Research On Deformation Prediction And Early Warning Of Large Section Tunnel Surrounding Rock Based On Deep Learning

To achieve the transformation of China’s economy to the stage of high-quality development,the building of transportation infrastructure is always a mountain that cannot be climbed.With the implementation of the national comprehensive development strategy,the percentage of large-section tunnel construction is continuing to increase.huge section tunnels sometimes encounter challenges during on-site construction,including unfamiliar design parameters,challenging construction techniques,and a complex construction environment,due to their huge design section and the complex mechanical characteristics of the surrounding rock.Therefore,in order to better direct onsite construction,avoid safety issues,and fundamentally improve the construction environment of large-section tunnels,it is necessary to conduct an in-depth study on the prediction and early warning of the deformation of the surrounding rock of tunnels with large sections based on engineering monitoring data combined with deep learning algorithms.Based on the Chengzai Mountain Tunnel,this study examines large section tunnel surrounding rock deformation prediction and early warning.Field monitoring and numerical simulation are used to study the spatial deformation law of large section tunnel surrounding rock;the SSA algorithm,ELM algorithm,and PSO algorithm are used as the theoretical underpinnings.The orthogonal experimental design is also utilized to establish the numerical computation.The surrounding rock parameters’ SSA-ELM-PSO inversion analysis model was built using the numerical calculation database as training data;The YK54+560 section constructed by the three-step method,the YK54+475 section constructed by the mid-bore method,and the YK54+475 section constructed by the double-sided method were predicted based on the real monitoring data of the YK54+565section,YK54+480 section,and the YK54+590 section and using the SSA-ELM-PSO model to invert the mechanical parameters of the surrounding rock.In order to realize the dynamic early warning of the excavation of the surrounding rock of the large section tunnel,the energy criterion based on the cusp mutation theory was introduced into the theory of strength reduction method.The ultimate displacement values of the vault settlement of the tunnel surrounding rock at each excavation stage of the study section were obtained with the aid of numerical calculation software.The following are the paper’s primary conclusions:1)The stabilization phase of the displacement field of the big section tunnel surrounding rock is born after the support structure is closed into a ring,according to the results of numerical simulation and field measurement.The surrounding rock’s deformation mode can be summed up as pre-deformation,abrupt deformation,regular deformation,and steady deformation after analyzing the deformation curves under various working methods.2)The inversion model of the surrounding rock parameters based on the SSA-ELMPSO algorithm was constructed,and the efficiency of the SSA algorithm to optimize the ELM model was tested in the forward training process.A numerical calculation training database was established by using the principle of orthogonal experimentation.3)Through the inversion of mechanical parameters,the prediction of the surrounding rock deformation of each study section is carried out,and the prediction results show that: the maximum relative error between the predicted and measured values of arch settlement is 2.69%,and the minimum relative error is 1.02%;the minimum relative error of arch waist measurement line and arch foot measurement line,which are influenced by the actual construction human factors,is 6.75%,and the maximum relative error does not exceed 17%.4)The limit displacement values of the arch settlement were calculated using the energy method criterion based on cusp mutation for each study section and excavation stage,and real-time curves of the predicted and warning values with excavation stage were drawn to realize the real-time follow-up of the warning values and to determine whether the predicted values of the section exceeded the threshold.