Abstract
Keywords: Shape optimization, crosswind stability, response surface, bus aerodynamics
The paper presents optimization of a double-deck bus with the aim to improve its aerodynamic stability in crosswind flows. An optimization algorithm based on the usage of either polynomial or radial basis neural network-based response surfaces was used for identification of the best bus design. The data required for construction of the response surface models was obtained with steady and transient Reynolds-averaged Navier-Stokes simulation. The transient simulations together with response surfaces based on radial basis neural networks were found to produce most reliable results for the optimization of crosswind stability. The bus design with largest radius of the front end vertical edges and medium angle of the windscreen was found to be optimal for crosswind stability at yawing angle of 300 in agreement with the previous experimental investigations.