Numerical and experimental validation of variance prediction in the statistical energy analysis of built-up systems

In the statistical energy analysis (SEA) approach to vibration modeling, a complex system is represented as an assembly of coupled subsystems, and the method leads to the prediction of the vibrational energy level of each subsystem. The averaging procedures implicit in the technique imply that the predicted energy is the mean value taken over an ensemble of random structures, such as a set of vehicles leaving a production line. Recently, a new method has been developed to allow the ensemble variance, in addition to the mean, to be predicted within the context of SEA, and the present paper concerns further extension and validation of this work. The theoretical extension concerns the variance of the energy density at a single point in any of the subsystems, and the validation includes both simulation and experimental studies. The simulation results concern plate assemblies, while experimental results are presented both for a single-plate and for a cylinder-plate structure. In each case an ensemble of random structures has been generated by adding small point masses at random locations on the structure. In general, good agreement between the predictions and the validation results is observed.