Study On Tunnel Construction Simulation Based On Advanced Classification Of Imbalanced Surrounding Rock Using Improved XGBoost

The classification of surrounding rock is the key factor affecting the construction of tunnel,which is directly related to the selection of construction parameters such as footage driving cycle,blasting parameters and support parameters.In addition,because very hard or soft rock often only accounts for a small part of the engineering rock mass,the type of surrounding rock is imbalanced.Therefore,it is the key to ensure the accuracy of tunnel construction simulation results to realize the effective judgment of surrounding rock classification under the condition of considering the class imbalance of surrounding rock classification.At present,most of the tunnel construction simulation studies directly determine the surrounding rock category based on the initial geological exploration results,and a few studies deduce the unrevealed surrounding rock category based on the properties of excavated rock mass,but the accuracy of the prediction results largely depends on the understanding of the properties of excavated rock mass.Although some scholars have carried out advanced classification of surrounding rock by taking advantage of advanced geological prediction,which can directly obtain the characteristics of rock mass in front of excavation,the existing methods mostly use a single weak classifier,and do not consider the class imbalance of tunnel surrounding rock,so the classification accuracy is low.In view of the above problems,this paper carried out tunnel construction simulation research based on the advanced classification of imbalanced surrounding rock using improved XGBoost,and achieved the following research results:(1)In view of the class imbalance of tunnel surrounding rock,which causes the learning of classier to lean to majority classes and poor classification accuracy,put forward the optimal method of surrounding rock classification imbalance based on KMeans-SC-SMOTE.In this paper,the absolute contour coefficient of evaluating index of K-means clustering algorithm is used as the root sample discrimination basis to improve the synthetic oversampling algorithm,in order to avoid the marginalization of the distribution of synthesized samples and relieve the intensification of the imbalance within the class.On this basis,the imbalance of surrounding rock classification is optimized.Firstly,the influencing factors of surrounding rock classification are determined according to the traditional classification method.Secondly,the correlation between TSP advanced geological forecast data and influential factors of surrounding rock classification is analyzed,and then the initial data set of surrounding rock advanced classification is constructed by selecting surrounding rock classification indexes.Finally,the KMeans-SC-SMOTE algorithm is used to make the synthetic oversampling of the minority classes in the initial data set of surrounding rock classification,so as to achieve the imbalance optimization of surrounding rock classification.(2)Considering the lack of single weak classifier and limited accuracy of existing advanced classification methods based on advanced geological prediction,the advanced classification model of surrounding rock based on improved XGBoost was established under the condition of considering the imbalance of surrounding rock classification.The improved XGBoost integration algorithm is used to learn the imbalanced optimized advanced classification data set of surrounding rock,and the high-precision mapping relationship between surrounding rock classification index and surrounding rock classification is established.Among them,the traditional XGBoost method adopts manual parameter tuning,which is time-consuming and difficult to obtain the optimal hyperparameters.Aiming at the average accuracy of cross validation that can reflect the model generalization performance,this paper adopts Harris Hawks optimization algorithm to automatically optimize the learning＿rate,n＿estimators,max＿depth and other nine hyperparameters of XGBoost.The optimal combination of hyperparameters is obtained to give play to the best classification performance of the model and improve the efficiency of parameter adjustment.(3)In view of the problem that most of the current tunnel construction simulation studies determine the surrounding rock category directly according to the initial geophysical exploration,and a few of the prediction methods based on the excavated rock mass are highly dependent on the mastery of the properties of the excavated rock mass,combined with the advanced geological prediction can directly obtain the advantages of the rock mass characteristics index,a tunnel construction simulation method based on advanced classification of surrounding rock using improved XGBoost is proposed.The advanced classification method of surrounding rock based on improved XGBoost is coupled with the tunnel construction simulation model,and the tunnel construction simulation model coupled with advanced classification of surrounding rock is established.Firstly,based on the classification index of surrounding rock of tunnel advanced geological forecast,the improved XGBoost model is used to carry out advanced classification of surrounding rock.Secondly,construction simulation parameters such as footage driving cycle,blasting parameters and support parameters are optimized based on the results of surrounding rock advanced classification,and then the simulation analysis of tunnel construction is carried out to effectively improve the accuracy of simulation.(4)Taking a diversion tunnel project as an example,the application research of tunnel construction simulation based on the improved XGBoost advanced classification method of unbalanced surrounding rock is carried out,and the superiority of the proposed method is proved by comparative analysis of the results.Based on a diversion tunnel project,the practical application of tunnel construction simulation based on advanced classification method of unbalanced surrounding rock using improved XGBoost is carried out.Firstly,the surrounding rock classification index was extracted based on the advanced geological prediction data and the KMeansSC-SMOTE synthetic oversampling algorithm was used to optimize the class of surrounding rock.Secondly,an advanced classification model for surrounding rock based on improved XGBoost is established under the condition of balanced data,and the classification of unexcavated rock mass is predicted.Finally,the simulation parameters of tunnel construction are optimized based on the advanced classification results of surrounding rock,and then the layered simulation model of tunnel construction is used to simulate the tunnel construction process.The results show that the class imbalance optimization of surrounding rock and the XGBoost model optimized by Harris Eagle algorithm can effectively improve the accuracy of surrounding rock advanced classification.In addition,the relative deviation between the simulation results based on the advanced classification of surrounding rock and the actual progress is reduced by 10.7% compared with the traditional simulation,which is more consistent with the engineering practice,proving the superiority of the method proposed in this paper.