About this Abstract |
Meeting |
2024 TMS Annual Meeting & Exhibition
|
Symposium
|
Additive Manufacturing Modeling, Simulation and Machine Learning
|
Presentation Title |
Planning and Adaptive Control of AM Processes via In Situ Characterization, Faster-than-real-time Simulations, and AI/ML Methods |
Author(s) |
Stephen J. DeWitt, Bruno Turcksin, James Haley, Ke An, Yousub Lee, Thomas Feldhausen, Venkatakrishnan Singanalllur, Ayana Ghosh |
On-Site Speaker (Planned) |
Stephen J. DeWitt |
Abstract Scope |
Improved understanding and control of additive manufacturing (AM) processes are important for accelerating the qualification of AM components. We present a new approach for planning, monitoring, and adaptively controlling AM processes that combines in situ characterization, simulations, and AI/ML methods. We use thermomechanical simulations and active learning methods to determine baseline process parameters that are expected to yield a target surface displacement field. As we print with those parameters, we combine in situ infrared camera data with faster-than-real-time simulations to estimate the state of the part via data assimilation. Starting from that state, forward-looking simulations explore parameter sets for upcoming layers, sending the best parameter sets to the printer. We present validation experiments of subsystems of the adaptive control system using in situ and ex situ neutron diffraction. Then, we present the results of experiments where builds with the adaptive control system are compared to builds with static expert-chosen parameters. |
Proceedings Inclusion? |
Planned: |
Keywords |
Additive Manufacturing, Computational Materials Science & Engineering, Machine Learning |