Comparison of Local and Global Constrained Aerodynamic Shape Optimization
Daniel J. Poole, Christian B Allen, Thomas Rendall · 32nd AIAA Applied Aerodynamics Conference · 2014
A comparison is presented of gradient-based and agent-based optimization schemes, applied to constrained aerodynamic shape optimization in two dimensions. The optimizers developed are a feasible sequential quadratic programming (FSQP) gradient-based algorithm and a modified gravitational search algorithm (GSA), extended to include a new separation-sub-swarm (3S) constraint handling method. Lift constrained drag minimization of the NACA 0012 and RAE 2822 aerofoils at different flow conditions are presented, and results show that the constraint handling global search algorithm is able to locate better optima than the gradient-based scheme, although at the cost of more objective function evaluations. The design variables used here are derived using a mathematical decomposition, which has produced a set of orthogonal design variables that are very efficient, and as few as six are required to obtain shock-free solutions using the global search algorithm.