A database of PIV measurements within a turbomachinery stage and sample comparisons with unsteady RANS


Uzol O. , Brzozowski D., Chow Y., Katz J., Meneveau C.

JOURNAL OF TURBULENCE, cilt.8, ss.1-20, 2007 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 8 Konu: 10
  • Basım Tarihi: 2007
  • Doi Numarası: 10.1080/14685240601142867
  • Dergi Adı: JOURNAL OF TURBULENCE
  • Sayfa Sayıları: ss.1-20

Özet

This paper describes an experimental database obtained using two-dimensional Particle Image Velocimetry (PIV) measurements within an axial turbomachinery stage, and presents sample unsteady Reynolds Averaged Navier - Stokes (RANS) simulations to illustrate its applicability for turbomachinery model validation. The experiments are performed in a refractive-index-matched facility that provides unobstructed view, and cover the entire second stage of a two-stage axial pump. The data were obtained at ten different rotor phases covering one rotor blade-passing period, and at mid-span. Several features of the data at selected phases have already been presented and discussed in prior publications. Here we present the complete database together with sample CFD results. Two-dimensional unsteady RANS simulations are performed using the commercial flow solver FLUENTTM, with two standard turbulence models, i.e. Renormalization Group (RNG) k-epsilon and Reynolds Stress Transport Model (RSM). The spatially non-uniform inlet velocity and turbulence boundary conditions are provided from the experimental data. Detailed side-by-side comparisons of measured and computed velocity as well as turbulence fields within the entire stage are presented using line distributions within the rotor - stator gap and the stator wake regions, as well as whole-field animations. The results show that, although there is reasonable agreement, in general, between the experimental results and the computational simulations, some critical flow features are not correctly predicted. The turbulent kinetic energy levels are generally too high in the simulations, with substantial amount of unphysical turbulence generation near the blade leading edges, especially in the case of RNG k-epsilon model. Also, wake diffusion is underestimated. The results highlight the usefulness of comparisons that cover the entire unsteady flow in a passage, as afforded by the present database and side-by-side animations.