Prediction on parameters optimized GA-ELM model for cast blasting in open-pit mine

In order to make a correct production plan effectively and improve the accuracy of cast blasting in open-pit mine,based on the analysis of cast blasting factors,the optimized hidden layer node parameters are determined by “tri- al-and-error” and then one kind prediction model is proposed with the combination of Genetic Algorithm ( GA) and Extreme Learning Machine (ELM). On the basis of network input and output parameters’ selection,the optimization and selection are made using GA for ELM model’s input weight matrix and hidden layer deviation. Using a cast blas- ting monitoring data in an open-pit mine,the model analysis was conducted,then the prediction results of RBF,BP, SVM,and GA-BP models were compared with the model results. In addition,Weibull model was introduced to simulate the coal blasting form by predicting control parameters a,β. The results show that:① The optimal hidden layer node number is 39,which reduces the system simulation error effectively,and the simulation error is 0. 137 7;② Compared with traditional ELM prediction model,after optimized by GA,Effective Casting Blasting Ratio,Loose Coefficient and Cast Distance all have lower Mean Square Error (MSE:0. 258 0,1. 748 5×10-4 ,3. 618 4) and higher Decision Coeffi- cients (R2 :0. 986 4,0. 995 3,0. 970 6). Improved GA-ELM model has a better fitting effect and generalization abili- ty;③ Compared with other intelligent algorithm,such as BP,RBF,SVM,GA-BP,the test set results (MSE,R2 ,SE) of improved GA-ELM model are much better and improve the prediction accuracy effectively;④ Using trained GA-ELM network to predict coal blasting form,the prediction error of control parameters a and β are not more than 5% and the prediction of blasting heap curve is close to the real blasting heap.