Studies in Informatics and Control, 26(2) 239-248, June 2017 ISSN: 1220-1766 eISSN: 1841-429X https://doi.org/10.24846/v26i2y201712 ICI Bucharest © Copyright 2012-2017. All rights reserved http://www.sic.ici.ro 239 1. Introduction Fault diagnosis in dynamic systems has been the subject of several research works. It is defined as a process of detecting any deviation or intended behaviour of a system and isolating the cause of this failure. There are many approach diagnosis and technical processes [1, 5, 10, 16, 17, 18, 23, 24, 29] that have been applied to fault diagnosis of dynamic systems. Their goals are focused on a timely diagnosis of defects in order to provide a precise judgment rule for distribution operators. Particularly when serious defects occur in a system, a lot of alarm information is transmitted to the control system. In such cases operators must quickly and accurately assess the causes, location and defect system components. Diagnosing faults with the right methods can provide accurate and effective diagnostic information for ship operators and help them take adequate action in certain faulty situations to ensure safety. In the context of diagnosis more model-based systems have been used: finite state machine [17, 8], Petri nets [9, 31, 34, 35, 38]. Despite the various researches in recent years and the proposed modeling approaches for the fault diagnosis, several unresolved problems have remained. The performances of the diagnostic approaches depend on the means of the model being used. Obtaining and using the model to construct a diagnosis system is a complex and difficult task more particularly for the uncertain systems because of unforeseen and uncontrollable events that characterize them. This paper focuses on fault diagnosis of uncertain discrete-event systems and tries to use Interval Fuzzy Constrained Petri Nets (IFCPN) as a tool of modelling, identification and isolation. The proposed tool IFCPN model is introduced in order to extend some properties of the Interval Constraint Petri Nets [13, 14] which is considered as an extension the P-temporal Petri Net [26] and a sub-class of High Level PN [21, 25]. Our main contribution is to extend the functional range of ICPN applications to fault diagnosis of uncertain systems where the validity intervals of any parameter are fuzzy and characterized by the propagation of uncertain events [4, 15, 30]. This paper is organized as follows: In the first part of this work, we studied the modelling of a mixing system: we propose an approach using Statistical Process Control in order to build the validity ranges of fuzzy intervals and we compute robust control laws of this model. In the second part we study the robustness of a system which is defined as its ability to maintain the concentration properties specified on the occurrence of disruptions. After having proved this robustness we use the tool built for the diagnosis of the defaults. The diagnosis method is based on fuzzy logic [37]. Fault Diagnosis of Uncertain Systems Based on Interval Fuzzy PETRI Net Fatma LAJMI, Achraf Jabeur TALMOUDI, Hedi DHOUIBI LARATSI, National Engineering School of Monastir 5000 Street Ibn El Jazzar – Monastir 5035, , Tunisia flajmi@yahoo.fr; achraf_telmoudi@yahoo.fr, hedi.dhouibi@laposte.net Abstract: The purpose of the following article is a novel approach to modeling, diagnosis and control of discrete event systems. This approach uses the Petri net Fuzzy Interval (IFPN) to describe the diagnosis system. The Petri network is used to model the system which needs to be controlled. This work is a description of a case study applied to an ingredient mixing system where the concentration of ingredients has to be held within a valid range. Keywords: Uncertain systems, Intervals Fuzzy Petri Nets, Modelling, Fault Diagnosis.