Dynamic Recurrent Neural Networks for Identification
of a Multivariable Nonlinear Evaporation System

F. Garces, C.Kambhampati and K. Warwick


Abstract
********

This paper proposes an indentification technique for a nonlinear
evaporator system.  Extending an early result to multi-input multi-
output (MIMO) systems [6], the technique approaches a nonlinear
multivariable identification based on a dynamic recurrent neural
network (DRNN).  The method results on a state space nonlinear
model of the evaporator system identified.  The network and weight-
updating algorithm proposed make the method remarkably easy to
implement.