Semantic equations are important pieces of the denotational semantics approach that enable the translation of syntactic constructs of models to semantic domains. In a previous work, we have extended the Model-Driven Architecture (MDA) four-layer proposal of architecture in order to incorporate formal semantics to its artifacts. Thus, semantic equations are key elements in this extended architecture. In this paper, we propose an approach to incorporate semantic equations described by model transformations in the MDA infrastructure. As a result, we automatically generate semantic models from the syntactic models involved in the MDA transformations. Therefore, the semantic equations play an important role in the MDA architecture since they allow the construction of tools for automated reasoning for several purposes, such as: (i) verification of semantics preservation in model transformations; (ii) choice of the best format to represent meaning and behavior of models and (iii) automatic generation and production of quality compilers. For instance, we claim that in the scenario of model transformations involving concurrent models described using Petri nets, semantic concepts can be more easily and appropriately handled in the MDA architecture. We have evaluated our approach by applying semantic equations to Petri nets models in a project for embedded systems codesign.