The blasting vibration velocity is one of the important factors that need to be considered in blasting design.However,in predicting blasting vibration velocity,the determination of hyperparameters in BP neural networks depends on empirical formulas and has subjectivity.To overcome this limitation and improve the accuracy of vibration velocity prediction,the hyperparameter optimization algorithm of TPE was used to optimize the hyperparameters of the BP neural network.A BP neural network(TPE-BP)prediction model with 31 hidden layers neurons was established using the maximum explosive charge,borehole depth,horizontal distance,vertical distance,and explosive consumption parameters as input.The average prediction error of the blasting vibration velocity of the model was 2.35％,with a maximum error of 6.29％.Compared with the BP neural network model based on empirical formulas to determine hyperparameters and the traditional BP neural network model,the average prediction error was reduced by 23.26 percentage points and 4.24 percentage points,respectively.The results indicate that the optimized parameters network of the TPE-BP prediction model can better fit the vibration data,and its prediction results are closer to the true values.The study can provide a reference basis for blasting parameter design,thereby further effectively control the blasting vibration.
关键词
爆破振动/振动速度预测/BP神经网络/TPE算法/超参数优化
Key words
Blasting vibration/Prediction of vibration velocity/BP neural network/TPE algorithm/Hyperparameter optimization
维普
万方数据