Estimation of Roughness Coefficient in Erodible Channels by ANNs and the ANFIS Methods

Estimating the roughness of erodible open channels plays an important role in their hydraulic design. This parameter is important for the development of numerical models and hydraulic design of these erodible channels. For this reason, several empirical methods have been presented so far to estimate the roughness coefficient while the previous study shows that these methods are not sufficiently accurate. These methods usually are based on empirical activities which are too time-consuming and expensive. Therefore, in this paper, the so-called Artificial Neural Networks (ANNs) and Adaptive Network Based Fuzzy Inference System (ANFIS) methods as soft computing methods are used to estimate the roughness coefficient in erodible open channels. To achieve this firstly, it is attempted effective parameter on the coefficient are extracted based on empirical methods and a dimensional analysis. Then effectiveness of the parameters on the coefficient is investigated via a sensitivity analysis versus the error of estimation. Following to the development of the models, they are implemented to estimate the coefficient. Based on the method none-dimensional water depth, Sheilds number, shear Reynolds number and none dimensional falling velocity are determined as input parameters of soft-computing models. Final results show that the employed methods are more accurate than empirical methods to estimate the parameter and these methods can be used as an alternative method for the estimation of parameters. In addition, the effectiveness of some other parameters such as shear Reynolds number and none dimensional water depth over the event are extracted by the sensitivity analysis.