The Research Of Simulink On Friction Compensation By Adapative Neural Network

Friction is a complex, non-linear and uncertain natural phenomena which will lead to bad performance of mechanical system. So friction compensation becomes a widely used strategy in control. Neural networks (NNs) can compensate for the non-linear uncertain effect in friction phenomena by its universal approximation ability. This thsis aims to use adaptive NNs as compensator in different engineering system for different friction model.At the beginning of this thesis, there is a detailed introduction to friction model. Then adaptive NNs (Neural Networks) are introduced for their structure and approximation ability. The models are presented and control schemes are designed for turn-table servo system, 2-DOF manipulator system and hydraulic pressure servo system respectively and simulateon are conducted at last.In the study of turn-table servo system, the model and structure of simulator are introduced at first. Then the controller is designed by PD controller and wavelet NNs that is proved a steady compensator for friction. Finally, the simulation in which there is static friction is achieved.In the study of 2-DOF manipulator system, the model and structure of manipulator are introduced firstly. Then the controller is designed by PD controller and extended multilayer NNs that is proved a steady compensator for friction. Finally, the simulation is achieved in which there is dynamic friction. The model and structure of hydraulic pressure servo system are introduced firstly in the study of hydraulic pressure servo system. Then PD controller and extended multilayer are composed to system controller. Finally, the simulation is achieved in which there is static friction.All results of simulation that are achieved are proved the availability of designed controller.