Hojjat Emami

Weirs are important hydraulic structures widely used to control the flow rates of open channels and rivers. The discharge coefficient is a vital parameter in computing flow discharge over weirs. In this work, we introduce the Walnut algorithm, a new nature-inspired optimization strategy. Then combine it with support vector regression (SVR) to predict the discharge coefficient parameter of triangular labyrinth weirs. The proposed Walnut-SVR method takes as input a set of observations characterized by five non-dimensional features and attempts to find the discharge coefficient of unseen records. We propose the Walnut algorithm to feature selection and find optimum values for SVR's parameters. The proposed method is evaluated employing the Kumar dataset and compared with several counterpart methods. The results show the superiority of the Walnut-SVR method compared to other counterparts with $R^2 =0.986$, $RMSE= 0.004$, $SI=0.006$, $\sigma=0.858$, and $NSE=0.981$ on test dataset. Feature analysis shows that the proposed method obtained the best results when three geometric parameters, the ratio of the weir crest length to the weir height ($L/w$), the ratio of head over the crest to the weir height ($h/w$), and the vertex angle ($\theta$) are used.