Measurement and Prediction of Blast-Induced Flyrock Distance Using Unmanned Aerial Vehicles and Metaheuristic-Optimized ANFIS Neural Networks

Abstract

Flyrock from blasting in open-pit mining is one of the most dangerous occurrences that can cause accidents to workers, damage to machinery and equipment and even fatalities. Therefore, quick and reliable prediction of blast-induced flyrock distance (BIFRD) in open-pit mines is very crucial to ensure the safety of the surrounding environment. In this study, unmanned aerial vehicle (UAV) technology combined with advanced artificial intelligence techniques was used to predict BIFRD in open-pit mines and improve safety. UAV was used to record blasting operations and the resulting flyrock. The distance of the flyrock was then measured from the recorded video footage and was analyzed using the ProAnalyst software. Then, various metaheuristics-optimized ANFIS (adaptive neuro-fuzzy inference system) was developed to predict BIFRD. These networks were optimized using adaptive differential evolution with optional external archive (JADE), genetic algorithm (GA), fireworks algorithm (FWA), and artificial bee colony (ABC) algorithms and resulted to JADE–ANFIS, GA–ANFIS, FWA–ANFIS, and ABC–ANFIS models. A dataset with 204 blasting events was gathered and analyzed, and finally, only four input variables were used for developing these models, including spacing, weight charge, stemming, and powder factor. The results showed that JADE–ANFIS is the best with high accuracy (97.8%), good generalizability (MAPE of 1.1%), and reasonable training time for predicting BIFRD in this study. The other models performed poorly with accuracy ranging from 88.7 to 96.5% and MAPE ranging from 1.4 to 3.0%. Sensitivity analysis also showed that the length of stemming is the most affecting factor to flyrock distance in blasting and thus careful consideration should be given in designing blast patterns to control flyrock distance in open-pit mines.