Neuronal Approach and the Use of KIK-NET Network to Generate Response Spectrum on the Surface

The determination of response spectrum of a seismic excitation recorded on a chosen site is necessary in the structural dynamic design. The response spectrum can be obtained directly from recorded seismic data but this operation is expensive. The aim of this work is to avoid this problem where the acceleration response spectrum at the free face of a classified site is generated from that of reference by using the method of Artificial Neural Networks (ANN). The recorded spectrum value and corresponding period represent the inputs of the RNA, while the output is only the spectrum response value on the free field. The seismic data and the sites of the KIK-NET accelerograph network are used for the training and the validation of the neuronal model. The promising results obtained by this approach open a new research orientation on the use of the ANN in the field of soil dynamics and signal treatment, thus allowing the enrichment of the paraseismic codes into force.