LXGB: a machine learning algorithm for estimating the discharge coefficient of pseudo-cosine labyrinth weir

Area of Science:

• Hydraulic Engineering
• Computational Fluid Dynamics
• Water Resource Management

Background:

• Weir efficiency is often improved by modifying plan geometry and increasing weir length.
• The discharge coefficient (Cd) is a key parameter for weir performance.
• Traditional methods may not fully capture complex hydraulic behaviors.

Purpose of the Study:

• Introduce a novel pseudo-cosine labyrinth weir (PCLW) design.
• Develop and evaluate a hybrid machine learning algorithm (LXGB) for estimating the discharge coefficient (Cd) of PCLWs.
• Identify optimal input parameters for accurate Cd estimation.

Main Methods:

• A hybrid LXGB algorithm, combining LSHADE and XGBoost, was employed.
• Seven input scenarios were tested using 132 data series from PCLW1 and PCLW2 models.
• Performance was evaluated using RMSE, RRMSE, and NSE metrics.

Main Results:

• The LXGB model achieved high accuracy, with average RMSE = 0.009, RRMSE = 0.010, and NSE = 0.977.
• The most effective input parameters were R/W, L/W, and H/W.
• The LXGB model demonstrated superior performance compared to SAELM, ANFIS-FFA, GEP, and ANN.

Conclusions:

• The proposed PCLW design offers a practical solution for enhancing weir efficiency.
• The LXGB hybrid intelligent approach is highly effective for estimating the discharge coefficient of PCLW weirs.
• Geometric and hydraulic ratios are crucial predictors for discharge coefficient estimation.