Blasting Fragmentation Prediction based on PSO-BPNN Model

The impact of fragmentation size and gradation on the stability and permeability of rockfill in hydraulic engineering is of great significance. Accurate prediction of fragmentation size has become a key focus in rock blasting research. In this study, a PSO-BPNN model is developed based on the Backpropagation Neural Networks(BPNN) with optimized network weights and biases using the Particle Swarm Optimization(PSO) algorithm. The model is trained and tested using representative blasting data, and its reliability and applicability are validated through its application in the Hunyuan Pumped Storage Power Station project in Shanxi. Results demonstrate that the PSO-BPNN model exhibits short computation time and high reliability for predicting fragmentation size, with a maximum relative error between the model output and actual average fragmentation size of 6.56%. Therefore, this model demonstrates high predictive accuracy and applicability, providing precise guidance for construction of rock-fill dams at the Hunyuan Pumped Storage Power Station in Shanxi province.