“…This framework has, until now, only been applied to simple 2d testcases with Reynolds numbers below 50, 000. Data-driven turbulence modelling is a recent development in the fluid-dynamics community and its merit has generally been resticted to relatively simple two-dimensional flows [15,16,17,18,19]. Data-driven approaches to turbulence modeling can be divided into two broad categories based on the underlying regression model: either using (a) extremely general models with a very large number of parameters, such as artificial neural networks and random forests [20,21,22,23,24,15]; or (b) and methods using symbolic algorithms such as sparse regression and Gene Expression Programming (GEP) which tend to result in concise, inspectable models [14,25,26,27,28].…”