关键词: 湍流模型, SST, 数据驱动技术, 可压缩, 分离流, 符号回归

Abstract:

The Reynolds-averaged turbulence model has high computational efficiency， and is of great significance in engineering applications. The purpose of the improvement of the traditional and the new data-driven modes of turbulence model is to improve the prediction accuracy. However， the new data-driven mode is mainly for the low-speed flow， and there are less reports on the application， evaluation and promotion of the mode for the high-speed flow. The lack of comparative study of the two modes also causes problems for their rational use. Under the frame of the standard Reynolds Shear Stress Transport （SST） model， the traditional mode is adopted to introduce the compressibility effect into the dissipative term of the transport equation， and the new data-driven mode is used to modify the dissipative term. The improvement in the two modes mainly acts on the shear layer in the flow field. The test results of supersonic compression corner and supersonic cavity ramp show that the new data-driven mode can obtain a preliminary and explainable nonlinear relationship between different physical characteristics， but can only be applied after adjustment and re-optimization of the traditional mode. It has a certain generalization and is better than the traditional mode in capturing some turbulence details， but the accuracy still needs to be improved.

Key words: turbulence model, SST, data-driven technology, compressible, separated flow, symbolic regression