishan Khadka, Nicholas Capps et al.

Digital Glass Forming (DGF) is a new manufacturing process for low-batch glass fabrication. The work zone temperature in DGF processes must be maintained in the glass's working range to ensure good fabrication. If the temperature is too low, the filament will not wet to the substrate or previously deposited material and, if the temperature is too high, the filament may disengage from the substrate or previously deposited material, or it may partially vaporize. In this work, a real-time temperature control system capable of synchronizing process parameter, thermal camera, and visual camera data for the DGF process is introduced. A process parameter map for a scan velocity of 0.5 mm/s is constructed, as is a data-driven dynamic temperature process model. A digital controller is designed to regulate the work zone temperature. The temperature controller is a closed loop tracking controller that adjusts the commanded laser power to regulate the measured temperature. Two sets of experiments are conducted to analyze the controller performance. In the first set of experiments, single tracks on a substrate are fabricated with constant laser power and with the closed loop temperature controller. It is seen that the closed loop controller is able to extend the process parameter map into regions where using a constant laser power will result in a failed build. In the second set of experiments, walls are fabricated. Using constant laser power results in a failed build (i.e., material vaporization at the corners and the filament prematurely detaching from the substrate) as the temperature process dynamics change with layer and at the corners. The closed loop controller successfully fabricated the wall without vaporization at the corners and premature filament detachment as the controller adjusts the laser power to account for the changing temperature process dynamics.

Grigorii Khvatskii, Yong Suk Lee, Corey Angst et al.

This paper examines the performance of Multimodal LLMs (MLLMs) in skilled production work, with a focus on welding. Using a novel data set of real-world and online weld images, annotated by a domain expert, we evaluate the performance of two state-of-the-art MLLMs in assessing weld acceptability across three contexts: RV \& Marine, Aeronautical, and Farming. While both models perform better on online images, likely due to prior exposure or memorization, they also perform relatively well on unseen, real-world weld images. Additionally, we introduce WeldPrompt, a prompting strategy that combines Chain-of-Thought generation with in-context learning to mitigate hallucinations and improve reasoning. WeldPrompt improves model recall in certain contexts but exhibits inconsistent performance across others. These results underscore the limitations and potentials of MLLMs in high-stakes technical domains and highlight the importance of fine-tuning, domain-specific data, and more sophisticated prompting strategies to improve model reliability. The study opens avenues for further research into multimodal learning in industry applications.

Sina Malakpour Estalaki, Cody S. Lough, Robert G. Landers et al.

Variation in the local thermal history during the laser powder bed fusion (LPBF) process in additive manufacturing (AM) can cause microporosity defects. in-situ sensing has been proposed to monitor the AM process to minimize defects, but the success requires establishing a quantitative relationship between the sensing data and the porosity, which is especially challenging for a large number of variables and computationally costly. In this work, we develop machine learning (ML) models that can use in-situ thermographic data to predict the microporosity of LPBF stainless steel materials. This work considers two identified key features from the thermal histories: the time above the apparent melting threshold (/tau) and the maximum radiance (T_{max}). These features are computed, stored for each voxel in the built material, are used as inputs. The binary state of each voxel, either defective or normal, is the output. Different ML models are trained and tested for the binary classification task. In addition to using the thermal features of each voxel to predict its own state, the thermal features of neighboring voxels are also included as inputs. This is shown to improve the prediction accuracy, which is consistent with thermal transport physics around each voxel contributing to its final state. Among the models trained, the F1 scores on test sets reach above 0.96 for random forests. Feature importance analysis based on the ML models shows that T_{max}is more important to the voxel state than /tau. The analysis also finds that the thermal history of the voxels above the present voxel is more influential than those beneath it.