Zastosowanie sieci neuronowej i filtru Kalmana do symulacji spektrów odpowiedzi budynków poddanych działaniu wymuszeń parasejsmicznych

Decoupled Extended Kalman Filter (DEKF) algorithm was used for the training of Feed-forward Layer Neural Network (FLNN). Data taken from [1] correspond to Displacement Response Spectra (DRS) computed on the base of vibration records measured on the ground level for paraseismic excitation (inputs to FLNN) and on the fourth floor of monitored buildings (outputs of FLNN). It was proved that the application of DEKF gives much more accurate predictions of DRS than standard NN discussed in [1].