In this article, we consider the problem of a digital twin intended to be a decision aid and a forecasting tool for production management. This work is conducted in collaboration with an industrial company that produces mechanical components for the hydraulic and electrical sector. For the two objectives, decision aid and previsions, we propose an architecture for the digital twin based on a hybrid simulation optimization model. This architecture has been tested with industrial instances on a scheduling problem where optimal solutions cannot be found in a reasonable time.