[en] This work intends to establish a model of knowledge representation based on a expert system to supervise either security or operating to be applied generally on monitoring and detecting faults of industrial processes. The model structure proposed here let the system represent the knowledge related to faults on a process using a combination of rules either basic or associative. Besides, the model proposed has a mechanism of propagation of events in real time that acts on this structure making it possible to have an intelligent alarm processing. The rules used by the system define faults from the data acquired by instrumentation (basic rules), or from the establishment of a conjunction of faults already existent (associate rules). The computing implementation of the model defined in this work was developed in OPS5. It was applied on an example consisting of the shutdown of the Angra-I's power plant and was called FDAX (FDA Extended). For the simulated tests the FDAX was connected to the SICA (Integrated System of Angra-I Computers). It results save validity to the model, confirming thus its performance to real time applications. (author)