A Novel Method of Correlation Analysis Between Ground Subsidence and Tunnelling Parameters Based on Model Fusion

Ground subsidence is important during shield construction. How to select reasonable tunnelling parameters is critical to control ground subsidence. However, the selection of shield tunnelling parameters for ground subsidence control usually relies on human experience. It is critical to study how to make judgements of ground subsidence adaptively during shield construction and give recommended values of tunnelling parameters to mitigate subsidence. To meet the above requirements, this paper innovatively proposes a method for analysing the correlation between ground subsidence and tunnelling parameters based on the fusion method of qualitative and quantitative models. First, the a priori algorithm based on MAX–MIN Ant System is adopted to analyse the qualitative relationship between ground subsidence and tunnelling parameters and provides the tunnelling parameters intervals under different ground subsidence intervals. Second, the improved Beetle Antennae Search-Deep Neural Networks is constructed to calculate the occurrence probability of different discrete internals of ground subsidence to correct the rougher results obtained by qualitative analysis. Finally, a qualitative and quantitative model fusion formula is proposed. The occurrence probability is used to modify the discrete interval value range of the tunnelling parameters intervals under different ground subsidence intervals to obtain the optimised values’ range of the key tunnelling parameters under different ground subsidence intervals. The proposed model is used in case analysis. The results of the model predict the probability values corresponding to each subsidence interval that occurred. The prediction accuracy of the quantitative model between ground subsidence and tunnelling parameters in different intervals is higher than 96%.