DMS Statistics and Data Science Seminar

Speaker: Xiaowei Yue, Virginia Tech

Title: Stochastic Surrogate Models: Method, Algorithm, and Engineering Applications

Abstract: Surrogate models have been widely used in advanced design and manufacturing to tackle the high computational cost in high-fidelity simulation and digital twin. Due to sensing errors, actuating errors, and computational errors, uncertainties inevitably exist in engineering systems. With incorporating the influence of uncertainties, stochastic surrogate modeling has become an emerging field. We developed two stochastic surrogates: (1) Neural Process Aided Ordinary Differential Equation (NP-ODE); (2) Neural network Gaussian process considering input uncertainty. We show the relationships between deep neural networks, Gaussian process, and differential equations, and use their relationship to develop new physics-informed data analytics methods. We also demonstrate their applications in engineering simulations such as Finite Element simulation, Digital Twin, and Materials Science simulation.

Location : ZOOM