During fault diagnostics of mechanical systems, information fusion methods were widely used to improve the reliability. However, diagnostic information comes from all life cycle of a mechanical device and they are always heterogeneous. Whereas heterogeneous information is difficult to be interconnecting, intercommunicating and inter-operating, so how to represent heterogeneous information is a key problem in integrated diagnostics. In this paper, a novel ontology-based representation method of heterogeneous diagnostic information is presented in detail. Firstly it introduces basic principles of ontology-based modeling. Then three methods of ontology-based information modeling are studied respectively. In the end an example of ontology-based modeling for a diesel engine is given.