Physics Informed Neural Networks for Engineering Systems
Abstract
This thesis explores the application of deep learning techniques to problems in fluid mechanics, with particular focus on physics informed neural networks. Physics informed neural networks leverage the information gathered over centuries in the form of physical laws mathematically represented in the form of partial differential equations to make up for the dearth of data associated with engineering and physical systems. To demonstrate the capability of physics informed neural networks, an inverse and a forward problem are considered. The inverse problem involves discovering a spatially varying concentration field from the observations of concentration of a passive scalar. A forward problem involving conjugate heat transfer is solved as well, where the boundary conditions on velocity and temperature are used to discover the velocity, pressure and temperature fields in the entire domain. The predictions of the physics informed neural networks are compared against simulated data generated using OpenFOAM.
Degree
M.Sc.
Advisors
Ardekani, Purdue University.
Subject Area
Engineering|Physics|Artificial intelligence|Fluid mechanics|Mathematics|Mechanics|Thermodynamics
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.