Prediction of blast-induced ground vibrations with the use of artificial neural networks. A case study in Greece.

Abstract

Creating models, capable of making accurate predictions of the Peak Particle Velocity (PPV) after a blast, is always one of the researchers’ most important goals. The use of such prediction can assist in assessing the intensity of the blast vibrations and more importantly in designing the whole blasting phase so as to mitigate potential problems to nearby structures. Main aim of this paper is to demonstrate the capabilities of artificial intelligence applications in geotechnology and more specifically to assess PPV and the characteristics of the blast wave attenuation in an underground construction case study in Greece, using multilayer feed forward artificial neural networks (ANNs). The results showed that the forecasting ability of the developed ANNs was, in almost every case, more accurate than the ones given by the use of traditional empirical formulas, as benchmarked using Root Mean Squared Error (RMSE) and coefficient correlation (R). In this manner, the ANNs proved to be a reliable and accurate method to assess PPV from underground blasting and once trained they become an efficient off-the-shelf tool to assist engineers both in the blast design and in the mitigation of blast wave induced problems.