Application of Radial Basis Function Neural Networks in Modeling of Nonlinear Systems with Deadband

Presence of dead-band in engineering process decreases performance of the system. Modeling of systems with such nonlinear properties is a key factor in model-based control of this phenomenon to mitigate its effect which is a challenging task in conventional mathematic methods. Because of capability in learning, adaptation, and classification, neural networks which can approximate any nonlinear continuous function arbitrarily well on a compact set is a good choice in this regard. In this paper application of radial basis neural networks in such systems is investigated. The nonlinear static part of the system first can be decoupled from linear dynamic part and then is modeled using radial basis function (RBF) network. While the dynamic linear part of the system can be identified using linear models. Results show that RBF can capture well the key model of the systems with dead band.

Application of Radial Basis Function Neural Networks in Modeling of Nonlinear Systems with Deadband Keywords:

Radial basis function , Deadband , modeling , static , dynamic