Materials Manufacturing Optimization via Coupled Artificial Intelligence, Simulation, and Experiments

Accelerated development and deployment of next generation materials is a critical component of technological solutions required for a clean energy transition. Materials manufacturing processes are often characterized by their complex, multi-scale physics and the numerous processing variables that affect product quality, yield, and process efficiency. This challenges the traditional trial and error driven development process that relies on repeated experimental trials driven by operator expertise. An opportunity exists to greatly reduce development timelines while simultaneously reducing costs and increasing product performance and quality by leveraging artificial intelligence methods to suggest optimal combinations of simulations and limited experiments. In this presentation, we share a methodology consisting of careful multi-objective cost function definition, Gaussian process surrogate modeling, and multi-fidelity process simulations. Example applications will be presented for processes including flame spray pyrolysis, atomic layer deposition, and induction pipe bending, in addition to an in-depth exploration of the optimization of the polymer melt blowing process.