Multiscale Materials
Mission statement:
Our mission is to deliver scalable simulation tools and physics based models to a broad range of energy, transportation, material synthesis, separation and advanced manufacturing applications. Through this research, we aim to solve critical challenges that will improve the performance, economics and efficiency of real world applications.
Group summary
The group develops and delivers multiscale models for energy, materials and manufacturing applications using novel and unique methodologies and scalable tools. The expertise and capabilities within the group span a wide range of scales, such as atomistic and molecular scale simulations, mesoscale phase field simulations, microstructure models, interface resolved porous material simulations, discrete element particle simulations, macroscale thermo-mechanics models and multiphase flows.
Our research focuses on numerical models for additive manufacturing (AM) and hybrid additive/subtractive manufacturing (AM/SM) that combines metal additive manufacturing with machining in a single system. The group develops validated predictive tools for controlling distortion and residual stresses in hybrid manufacturing, paving the way for design optimization, qualification, and certification of complex metal components. The group has developed the first fully-coupled finite element simulation framework for hybrid AM/SM, capable of predicting distortion and residual strain/stress evolution throughout the interleaved process. The model was validated using neutron diffraction measurements at ORNL’s High Flux Isotope Reactor, ensuring high-fidelity strain mapping in three dimensions.
Our research has also advanced laser directed energy deposition (DED) of Nb-based refractory alloy by integrating high-fidelity multiphysics melt pool modeling with experimental validation. Using computational fluid dynamics (CFD) simulations and high-performance computing (HPC), process maps were developed to optimize geometric precision and minimize defects, thereby enabling defect-free, dimensionally accurate builds, reducing trial-and-error and accelerating adoption for aerospace and high-temperature applications.
The group is actively developing artificial intelligence (AI) solutions, such as deep-learning-based AI tools for rapid segmentation of melt pools and defects in additive manufacturing images. The group has developed methods using generative adversarial neural network (GAN) architecture to enable accurate identification of melt pool boundaries and defect features, reducing manual intervention and accelerating part qualification. Statistical analysis reveals strong spatial correlations between defects and melt pools, and sensitivity to thermal effects from laser passes. The approach is adaptable across materials and imaging modalities, supporting process optimization and improved reliability in metal AM.
The group capabilities allow modeling of complex physical and chemical processes in material synthesis such as the manufacturing of ceramic matrix composites (CMC) advancing predictive design for high temperature materials manufacturing. The group has developed a hybrid reduced-order model for silicon carbide (SiC) deposition using machine learning (ML) techniques, namely, principal component analysis (PCA) and autoencoder neural networks (AE). The approach enables efficient transport and accurate reconstruction of gas-phase and surface kinetics. Validated against detailed chemical vapor deposition simulations, the model achieves 99% accuracy and an eightfold computational cost reduction.
