Ensemble Learning Model And Engineering Application Of Environmental Effect Prediction Of Rock Mass Blasting

Blasting is still an economical and feasible method of rock excavation in mining and civil engineering.However,the blasting vibration,flying stone,air shock wave and seismic wave are produced during excavation and blasting,they often cause cracks and even collapses in surrounding structures,and causing serious and even casualties.Moreover,because of universality,complexity and easy to cause civil disputes,it not only makes the construction more difficult,but also affects the social stability.Due to the complexity of its nonlinear evolution,how to accurately evaluate the environmental effects of blasting has been a difficult problem for the engineering and scientific communities.On the basis of systematically summarizing and analyzing the advantages and disadvantages of blasting environmental effect assessment methods at home and abroad,and based on the theory of integrated learning,the topic of predicting the harmful effects of blasting is discussed in depth.An integrated learning model of blasting vibration peak velocity type,blasting induced structure failure and blasting flying rock prediction is constructed.The main contents and conclusions are as follows:(1)For evaluation of the effect of blasting environment uncertainty and the problem of low precision,constructed the blasting environmental effect prediction model library based on the kinetic characteristics of blasting vibration and flying rock,including the blasting vibration peak velocity prediction of PSO-XGBoost integrated learning model,damage prediction of blasting induced houses structure integrated Adaboost learning model and blasting flying rock GBDT integration model.The integrated learning model constructed in this paper has the pertinence in the blasting environmental effect prediction and the union in the engineering application.In the specific engineering application,it can be used independently to solve the specific forecasting problems.And it can be combined arbitrarily to get the comprehensive prediction and evaluation results according to different engineering requirements.The results show that the proposed integrated learning model is very reliable,has good evaluation effect,and has high prediction accuracy,and achieves the accurate qualitative and quantitative prediction of blasting environmental effects.(3)The contribution of each input parameter in the model to the prediction model is revealed,the verification strategies of internal and external validation of the prediction model are established,and the k-fold cross validation method is used to optimize the super parameters in the model.A joint evaluation method of R2+MAE+RMSE is established to solve the regression problem;Aiming at the classification problem,the joint evaluation method of the Error matrix+User precision+Producer precision+Total precision+Kappa coefficient was established,which improved the comprehensiveness and correctness of the evaluation effect of the model.The problem of overlearning or under learning of the integrated learning model for the environmental effects of rock blasting is solved.(4)In order to further verify the integrated learning model of blasting environmental effect prediction constructed in this paper,taking the Heshangqiao Iron Mine as an example,and making the PSO-XGBoost integrated learning model of PPV for predicting blasting vibration,Adaboost integrated learning model for predicting blasting vibration damage to civil buildings and gradient hoist model for predicting blasting flying rock distance were verified by engineering verification.Good results are obtained and the validity of the model database is verified.