Abstract: A genetic algorithm combined with a neural network technique is proposed to search for turbomachinery diffuser blade profile under the condition of a given velocity distribution on its surface.Such a new inverse method works through minimizing the error between surface velocity distribution of candidate blades and the target velocity distribution.For ease in employing genetic algorithm,the blade profiles searched are parameterized with Bezier spline curves.The flow analysis of the candidate blades is carried out by BP neural networks which have been trained with a set of samples.The proposed approach is demonstrated through the application to a diffuser with two-dimensional height-and thickness-constant blades.The simulations show that the new method is not only feasible,but also reliable and effective.