Development and Validation of a Machine Learned Turbulence Model

A stand-alone machine learned turbulence model is developed and applied for the solution of steady unsteady boundary layer equations, issues constraints associated with are investigated. The results demonstrate that an accurately trained can provide grid convergent, smooth solutions, work in extrapolation mode, converge to a correct from ill-posed flow conditions. accuracy response surface depends on choice variables, training approach minimize overlap datasets. For former, grouping variables into problem relevant parameter input features desirable. latter, incorporation physics-based during helpful. Data clustering also identified be useful tool as it avoids skewness towards dominant feature.