POD-interpolation based prediction of indoor airflows

Abstract

This work reports a proper orthogonal decomposition (POD)-interpolation based prediction of indoor airows related to displacement ventilation. Steadystate computational uid dynamics (CFD) solution snapshots with varying relevant non-dimensional number are used to estimate the dominant POD coe _cients/modal amplitudes and POD modes. A cubic spline interpolation of the POD coe_cients is used to compute the solution for desired value of the non-dimensional number of interests. The veri_cation and validation of this data-driven procedure is performed considering a 2D mixed convection problem involving a horizontal channel with cavity heated from below for a range of Richardson numbers. On the other hand, CFD solutions for a standard displacement ventilation con_guration is used to decompose the ow _eld variables in terms of Archimedes number dependent POD coe_cients and associated space dependent POD bases. A detailed analysis of the CFD and POD-interpolated predicted ow-_eld variables for displacement ventilation cases, error estimates and the spatial structures of the POD modes are presented.