Factorial Design Applied to CFD

Abstract

Factorial design, a statistical method widely used for experiments, and its application to CFD are discussed. The aim is to propose a systematic, objective, and quantitative method for engineers to design a set of simulations in order to evaluate main and joint effects of input parameters on the numerical solution. The input parameters may be experimental uncertainty on boundary conditions, unknown boundary conditions, grid, differencing schemes, and turbulence models. The complex flow of the Turbine-99 test case, a hydropower draft tube flow, is used to illustrate the method, where four factors are chosen to perform a 24 factorial design. The radial velocity at the inlet (not measured) is shown to have an important influence on the pressure recovery (7%) and the energy loss factor (49%).