Document Type : Technical Note
In recent decades, various formulas have been developed to achieve a better evaluation of the characteristics of rivers and natural channels. In this article we compare some of well-known classical methods in estimating stable sections are compared. Also two non-linear functions have been used to design the width and depth of canals and sand bed rivers with the formulations which have been obtained from 70 natural canals in North-America. Genetic Programming (GP) and Group Method of Data Handling (GMDH), two methods in soft computing family, have been used in this research.
Finally these methods have been compared with three general methods in designing stable canals. In this modeling, in addition to flow parameters like discharge (Q), sediment quality as average bed particles diameters (D50), and channel slope (S) have been used. The results show that GMDH is more accurate than other methods. Finally sensitivity analysis has shown the effect of each parameter used in the best model.