PREDICTION OF BLAST INDUCED GROUND VIBRATIONS BY USING ARTIFICIAL NEURAL NETWORKS

Abstract

In this study, artificial neural networks (ANN) were used as a tool for estimation of blast-induced vibrations. For this purpose, the blast shots carried out in a quarry in Istanbul were monitored and the blast-induced vibrations were recorded. Peak Particle Velocities (PPV) and Scaled Distances (SD) of 24 events were recorded in the first 12 shots, subjected to statistical analysis and the site-specific ground vibration propagation equation was obtained. This data set was also used to train an ANN model while SD was an input and PPV was an output; and a new model, that used to estimate blast-induced vibrations in the related field, was developed. Using the vibration propagation equation and the developed ANN model, blast-induced vibrations were estimated for 19 shots performed subsequently, and the results were compared with 37 recorded vibration data. It was seen that there was linear relationship with a high correlation between the values calculated with the equation and recorded data; and there was linear relationship with a higher correlation between outputs of ANN model and recorded data.