Application of ANN and metaheuristic algorithm optimized ANNs for the prediction of blasting-induced ground vibration in urban areas

Prediction and minimization of the adverse consequences of blasting, particularly blast-induced ground vibration (BIGV), are important tasks that largely determine the success of mining, tunneling, and civil engineering projects. However, most previous studies on the prediction of the BIGV have focused on mining and tunneling, whereas the BIGV of construction excavation blasting, including urban and rural highway construction and urban residential land development and redevelopment projects, have rarely been predicted. BIGV from construction project sites are critical because of their proximity to urban and rural dwellings, and important public utilities. This study introduces a novel hybrid machine learning (ML) model of an artificial neural network (ANN) optimized using the slime mould algorithm (SMA) and grasshopper optimization algorithm (GOA) to forecast and minimize BIGV generation from multiple highways and urban residential land development and redevelopment project sites in South Korea. In this study, 115 blasting events from construction sites in South Korea were monitored and their parameters, peak particle velocity, and rock mass rating were recorded. A comparison was made between the newly introduced ANN-SMA and ANN-GOA models, other developed machine learning (ML) models, and three empirical models. The newly introduced models significantly outperformed other models. The suggested hybridized models were transformed into adjustable and user-friendly explicit equations that can assist field and blasting engineers working at construction sites in accurately predicting BIGV during construction. The models were validated for practical engineering applications using 20 separate datasets that were not employed for model development. Finally, the relevance importance appraisal of the model inputs was performed using the cosine amplitude method (CAM), and the rock mass rating (RMR) had a significant influence on the forecasted BIGV. The findings of this study could aid engineers and researchers in accurately estimating the potential PPV values in construction excavation projects before actual blasting, which will help in advance planning to enhance blast design and mitigate conflicts and disagreements between mines and local communities.