Since 2001, we are involved in the development of a system named SABRE [1, 2, 3, 4] for supporting the training of military students in the French Army; we expressed the knowledge of the environment and concepts of military work as ontologies using Protégé. We consider that the ontology contains explicit descriptions of the concepts (represented by words: e.g. " to serve one's homeland ") used in our military domain. The knowledge is also at stake in the military training and then must be understandable by the students. Finally, not only our ontology must produce usable data structures, but it also must be accepted by the French Army instructors. The SABRE system must consequently possess the necessary data structures so that this knowledge can be edited, filed, recalled, checked with that of the students', stored for future reference, and be used for a training session. However, building ontologies and designing the data structures for such a knowledge-based system are not easy tasks; we recently found that some set-theory-based characteristics extracted from the " under work " ontology could considerably help the ontology builder. This paper describes our experience on building the ontology of the French Military training using Protégé and how this step-by-step process could be supported and guided on-the-fly by systematically checking the theoretical properties of the instance sets of the ontology. During the design of an ontology for military training [1, 2], we partially solved the ambiguity (getting different interpretations for identical values of different slots), completeness (adding interpretations for all values) and minimality (avoiding intractability) problems with extra design and rules [4]. We here present several (polynomial-time) functions which can automatically detect where to add relevant interpretation rules or complete our ontology (adding classes and slots) for assisting the ontology design.