Novel Grain and Form Roughness Estimator Scheme Incorporating Artificial Intelligence Models

Determination of flow resistance in open channel flows is not only important for practical engineering applications but also challenging because of multiple factors involved. The literature review reveals that despite of various data-driven formulas and schemes, only classic Manning's resistance equation and Keulegan's formula have been utilized in practice. It also indicates that sole application of Artificial Intelligence (AI) models facilitates roughness estimation while they have not been used within a systematic roughness estimator scheme. In this study, a new eight-step scheme is developed to predict grain and total Manning's coefficients when grain and form roughness are the major sources of friction, respectively. The new scheme not only uses a new explicit equation for computing hydraulic radius related to bed for estimating grain roughness coefficient but also utilizes AI models named artificial neural network and genetic programming in the seventh step for estimating form roughness coefficient. It improves R-2 for estimating Manning's grain coefficient and RMSE for estimating discharge by 21% and 64% comparing with that of one of common formulas available in the literature, respectively. Moreover, the new scheme incorporating AI models significantly enhances the accuracy of estimation results for predicting roughness coefficient and discharge comparing with the new scheme using new developed empirical formula based on RMSE, MARE and R-2 criteria. The obtained improvement demonstrates that application of AI models as a part of a data-based roughness estimator scheme, like the one suggested, may considerably improve the precision of prediction results of flow resistance and discharge.