(389g) Keynote Talk: Faster Prediction with Artificial Intelligence: Using Process Simulations and Production Data to Develop Fast-Running Inference Models for Manufacturing

Navigating a large-dimensional design space with simulations is a significant improvement over doing so with physical experiments but can still be very time consuming. Deep learning methods are often used to develop fast-running surrogate models of more computationally expensive simulations in order to get near-real-time predictions and to optimize complex systems such as manufacturing processes, energy systems, and fusion reactors. In this talk, I will discuss an application of machine learning to develop a fast-running surrogate models that captures the dynamics of industrial multiphase fluid flows. I will also discuss an improved population search method that can help the analyst explore a high-dimensional parameter space to optimize production while reducing the model uncertainty. Results from demonstration problems and from real-world Inertial Confinement Fusion simulations will be presented.