Predicting the deformation of roller compacted concrete dam using least squares support vector machine

As a new type of dam, roller compacted concrete dam (RCCD) develops very fast in recent years. Deformation plays an important role in the RCCD’s safety. The deformation of RCCD is influenced mainly by three parts: water pressure, temperature and time effect. As to any of the three parts, there are many factors. Therefore, the deformation of RCCD is a complicated system. The least squares support vector machine algorithm is used to predict the deformation of RCCD. The applicability of this algorithm is illustrated with Jinanqiao dam, which is located in the southwest of China. There are totally 80 deformation monitoring data from November 11, 2010 to April 24, 2012. The least squares support vector machine is firstly trained with 70 data, and then the deformation prediction results are given by the trained least squares support vector machine. The other 10 monitoring data are used to test the prediction efficiency of the least squares support vector machine. The performance of the least squares support vector machine is compared with the neural network algorithm. The prediction results indicate that the least squares support vector machine is superior to the neural network, and the mean absolute percentage error MAPE of the least squares support vector machine is only13.12% while that of the neural network is 20.78%. The prediction values of the least squares support vector machine fit the monitoring data with a high accuracy. Thus the least squares support vector machine algorithm has great potential to prediction of deformation of RCCD.