Different control systems parametric identification using neural networks

Abstract EN

-The information is to used for system identification which involves the determination of model structure, parameter estimation and also signal estimation of non-measurable signals such as noise signals in stochastic systems.

The determination of the model structure requires a construction of a mathematical representation of the system relevant to the problem at hand.

Identification is classified into parametric and non parametric procedures.

Parametric identification is used to identify the system parameters.

On the other hand, non parametric identification method is used to map the input-output properties of the system.

In this work different types of neural networks architectures, namely, feed forward neural network and Hopfield network are used as parametric identifier for both continuous and discrete control systems with different orders and Types assisted by an off-line computer simulation using C++ language.

Hopfield Neural Network is used as a parametric identifier with assumption the availability of system states for measurement then, identification is performed based on input / output measurement only.

Simulation results show the capability of these types of NNs to identify the unknown parameters of the plants.