Intelligent Identification of Stratum Characteristics During Shield Tunnelling Based on PCA-GMM Model

Real-time identification of geological conditions ahead the tunnel excavation face during shield tunnelling can help to set appropriate operational parameters,which guarantees tunnel quality via avoiding construction accidents and ensuring construction safety.The traditional geological prediction methods require additional testing equipment,construction space and time.This paper proposed an identification method based on PCA-GMM (Principal Component Analysis-Gaussian Mixture Model) that identifies geological conditions by using shield tunnelling parameters.The research was conducted via relying on Zhuhai Sanzao tunnelling project.The relationship between tunnelling parameters and geological features was explored using the proposed PCA-GMM model and then geological characteristics was identified based the clustering results using silhouette coefficients.Since this identification from PCA-GMM model is based on the principle of mechanical correlation between tunnelling parameters and geological features,PCA-GMM model can identify soil layer with large mechanical property gaps.The results indicate that PCA-GMM model can identify silty sand layer from silty layer.For those layers with less difference of mechanical property,such as clayey silt layer and the silty clay layer,the identified result is not satisfactory.Even though,the proposed identification method can guide the adjustment of shield operational parameter.