Vibroacoustic model’s likelihood computation based on a statistical reduction of random FRF matrices 2019-01-1593
The likelihood appears as the natural tool to perform such comparisons as soon as the probability of a given result may be estimated. Vibroacoustic analysis mainly relies on complex matrix-valued
Frequency Response Functions that can be easily measured and computed. The likelihood of such complex and frequency dependent matrices is investigated.
A two stage statistical reduction, based on Independant Components Analysis, is proposed in order to separate statisticaly independent components with random complex amplitudes. Their probability may be computed independently from one to another one.
The joint probability density fonction of the real part and of the imaginary part of each independent complex-valued random variables is estimated using a nonparametric stochastic model of model uncertainties implemented in MSC/NASTRAN and the Monte Carlo simulation method as a stochastic solver.
The proposed statistical reduction presents many interesting properties including data reduction or related to the physical behavior of the studied system.
Laurent Gagliardini, Christian Soize, Justin Reyes
PSA Group, Université Paris-Est, Groupe PSA
Noise and Vibration Conference & Exhibition