Physics simulation, at ever increasing computational expense, has been a core component of academic and industrial engineering activity for decades. More recently with the meteoric rise of the machine learning and data science fields has come a panoply of tools and methods that can be used to organize and infer relationships within datasets. When we combine simulation and ML techniques, we can improve the accuracy as well as the utility of simulation results either through calibration using experimental observations or creating surrogate models of their predictions (or both). In this talk I’ll review a series of off-the-shelf ML tools (clustering, deep neural network regression, Gaussian processes) deployed in both novel and excruciatingly practical use cases spanning aerodynamics to real time process optimization and control. No background in machine learning will be needed to follow along.
Kyle Mooney Ph.D. is a Senior Applications Engineer at Geminus.AI, a Silicon Valley startup from Palo Alto, California. His career has primarily involved the use of computational physics in industrial applications such as automotive aerodynamics, bio-medical device design, ag-tech, and energy sector process optimization. He received his doctorate in Mechanical and Industrial Engineering in 2016 with dissertation research focused on computational fluids dynamics simulation of multiphase droplet dynamics.