Energy sensitivity based method for identification of statistical energy analysis parameters

The Statistical Energy Analysis (SEA) has currently been widely recognized as a method of high frequency vibration analysis for aerospace structures. However, numerous difficulties may be encountered for SEA modeling of complex structure due to the theoretical assumptions and especially the requirement for proper estimates of the essential SEA parameters. The experimental SEA techniques can be used to significantly enhance the analytical SEA procedure. An acknowledged one of these techniques is the power injection method (PIM), which requires yet exciting every single subsystem sequentially and measuring the input power and energy levels of every subsystem completely. This paper aims at identifying SEA parameters from incomplete measured data. An energy-sensitivity- bascd SEA parameters adjusting scheme using the Levenberg-Marquardt algorithm is developed. The objective of the methodology is to minimize the residuals between measured and theoretical predicted energy levels. The improved parameters are determined iteratively by solving a nonlinear constrained optimization problem as the energy sensitivities with respect to the selected damping and coupling loss factors have been computed. In addition, a series of formulas of error sensitivity are derived from energy sensitivity to assess the accuracy of the SEA parameters with the variances of measured data and analytical SEA parameters which are not chosen for adjusting. The methodology is validated through simulated data of two numerical SEA models. Also, an experimental investigation of a two-plate structure is carried out. The SEA parameters are successfully identified from the test data in the case of one excitation subsystem and two measured ones, which indicates that the methodology has the potential to be implemented in a complex structure.