The use of simulation for a pedagogical purpose is very interesting. One of the essential repercussions of the construction of a pedagogical simulator is the perennisation of the ‘know’ and especially ‘know-how’ held by the human experts. This deals with an expert system, ` kernel of the simulator´, able to diagnose and detect faults, as well as to describe scenarios of maintenance. Our major concern is to take into account the diversity of the types of knowledge held by the human expert, because they tend to be procedural (functional) or declarative, founded on a confirmed theory or a simple experience lived by the expert. Unfortunately, the more the structures of knowledge are complex, the more it is difficult to choose one of them. This difficulty still increases with the next setting in œuvre. This is really related to the different interactions of this expert system in the architecture of our pedagogical simulator .