Document Type : Research Paper

Authors

• Kiyoumars Roushangar ¹
• Reza Saadatjoo ²
• Hamidreza Abbaszadeh ³
• Aydin Panahi ³

¹ Department of water engineering ,faculty of Civil Engineering, University of Tabriz, Tabriz, Iran

² M.Sc. Student, Department of Civil Engineering, Faculty of Civil Engineering, University of Tabriz, Tabriz, Iran

³ Department of Civil Engineering, Faculty of Civil Engineering, University of Tabriz, Tabriz, Iran

Abstract

The high flow velocity and the pressure reduction in the spillways cause damage to the spillways. In the present research, the application of Gaussian Process Regression (GPR) and Support Vector Machine (SVM) was investigated for predicting air concentration in stepped spillways. For this purpose, a comprehensive set of experimental data obtained from hydraulic models of stepped spillways in the modeling process was utilized. Input models were defined based on various combinations of measured parameters. In predicting the air concentration in the stepped spillway under natural aeration conditions, parameters od discharge (qw), the ratio of flow depth (normal to spillway step) to channel width (Z/W), the ratio of longitudinal distance from the beginning of the step to the length of the step (x/L), and the ratio of distance from the midpoint line of the spillway step to the step width (Y=2y/w) had a significant impact. The results obtained demonstrate the high capability of both methods in estimating the required air concentration on spillways. The results revealed that the Radial Basis Function (RBF) kernel performs favorable results. The R2, DC and RMSE for the GPR were 0.79, 0.79, and 0.12, respectively, and in the SVM were 0.86, 0.86, and 0.098, respectively.

Keywords

• Gaussian Process Regression
• Sensitivity analysis
• Stepped spillway
• Support Vector Machine

Main Subjects

• Hydraulic and river engineering