Underground blast induced ground shock and its modelling using artificial neural network

The ground shock motions induced by underground explosions are of major concerns to designers in mining, construction and defence engineering. The ground shocks are usually measured by the peak particle velocity (PPV), and various empirical formulae exist for the prediction of PPV as a function of the scaled distance for given geological site conditions. In fact, apart from the PPV, the frequency content and the relative amplitude of horizontal and vertical components can also play important roles with regard to the response of built structures in nearby areas. On the other hand, various variables such as the charge loading density, site geology, charge chamber geometry, can also affect the ground shock at a given scaled distance. Hence, the prediction of the ground shock parameters effectively involves a multi-input and multi-output system, which is difficult to describe fully using simple regression formulae. In this paper, a comprehensive research programme on the effect of the various input variables on the ground shock parameters is summarized. The data obtained from this programme are processed to form a suitable dataset to represent the underlying physical problem. The artificial neural network (ANN) technique is then applied to identify the system pattern and serve as a function for the prediction of the ground shock. The neural network approach proves to be successful as the trained neural network can predict the unseen test data consistently and with satisfactory accuracy. The results also provide new insight into the significance of some influencing variables such as the path orientation on the ground shock parameters.