This paper presents a numerical investigation of unsteady flow phenomena in a two-stage partial admission axial steam turbine. Results from unsteady three-dimensional computations are analyzed and compared with the available experimental data. Partial admission in the present study is introduced into the model by blocking only one segmental arc of the inlet guide vanes. Blocking only one segment (which corresponds to the experimental setup) makes the model unsymmetrical; therefore it is necessary to model the whole annulus of the turbine. The first stage rotor blades experience large static pressure change on their surface while passing the blocked channel. The effect of blockage on the rotor blades’ surface pressure can be seen few passages around the blocked channel. Strong changes of the blades’ surface pressure impose large unsteady forces on the blades of first stage rotor row. The circumferential static pressure plots at different cross sections along the domain indicate how the non-uniformity propagates in the domain. A peak pressure drop is seen at the cross section downstream of the first stage stator row. At further downstream cross sections, the static pressure becomes more evenly distributed. Entropy generation is higher behind the blockage due to the strong mixing and other loss mechanisms involved with partial admission. Analysis of the entropy plots at different cross sections indicates that the peak entropy moves in a tangential direction while traveling to the downstream stages. Comparisons of the unsteady three-dimensional numerical results and the experimental measurement data show good agreement in tendency. However some differences are seen in the absolute values especially behind the blockage.

No abstract

Prediction of boundary-layer transition is important to turbomachine design. Various experimental correlations are still the most practical models used in engineering calculations. In an acceleration or deceleration flow field, however, the predictions depend on the free-stream turbulence intensity incorporated in the correlations. A numerical analysis is therefore presented in this paper to investigate how much the different implementations of the turbulence levels can affect the numerical results. The main emphasis is on the importance of such influence in engineering calculations in comparison with the importance of correlation and grid resolution effects. The analysis is performed using an industrial Reynolds-averaged Navier-Stokes solver with different transition correlations and models of free-stream turbulence intensity. Eight benchmark test cases, including basic incompressible flat-plate cases and more realistic transonic cascade cases, have been calculated. The study reveals that the effects of grid resolution and the choice of correlations on the transition results are relatively small. But a proper presentation of free-stream turbulence intensity in the correlations plays a vital role in this sort of calculation. The use of a computed local turbulence level with the correlations appears to be the most flexible choice among the approaches considered.

A numerical study has been performed to compare the overall performance of three transition models when used with an industrial Navier-Stokes solver. The three models investigated include two experimental correlations and an integrated eN method. Twelve test cases in realistic turbomachinery flow conditions have been calculated. The study reveals that all the three models can work numerically well with an industrial Navier-Stokes code, but the prediction accuracy of the models depends on flow conditions. In general, all the three models perform comparably well to predict the transition in weak or moderate adverse pressure-gradient regions. The two correlations have the merit if the transition starts in strong favorable pressure-gradient region under high Reynolds number condition. But only the eN method works well to predict the transition controlled by strong adverse pressure gradients. The three models also demonstrate different capabilities to model the effects of turbulence intensity and Reynolds number.

No abstract