Research On Fatigue Reliability Evaluation Method Of Orthotropic Steel Bridge Deck Based On Improved Active Learning

Orthotropic steel bridge deck(OSD)has good development potential and broad application prospects.However,due to the coupling action of multiple factors such as load,material,structure and environment,the fatigue cracking phenomenon has significant randomness,which is essentially a reliability problem.Scientific and accurate evaluation of fatigue reliability is essential for evaluating the performance of OSD,and can provide an important reference for the design and manufacture of new bridges and the maintenance and reinforcement of existing bridges.However,the problem of fatigue reliability evaluation of orthotropic bridge decks is very complicated,and there are the following difficulties:(i)How to accurately describe the external load effect by using random process theory;(ii)Complex fatigue function functions generally have implicit characteristics,and traditional reliability evaluation methods are inefficient;(iii)The fatigue reliability of the structure in the early stage of operation is generally a problem of small failure probability,and the traditional reliability evaluation method has a large amount of calculation.Therefore,it is still a huge challenge for the academic and engineering circles to efficiently and accurately assess the fatigue reliability of OSD.Aiming at the shortcomings of existing research,systematic research is carried out on key issues such as fatigue load,crack growth simulation and reliability theory.Introduce the basic concepts and framework of the active learning surrogate model.The fatigue stress spectrum simulation and fatigue reliability evaluation methods based on probabilistic fracture mechanics are studied.And it is applied to the evaluation of fatigue reliability of structural details.The main research contents and conclusions are as follows:(1)First,through the segmental finite element model and moving load step loading,the stress response characteristics of the fatigue details of the first domestic road-to-rail steel box girder orthotropic bridge deck on the same layer are studied.It is found that the cantilever steel box girder increases the lateral influence range of the structural details due to the overall bending and shear effect of the beam.Therefore,the effect of multi-lane loading needs to be considered when analyzing and evaluating its fatigue performance.Then a full-scale segmental test model is designed and a fatigue model test is carried out.The test results verify the validity and accuracy of the segmental finite element analysis.(2)The fatigue stress spectrum under random traffic flow provides strong technical support and theoretical basis for reliability assessment.At present,the random traffic flow simulation results based on inverse functions and Monte Carlo sampling have a limited scope of application and have certain deviations from real traffic.Therefore,based on the DRAM sampling method,a random traffic flow simulation program under complex probability distribution parameters is developed.This method can traverse the entire space,has high sampling efficiency and accuracy,and has a wide range of applications,which can be better applied to the simulation of random traffic flow.Combined with random traffic flow and stress response characteristics,a fatigue stress spectrum simulation method based on measured random traffic flow data is constructed,and the effect of fatigue load is statistically analyzed.(3)The theory of fracture mechanics can reasonably describe the crack expansion law and fatigue failure mechanism,and is widely used in fatigue reliability assessment.However,the stress intensity factor needs to be solved using numerical methods(i.e.,the functional function has implicit characteristics),which leads to the barrier of computational inefficiency in fatigue reliability assessment using probabilistic fracture mechanics.Therefore,this Dissertation derives a new active learning function JSD based on the active learning mechanism,adaptively selects training sample points to update the Kriging surrogate model,and introduces a termination criterion based on the uncertainty function,proposing a new reliability analysis based on the active learning Kriging model and Monte Carlo Simulation(AK-MCS-JSD).The JSD not only ensures that the selected training sampling points are close to the limit state boundary region,but also has a large dispersion,so that AK-MCS-JSD can effectively reduce the number of calls to complex functional functions,thus improving the efficiency of the active learning reliability analysis method.The fatigue reliability of detail of OSD was evaluated using the AK-MCS-JSD by combining the fatigue stress spectra,and a universal fatigue reliability evaluation framework based on improved active learning and probabilistic fracture mechanics was formed.(4)In order to ensure the safety and durability of the structure,the structural design has a high safety margin,which results in a small fatigue failure probability of the structural details of the orthotropic bridge deck in the early stage of operation,which is a typical small failure probability problem.In order to accurately evaluate the fatigue reliability of the orthotropic bridge deck with the characteristics of small failure probability in the early use,this Dissertation derives the importance probability density function of multiple failure points based on the Gaussian mixture model on the basis of the importance sampling method(Importance Sampling,IS)Combining the concept of variable credibility agency model,a reliability analysis based on Active learning Kriging and IS(AK-IS-JSD)is proposed.This method can be effectively applied to the evaluation of the problem of small failure probability and the reliability of multiple failure mode systems.Compared with Monte Carlo simulation,the number of candidate sample groups is greatly reduced.Finally,the fatigue reliability of OSD at the early stage of operation of each construction detail was assessed using the AK-ISJSD method in conjunction with the fatigue stress spectrum simulation results,forming a framework for fatigue reliability assessment of small failure probabilities based on AK-ISJSD and probabilistic fracture mechanics.Finally,this article is summarized,and the next research work is prospected.