Abstract The present study investigated the flow discharge coefficient (Cdt) in the combined rectangular broad crested weir-gate structure. To this end, the effect of the following dimensionless parameters on the Cdt were investigated: the width ratio of the central weir to the width of the total structure (B/Bo), the height ratio of the central weir to the height of the central weir floor (Z/P), the ratio of the gate width to the width of the total structure (b/Bo), the ratio of the gate opening height to the height of the central weir floor (d/P), and the ratio of the head on central weir to the total head behind the structure (h1/H). The Flow-3D numerical model, artificial intelligence models such as linear multilayer perceptron (MLP), Canfis network (CNN), recurrent network (RNN), modular neural network (MNN), and regression equation, were used to estimate the Cdt. The results indicated that increasing d/P and b/Bo ratios led to a decline in this coefficient. In the case of h1/H ≤ 0.4, an increase in B/Bo ratio resulted in decreasing the turbulence intensity and Cdt while the impact of enhancing the size of B/Bo was not significant if h1/H > 0.4. Besides, increasing Z/P ratio caused an increase in resistance against the flow and thus a decline in Cdt. Further, the results of artificial intelligence models and regression equation demonstrated that the MNN model with an RMSE and R2 of 0.03 and 0.97, respectively, could have an accurate estimate of the Cdt values.

中文翻译：

---

组合堰闸结构中的流量系数

摘要 本研究研究了组合矩形宽顶堰闸结构中的流量系数（Cdt）。为此，研究了以下无量纲参数对 Cdt 的影响：中央堰宽与总结构宽度比（B/Bo）、中央堰高与中央高堰底(Z/P)、闸门宽度与总结构宽度之比(b/Bo)、闸门开口高度与中央堰底高度之比(d/P)、中央堰上水头与结构后面总水头的比值 (h1/H)。Flow-3D数值模型、线性多层感知器（MLP）、Canfis网络（CNN）、循环网络（RNN）、模块化神经网络（MNN）、回归方程等人工智能模型，用于估计 Cdt。结果表明，增加 d/P 和 b/Bo 比率导致该系数下降。在 h1/H ≤ 0.4 的情况下，B/Bo 比的增加导致湍流强度和 Cdt 降低，而当 h1/H > 0.4 时，B/Bo 尺寸增大的影响不显着。此外，增加 Z/P 比会导致流动阻力增加，从而导致 Cdt 下降。此外，人工智能模型和回归方程的结果表明，RMSE 和 R2 分别为 0.03 和 0.97 的 MNN 模型可以准确估计 Cdt 值。B/Bo 比的增加导致湍流强度和 Cdt 降低，而如果 h1/H > 0.4，B/Bo 大小的增加影响不显着。此外，增加 Z/P 比会导致流动阻力增加，从而导致 Cdt 下降。此外，人工智能模型和回归方程的结果表明，RMSE 和 R2 分别为 0.03 和 0.97 的 MNN 模型可以准确估计 Cdt 值。B/Bo 比的增加导致湍流强度和 Cdt 降低，而如果 h1/H > 0.4，B/Bo 大小的增加影响不显着。此外，增加 Z/P 比会导致流动阻力增加，从而导致 Cdt 下降。此外，人工智能模型和回归方程的结果表明，RMSE 和 R2 分别为 0.03 和 0.97 的 MNN 模型可以准确估计 Cdt 值。