Study on Intelligent Recognition of Geological Types in Shield Excavation Face Using Machine Learning
The shield method is widely applied in subway tunnel projects,but complex geological environments can lead to issues like jamming,cutterhead wear,over-excavation,and ground subsidence during shield construction.Accurate and timely detection of geological condition changes is crucial.In China,geological surveys are primarily conducted through drilling,and existing advanced geological prediction techniques often fail to meet the demands of rapid shield tunneling.A machine learning model is proposed to identify geological types at the shield excavation face.This model integrates Genetic Algorithm(GA)with the eXtreme Gradient Boosting(XGBoost)library to determine optimal hyperparameters for the XGBoost model.The model is validated using shield and geological data from the third section of the Fuzhou Binhai Express Line and is compared with three classical machine learning models—LightGBM,Random Forest,and SVR—to assess its robustness.Results demonstrate that the GA-XGBoost model significantly outperforms the alternatives in accuracy,with GA effectively enhancing model precision by optimizing hyperparameters.
万方数据
维普
