Optimizing vertical wind tunnel design through integrated numerical and statistical modeling

This paper presents a mathematical optimization (meta-modeling) approach for wind tunnel design, focusing on a vertical wind tunnel for droplet levitation. A robust, interconnected feedback methodology integrates component sizing and layout within a numerical tool for computational flow analysis. Numerical analysis of guide vanes minimizes the loss coefficient (as low as 0.03 for $Re\approx 0.2\times 10^6$ and 0.18 for $Re\approx 15000$) and optimizes wind tunnel dimensions. Rather than relying on intuitive or literature-based parameter choices, the method solves a multi-variable design problem using response surface approximations. Statistical models of expensive functions – derived from Navier–Stokes solutions – account for non-linear and discontinuous behavior. The proposed methodology identifies optimal tunnel designs based on user-defined targets (e.g., test section size, turbulence intensity, pressure loss) and validates them via numerical simulation. Moreover, the approach significantly reduces the need for full Navier–Stokes simulations by using only 26 numerical runs to solve four-variable constrained multi-objective functions. This enhances computational efficiency while enabling optimization across a wide range of design targets using a single set of simulations.