Prediction Of Blasting Vibration Velocity Peak Value Based On BP Neural Network Optimized By Sparrow Search Algorithm

As one of the main blasting hazards,blasting vibration requires strict control and accurate prediction.In this paper,taking the blasting mining of Wujiata open-pit coal mine as the background,and aiming at the safe mining of the western side to the boundary,using the field measured data,different nonlinear prediction methods are used to predict the blasting particle vibration velocity peak value of the western side.Provide reference and support for blasting parameter design in the area.Through the long-term monitoring of the vibration generated by the blasting of the Wujiata open-pit coal mine,several sets of field measured data were obtained,the least squares method was used to fit and analyze the Sadowski formula,and the elevation correction formula was used to measure the blasting vibration of the positive elevation bench.The peak value is fitted and calculated,the attenuation coefficient and β value in the formula are determined,and the prediction formula of blasting vibration peak velocity suitable for this area is obtained,and the average relative errors in the horizontal radial,horizontal tangential and vertical directions are 30.226 %,35.415% and 37.032%.The sparrow search algorithm(SSA)improved BP neural network is used to predict and compare the blasting peak vibration velocity actually measured on site.Line length as input variable.The vibration velocity peaks in the horizontal radial,horizontal tangential and vertical directions are used as output variables.The output value obtained after repeated training is used as the predicted value,which is compared with the actual blasting vibration peak velocity measured on site.The average relative errors of the BP neural network improved by the Sparrow Search Algorithm(SSA)in three directions are 6.338%,7.152% and 7.228%,respectively,which greatly reduces the prediction error compared with the empirical formula of step blasting vibration.In order to further verify the SSA-BP neural network prediction model,the ANSYS/LSDYNA numerical simulation software was used to simulate the step blasting process of the openpit mine,and the peak particle velocity of the corresponding sample points was obtained.The average relative error of the three directions in the direction and the vertical direction is lower than12%,which verifies the accuracy and reliability of the numerical simulation.After that,the step blasting under different parameter conditions is simulated,and the peak velocity of blasting vibration at different positions is counted.The SSA-BP neural network model is used again to predict the preset samples,and the logarithmic value of the SSA-BP neural network model is obtained.The average relative error of the simulated vibration velocity peak is less than 11%,which fully verifies that the SSA-BP neural network prediction model can accurately and nonlinearly predict the peak velocity of blasting vibration.This algorithm has a good reference and promotion for blasting vibration in other projects.value.