An Assessment of Extreme Learning Machine Model for Estimation of Flow Variables in Curved Irrigation Channels

The bend existence in river and artificial channels irrigation channel is an imminent act. So, the hydraulic flow variables investigation using recording laboratory instruments and computational models has considerable importance. In the present paper, the velocity and flow depth variables in 60° open channel bend are measured in laboratory flume by sensor instrumentation. Furthermore, the main objective of the current study is to evaluate the performance of the ELM model because of its significant advantages such as high learning speed. Accordingly, a robust Extreme Learning Machine (ELM) model is designed and trained based on available experimental data. The present paper is the first application of the ELM model in estimating flow variables in curved channels. The results show that the variables measurement instrumentations act more accurately and the laboratory model can predict bend flow patterns well. Furthermore, the ELM model can predict the flow velocity and depth with low error indices and has an acceptable agreement level with experimental values (Mean Absolute Relative Error (MARE) = 0.020 and 0.034 in depth and velocity prediction model, respectively). Both the laboratory and computational ELM models have good efficiency in different passing flow discharges. The proposed ELM model can be used in the implementation and design of a curved channel in practical cases. Also, the high accuracy measurement instruments can be applied in measuring and controlling flow variables in various laboratory fields.