Exploring Evolutionary Features Of Safety Risk In The Subway Construction Based On Complex Network Theory

As a modern means of transportation,the subway has developed rapidly in China and effectively relieved the increasing traffic congestion caused by the surge of populations due to its advantages of safety,high capacity,intensification,low energy consumption,low pollution,comfort and to the environment,high speed and on time performance.However,the subway construction has some disadvantages of long duration,large investment,and complex technical issues.As a dynamic open system,the subway construction is affected by inherent uncertainties and numerous safety risk factors,such as interacting roles of human,technology,materials,and complex environment,causing a particular high probability of hidden dangers and even accidents with serious consequence.At present,the research on safety risk mainly focuses on safety risk monitoring and evaluation,lacking the holistic perspective to recognize the evolution mechanism of safety risk in the subway construction system.Therefore,it has great theoretical significance and application value to control the evolution of safety risk and prevent accidents by taking the safety risk evolution mechanism of subway construction as the research object,exploring the process of accidents,analyzing the evolution law of safety risk and obtaining key safety risk factors.In this paper,the evolutionary features of safety risk in the subway construction is studied.First,accident cases of subway construction are analyzed statistically.Second,from the perspective of complex network theory,the evolutionary features and rules of directed weighted hazard network of the subway construction,safety risk networks and time series networks of shield construction are analyzed.Finally,an PSO-LS-SVM model for safety risk evolution control is established in the subway construction.Taking controlling the collapse accident evolution as an example,ground settlement control is realized by optimizing tunneling parameters of subway shield construction.The main research work and conclusions are as follows.(1)Statistical analysis of subway construction accidents in 2003-2018161 subway construction accidents from 2003 to 2018 are statistical analyzed from aspects of types,occurance time,the location,factors,causes and grades of accidents.The characteristics and rules of subway construction accidents is revealed from different aspects,which provided data support for the analysis of safety risk evolution.(2)Evolution analysis of directed weighted hazard network in the subway constructionThe complex network theory is adopted to construct the directed and weighted hazard network of the subway construction,composed of 53 nodes and 213 edges.Indicators for analyzing the directed weighted network are employed to capture the complex features of the network.The directed weighted network has both the property of small-world and scale-free.The results can assist the decision-making of safety risk management in the subway construction and prevent similar accidents in the future.(3)Evolution analysis of safety risk networks in the subway shield constructionThe complex network theory is used to construct an overall safety risk network and six categories of derivative accident subnetworks of subway shield construction.The networks possess the typical properties of small world network,Three different attack strategies including out-degree attacks,betweenness centrality attacks and random attacks are used to measure the changes of the network global efficiency and identify the major safety risk factors in the subway shield construction.(4)Evolution analysis of time series networks in the subway shield constructionTime series networks of subway shield construction are constructed by the visible graph method.By analyzing the inner structural information of networks,the evolution features of the time series are explored.The networks obey power law distribution and have scale-free characteristics,which better accord with the characteristics of actual time series data.Meanwhile,the time series networks have properties of small-world network,meaning that the prediction of time series will always have errors and the evolution process of time series may have mutations,which will deviate from the trajectory of the evolution of time series fitted by the prediction method.(5)The control of safety risk evolution in the subway shield constructionThe PSO-LS-SVM model for controlling safety risk evolution is constructed.Taking control of the evolution of collapse accident as an example,the comprehensive consideration of factors such as design,geological and construction have influence on ground settlement.The prediction error of the established LS-SVM model for ground settlement prediction is in a reasonable range,which can satisfy the need of prediction.That is to say,the established model can be used for the prediction of ground settlement.Meanwhile,the PSO-LS-SVM model can achieve the purpose of controlling the ground settlement by setting and optimizing the tunneling parameters and prevent the ground settlement evolution into collapse accident.Therefore,the model can be used to guide the construction.In this paper,the evolution characteristics of safety risk in the subway construction is explored.The feasibility of the safety risk evolution model is verified by the analysis of actual cases.This study provides reference value for the safe and well-organized subway construction.