Application of DOE to estimate the variability of SEA solution

Statistical Energy Analysis (SEA) is the most acknowledged method to predict the averaged sound and vibration levels in mechanical systems in the high frequency range. A limit of this analysis is that of providing only the mean value of the variables of interest. The mean value provided by SEA equations is the mean of the responses of a set of similar systems, averaged on frequency bands. Two systems are considered similar if their physical parameters are slightly different. No information on the standard deviation is obtained by SEA as it would be expected by a true statistical approach. In this paper, the variability of SEA parameters (coupling loss factors, internal loss factor and injected powers) to uncertainties in the physical properties of the considered mechanical system (Young modulus, material density, geometry, ...) is investigated using a Design of Experiment (DoE) approach. This is done in order to take into account the idea of similar systems. Subsequently, the variability of SEA solution to the uncertainties on SEA parameters found at the previous step is investigated by using again a DoE approach.