<p>To solve the problem that traditional inversion methods cannot estimate the variation trend of parameters over time, a Bayesian updating back analysis model is proposed. On the basis of displacement and prior distribution of material parameters, updating back analyses of material parameters are performed. The deformation of concrete arch dams is calculated via a stacked ensemble model. Given the dam displacement and prior distribution on different dates, the posterior distribution information for the dam material parameters is obtained, and the evolution process of the material parameters is tracked. With a concrete arch dam as an example, the surrogate model used in this paper yields better prediction performance than a single model. The R² between the predictions of the surrogate model and the finite element results is 0.981, the MAE is 0.912, the MAPE is 1.48, and the RMSE is 0.981. The comprehensive elastic modulus of the dam body and bedrock obtained via the Bayesian updating back analysis method has an R² above 0.95 when establishing a mixed model of radial displacement and measurement of the dam, demonstrating excellent predictive performance.</p>