Intelligent Prediction Of Blasting Vibration Based On Artificial Neural Network:Optimization Algorithm Comparison And Improvement

Blasting technology brings economic benefits and also brings many harmful effects.Blasting vibration is the most concerned hazard in engineering.How to predict and control more accurately is a hot issue in engineering.At present,the blasting vibration control method in practical engineering is still based on empirical formula prediction to optimize blasting design and reduce the harm caused by blasting vibration.However,there are many factors affecting blasting vibration in complex sites,and there are complex nonlinear relationships.The prediction accuracy of empirical formulas is low,and the prediction results can no longer meet the needs.In order to better predict the harm caused by blasting vibration control,this paper combines the ability of artificial neural network to solve complex nonlinear function approximation and the global optimization ability of swarm intelligence optimization algorithm to establish a variety of improved artificial neural network prediction models.The prediction model is used for blasting vibration research to find the best prediction model and improve the prediction results to meet the engineering needs.The main research and results are as follows:(1)Based on the blasting vibration data monitored by blasting excavation of the left abutment groove of the Baihetan Hydropower Station,the max charge per delay,distance from the blast face,height difference and longitudinal wave velocity were selected as input parameters.Through the comprehensive evaluation of the results using the running time and the Rooot Mean Square Error(RMSE),Mean Absolute Error(MAE)and Determinant coefficient(R~2)evaluation indexes,a new algorithm,Grasshopper Optimization Algorithm(GOA),suitable for optimizing an Artificial Neural Network(ANN)to predict the peak particle velocity was obtained.Comparing the Mean Absolute Percentage Error results of the empirical formula,ANN and GOA-ANN models,it is verified that the model has good generalization ability.Combined with sensitivity analysis and prediction results,it can be concluded that when the main factors affecting blasting vibration in complex sites change,the prediction results of GOA-ANN model are better matches with actual monitoring values,and more reliable and scientific.(2)Aiming at the shortcomings of the GOA algorithm,this paper proposes a multi-strategy collaborative improvement of the GOA algorithm,which combines Sobol sequence initialization population,piecewise adaptive coefficient C adjustment strategy,Cauchy mutation strategy and limit threshold idea,to improve the global search ability of the algorithm and ensure the diversity of the population.The experimental results show that the improved algorithm is far better than the standard GOA algorithm.The performance of the improved GOA algorithm to optimize ANN is compared with the standard GOA-ANN model,and it can be found that its accuracy has further improved on 50,100,150 populations,which verifies that the improved algorithm is effective.The computational efficiency can be increased by reducing the number of population,and good results can be achieved.(3)According to the damage principle of blasting vibration to buildings,the improved prediction model is used to predict the peak particle velocity and the main frequency.Based on the collected monitoring data of earthwork blasting engineering of the approach channel of Honghua Water Conservancy Project in Liujiang,Guangxi.The predicted results can meet the engineering requirements well.