Collapse Risk Assessment And Early Warning Research For Mountain Tunnelling Based On Bayesian Theory

At present,the drilling and blasting method is widely used in mountain tunnel excavation in China,which leads to the frequent occurrence of collapse accidents during construction due to the complex construction process and more risk factors involved,and effective safety risk management during construction can prevent and reduce such accidents.At present,risk management in tunnel construction is mainly based on qualitative and static risk assessment by expert experience,which has problems such as lack of case data,insufficient dynamic risk feedback and over-reliance on expert experience.In order to address the above problems,based on the Wenbishan Tunnel Project of the Fujian Puyan Expressway,this paper investigates the collapse risk assessment and early warning in mountain tunnel construction by combining theoretical analysis,field measurements and numerical simulations.The main research contents and results are as follows:(1)The collapse risk factors in mountain tunnelling are identified.Based on the literature review and accident statistics,the risk factors of tunnel collapse are identified,the types of tunnel collapse are classified,the main risk factors affecting the occurrence of collapse are summarised and analysed,the static evaluation index of tunnel collapse is obtained,and the monitoring data and overcasting results are considered as the dynamic risk evaluation index,and a risk evaluation index system is established.(2)A risk assessment model based on Bayesian networks is developed.Taking both the network structure and the network parameters of Bayesian networks as the starting point of the study,the degree of correlation between risk factors was considered,the network structure was modelled hybridly by the fuzzy explanatory structure model and the causal graph method,and the Bayesian network prior probabilities and conditional probability tables were calculated by combining data learning with the affiliation cloud model,and the above methods were integrated to construct a dynamic assessment model for tunnel collapse risk.(3)A collapse risk early warning method based on the prediction of the surrounding rock displacement is proposed.In perimeter rock displacement prediction,training samples for the proxy model are obtained through orthogonal tests and numerical simulations.The perimeter rock parameters with high sensitivity to the displacement calculation results are defined as random variables,and the posterior probability distribution of the perimeter rock parameters is updated from the displacement monitoring data obtained each time by a Bayesian updating framework,and the displacement values during tunnel excavation are calculated by the GA-BP neural network model in the forward direction.The predicted values of the perimeter rock displacements are fed into the dynamic risk assessment model to predict the level of construction risk in the coming days and to achieve the purpose of risk warning.