Abstract: The monitored displacements of an arch dam body are decomposed into two parts: one caused by dam body loading and the other by the constrained effect of dam foundation. Then, a set of equations that take the minimum systematic error of the monitored and predicted values as the objective function, can be established for the two unknowns－constrained foundation displacement and dam body elastic modulus. And they are solved simultaneously from a large body of monitored displacement data combined with PSO to achieve data-driven effect. A case study of a dam body on a complicated dam foundation shows that relative to FEM calculations, the dam body displacements calculated using this new method have relative errors in the range of 1.5%, with the exception of the nodes (with 3% errors) near its interface with the foundation, proving our method is effective and accurate. Application to the Baihetan arch dam reveals that the displacement calculations feature a reasonable spatial pattern of relative errors, and the deformation separation results are close to the real case, which validates our new displacement model.

Key words: data-driven, deformation separation, arch dam, safety monitoring, inversion