The fuzzy neural network control in the control domain has already become a hot topic at present. Applying the neural network in fuzzy systems may solve the fuzzy systems' knowledge extraction problem. Applying the fuzzy system in neural networks, the neural network is no longer a black box, and humanity's knowledge is very easy to fuse in the neural network. We propose a new fuzzy neural network structure and its learning algorithm based on the expansion radial direction base network. Its learning algorithm's most major characteristics are the parameter adjustment and the structure identification perform concurrently, and the study speed is high, which may be used in realtime modelling and controling. We develop the related algorithm routines, and finally carry out a simulation analysis as for real cases. The simulation result indicates that the dynamic fuzzy neural network is fast, compact,capable in generalization.