Neural networks are applied to the identification of non-linear structural
dynamic systems. Two complementary problems inspired from customer surveys
are successively considered. Each of them calls for a different neural appr
oach. First, the mass of the system is identified based on acceleration rec
ordings. Statistical experiments are carried out to simultaneously characte
rize optimal pre-processing of the accelerations and optimal neural network
models. It is found that key features for mass identification are the four
th statistical moment and the normalized power spectral density of the acce
leration. Second, two architectures of recurrent neural networks, an autore
gressive and a state-space model, are derived and tested for dynamic simula
tions, showing higher robustness of the autoregressive form. Discussion is
first based on a non-linear two-degree-of-freedom problem. Neural identific
ation is then used to calculate the load from seven acceleration measuremen
ts on a car. Eighty three per cent of network estimations show below 5% err
or. (C) 2001 Academic Press.