A Neural Network Approach for the Prediction of Ground Vibrations Induced Due to Blasting

Abstract

This project presents the application of neural networks as well as statistical techniques for prediction of ground vibration by major influencing parameters of blast design. The predictions by artificial neural network (ANN) is compared with the predictions of conventional statistical relation. Ground vibrations and frequency induced due to blasting were monitored at Indian Detonators Limited Rourkela (IDL), Balphimali Bauxite mine (UAIL) and Dunguri Limestone mine (ACC). The neural network was trained by the data sets recorded at the various mine sites. From the analysis it was observed that the correlation coefficient determined for PPV and frequency by ANN was higher than the correlation coefficient of statistical analysis. The correlation coefficient determined for PPV and frequency by ANN for Balphimali Bauxite mine (UAIL) was 0.9563 and 0.9721 respectively and correlation coefficient determined for PPV and frequency by ANN for IDL was 0.9053 and 0.9136 while correlation coefficient determined for PPV and frequency by ANN for Dunguri Limestone mine (ACC) was 0.9322 and 0.9301. The difference in correlation coefficient of PPV and frequency in different mines is due to different number of input parameters for the neural network and number of datasets used for the training of network. The number of datasets and input parameters were more for Balphimali Bauxite mine (UAIL), thus it showed higher correlation coefficient between the recorded and predicted data by ANN than other mines.