<p>Increasing the energy dissipation in stepped weirs requires more study to lower the risk for hydraulic structures. This research investigates stepped weirs with various geometries under flat, fully pooled, and zigzag pooled conditions. Simulations were performed using FLOW-3D with flow-rates of 0.007–0.025&#xa0;m<sup>3</sup>/s and nested mesh sizes of 0.01 and 0.005&#xa0;m. The results indicated that RNG turbulence model with Root Mean Square Error (RMSE) = 0.011 has higher accuracy compared to k–ε, k–ω and LES. Fully pooled steps dissipate more energy than flat and zigzag pooled steps, with a 5% increase when a barrier is applied at the step’s end. Artificial neural network and support vector machine analyses were used to predict the energy loss along with a non-linear regression equation. Statistical analysis (R = Correlation Coefficient, RMSE, KGE = Kling Gupta Efficiency, and Avg RE = Average Relative Error) indicated that the Radial Basis Function (RBF) kernel (R = 0.971, RMSE = 1.17%, Avg RE = 1.12% and KGE = 0.865) outperformed linear, polynomial, and sigmoid. Among ANN models, the Multi-Layer Perceptron (MLP) network demonstrated higher accuracy (R = 0.962, RMSE = 0.99%, Avg RE = 0.92% and KGE = 0.954) than the RBF. The results indicated that the non-linear regression equation has a good overlap with the observation results with R = 0.91, RMSE = 1.91%, Avg RE = 1.94%, and KGE = 0.809. This research contributes to the sustainable design of hydraulic structures by improving energy dissipation methods, thereby reducing erosion risks, protecting aquatic ecosystems, and enhancing the resilience of water management infrastructure in the face of increasing hydrological challenges.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Enhancing energy dissipation in stepped weirs: numerical analysis and machine learning of ANN, SVM and non-linear regression predictions

  • Parisa Ebadzadeh,
  • Hamidreza Abbaszadeh,
  • Rasoul Daneshfaraz,
  • John Abraham

摘要

Increasing the energy dissipation in stepped weirs requires more study to lower the risk for hydraulic structures. This research investigates stepped weirs with various geometries under flat, fully pooled, and zigzag pooled conditions. Simulations were performed using FLOW-3D with flow-rates of 0.007–0.025 m3/s and nested mesh sizes of 0.01 and 0.005 m. The results indicated that RNG turbulence model with Root Mean Square Error (RMSE) = 0.011 has higher accuracy compared to k–ε, k–ω and LES. Fully pooled steps dissipate more energy than flat and zigzag pooled steps, with a 5% increase when a barrier is applied at the step’s end. Artificial neural network and support vector machine analyses were used to predict the energy loss along with a non-linear regression equation. Statistical analysis (R = Correlation Coefficient, RMSE, KGE = Kling Gupta Efficiency, and Avg RE = Average Relative Error) indicated that the Radial Basis Function (RBF) kernel (R = 0.971, RMSE = 1.17%, Avg RE = 1.12% and KGE = 0.865) outperformed linear, polynomial, and sigmoid. Among ANN models, the Multi-Layer Perceptron (MLP) network demonstrated higher accuracy (R = 0.962, RMSE = 0.99%, Avg RE = 0.92% and KGE = 0.954) than the RBF. The results indicated that the non-linear regression equation has a good overlap with the observation results with R = 0.91, RMSE = 1.91%, Avg RE = 1.94%, and KGE = 0.809. This research contributes to the sustainable design of hydraulic structures by improving energy dissipation methods, thereby reducing erosion risks, protecting aquatic ecosystems, and enhancing the resilience of water management infrastructure in the face of increasing hydrological challenges.