Void Identification Of The Cement-emulsified Asphalt Mortar Layer Of The Slab Track System Based On Time-domain Bayesian Method

For the concealment of the CA mortar and the difficulty in detecting void on the CA mortar layer of the slab track,a void identification method of CA mortar layer of slab track structure was proposed based on the measured vibration data by using Bayesian model updating and model class selection approach.The void identification methods of CA mortar layer based on Laplace’s method for asymptotic approximation and Markov chain Monte Carlo Bayesian are systematically studied via numerical analysis and experimental verifications,which lays a theoretical foundation for the application of the undeterministic damage identification method in the field of high-speed railway.The main research contents and contributions are as follows:(1)The Bayesian model class selection algorithm for the beam-type structure was extended,and a new two-phase model class selection algorithm was developed for the plate-type structure.(2)A void identification method of CA mortar layer based on Laplace’s method for asymptotic approximation is adopted in this thesis.Firstly,by identifying the most probable values of the mortar stiffness scaling factors of CA mortar layer,the void location and severity of CA mortar are identified.In addition,the uncertainties of the model parameters are evaluated by using the covariance matrix obtained from the Hessian matrix.This thesis utilizes the elastic modulus reduction method to simulate the CA void,and comprehensive numerical analysis are carried out on different damage cases.The results show that the proposed method can successfully identify the CA mortar void.In addition,experimental verification of the proposed void identification method using the measured vibration data from the laboratory model was carried out in this thesis,and the posterior uncertainties of the model parameters under different number of sensors are quantitatively studied.The results show that the uncertainty level of the identification results was kept at an acceptable level even in the situation with only one sensor(installed near the impact location).(3)Then a void identification method of CA mortar layer based on Markov chain Monte Carlo Bayesian framework is adopted in this thesis for possible unidentifiable case.The discrete samples of the model parameters loacated in the importance region was obtained by following Markov chain Monte Carlo Bayesian model updating,and the damage probability of the void severity in each region of CA mortar layer was calculated,and the void location and severity of CA mortar layer can be identified by using the cumulative distribution function of the void severity.In addition,the posterior probability density function of the model parameters was obtained by the kernel density to quantitatively evaluate the uncertainties of model parameters.