| In many industry processes and national defence areas, often, it is very difficult to ascertain model of the object controlled in advance, even if the model has been ascertained under some condition, its dynamic parameters therefore as for its structure often will vary while the working conditions of the object controlled vary, so that good control quality can not be gained by general controller. In these complicated processes, also there are time-delay phenomena widely, inhere time-delay of the object controlled bring quite difficulty for analyzing and synthesizing system. Furthermore, the property of time-varying delay make problem more complexity, so it is very important for theory and practice to study the problem.Single input single output system with time-varying delay is studied in this dissertation, their process dynamic property vary along with working conditions of system varying, in addition, the process dynamic property can be defined using a set of variables which also can be measured and controlled. We use modeling method in multi-model adaptive control, sliding model variable structure control strategy, modeling method using BP linear neural network and Smith predictor to implement adaptive control of the time-varying delay systems.While structure of a system is unknown and constant, BP linear neural network will be used to identify model of the system. BP linear neural network is a multiple layers feedforward neural network, and all neuron fire function of the network are linear functions. Thus the nonlinear mapping relation approximated by neural network not only can be obtained through learning, but also the mapping relation can be expressed by analytic equation. It is very important for computing control input using sliding model variable structure controller.While structure of a time-varying system is known, to fuzzy partition variables that can ascertain working space of the system, so generating...
|
PDF Full Text Request