Numerical modeling of transition to turbulence in low-pressure axial gas turbines.

摘要

Experimental data from modern turbofan engines indicate that the low-pressure turbine stages experience a significant drop in efficiency as the aircraft reaches its cruise conditions at high altitude. Under these circumstances, the low Reynolds number flow allows the apparition of a boundary layer which is no longer turbulent but transitional in nature. A further decrease in velocity may lead to the separation of the highly unstable laminar portion accompanied by a dramatic growth in aerodynamic losses.; The methods for numerically simulating the transitional flows occurring over turbine blades were reviewed. Two large categories were identified as suitable for numerical implementation into a fully-implicit, finite-difference, Navier-Stokes code. The first involved a Baldwin-Lomax turbulence model corrected for attached flow transition with an intermittency factor distribution. The general expression of Solomon, Walker and Gostelow was added to the code, in parallel with the zero-pressure gradient form of Narasimha, used for reference. In both cases transition inception is detected with the Abu-Ghannam Shaw correlation. Whenever laminar separation takes place, Robert's correlation for short bubble transition is activated. The second category comprised the two-equation, low Reynolds number turbulence models of Chien and Launder-Sharma. They have a certain ability to predict bypass transition and seem to better comprehend the physics of wake-induced transition. For the approximate factorization solution algorithm, the implicit part of the Launder-Sharma system was expressed in an original form. Also, the Kato-Launder correction was added to be used as an option.; Numerical investigations of attached flow bypass transition and separated flow short bubble transitions were performed on two cascade geometries. The Abu-Ghannam Shaw criterion proved to be inaccurate for curved surfaces. The Solomon, Walker Gostelow distribution did not perform better than Narasimha's. Launder-Sharma model was more dependent than Chien's model upon the turbulence boundary conditions in the attached flow case. Both underpredict the transition length. The Baldwin-Lomax turbulence closure in combination with Robert's correlation gave accurate results for the short bubble predictions. Launder-Sharma model was extremely reliable for all separated flow test cases but special precautions had to be taken to achieve convergence.