Identification of rail-sleeper-ballast system through time-domain Markov chain Monte Carlo-based Bayesian approach

This paper reports the step-by-step procedures for identification of the rail-sleeper-ballast system with the use of the measured vibration data of an in situ sleeper on an existing ballasted track. The rail-sleeper-ballast modeling method, which has been used for modal-based model updating, was used to fit the measured time-domain vibration from the field test. However, the match between the measured and model-predicted responses was not good at some measured locations. Based on the observed discrepancy, the rail-sleeper-ballast modeling method was modified in this paper for suitable use in time-domain model updating. Based on the field test data and the modified modeling method, this study puts forward the time-domain Markov chain Monte Carlo (MCMC)-based Bayesian model updating and model class selection method for identification of the rail-sleeper-ballast system. MCMC was used to ensure that the proposed method can be applied even when the problem is unidentifiable. The proposed method identified the distribution of railway ballast stiffness under low-amplitude vibration and the “equivalent” rail stiffness and mass using impact hammer test data. The model updating results confirmed that the ballast stiffness under the sleeper was uniform, which implies that there was no ballast damage under the tested sleeper. Based on the proposed method, a comprehensive study was carried out to quantify the posterior uncertainties of the identified ballast stiffness when different amounts of measured information were used for model updating. The results showed that the uncertainty of the identified ballast stiffness was at an acceptable level even when using the measured data from only one sensor.