Advanced gas turbines operate at very high temperatures. Optimal cooling is necessary to ensure that the material properties of the turbine blade do not deteriorate under these harsh conditions. A methodology is developed for improved prediction of flow and heat transfer in turbine blade passages. A Reynolds-averaged thin-layer Navier-Stokes flow solver for structured grids entitled CFL3D is used. The code is modified to allow solution in a rotating frame of reference. This modification aids in circumventing the need to use computationally intensive time-accurate numerical methods. A good resolution of the region near the wall is paramount because this is the region which dictates the actual heat transfer between the fluid and the solid in contact. Menter's k-ω model and the Explicit Algebraic Stress Model (EASM) within the k-ω formulation are applied. Menter's k-ω model accounts for the near wall effects without any ad-hoc treatment and is not unrealistically too sensitive to the free-stream conditions. EASM accounts for the complex strain field and is superior when turbulent anisotropy is dominant. Special attention is given to turbulence anisotropy, buoyancy, and Coriolis effects on heat transfer. In stiff problems, fuzzy logic is used to automate the choice of CFL number. The proposed approach is tested for a stationary duct, a rotating duct and a U-bend case. For the first set of test data from Pratt & Whitney, the predicted solutions are compared to the experimental data for one stationary and three rotating cases. For the second set of test data from General Electric Company, the numerical solutions are compared to the experimental data for three rotating cases. For this set of data, the operating conditions are analogous to true gas turbine operating conditions. For the third set of data, which is the U-bend case from the Von Karman Institute for Fluid Dynamics, comparisons are made to the experimental data for two different conditions. For one case, comparison is also made to a previously published numerical solution. CFL3D predicted heat transfer qualitatively and quantitatively reasonably well. Fuzzy Logic proved to be a fairly useful tool in a few cases, at least in the initial part of the solution process. CFL3D code encountered difficulties in convergence when the speeds of rotations for the straight duct cases were increased. The convergence behavior was less than desirable for the U-bend cases wherein the physics was very complex. Even the use of fuzzy routine did not alleviate the convergence problems in such more involved cases.
展开▼