No abstract

A computational study was performed for the flow and heat transfer in rotating coolant passages with two legs connected with a U-bend. The dimensionless flow conditions and the rotational speed were typical of those in the internal cooling passages of turbine blades. The calculations were performed for two geometries and flow conditions for which experimental heat transfer data were obtained under the NASA HOST project. The first model had smooth surfaces on all walls. The second model had opposing ribs staggered and angled at 45 deg. to the main flow direction on two walls of the legs, corresponding to the coolant passage surfaces adjacent to the pressure and suction surfaces of a turbine airfoil. Results from these calculations were compared with the previous measurements as well as with previous calculations for the nonrotating models at a Reynolds number of 25,000 and a rotation number of 0.24. At these conditions, the predicted heat transfer is known to be strongly influenced by the turbulence and wall models. The differential Reynolds-stress model (RSM) was used for the calculation. Local heat transfer results are presented as well as results averaged over wall segments. The averaged heat transfer predictions were close to the experimental results in the first leg of the channel, while the heat transfer in the second leg was overestimated by RSM. The flow field results showed a large amount of secondary flow in the channels with rotational velocities as large as 90 percent of the mean value. These secondary flows were attributed to the buoyancy effects, the Coriolis forces, the curvature of the bend and the orientation of the skewed ribs. Details of the flow field are discussed. Both the magnitude and the change of the heat transfer were captured well with the calculations for the rotating cases.

Effective turbine blade cooling is necessary to enhance the efficiency of gas turbines. Usually the coolant is mainly ejected through the trailing edge of the vanes. In addition to the desired temperature reduction at the trailing edge there is a 3D-aerodynamical interaction between the hot gas and the coolant. The complex mechanisms of the mixture are a main problem in the numerical prediction of the flow situation in this region. This paper presents the experimental and numerical results of investigations of annular guide vanes. The experiments were conducted in a scaled turbine test rig. The mixing flow of coolant and hot gas was analyzed by measurement of the distribution of both velocity and turbulence very close to the trailing edge using a 2D-LDA measurement technique at different radial positions. The experimental results show that the radial and circumferential distribution of the coolant depends on the pressure gradient in both directions. Inside of the mixture region the turbulence was found to be anisotropic resulting in a non-symmetrical distribution of the coolant. For the numerical calculations a Navier-Stokes-Code was used. The numerical scheme works on the basis of an implicit finite volume method combined with a multi block technique. In order to simulate the aerodynamical effects near the injection slot of the vane it was nessessary to include the coolant flow inside the guide vane.

A computational study was performed for the flow and heat transfer in coolant passages with two legs connected with a Ubend and with dimensionless flow conditions typical of those in the internal cooling passages of turbine blades. The first model had smooth surfaces on all walls. The second model had opposing ribs staggered and angled at 45 0 to the main flow direction on two walls of the legs, corresponding to the coolant passage surfaces adjacent to the pressure and suction surfaces of a turbine airfoil. For the ribbed model, the ratio of rib height to duct hydraulic diameter equaled 0.1, and the ratio of rib spacing to rib height equaled 10. Comparisons of calculations with previous measurements are made for a Reynolds number of 25,000. With these conditions, the predicted heat transfer is known to be strongly influenced by the turbulence and wall models. The k-e model, the low Reynolds number RNG k-e and the differential Reynolds-stress model (RSM) were used for the smooth wall model calculation. Based on the results with the smooth walls, the calculations for the ribbed walls were performed using the RSM and k-e turbulence models. The high secondary flow induced by the ribs leads to an increased heat transfer in both legs. However, the heat transfer was nearly unchanged between the smooth wall model and the ribbed model within the bend region. The agreement between the predicted segment -averaged and previously -measured Nusselt numbers was good for both cases.