Modular plants provide the opportunity to increase flexibility and reduce time-to-market in the process industry. Process validation is time-consuming and limits the realization of this potential. Therefore, it should be supported efficiently. Smart process equipment assemblies (sPEAs) built from the real module, a digital twin (DT), and suitable methods and algorithms provide high potential to do so. In this paper, we investigate how a suitable DT should look like to support the process design phase in process validation. Therefore, the semantics and information demand of different relevant simulation and optimization problems are analyzed. We reason that the DT should combine structural information like engineering data, e.g. through DEXPI, and behavioral models in form of simulation models. Furthermore, it should provide the descriptive capability to capture the semantics of different application cases like e.g. design of experiments. We suggest a linked-databased architecture to meet these requirements. The simulation models are semantically lifted into linked data through an information model describing its purpose, quality and variables. The approach provides the potential to reduce manual effort of the user since information is interconnected, accessible and processible automatically.