The increasing complexity of communication protocols for asynchronous multiplex systems requires the use of simulation during the optimisation of these protocols or the integration of other control units. Consideration of realistic communication behaviour of the connected control units is essential for performance analysis of multiplex systems.For a first pass, the use of simple statistical distributions (e.g. Poisson distribution) is suitable to get some simulation results. A better way to get realistic results is the approximation of empirical communication data through the use of more complex statistical distribution (e.g. mixed Erlang distributions).In this paper several approaches for the approximation of empirical data are presented. Beside simple statistical distributions (with one parameter), the use of more complex statistical distributions is discussed and methods for the identification of their parameters are presented. The methods are compared with regard to their efficiency and their quality of approximation.Finally, the approaches are applied to log-files of the communication traffic recorded during test drives to get analytical models of the communication behaviour of the local control units. The results of the methods are presented and their integration into a simulation environment is discussed.