The performances of controlled system are strongly linked to performances of the control architecture timed. To manage these performances, the control engineer has to evaluate them at each stage of the life cycle: from the project requirements to the detailed design and at the setup stage. However, the engineer expectations levels concerning the estimated performances are variables. Approximated results quickly obtained with rough data are enough at the beginning of the life cycle. But he needs reliable results at the detailed design stage even if the performances are longer and more difficult to obtain. Our approach takes into account theses various engineer needs. The proposed method evaluates timed performances of networked control architecture and guides the engineer throughout the architecture development. A modular design has been retained to model the dynamic behavior of the control architecture using Timed Colored Petri Nets. Three points of view are considered: The dynamic behavior model of functional architecture is obtained by modeling the control tasks and their interactions; The dynamic behavior model of hardware architecture is obtained by modeling the topology and the connections between hardware components (PLC, communication networks ...); The dynamic behavior model of operational architecture is obtained by modeling the mapping of the control tasks and data flows to the hardware components. Then the timed performances evaluation is obtained by simulation of the Petri net. To validate our approach, we treat a relevant example using the software platform Design/CPN. With this case study, we present different studies concerning the dynamic behavior models convergence, the simulation results sensitivity according to the parameters settings error and the simulation performances accuracy by confrontation with the performances measured on the real system.
