Intelligent technique for prediction of blast fragmentation due to the blasting in tropically weathered limestone

Abstract

In the rock blasting scenario, the success of fragmentation plays a major role. Prediction of blasted rock mass fragmentation has a significant role in the overall economics of opencast mines. Blast fragmentation has a direct impact on efficiency and cost of operation consisting of loading, transport and crushing. Tropical weathering has a direct impact on rock mass properties and strength of rock. Thus, challenging issues are created for blastability of tropically weathered limestone. It is necessary to analyze the blast fragmentation and optimize the blasting parameters. Selected limestone mine for this study is in Thailand. Various rock mass properties such as GSI, RMR, and stemming length were studied to find out the correlation with rock fragmentation. Stiffness ratio, hole diameter to burden ratio, powder factor, maximum charge per delay, RQD, blastability index (BI), weathering index (WI) and mean block size were input parameters to analyse mean rock fragment size. Total 105 data sets were collected. In this paper, a hybrid model with Artificial neural network (ANN), Particle swarm optimization (PSO) known as PSO-ANN was implemented to analyse the blast fragmentation. Multivariate regression Analysis (MVRA) was also performed. 83 datasets were trained and balance data sets were utilised for testing data. R2 values for training with PSO-ANN and MVRA showed 0.818 and 0.657 respectively. R2 values for testing with PSO-ANN and MVRA showed 0.70 and 0.694 respectively. Thus the hybrid model PSO-ANN was found useful to predict fragmentation.