Abstract: To solve the problem that the scale of complex product domain ontology is too large to be applied efficiently, a method of domain sub-ontology extraction for complex product knowledge management was proposed. Initial entities in domain ontology were identified by end-users. Relevancy between entities was employed to replace the similarity. Relevancy matrix was built and updated by expanding the relevancies iteratively. Expansion type domain sub-ontology could be acquired by restoring the relations of entities. Entities were expanded with heuristic search at the dimensions of structured relations and restrictive relations in ontology, and the relevancies between entities were attenuated by expansion factors. Maximum expansionism and harmonic expansionism were presented to update relevancy matrix. Finally, robotics domain was taken as the example. Domain ontology was built to evaluate the complexities of domain sub-ontologies, the scales and complexities of domain sub-ontologies were showed at different parameter conditions.