By using physical models and continuously updated sensor data, digital-twin can map industrial physical space entities to cyber space model system to realize the whole life cycle simulation and evaluation of complex industrial entities, which is an important key technology to promote industry 4.0. The management of complex heterogeneous models of digital-twin systems in an intelligent factory is facing severe challenges. This paper presents a multi-level heterogeneous model data framework for intelligent factory digital-twin systems. The multi-level integration framework is established for all levels of unit equipment, production lines, workshops and factories, as well as cross-domain product design, manufacturing, operation and maintenance. The paper presents an industrial model data management framework for multi-level digital-twin systems, and designs an data interoperability mechanism for cross-domain heterogeneous models based on knowledge ontology semantic networks. The proposed framework can provide an important theoretical framework for the management of complex model systems of digital-twin systems in intelligent factory.