Abstract

Sediment transport in open channels has complicated nature and finding the analytical models applicable for channel design in practice is a quite difficult task. To this end, behind theoretical consideration of the open channel sediment transport through incorporating of four fundamental characteristics of fluid, flow, sediment and channel, recently, machine learning techniques are used for modeling of sediment transport in open channels. However, most of the studies in the literature used limited number of data for model development neglecting some effective parameters involved which may affect their performances. Moreover, most of this studies had not provided a comprehensive explicit equation for future use. Accordingly, this study applied four machine learning techniques of Gene Expression Programming (GEP), Extreme Learning Machine (ELM), Generalized Structure Group Method of Data Handling (GS-GMDH) and Fuzzy c-means based Adaptive Neuro-Fuzzy Inference System (FCM-ANFIS) to model sediment transport in open channels. Four existing data sets in the literature with wide ranges of pipe size, sediment size, sediment volumetric concentration, channel bed slope and flow depth are used for the model development. The recommended models are compared with their corresponding conventional regression models taken from the literature in terms of different statistical performance indices. Results indicate superiority of the machine leaning techniques to the conventional multiple non-linear regression models. Although, developed GEP, ELM, GS-GMDH and FCM-ANFIS models have almost same performances, GS-GMDH gives slightly better performance which can be linked to the generalized structure of this approach. A MATLAB code is provided to calculate the sediment transport in open channel for practical engineering.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call