Evaluation And Prediction Of Blast Induced Ground Vibration Using Artificial Neural Network

Ground vibration caused by blasting is one of the most severe environmental ill effects of mining operations and civil engineering projects. Ground vibration not only damages nearby structures but it can also cause mine slope failures and affects the overall economy of the mine. Hence, prediction and controlling of ground vibration is very important in the mining industry. The primary aim of this study is to predict blast induced ground vibration and evaluate the performance of the artificial neural network method of prediction. 34 ground vibration input data collected from one of an open pit limestone mine were predicted using ANN and empirical predicators. In the ANN model the network was trained by a gradient decent back propagation algorithm. A three layer feed forward BP neural network with an architecture of 7-8-2 was found to be optimum. While in the empirical methods both the Sardolfski's and Ambrasey-Hendron predicator equations were used. The peak particle velocity of the monitored vibration was compared with PPV predicted by the ANN and the empirical models. Results have shown that predicting using simple application of site constants k,?(A-H formula) and ? in the case of Sardolfski's formula may lead to an error due to many factors. Besides, the empirical method has a limitation in that it is not setup to predict the important parameter, frequency. In contrast, prediction of both PPV and frequency can be achieved with high accuracy and reliability by the ANN model. The coefficient of correlation(r) for both PPV and frequency is well above 0.9 for all the components, while the mean square of errors(MSE) is less than 0.1 in the case of the ANN model. Results of the sensitivity analysis shows that the horizontal distance between the point of blasting and point of monitoring, is very important input parameter for both PPV and frequency of the blast vibration.