STATISTICAL-ANALYSIS OF THE POWER INJECTION METHOD

The power injection method (PIM) has become widely accepted as a valua ble technique to predict the statistical energy analysis (SEA) paramet ers of a structure in situ. PIM requires the measurement of the power input into every single subsystem and the measurement of the energy le vel of every subsystem. By inverting the appropriate measured energy m atrix, the SEA parameters can be obtained. Conventional PIM suggests t o use a more or less predefined number of excitation and response loca tions. Most often three excitation and five response locations are tak en. Due to the spatial variation of the energy density in a subsystem, there will always be a substantial variability associated with the sp ace-averaged energy of vibration and hence associated with the SEA par ameters. An accurate SEA model requires accurately measured space aver aged energies of vibration and power input levels. The accuracy of the SEA parameters can be assessed by evaluating the confidence levels of the so-called ''normalized energy levels'' and by converting these va lues into confidence levels of the SEA parameters. This paper aims at presenting the identification procedure of the statistics of the SEA p arameters as a function of the number of excitation and response point s taken into account during the PIM measurements. In the first place, emphasis will be given to the presentation of the equations dealing wi th the essential statistical properties of the SEA parameters. It will be shown that by applying straightforward formulas it is possible to assess the accuracy of the estimated SEA parameters and to assess the accuracy of the SEA applications, i.e., vibration prediction, source l ocalization, power flow analysis, sensitivity analysis, etc. Some form ulas which are given in this respect will be illustrated by means of S EA results on an irregular box-type structure. (C) 1996 Acoustical Soc iety of America.