Abstract: The construction of S_p (perfect slip ratio) identifier with back-propagation neural network was proposed to prevent skidding and have the best braking effect in the aircraft braking process. In order to improve the learning ability of the network, a type of self-adaptive learning rate method, second-order learning rate method, was introduced. Some problems in the practice of this method were discussed and the solutions were presented. A third-order learning rate method was deduced based on the method. The method of reasonable configuration of active functions was proposed. The combination of these methods renders better learning precision and speed.