Fuzzy Control Rules To Automatically Generate The Fuzzy Control System Development Environment

With the rapid development of fuzzy logic and fuzzy control, the extraction of fuzzy rules has become an important step for the application of fuzzy technology. This thesis focus on the automatic generation and tuning of fuzzy control rules.The problem confronted is how to generate a Mamdani Fuzzy Controller in learning processing. The subsequence of rule is fuzzy language variable. In addition, the generating method should work on-line.In order to satisfy the requirements of application, a method, named Model Reference Fuzzy Neural Network Learning Control Method, is presented in this thesis to generate fuzzy control rules automatically. And it can dynamically modify fuzzy rules while the system is running. The learning system is composed of three components: Reference Model, Fuzzy Inverse Model, and Fuzzy Neural Network.In this system, Reference Model gives an ideal response. After comparing the Reference Model's output and the system's actual output, Fuzzy Inverse Model produces necessary modifications of control value, which then feedback to the Fuzzy Neural Network as the teacher signal. The Fuzzy Neural Network is tuned by the combining of Self-Organization algorithm and BP algorithm.Based on this method, some simulation experiments have been finished. And the results show that the theory is correct.In this thesis, Fuzzy Control Development System FDS2.0 is also described. The function of FDS2. 0 is to design the fuzzy control system, include constructing a fuzzy control system, designing Mamdani and T-S Fuzzy Controller, simulating the control system. In addition, FDS2. 0 can also compile the Fuzzy Controller to the code of F200, which is a kind of Fuzzy Control Chip.