Nowadays, system configuration helps to achieve mass customization and manage a wide variety of systems. System configuration is based on a model that gathers all relevant knowledge for a family of systems. This knowledge model can be difficult to formalize and keep up to date; indeed, the knowledge must be made explicit, and can come from different departments and experts within an organization. In addition, it must reflect the different variants and options of a family of systems. Therefore, we try to answer the following question: how to better formalize and structure knowledge for system configuration to define configuration models and use the benefits of specialization in terms of modeling? Thus, in our proposal, we introduce the elements required to formalize the knowledge and define configuration models. This formalization will be done in a structured way using the association of ontology and Constraint Satisfaction Problem (CSP). We use the abilities of ontology to model knowledge of the different artifacts, their characteristics, and composition, and the abilities of CSP to model the relations between artifacts. In our proposal, we also define artifacts at different levels of abstraction using specialization which allows experts to detail or refine the formalized knowledge. We illustrate our proposals on a simplified but realistic example of a bicycle.