Surrogate-model-based Active Drag Reduction of High Speed Trains

Rapidly climbing aerodynamic drag is one of the major obstacles for developing commercially viable high speed trains (HSTs) up to 400 km/h. This paper proposes an active flow control (AFC) technique based on surrogate model and machine learning algorithm for achieving enhanced HST aerodynamic performance. Five pairs of distributed blowing jets are deployed on the tail car of a HST model, each associated with its own blowing velocity and angle. The ten independent control parameters are optimized in terms of minimizing the drag of the HST model and meanwhile the control power input based on a non-dominate sorting genetic algorithm-II incorporated with an innovative modification of the Kriging model. The results show that the modified Kriging model not only results in a substantially accelerated optimization process, but also proves to be more effective and accurate in global optimization compared to the conventional Kriging model. For a scaled maglev train model which is highly streamlined, a remarkable drag reduction up to 7.3% and net energy savings of 4.6% are achieved at a speed of 108 km/h.