Abstract: During the operation of a concrete dam, various uncertainties－such as sudden events, natural disasters, and changes in human management－are possible to impose an impact on it, potentially deviating its structure deformation from the conventional patterns. An accurate identification of such changes is crucial for raising the level of concrete dam warning and forecasting. This paper presents an intelligent method for identifying dam deformation under uncertainties. First, we use a spatial clustering method to categorize measurement points that are located in different regions of the concrete dam structure but share certain similarity. Then, a fuzzy clustering (Gath-Geva) algorithm is used to segment a multivariate time series into different phases, allowing its data points to belong to multiple periods based on the membership degree, to measure the homogeneity of segments and detect changes in its hidden structure. Last, we use a fuzzy decision algorithm based on the cluster compatibility criteria to determine the number of segments required, and adopts the principal component analysis (PCA) to identify the number of principal components, further improving the accuracy of the Gath-Geva algorithm. This intelligent method has been applied in a case study of a concrete arch dam structure to identify the changes hidden in the time series of its displacement measurements. Comparison of its results with those of single-period data shows that it is effective in extracting sudden anomalous changes during the operational phase of the dam, and that it is a valuable approach for assessing the operational conditions of concrete dams.

Key words: hydro-engineering, dam safety monitoring, Gath-Geva, time series analysis, statistical model