Abstract. Simulation models are important parts of industrial system analysis and control system design. Despite the wide range of possible usage, their formalization, integration and design have not been satisfactorily solved. This paper contributes to the design phase of simulation models for large-scale industrial systems, consisting of a large number of heterogeneous components. Nowadays, it is the simulation expert who assembles the simulation model for a particular industrial plant manually, which is time-consuming and error-prone. We propose to use a semi-automated semantic engine that assembles the simulation model. We represent a structure of a real industrial plant in a plant ontology and available simulation blocks in a simulation ontology. Signals of each simulation block are specified via signal ontology. As the knowledge is formalized, the simulation model can be assembled automatically, based on the ontologies and SPARQL queries. Since each real plant device can be represented by more than one simulation blocks, the selection of suitable simulation candidates is based on matching interfaces of neighboring blocks. In the presented case, simulation models can be efficiently redesigned and their components can be reused and shared; the methodology contributes to avoiding errors in both run-time and design phases. Evaluation on a real-life industrial use-case, dealing with design of simulation models of passive houses shows improvements in both reducing development time and avoiding design errors. Major results of this paper are the proposed structures of the ontologies and the SPARQL query realizing the selection of the appropriate simulation blocks.