Concrete arch dams are three-dimensional structures which are statically indeterminate due to integrity and arching performance. Hence, the spatial and temporal temperature gradients in concrete arch dams affect the volume of the structures and generated internal stresses threaten stability of the structures. Accordingly, estimation of long-term thermal behavior of these structures for proper serviceability with considering dam crest displacement is necessary, and this issue requires the application of appropriate prediction models. The goal of this study is to implement the support vector regression (SVR) and artificial neural network (ANN) models for prediction of the intermediate block displacement of the dam crest. For this purpose, displacement of dam crest is investigated with ABAQUS simulation model over a period of 8 years, and then, the results of the simulation are used in the soft models (SVR and ANN) as the input data. The analysis of the results of two models with five error indicators shows that the error reduction in the SVR model is about 32% less than the ANN model in the testing stage. Also, investigation of the normal cumulative probability distribution related to the outputs of two models indicates high degree of deviation on cumulative probability distribution of the ANN model. This is due to the fact that the ANN model ignores fundamental errors in the training process. Therefore, based on the SVR model one can predict the dam stability in an acceptable accuracy range, only by measuring two different parameters including reservoir water level and the air temperature.
