This paper deals with the application of adaptive fading Kalman filtering (AFKF) to the design of an observer used to predict the state of an counter-current heat exchanger. In order to implement the AFKF, an accurate mathematical model, based on the approximation of temperatures along the process by first order dynamic model, is considered. This method could be applied when the heat exchanger is solicited by the flow rate changes. Experimental validation of this method is illustrated by an example. (C) 2003 Elsevier Ltd. All rights reserved.