Method for geological characteristics prediction during shield tunnelling: SCA-GS

Geological characteristic (GC) is one of the most essential factors influencing setting earth pressure balance (EPB) shield parameters and cutterhead wear. Identification of GC has crucial significance to shield tunnelling efficiency and safety. Stacking classification algorithm (SCA) is widely applied in engineering with the identification and classification. Grid search (GS) is designed to tune hyper-parameter and optimize non-linear problems with K-folds cross-validation (K-CV), which is commonly used to change validation set in the training set. The performance of SCA can be improved by GS and K-CV. The types of GC during shield advance can be identified by integrating K-means++ with silhouette coefficient (Si) and elbow method (EM). The results of K-means++ and shield parameters severed as a database for SCA. The approach was applied in Guangzhou mixed ground. The results showed that the proposed framework could predict the geological characteristics well. The method article is a companion paper with the original article [1]. The proposed method enables: • Developed approach merges SCA and GS method. • Application of SCA-GS method in geological characteristics classification. • It can increase the reliability of classification results.