Abstract: Concrete arch dams, as important hydraulic structures, frequently have missing measurement data due to monitoring equipment failures, human factors and other influences, which may reduce the effectiveness and accuracy of dam safety assessment and prediction. Previous methods mostly rely on single-point interpolation, neglecting the correlation and heterogeneity between measurement points. This paper develops a new method for interpolating the missing deformation data based on a panel data model. First, the incremental speed index of traditional deformation similarity is improved to solve the problem that its denominator may be equal to zero. Then, a combined weighting method is formulated to calculate a composite deformation similarity indicator, and an improved density-based clustering method is used to categorize the deformation monitoring points. Next, a panel model is developed to fill in the missing data in different intervals of the data sequence. This new method fills in the missing monitored deformation data of concrete arch dams more accurately, and thus can effectively solve the missing data problem.

Key words: missing data imputation, deformation similarity index, clustering method, panel data model, concrete arch dams