Data-Driven Algebraic Models Tuned with a Vast Experimental Database of Separated Flows

Abstract

This work provides an extended experimental database characterizing the transition process induced by flow separation on a flat plate with variable adverse pressure gradient. The current database has been used to tune data-driven turbulence models for laminar separation bubbles. The database collects the 2D components of the Reynolds stress tensor for 87 flow cases characterized by different flow Reynolds numbers, freestream turbulence intensities and adverse pressure gradients. The ranges of variation of parameters (15 k\(\le\)Re\(\le\)80 k, 1.5%\(\le\)Tu\(\le\)3.5%, \(-\)0.41\(\le\)AP\(\le\) \(-\)0.18) covers a wide area of internal flow conditions characterizing turbomachinery operation. Within this range bursting process occurs, with the consequent formation of separation bubbles of both short and long types. Sparse Bayesian Learning has been exploited to tune two models of the Reynolds stress tensor for short and long bubbles. The models have been formulated in terms of the invariants of the strain and rotation tensors with the addition of different physical-based flow features to account for the strong anisotropy that can be observed in the transitional region. The capability of the tuned relations in reproducing the anisotropy tensor along the separated shear layer is here shown by predicting flow cases that did not participate in the tuning process. The accuracy of the present model is compared with previous models available in the literature. The paper aims at providing a thorough database for future researches in the field and to improve turbulence models and RANS capabilities in the simulation of separated flows. Data presented in the paper can be downloaded at: https://github.com/danielepetronio/LSB_unige_data.