Reconstruction of Gappy Data for Cavity Flow Using Gappy Proper Orthogonal Decomposition

The present study examines the possibility of completing gappy flow fields with the method of gappy proper orthogonal decomposition (GPOD). The procedure is performed on a numerically simulated cavity flow. The DNS data is artificially made incomplete by randomly omitting localized data. Two levels of gappiness are examined to evaluate the GPOD procedure. The results indicate that the relative error between the GPOD estimation and the real field directly depends on the level of gappiness. As the gappiness increases, the prediction accuracy decreases. It is shown that the relative error does not monotonically decrease because of inherent noise in higher-level modes of energy. The optimal count of modes in GPOD procedure is obtained and discussed. The contribution of GPOD procedure on spatial experimental data is also addressed.

Keywords:

Proper Orthogonal Decomposition , Cavity Flow , Gappy Data , reconstruction , Energy mode