Some method are proposed f o r m rula acquisition and J%ZTY system parameters tuning baed on j k y neural networks. The feasibility of the proposed methoak is tested with an experiment of automatic train operation simulation. Thk experiment is also used to compare the Ieaming and control of-inference system with those of standard BP network and basic-system. A summrrry is made of the characteristics of the method. The final result indicates that the method of-rules generation and j k z y system tuning are very d a t i v e .Lu Jun, Hua Keqiang, Auto-generation of fuzzy control rule bases. Proc. of the 1997 IEEE Internat Conf. on Intelligent Wu Yan, Shi Hongbao, A learning method of guaranteeing the completeness of membership functions and consistence of fuzzy sets,