Stochastic Inversion of Soil Dynamic Parameters Based on Non-Intrusive Data
Yanmei Cao,Zhaoyang Li,Jialiang Chen,Chaoyi Xia
DOI: https://doi.org/10.1016/j.soildyn.2024.108640
IF: 4
2024-01-01
Soil Dynamics and Earthquake Engineering
Abstract:The prediction and evaluation of environmental vibrations generated by traffic loads have been playing more and more important role during the planning and designing of high-quality traffic system. From the traffic source to sensitive building, the subsoil is the one of key propagation media, so the determination of soil dynamic parameters is very crucial in studying the transmitting and attenuation of environmental vibration. Due to the spatial variability of natural soil layers, the soil dynamic parameters are actually not certain but probabilistic with stochastic characteristics. However, most current inversion methods of soil parameters cannot consider the uncertainty. In this paper, an efficient inversion method is proposed to obtain the random distribution of soil dynamic parameters based on the combination of field experiment, forward soil modelling, and Bayesian theory. Take the shear modulus and damping ratio of subsoil as the typical parameters, the experimental dispersion curve and spatial attenuation curve can be obtained by field experiment, while the corresponding theoretical ones can be calculated by the thin-layered soil model with perfectly matched layer. The combining of experimental and theoretical curves generates the likelihood function. A prior probability distribution model is firstly established, and subsequently the posterior probability distribution model is deduced by genetic algorithm, Monte Carlo-Markov chain algorithm, and Bayesian theory. In order to verify the proposed inversion method, an on-site experiment based on Multichannel Analysis of Surface Waves(MASW) was conducted in an open field in Beijing. The results show the mean values of inversed soil parameters were close to the determined parameters provided by geological exploration. Compared to the traditional determined inversion, the proposed stochastic inversion method can consider the variety of soil parameters within small depth by providing probability distribution, mean and root-mean-square error, which has greater applicability in the further probabilistic prediction of environmental vibration.