Artificial neural network and regression models for prediction of free-field ground vibration parameters induced from vibroflotation

Abstract

The prediction equation of various construction induced ground vibration have been documented in the literature. However, no research is available in the literature to model the vibroflotation induced vibration. The major objective of this paper is to propose a methodology to predict the vibroflotation induced free-field ground vibration using large recorded data. The major contribution of this work is to establish relations between the vibration parameters induced due to vibroflotation and the source to receiver distance using regression and artificial neural network modelling. Two different sites have been considered in this study to incorporate the effect of soil parameters, site characteristics and the vibroflotation design parameters. The numbers of model input parameters have been reduced based on sensitivity analysis to improve the efficiency and reduce the error in the developed ANN model (approximately 6% and 4% error reduction for PPV and PGA prediction respectively). It has been found from the sensitivity analysis that the source to receiver diagonal distance, effective overburden pressure and the treatment depth have significant importance in the proposed PPV attenuation correlation. However, the location of the vibroflotation point and the distance between source and receiver have a significant impact on the proposed PGA attenuation model.