Predicting the wall-shear stress and wall pressure through convolutional neural networks

The objective of this study is to assess the capability convolution-based neural networks predict wall quantities in a turbulent open channel flow. first tests are performed by training fully-convolutional network (FCN) 2D velocity-fluctuation fields at inner-scaled wall-normal location $y^{+}_{\rm target}$, using sampled velocity fluctuations wall-parallel planes located farther from wall, input}$. predictions FCN compared against proposed R-Net architecture. Since model found perform better than model, former architecture optimized streamwise and spanwise wall-shear-stress components pressure wall. dataset obtained DNS flow $Re_{\tau} = 180$ $550$. various locations, along with wall-shear stress pressure. At $Re_{\tau}=550$, both can take advantage self-similarity logarithmic region $y^{+} 50$ 100$ as input about 10% error prediction streamwise-fluctuations intensity. Further, also able wall-pressure $y^+ around intensity corresponding These results an encouraging starting point develop neural-network-based approaches for modelling turbulence near large-eddy simulations.