Abstract : While a system is in operation, one of the aims of the process safety analysis is to guaranty that the physical variables of the system follow a certain trajectory and stay in a given zone considering all the perturbations and the uncertainty on the real values of the parameters of the model and the measurements. An original solution for this problem which has been proposed in this document is to consider the environmental perturbations and the ignorance of the exact values of the parameters of the model as uncertainty on the parameters of the model, characterize the support of the parameters of the model by intervals and then to use the interval arithmetic to calculate all the possible trajectories of the system. In the other hand, to represent the behaviour of a complex system (a multi-mode system or a non-linear system), instead of using a complex single-mode model, a classical approach is to split the operating regime of the system into local zones and to represent every zone by a linear model which has a simple structure. In this work, we try to propose a solution to the identification problem of uncertain dynamic systems in the first time by a single-mode model and then by a multi-mode model in the process safety analysis context. This document consists of five chapters. In the first chapter, the objective of this work and the problem to be solved is explained. In the second chapter, the study of the literature of the problem is given. We then deal with the problem of the identification and the simulation of interval model in chapter three. To improve modelling quality and to increase the prediction precision, a methodological approach is proposed in chapter four to identify and to simulate multi-mode interval models. In the last chapter, the conclusion of this work is given.