Feedforward neural networks have been extensively applied to modeling and control of nonlinear systems. It has been known that using only one NN model to approximate accurately a highly nonlinear plant within a large domain is very difficult, and the controller based on the model often fail when the operating point changes greatly. This paper proposes a nonlinear direct adaptive control strategy based on radial basis function (RBF) neural networks and multi-models. An online adaptive algorithm and several effective model switching methods are given. The adaptive control strategy based on a single NN model has been proved to be robust, reliable, efficient and simple. The strategy based on multi-model proposed in this work can trace an expected output accurately without oscillation within a large domain. The control strategy is also applied to a pH continuously stirred tank reactor and the simulation results demonstrate the advantages.
[1]
David S. Broomhead,et al.
Multivariable Functional Interpolation and Adaptive Networks
,
1988,
Complex Syst..
[2]
D. Broomhead,et al.
Radial Basis Functions, Multi-Variable Functional Interpolation and Adaptive Networks
,
1988
.
[3]
Chen Xiaohong,et al.
Adaptive control based on RBF networks
,
1996,
Proceedings of 35th IEEE Conference on Decision and Control.
[4]
Daniel Sbarbaro,et al.
Neural Networks for Nonlinear Internal Model Control
,
1991
.
[5]
Thomas J. McAvoy,et al.
Neural net based model predictive control
,
1991
.
[6]
Shang-Liang Chen,et al.
Orthogonal least squares learning algorithm for radial basis function networks
,
1991,
IEEE Trans. Neural Networks.