Identification of dynamic object using Z-transform artificial neural network

Abstract The aim of the paper is to present the method of identification of dynamic discreet objects using neural networks with Z-transform. The process of network learning consists of determining approximation of object Z-transform, with increasing precision in each iteration, based on all the previous input and output signals. This method of identification may be used also in non-stationary objects. It allows to modify the Z-transform on line. The proposed method of identification is based on iterative modification of the transmittance of a neural network with Z-transform. The modification is calculated after every step which provides new coefficients. The exact transmittance of the object is defined after evaluation of the last coefficient which has the longest delay time. In the beginning of the article, theoretical solutions describing the identification method have been described. Next part shows the application of those solutions on four examples. The last part presents the results and proposed future development. The article also includes an appendix with Z-transform basics.