The Brazilian Electrical Energy supply is almost entirely based on the energy produced by the Hydroeletric Power Station Machines. These Hydroeletric Machines own particular behavior in comparison to the turbogenerator behavior. This work investigates the use of Knowledge based system Hydroeletric Machines fault diagnosis. The system was modeled using information obtained by: the maintenance s systematic executed Hydroeletric Power Stations, though Furnas (Minas Gerais) maintenance team interviews; the Machine electromechanical behavior; and a Case Based study. The system is able to identify, from machine located sensors data analysis, eventual faults, allowing the execution of programmed operational interrupts in the machine in a optimized manner. A computational prototype and, mainly, the interface explain engine in addition to the knowledge modification through acquisition, are the innovative contributions of this work. The machine fault diagnosis problem domain is described, identified in the information, collected from the maintenance team and the electromechanical behavior experts, analysis. It is also described the projected symbolic model, the domain representation, using graphical and friendly interface, aiming its practical implementation in real Power Stations. It is shown a sensor information detach solution, through a Fuzzy Logic based module which converts the numerical data in a symbolic one, known by the diagnosis system, allowing its use, without any modification, in a sort of different machines. Finally, it is shown the test methodology adopted for the prototype validation through oscillation data simulation, which relationship with machine faults is partially known, and the symbolic model praticality and feasibility in the Hidrogenerator Diagnosis solution. Through the thesis development, it was verified that the Hydrogenerator fault knowledge wasn t still consolidated. So, the Knowledge Based system with knowledge acquisition became an excelent modeling tool for the domain experts.