Aging aircraft and combat aircraft that carry heavy
externals stores potentially face problems arising from nonliniarities in the
structure. The limit cycle oscillations (LCO) are one of the most common
consequences of these structural nonlinearies. Predicting the amplitude and the
frequency of the LCO helps to control the effects of these undesirable
phenomena. In this paper we propose an expert data mining system (EDMS) capable
to predict the asymptotic behavior of a system witch exhibits cubic or bilinear
concentrated structural nonlinearities. In the prediction module we have
implemented two methods, one based on nonlinear time series models and another
based on the unscented Kalman filter. The results obtained with these methods
are compared and if they are similar this will be the final output of the EDMS.
The main advantage of this new approach is that we don't need to know the
system parameters. The input of the EDMS is only a set of transient data (e.g.
obtained from a flight test). The performances of the EDMS are illustrated for
a two-degree-of-freedom airfoil oscillating in pitch and plunge.