Prediction of Total and bedform Roughness Coefficient in Alluvial Channels Based on Experimental Data via Gaussian Process Regression Method

Understanding the flow and turbulence characteristics in movable open channels with bedforms is of substantial importance for the management of rivers as well as design and operation of hydraulic structures. Dunes are one of the most important bedforms and have significant impact on the characteristics of turbulence. So far numerous studies have been down about hydraulic resistance in open channels with bedforms, however, due to the impact of various parameters on the roughness coefficient, the exact estimation of this parameter is difficult. In this research, using the data of experiments carried out at the Hydraulic Laboratory of University of Tabriz with two different gradation and two channel widths, flow resistance due to bedform was studied. Also, by combining these data with laboratory data from other researchers, using Gaussian Process Regression (GPR) different models were defined and investigated. The obtained results from the experiments showed that in investigating the effect of hydraulic parameters on flow resistance, Reynolds number showed a better correlation with flow resistance in comparison with other hydraulic parameters. Also, the obtained results from the developed models proved desired capability of GPR method in predicting roughness coefficient and it was observed that both flow and sediment particles characteristics are effective in estimating roughness coefficient. The results of the superior model sensitivity analysis showed that the Reynolds number has the most significant impact in predicting the roughness coefficient.