Enhanced site-specific constant estimation in blast-induced ground vibration models using stochastic fractal search optimisation

Traditional nonlinear least square (NLLS) techniques have commonly been used to estimate site-specific constants in blast-induced ground vibration models. However, models based on NLLS often produce suboptimal predictions. This study proposes using the stochastic fractal search (SFS) algorithm, a powerful metaheuristic optimisation method, to estimate these constants for four commonly used empirical models: the United States Bureau of Mines (USBM), Indian Standard (IS), Ambraseys–Hendron (AH), and Langefors–Kihlstrom (LK) models. A dataset of 294 blast events was used for model development and validation. Comparative analysis revealed that SFS significantly outperformed NLLS in terms of prediction accuracy by lower root mean squared error, lower scatter index, higher variance accounted for, and higher correlation coefficients (R) values. The findings underscore the potential of SFS as a superior optimisation tool for predicting blast-induced ground vibration. This is because it reduced the sum of squared error, objective function of the USBM, IS, LK and AH models developed by the NLLS technique, by a significant percentage of 55.93%, 15.03%, 46.25%, and 61.21% respectively.