基于神经元网络的预见控制器设计及其硬件在回路仿真-控制理论与控制工程专业论文.docxVIP

沈阳工业大学硕士学位论文The 沈阳工业大学硕士学位论文 The Design of Neural Network Based Preview Controller and the Simulation of Hardware-in—the-loop Abstract A neural network preview controller was designed，based the combination strategy of neural network and preview control theory，to improve the tracking accuracy for series of special plant whose trajectory is previewed，such as machine tools and manipulators．Simulation analyses and hardware-in-the—loop experiments show that the controller has the characteristics of self-learning，rapid response，strong robustness The preview controller tries utilize the future information of reference and disturbance effectively improve the control performance．Structurally，the preview system is derived from the normal seⅣo systems．which the preview feed—forward is added for compensation．Consequently，the stability is just the as usual．Furthermore，the following performance is much better just because the future information Was utilized besides the information In former methods，the compensating accuracy heavily depends the accuracy of mathematical model．While the neural network has lot of advantage that applicable automation．So，utilizing the neural network realize the function ofpmview controller will be propitious improve the following performance，self-learning ability and robustness． Above idea Was verified by MATLAB／SIMULINK，and the results show that the neural network preview controller possessed strong self-leaming ability and compensated the dynamic delay，improved the following performance and response speed greatly Considering the difference between simulation and reality，a real—time hardware·in-the- loop experiment was carried out，i．e．the proposed strategy was verified by connecting the prototype controller the physical plant directly．In this experiment，A single computer was used host and target control the Velocity of AC motor in the environment of Real—Time WindOWS Target． Key Words：Neural Network，Preview Control，Feed-Forward Compensation， Hardw