Abstract: Based on Bayesian theory, dynamic Bayesian identification method of Winkler subgrade parameter is put forward. The differential equations of Mindlin plate on Winkler subgrade are derived by means of Mindlin theory as well as the governing equations of the plate which allow the transverse shear deformation to be existed. Through applying Fourier transformative technology, the corresponding solutions for the pinned plate on Winkler subgrade are obtained. Dynamic Bayesian error function of Winkler subgrade parameter is established on the first time. The formulas of dynamic Bayesian expectation and variance are deduced. After the approach of automatic step-length search is developed, the identification computing steps are given by adapting conjugate gradient method. Research shows that Winkler subgrade parameter can be efficiently identified by applying dynamic Bayesian identification method and the convergence property of Winkler subgrade parameter is depended on both the precision of pre-known information of Winkler subgrade parameter and that of the measured displacement data at the interested nodes. And this dynamic identification method can also be applied to the problem of identification for other kinds of subgrade parameters.