Real-Time embedded systems must enforce strict timing constraints. In this context, achieving precise Worst Case Execution Time is a prerequisite to apply scheduling analysis and verify system viability. WCET analysis is usually a complex and time-consuming activity. It becomes increasingly complex when one also considers code generation strategies from high-level models. In this paper, we present an experiment made on the coupling of the WCET analysis tool Bound-T and our AADL to code generator OCARINA. We list the different steps to successfully apply WCET analysis directly from model, to limit user intervention.