The Study Of Parameter Estimation For Turbulence Model

Turbulence model exhibit a significant influence on computational fluid dynamic simulation. Some turbulence models have been established for engineering. But some empiric parameters were more or less contained in these turbulence models, and the value of the most empiric parameters were demarcated in the experiment with special condition, so they may not work on all condition. And the empiric parameters should be adjusted for the given flow condition in actual application. But the rule of the influence of the empiric parameters should be analyzed on computational fluid dynamic simulation before the empiric parameters were adjusted. Then Menter's k-ωSST turbulence model and SA one equation model used abroad in engineering were studied in the article.Firstly, the parameter calibration method of the standard k-εtwo-equation model is introduced, and, based on the standard k-εtwo-equation model the values of the parameters in the k-ωSST two-equation model are assigned. Another prevailing used turbulence model, SA one-equation model is studied.Secondly, by using the uniform design of experiments, the influences of the eight parameters of k-ωSST two-equation model on the calculated turbulent flow field are analyzed. The grid number is seen as a "parameter" and used together with model parameters of the k-ωSST two-equation model to carry out the parametric influence analyses. The results show that the influence of 10% variation of the parameter a1 on the calculated turbulent flow field is greater than the influence of grid number quadrupled. Then, when the grid size requirement is not fully satisfied in the numerical simulation, it's meaningful to adjust the parameter value of a1 to compensate. Thereafter, by using the uniform design of experiments, the influences of eight model parameters in Menter's k-ωSST turbulence model on the 2D airfoil flow field with separation are analyzed. The results show that for separated flow, because of exhibiting a significant influence on the calculated turbulent flow field, four parametersσω2,β2,β* and a1 should be attracted attention in study of parameter estimated. And then the influences of model parameters in Menter's k-ωSST turbulence model and grid number seen as a "parameter" for the 3D attached flow field are analyzed. The results show that for 3D attached flow, the influence of the parameter a1 is also greater than the influence of other parameters and grid number. Finally, the influences of the eight parameters of k-ωSST two-equation model on different reynolds number are analyzed.Thirdly, by using the uniform design of experiments, the influences of the eight parameters of SA one-equation model on the calculated turbulent flow fields are analyzed for two typical cases of transonic flow with low incidence, and subsonic flow with high incidence. The results show that for attached flow, the influence of the parameter Cb1 is much greater than the influence of other parameters. But for separated flow, the influence of the parameter Cb1 on the calculated turbulent flow field is not greatest, the parameterσof turbulence diffusion term and the parameter Ct3 of turbulence control transition term exhibit a significant influence on the calculated flow field with separation. So three parameters Cb1,σ,Ct3 should be attracted attention in study of parameter estimated.Finally, a preliminary parameter estimation example is given. And parameter estimation method for high-Reynolds number turbulence model with measured noise is carried out. And then parameter estimation method is used for optimizing parameter a1 on flows around DLR-F4 wing-body configuration. The estimated parameter for one incidence is used for computations of the other cases with different incidences. The prediction of aerodynamic coefficient of the most cases is better than original result, so, the results show that the estimated parameter exhibit universality. And then the interpolation technology was used for decreasing the influence of the grid error for parameter estimation. At last, a new parameter estimation method based on Genetic Algorithms and Kriging response surface model is constructed.