Reduced Order Modeling and Compression of Data Produced by Simulations of Transient and Periodic Heat Transfer Processes

Abstract

Reduced Order Modeling may be used to obtain compact and computationally efficient representations of complex dynamic systems. The objective of this paper is to demonstrate the application of reduced order modeling techniques to systems undergoing thermal transients. In this paper, a reduced order model is defined as a spectral method in which the dominant features of a spatially and temporally varying temperature profile are represented using a relatively small set of basis vectors. Although various approaches are possible, reduced order modeling generally relies on the use the singular value decomposition of a matrix containing representative data to generate an orthonormal basis for the process to be modeled. The results presented in this paper illustrate reduced order modeling of periodic and transient heat transfer in an axisymmetric system. Measures demonstrating the accuracy and computational savings associated with the use of reduced order modeling are presented.