Perspectives on ZENN AI Framework for Advanced Manufacturing

Manufacturing represents the changes of states of materials systems. Thermodynamics fundamentally describes the state of a system and how that state is established, maintained, and altered by external forcing through the derivatives of the system’s Helmholtz energy with respect to external and internal variables. The fundamental challenge is to construct this multidimensional Helmholtz-energy landscape. Based on statistical mechanics, a system is composed of configurations. For simple systems, one can use DFT to explore configurations and quantitively predict their Helmholtz energies, resulting in the development of zentropy theory (https://doi.org/10.1007/s11669-022-00942-z). For more complex systems, the landscape can be computed from MD simulations based on zentropy theory (https://doi.org/10.1103/physrevresearch.7.l012030). For systems intractable by current simulation tools, ZENN AI framework (zentropy-enhanced neural network) was recently developed to learn the configurations and their internal energies and entropies (https://doi.org/10.1073/pnas.2511227122). Zentropy theory, ZENN, and their potential applications for manufacturing will be discussed in this presentation.