|
Abstract
|
A labyrinth weir is a polygonal overflow weir that is characterized by its hydraulic performance and distinct geometric shape (triangular, trapezoidal or rectangular cycles). The present study provides new knowledge and design information on the performance and operation of adaptive neuro-fuzzy inference system (ANFIS) and gene expression programming (GEP) techniques for predicting discharge coefficient of labyrinth weirs in a flume. Discharge coefficient, Cd, was considered as a function of headwater ratio, upstream Froude number, magnification ratio, sidewall angle, apex ratio, and cycle width ratio (dimensionless parameters). The records obtained through 300 laboratory data sets were used to determine the Cd of labyrinth weirs. The results showed that the best GEP model determined the Cd values of the normal and inverted orientation labyrinth weir with a RMSE of 0.0139, 0.590, DC of 0.977, 0.852 and R of 0.996, 0.923, respectively by using the four dimensionless parameters of Fr, HT/P, Lc/W and A/w as input variables. The best ANFIS model determined the Cd value of the normal orientation labyrinth weir by using all six dimensionless parameters (RMSE = 0.0110, DC = 0.964, R = 0.987) and inverted orientation labyrinth weir by using three dimensionless parameters of HT/P, α, and w/p as input variables (RMSE = 0.0290, DC = 0.952, R = 0.987). The results of sensitivity analysis showed that Fr in the GEP model and Fr and HT/P in the ANFIS model are the most effective variables for determining Cd for normal and inverted orientation labyrinth weirs, respectively. It was found that when low headwater ratios were eliminated from analysis, performance accuracy of the models was improved
|