SEA (Statistical Energy Analysis) permits prediction of energy sharing between subsystems of an excited structure. SEA is theoretically easy to apply and permits parametric variations on a model of the structure during conception phase. This method is usefully used in building works but pains to be established in transportation sector. Some difficulties appear in that case and prevent a good modelling of the structure. One critical step is to define proper SEA elements. This partitioning, almost obvious in building works, becomes more complex in the case of car body for example. Evaluation of power injected by sources into the structure is a second difficulty. Indeed, subsystems energies are directly proportional to injected power. Thus, wrong evaluation of injected power leads to wrong description of energy sharing. The first part of this work deals with a model of input power injected into a rectangular plate excited by turbulent boundary layer (TLB). This simple model permits to underline some parameters influences on input power and quickly estimate it, averaged on frequency bands. An experimental validation is also exposed. In the second part, an automatic method for the identification of proper SEA subsystems based on analysis and classification of energy transfer functions simulated by FEM is described. A classical algorithm divides database into N subsystems (N=2 to Nmax). An index, validated on academic structures, indicates optimal sub-structuring for SEA application. Industrial structures, showing applicability of the method to complex systems, are also presented.
