Safety Prognosis Method For Long Span Cable-stayed Bridge Based On Multivariate Mixed Model

Operation condition and safety prognosis of bridge structures has always been taken more attention for many years.It is possible to employ bridge structural health monitoring（SHM）system to solve this problem.However,bridge structural health monitoring faces technical bottlenecks of damage alarming and safety prognosis.In this paper,Guanhe Bridge is taken as the engineering background and safety prognosis methods of long span cable-stayed bridge based on finite element model（FEM）and data-driven model are developed respectively by employing the massive data of its SHM system.Then the safety prognosis method based on multivariate mixed which combines FEM and data-driven model is developed to provide technical support for bridge maintenance and repair so the corresponding research has obvious theoretical significance and engineering application value.The main works in this paper are as followes:（1）The uncertainty of environment and vehicle load are analyzed and quantified in order to establish temperature load model and vehicle load model.（2）Kriging response surface method is adopted to update and validate the finite element model in order to provide the accurate finite element model for bridge safety prognosis.（3）The gray neural network method based on genetic algorithm and particle swarm optimization algorithm are used to predict the vehicle load model and obtain the future vehicle load model of the bridge structure.（4）Then the method which combines Kriging and important sampling（IS）method is adopted to obtain the failure probability of the middle of Guanhe bridge.The actual vehicle loads of Guanhe bridge are employed to assess fatigue reliability and fatigue remaining service life.（5）Gaussian Sum Particle Filtering algorithm and information granulation method based on support vector machine are used to predict and analyze the structural response of Guanhe Bridge,and the structural safety condition under the future vehicle load of Guanhe Bridge is evaluated.（6）The safety prognosis method based on multivariate mixed model is developed by combing the validated FEM and the data-driven model.The main conclusions contain:（1）A reliable temperature load model can be obtained by using the sum of the daily temperature difference and the daily mean value expressed by the Fourier series.Based on the generalized extreme value distribution theory,the monthly vehicle load model of the Guanhe Bridge is established.The average value of the vehicle load model is roughly 0.52 times of the load effect of highway I vehicles.The final monthly vehicle load model of Guanhe bridge based on Gray neural network prediction method is 0.49 times of the load effect of highway I vehicles.（2）The finite element model（FEM）updating and validation method for long-span cable-stayed bridge based on Kriging response surface method is established.The maximum error between the frequencies obtained from updated FEM and the measured frequencies is less than4%,and the Modal Assurance Criteria（MAC）of the corresponding mode shapes is above 95%.The maximum error of the frequencies based on the validated FEM of the Guanhe Bridge and measured ones does not exceed 7%.Therefore,this validated finite element model of Guanhe bridge can be further adopted for its safety prognosis.（3）The reliability analysis method of long-span cable-stayed bridge based on Kriging and important sampling method（AK-IS）is established,and the results are verified by the method which combines the least squares support vector machine and the important sampling method.If the maximum vertical displacement of Guanhe Bridge in the vehicle load is set to be 0.3m,the failure probability is 2.1663×10^-9,and the corresponding reliability is 5.871.It indicats that Guanhe Bridge is very safe under the vehicle load level.（4）The fatigue remaining service life prediction method for long-span cable-stayed bridges based on actual vehicle load is established.Under the current load level,the fatigue failure probability of Guanhe Bridge is 3.6900×10^-5 and the corresponding reliability is 3.9637,which is larger than the target reliability 3.5.In this condition,the remaining service life is 139years.When the growth rate of the traffic increases to 1%,the fatigue reliability of Guanhe Bridge is lower than the target reliability,and the fatigue remaining life is only 93 years.（5）The safety prognosis method based on data-driven model for long-span cable-stayed bridges is proposed.The results show that the dynamic linear equations of strain/vertical displacement in the middle of Guanhe Bridge under vehicle load based on Gaussian sum particle filtering method are reliable.According to the comparision of the prognosis of strain/vertical displacement in the middle of Guanhe Bridge under vehicle load/cable force of Guanhe Bridge and the corresponding usual value,the safety condition of Guanhe Bridge is good during the forecast period.（6）The safety prognosis method based on multivariate mixed model model for long-span cable-stayed bridges is established.It can be shown from the results that the predicted values of strain/vertical displacement in the middle of Guanhe Bridge under vehicle load based on data-driven model and multi-mixed model are in good agreement.According to the comparision of the prognosis of strain/vertical displacement in the middle of Guanhe Bridge under vehicle load and the corresponding usual value,Guanhe Bridge is safe during the forecast period.