Recent Developments in Machine Learning and Flyrock Prediction

Abstract

AbstractThe blasting techniques are employed in mining and underground works to loosen the rock mass and ease the excavation. The blasting practices are economical and swifter in terms of their engineering application, however, they are of major environmental and safety concerns. The major issues related to blasting are flyrock, air over pressure, and ground vibrations etc. The rock fragments of rockmass are thrown outward after blasting, which can be threat to workers and machineries involved in the work, and sometimes nearby human settlements can be its victim. Therefore, an accurate prediction of the flyrock distance is the needed by mining practitioners. Earlier, experts have developed several empirical methods based on certain known parameters to assess flyrock distance. However, with time they become irrelevant and were easily replaced with advanced machine learning algorithm. The present study reviews some of these latest publications (2019–2021) examining flyrocks through artificial intelligent technique. The study incorporates types of machine learning models employed, input parameters used and number of datasets supporting the models. The input parameters were further classified according to rock-mass properties, blast design at site, and explosives responsible for blasting. Moreover, to compare the reliability of the model coefficient of correlation of the testing data of the all the documented model were evaluated. Rock density, rock mass rating and Shmidt hammer rebound number (SHRN) were found to be uncertain parameters. Artificial Neural Network (ANN) and other hybrid models for prediction of flyrock were compared.KeywordsMachine learningOptimization algorithmsFlyrock predictionBlasting