Two-Dimensional Airfoil Shape Optimization Using Highly Differentiable Splines and Evolution Strategies
An optimization method based on Covariance Matrix Adaptation Evolution Strategies (CMA-ES) is applied on a parametric design tool for the automated generation of two-dimensional turbomachinery airfoil sections. Highly differentiable curves are managed to ensure continuity in the slope of the curvature on the blade surface to avoid undesired anomalies in the Mach number distributions. An Euler solver coupled with an integral boundary layer method is employed to assess the aerodynamic behavior of the geometries. Special care has been made defining several cost functions to allow the algorithm handle unfeasible geometries that can appear during the evolutionary process. The fitness function of feasible individuals can be set up to fulfill several geometric and aerodynamic constraints. To show the potential of the method, several optimization problems have been solved, tracing existing geometries originally defined in a point wise fashion, and applying inverse design to match target Mach number distributions. This method can facilitate the two-dimensional airfoil design and can be used to import external data defined with a set of points. This optimization approach could be employed as well to generate an initial blading geometry which could feed a more sophisticated optimization method based on a three-dimensional CFD solver.