With regard to the characteristics of flight data, a new identification method of flight data model based on dynamic fuzzy neural network is presented. By on-line learning, the proposed DFNN is learned for a compact network with better generalization ability. The network structure is learned by means of adding or pruning a new neuron, furthermore, the linear parameters as network weights are gained based on the recursive least squares algorithm. Through a great number of observations in a certain sortie, the DFNN method is applied to the identification of the association model of flight data. The test results show that the method is of faster constringency and better generalization.