This paper presents a modelling method based on the neural network technique. The main aim of this study is to establish a reference model of an industrial process behavior under the normal operating conditions. The use of this reference model should reflect the true behavior of the process in the whole way and thus distinguish a normal mode from the abnormal modes. This paper shows the choice and the performances of the neural network in the phases of the training and the validation. A study is related to the number of inputs, and of neurons used and their influence on the behavior of the neural predictor. In order to illustrate the ideas proposed concerning the dynamics modelling, a reactor-exchanger is used. The outlet temperature in the reactor-exchanger is modeled according to the inlet temperature by using the Nonlinear AutoRegressive Moving Average model with eXogenous variables (NARMAX).