Risk Assessment And Warning Research Of Water Inrush In Karst Tunnels Based On Data Learning

The karst tunnel water inrush disaster is a bottleneck that severely restricts the safe construction of karst tunnels.Once triggered,it will cause serious consequences such as economic losses,delays in construction periods,and casualties.Due to the variability and complexity of the underground engineering environment,the probability of karst tunnel gushing water disasters has not been accurately calculated.Therefore,it is of great engineering value to establish a novel risk assessment model for karst tunnel water inrush that can accurately predict the probability of water burst and the consequences of the disaster and propose an early warning mechanism based on the karst tunnel disaster prevention design.In this paper,the important water inrush cases of karst tunnels in recent years have been counted,and the existing risk classification standards have also been improved.Reliability theory,GA-BP neural network,and Bayesian network were introduced to build the novel risk assessment model.Quantitative risk assessment was achieved through self-developed procedures,and the relationship between risk level and disaster prevention measures was determined through the self-built database.The main research results obtained in this paper are as follows.(1)The primary and secondary rankings of 11 influencing factors in three categories that induced the karst tunnel water inrush were determined,and the four most controllable factors were obtained,namely unfavorable geology,topography and geomorphology,the contact zones of dissolvable and insoluble rock and groundwater level.Probability grading standards were introduced,and the risk grading system was also improved.Sub-level grading was determined on the basis of the existing karst tunnel water inflow grading to further improve the accuracy of the evaluation results.(2)A novel quantitative evaluation model based on the reliability theory and GA-BP neural network was established for the water inrush disaster of the karst tunnel with highpressure water-rich cave.The minimum outburst prevention thickness of the rock disk was selected as the dominant function,and the probability of water inrush was calculated by determining the probability density distribution of each random variable through the reliability theory.The GA-BP neural network was applied to predict the disaster consequence caused by water inrush.Six factors including water pressure,hydraulic replenishment,type of fractures,filling condition,water-rich degree and reserves of cave were selected as the input layer of the neural network and the disaster consequence was used as the output layer.Similar projects were screened to obtain statistical information for indicators,and the Normand function in MATLAB was used to transform this information into the quantitative data.Based on the possibility of water inrush and the consequences of the disaster,a risk assessment was performed on the 602 cave of Yesanguan Tunnel,and the actual situation was compared with the evaluation results of the PASM method to verify the feasibility and reliability of the model.(3)A Bayesian network risk assessment model was established in order to cope with unidentified disaster sources or insufficient accuracy of index information.The model was manually established by the interpretive structural model,and then modified by the causality diagram to determine the subordinate relationship between the network nodes.And the corresponding non-inrush samples were generated based on the adversarial network and analytic hierarchy process to enrich the network database.The model was trained and verified,and four indicators including the overall accuracy(ACC)were used to evaluate the training effect of the model.Finally,the Bayesian network risk assessment model was applied to the DK490+373 water inrush case of the Shanggao Mountain Tunnel to verify the feasibility and accuracy.(4)Based on the Visual Basic programming tool,two new risk assessment models were converted into applications.Through the induction and statistics of the principles and measures for prevention and control of water inrush disasters,the relationship between the level of warning risk and guidance for disaster prevention and control was established.The database of karst tunnel water inrush disaster cases was established based on the self-developed program,which is convenient for querying similar engineering cases after risk assessment,and providing guidance and reference for similar projects.The paper has 48 figures,41 tables,and 166 references.