Seite - 912 - in Book of Full Papers - Symposium Hydro Engineering

It is shown in Fig. 2 that the displacement prediction line of the PCA-RBF neural network model is more close to the measured displacement line, and the prediction effect is better than that of the statistical model. As shown in Table 3, the RMSE and MAE of PCA-RBF neural network at training stage are 0.288 and 0.201 respectively, and are 0.702 and 0.548 at the prediction stage respectively, which are all smaller than those of statistical models. The R2 of the PCA- RBF neural network model at the training stage and the prediction stage are 0.959 and 0.910 respectively, which are both larger than those of the statistical models. In addition, with the comparison of the residual results of Fig. 3, the prediction effect of the PCA-RBF neural network model is more accurate. Fig. 2 Comparison of training and prediction effect Fig. 3 Comparison of the residual 912