Research On Fuzzy-Neuro Model-free Control For Systems With Uncertainties And Applications

The fuzzy-neuro model-free control is one of the important areas in intelligent control. The system designs and applications of the fuzzy-neuro model-free control are studied in this thesis. Several fuzzy-neuro model-free control methods are proposed for complicated plants with uncertainties. The simulation tests are made and the results demonstrate the efficiency and advantage of the proposed methods. The main content of this thesis includes the following:1. To the level control problem of a spherical tank, two model-free control methods are proposed. In the former method, the Takagi-Sugeno fuzzy model is used to tune the neuron controller gain. In the latter method, the model-free control method using the neural network model proposed for nonlinear plants is presented. Simulation results show that both of them have satisfactory performance and strong robustness.2. To pH processes, which are nonlinear and time varying, the neural network model is structured and the learning algorithm is presented, based on which the model-free controller is designed, while the controller gain is scheduled by a fuzzy method. The simulation results demonstrate its efficiency and strong robustness.3. The model-free PID controllers with fuzzy-neuron and neuron-fuzzy gain scheduling are separately proposed for turning process. Simulation results demonstrate its satisfactory performance and strong robustness.4. The double-layered model-free control method is proposed for the independent joint of the direct drive robot whose dynamics is highly uncertain. The inner loop of the PID model-free control is used to overcome the uncertainty of the plant and the outer loop of the variable structure neuron is used to improve the system performance. The efficiency of the method is verified by simulation tests.