Detection and localization of nonlinear distortions

In this article a method to detect and localize nonlinear distortions in the dynamic behaviour of plate-like structures is introduced. The aim of the paper is not to determine a nonlinear model of the structure (as could be done with for instance Volterra kernels), but to use linearized FRFs with a different uncertainty level for both noise and nonlinear distortions. We provide an alternative for the so-called coherence function, which does not separate noise and nonlinear distortions and also requires a large measurement time. The introduced technique uses a scanning laser vibrometer to measure the structure with a high spatial resolution. By making use of a so-called random multisine (i.e. a sum of sines with equal amplitude and random phase) as excitation signal, the noise can be separated from the nonlinear distortions. To avoid the need for repeated measurements a spatial spline fit of the vibration patterns is used to estimate the noise and nonlinear distortion levels. The proposed technique will be validated on simulations and on scanning laser measurements of a car door.