On The Design And Simulation Of Neuron Self-tuning Pid Controller For Networked Control Systems

Networked Control System (NCS) is an integration of network communication technology and cybernetic theory. An NCS connects the controllers, the sensors and actuators through one shared network to form feedback control loops. Recent years, as the rapid growth in the field of industrial automatics, navigation technology and aerospace science, the study of NCS has attracted more and more attentions of the researchers. Uncertain delays are inserted into NCS loops by the shared network, and thus it is hard to apply traditional analytical methodology to an NCS directly. So it is rewarding to study the NCS academically and practically. This thesis does some research works on analysis of the NCS performance, the system's property analysis, the simulation of NCS, design and improvement of controller. The main research work is as follows:First, the background, concept and basic problems of NCS are introduced and a survey of NCS research work is given, then the main work of the thesis is proposed.Second, a Matalb/Simulink simulation platform is designed and realized. The effect of the two driven means of NCS has been researched. The result of the simulation shows that with the existence of the random time-delay, the event-driven type controller will behave a better performance as long as the delay is in some scope. However, when the time delay is much bigger than the sample period, the time-driven type controller will behave much better performance.Third, a single neuron self-tuning controller based on the control theory is designed. The simulation result shows that with the existence of time-delay, disturbance, the increase of package lost rate and the increase of the bandwidth of the competition node, the newly designed controller possesses better stability and robustlity, compared to the conventional PID controller.Finally, a neuron self-tuning PID controller based on genetic algorithm (GA) is designed to make up for the deficiency of traditional neuron self-tuning PID controller. The conventional GA algorithm is improved in the number of the population, the possibility of the crossover and the rule of the termination for the practical control objective. The simulation demonstrates the effectiveness of the improved algorithm.