Argonne’s Laboratory Computing Resource Center (LCRC) operates Blues, a 330-node computing cluster and Fusion, a 320-node computing cluster, as production systems and software development platforms. LCRC also provides user support and training, and maintains a wide range of scalable applications and tools.
Argonne’s interests in metal additive manufacturing computation research are focused on developing ways of efficiently coupling computation solvers (please see list below) so that molecular dynamics simulations can be used to generate the input data required to accurately solve problems at larger length scales. In particular, the following software tools (solvers) can be run on Argonne’s LCRC Blues and Fusion clusters to develop metal additive manufacturing models:
- LAMMPS to carry out molecular dynamics simulations. LAMMPS is currently used to extract at the atomic level the dynamics of the melt pool and solidification, melting, and sublimation of different alloys, and the impact of porosity at the atomic level.
- SPPARKS to carry out kinetic Monte Carlo simulations. It is a fantastic test bed to develop coarse-scale simulations, for instance to incorporate simplified kinetics to understand the impact of different transport parameters on part shape.
- OpenFOAM and FeNiCs to carry out simulations at the continuum level, including understanding the impact of energy transport and heat dissipation on the porosity of manufactured parts, and the validation and calibration of simpler models.
Researchers can also run finite volume, Monte Carlo, and molecular dynamics simulations using two complementary approaches: undertake complex problems that are computationally expensive, or run simultaneously thousands of simulations to explore the parameter space in a simpler representation of aspects of additive manufacturing. These can be used to efficiently explore the space available for additive manufacturing, go beyond hardware limitations to predict what the impact of some new conditions would be on the additive manufacturing process, and for optimization of the parameter conditions.
Additionally, Argonne’s analytical capabilities in reduced order modeling include deep learning; agent-based methods using tools such as the Repast Suite software systems and multi-scale model integration also using the Repast Suite and machine learning for pattern and anomaly detection. Expertise in reduced order modeling can be used to develop lower fidelity models that can serve as computational efficient, everyday modeling and simulation tools.