Engineering systems are increasingly monitored using sensor networks, motivating the need for predictive models that can consistently integrate observational data. Conventional finite element methods (FEM) rely on homogeneous material models and deterministic boundary conditions, and their predictions often deviate from observed system responses. Statistical finite element methods (statFEM) address this limitation by learning model–data discrepancies within a Bayesian framework and have primarily been developed for static problems [1,2]. In this work, we extend the statFEM framework to elastodynamics. Material heterogeneity is incorporated explicitly through spatially correlated random fields, while remaining model discrepancies are represented as Gaussian white-noise forcing. A stochastic dynamic forward model is developed that propagates both displacement and velocity states, accounting for uncertainty arising from material variability and stochastic loading. Based on this formulation, we derive prediction and filtering equations for sequential state estimation using a Bayesian filtering approach [3]. The material property field is treated as a time-invariant but unknown state and is inferred alongside the dynamic response through an augmented state formulation. The resulting framework enables simultaneous updating of displacement, velocity, material properties and discrepancy parameters in a unified probabilistic setting. The proposed methodology is demonstrated on one- and two-dimensional elastodynamic problems, including a bar under harmonic loading and an anti-plane shear problem with spatially varying material properties. The results show accurate state estimation, meaningful recovery of material fields from sparse measurements, and robust identification of stochastic forcing characteristics, highlighting the potential of statFEM for data-informed dynamic modeling of engineering systems. References: 1. Girolami M., Febrianto E., Yin G. and Cirak F., The statistical finite element method (statFEM) for coherent synthesis of observation data and model predictions, Computer Methods in Applied Mechanics and Engineering, 375, 113533, 2021. 2. Febrianto E., Butler L., Girolami M. and Cirak F., Digital twinning of self-sensing structures using the statistical finite element method, Data-Centric Engineering, 3, e31, 2022. 3. Särkkä, Simo, and Lennart Svensson. Bayesian filtering and smoothing. Vol. 17. Cambridge University Press, 2023.