Font Size: a A A

Adaptive Iterative Learning Control Of Nonlinear Parametric Systems

Posted on:2012-05-03Degree:MasterType:Thesis
Country:ChinaCandidate:J LiFull Text:PDF
GTID:2218330371962300Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Iterative Learning Control (ILC) can be used to correct control input by I / O data measured in the repeat process to achieve perfect tracking on a limited time span without precisely known dynamics of controlled systems. Now, ILC has become an important field of intelligent control direction and is widely used in industrial processes.The traditional method of ILC because of its fixed form of learning control law can not well response to changes in the systems and external interference. Although the learning control law of adaptive iterative learning control has the advantages of self-learning, Most of them need to know linearization structure of the nonlinear parametes. In other words, we have to know the model information of systems. That's defferent from iterative learning control methods that essentially belongs to method of model-free.(1) Considering a class of nonlinear parametric uncertain systems with known structures, we present an adaptive iterative learning control scheme with dead zone, which does not require the same initial conditions and the same reference trajectory in the controller design and analysis. We also give the convergence analysis of systems and theoretically prove the feasibility of the scheme. The effectiveness of the scheme is shown in the simulation results. It makes the adaptive iterative learning control scheme of a wider range of applications.(2) Considering the nonlinear parametric systems, we proposed adaptive ILC method based multiple model switching according to the characteristic of neural network that iteratively approximate the nonlinear part of systems. Using this scheme we can implement non-linear compensation, tracking the desired trajectory more quickly and accurately, ensuring the boundedness of the signal and making the systems of better stability and performance by setting the switching rules.(3) When considering a class of nonlinear discret time systems where both the structure and the order are unknown, a novel model-free adaptive iterative learning control based BP neural network is developed. The scheme only require I / O data of controlled systems. It doesn't require system modeling and also doesn't need to know any information of system model when we analysis and design the controller. It can achieve the perfect tracking task in a limited time interval only depend on the systems input and output data. The method is to solve the problem that using neural networks to search for a best "Mimetic Pseudo Partial Derivative", so it has virtue of the fasting and real-time traits. Simulation results show that this control scheme can achieve online control to the unknown systems, and have a better ability of self-learning and adapting to changes occurred in control environment.
Keywords/Search Tags:Iterative learning control, adaptive control, Model-free adaptive control, Time-varying neural network, Dead zone design, I/O data
PDF Full Text Request
Related items