We present a new finite difference-based thermo-fluid simulation method by using reduced order models. Reduced order modeling technique is to project a high dimensional problem onto a lower dimensional subspace and then solve the problem in the reduced subspace. We solve the incompressible Navier-Stokes equations and the heat equation in their respective subspaces and join the two equations by projecting the advection term onto the joint reduced space. Our algorithm allows for fast thermo-fluid simulation of the large-scale computational models at a fraction of computational load. Our method shows more than one thousand times faster speed than usual mesh-based fluid simulation, while maintaining acceptable engineering accuracy. The real-time feature of our method enables interactive user interfaces for, e.g., assessing an immediate thermal field change in response to consecutive control of air-conditioning equipment in a data center.

This content is only available via PDF.
You do not currently have access to this content.