Fuzzy logic theory has provided a model-free tool to develop intelligent control system for complex industrial processes by means of simulating the fuzzy reasoning process of human being. However, the performance of such a control system depends on the knowledge base (control rules and membership functions of fuzzy sets). For the control of complex industrial process in which the dynamic parameters of process is time-varying and non-linear, it is necessary to modify and optimize the knowledge base on-line. Adaptive fuzzy control provides a efficient approach for this objective. In this paper, a new fuzzy neural network (FNN) and an adaptive learning mechanism based on genetic algorithm has been proposed for modeling the fuzzy reasoning process and constructing an efficient adaptive fuzzy control systems. Experiment results show that the FNN is capable of modeling complex functions and simulating fuzzy reasoning process of human being.