Abstract: During blasting quarry of dam materials, fragmentation control is a key means to ensure dam construction quality. In previous studies of the prediction of blasting material fragmentation, several problems are left unsolved such as low prediction accuracy and poor generalization ability, and difficulty still remains in accurate control on the fragmentation of rockfill dam materials to meet the requirement of blasting quarry. This study develops a blasting fragmentation prediction model based on a random forest (RF) regression method to overcome the shortcomings of previous blasting prediction models and improve the control on blasting fragmentation. Cross-verification and comparison with other RF prediction models shows this RF model is superior. We verify its calculations and applicability using a blasting fragmentation prediction system of a practical project, and demonstrate its usefulness in management and control of the blasting construction of rockfill dams.

Key words: hydraulic engineering, dam material quarry, blasting fragmentation, random forest regression, prediction