PSI - Issue 17

Ahmed Belmokre et al. / Procedia Structural Integrity 17 (2019) 698–703 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

702

5

(b)

Fig. 3. RFR and SVR models prediction for drains: (a) FL-175 (b) FR-175

We try to measure the importance of input variables by applying equation (4). The values of N tree and m try are fixed at 200 and 2 respectively. Results allows us to identify which inputs dominate the flow rate. Fig. 4 shows that water temperature is the most important parameter; this can be explained by the impact of temperature on the physical properties of water. The viscosity of water depends on its temperature as described by equation (6). Additionally, the hydraulic conductivity of concrete is inversely proportional to the viscosity of the fluid (Philip 1957). ( ) 2 0.01775 / 1 0.033 0.000221 T T  = + + (6)

/ K k  =

(7)

(a)

(b)

Fig. 4. Variable importance measures for drains: (a) FL-175; (b) FR-175

4. Conclusion

In this paper, we apply random forest regression and support vector regression models to predict seepage flow rates in an Algerian roller compact concrete gravity dam. Water level variation, water temperature and time effect are the inputs to the models. The performance of both models is assessed through the mean square error and the mean absolute error. The predicted results suggest that random forest regression model provide better performance than the support vector regression model.