AUTOMATIC ANALYSIS OF SEISMIC DATA BY USING NEURAL NETWORKS: APPLICATIONS TO ITALIAN VOLCANOES.

Abstract

The availability of the new computing techniques allows to perform advanced analysis in near real time, improving the seismological monitoring systems, which can extract more significant information from the raw data in a really short time. However, the correct identification of the events remains a critical aspect for the reliability of near real time automatic analysis. We approach this problem by using Neural Networks (NN) for discriminating among the seismic signals recorded in the Neapolitan volcanic area (Vesuvius, Phlegraean Fields). The proposed neural techniques have been also applied to other sets of seismic data recorded in Stromboli volcano. The obtained results are very encouraging, giving 100% of correct classification for some transient signals recorded at Vesuvius and allowing the clustering of the large dataset of VLP events recorded at Stromboli volcano.