A method for identification of aircraft 6DOF dynamic model based on neural network is presented. In modeling via neural networks، there is not too much need to prior knowledge about system and identification procedure can be applied just by using a recorded set of system inputs and outputs. Thus this method is an appropriate technique for building aircraft model because it’s often very hard to exactly determine the value of aircraft mass، inertial moments، stability and control derivatives and so on. To train networks، data obtained from aircraft flight is used. After training، these networks can replace dynamic equations of motion and in this way remove uncertainties and simplifications of analytical expressions. Various tests on model confirm good performance and generalization of it.