Numerical and physical experiments in viscous separated flows

From this study of three quite different viscous separated flow problems using the same numerical techniques, it is apparent that experimentation is as important in the numerical sense as it is in the physical sense. The time-depen dent conservative equations with upwind differencing for advection terms produced stable solutions over a wide range of Reynolds numbers. The actual accuracy of the numerical solutions varied with the flow problem and mesh geometry. A variable mesh, concentrating mesh points in regions of steepest gradients, appears to be a convenient method of reducing the, ever present, artificial viscosity effect. Numerical solutions with engineering accuracy can be obtained if great care is taken to control this artificial viscosity effect. Furthermore, they allow the study of basic flow phenomena which would be more difficult to produce and study in physical experiments. These methods, however, are still in the developmental stages and are not available in foolproof form for use by design engineers.