The curse of numerical noise and implications for CFD-based design optimization

Numerical noise is an unavoidable by-product of Computational Fluid Dynamics (CFD) simulations which, in the context of design optimization, may lead to challenges in finding optimum designs. This article draws attention to this issue by illustrating the difficulties it can cause for road vehicle aerodynamics simulations. Firstly, a benchmark problem, flow past the Ahmed body, is used to highlight the effect of numerical noise on the calculation of aerodynamic drag. A series of simulations are conducted using three commonly employed Reynolds-Averaged Navier-Stokes (RANS) based turbulence models. Noise amplitudes of up to 22% are evident and the level of noise depends on the combination of turbulence model and grid used. Overall the Spalart Allmaras model is shown to be the least susceptible to noise levels for this particular application. Secondly, multi-objective aerodynamic shape optimization is applied to a low-drag aerodynamic fairing for a livestock trailer. The fairing is parameterised in terms of three design variables. Moving Least Squares (MLS) metamodels are constructed from 50 high-fidelity CFD solutions for two objective functions. Subsequent optimization is successful for the first objective, however numerical noise levels in excess of 7% are found to be problematic for the second one. A deliberate revision to the problem reduces the amount of noise present and leads to success with the construction of a small Pareto Front. Further analysis underlines the inherent capability of MLS metamodels in dealing with noisy CFD responses. Suggestions are also made to improve the chances of success for future investigations.