Neural network based selection of design parameters governing shape and powder factor of blasted muck piles: a case study

Abstract

This paper presents an application of the artificial neural network (ANN) technique to reduce the number of blast design parameters, which affect the shape characteristics and powder factor values of blasted muck piles. A scientific feature selection approach was used to establish the minimum number of input parameters for an ANN model. Blast design data already implemented in a limestone quarry were used in the ANN analysis. The input features were selected based on the least mean squared error value of the data set after proper training. The parameter elimination technique of the ANN platform shows promise for eliminating input parameters and selecting the optimum number of blast design parameters. Two parameters for throw and one parameter for drop were eliminated; four parameters were eliminated for powder factor. The results were tested and validated with an actual data set at acceptable correlation levels, and have been properly illustrated.