Abstract: A model of support vector machine for classification (C-SVC) and its application to the evaluation on crack damages of earth and rock-fill dams are systemically described herein. An analytic hierarchy process was used to obtain six training factors of the damage, i.e. dam type, dam height, storage capacity, construction quality, seismic intensity, and epicentral distance, and the training tab, i.e. dam crack. A C-SVC evaluation system was developed by determining the factors and tab in a certain way from qualitative to the quantitative. This system uses the field data of 100 earth and rock-fill dams damaged by the Wenchuan earthquake as the training set, and adopts optimal normalization preprocessing obtained through a comparison of different classification accuracies in predictions by different normalizations and the best kernel function selected by a K-fold cross-validation technique. The results show that the normalized preprocessing of [-1,1] and the polynomial kernel function are applicable to prediction of dam crack. Application of the system to 28 reservoirs in Mianyang city shows that the predictions for 25 of them are in good agreement with in-situ observations. This indicates that the C-SVC evaluation model can provide useful information for reinforcement of earth and rock-fill dams.