Scholarlink

Abstract <p>This research is concentrated on analyzing the discharge coefficient (<i>C</i><sub><i>d</i></sub>) of labyrinth gates with varying cycles and internal angles, utilizing soft computing techniques (SVM, ANN and MARS models). Through the analysis, significant variables, called cycle numbers (<i>N</i>), upstream depth to the gate opening (<i>H</i>/<i>G</i>), and internal angle (θ) were determined to predict the <i>C</i><sub><i>d</i></sub>. Data partitioning (210) involved allocating 70% for training and the rest for testing across all models. Results revealed that within the SVM model, the Radial Basis Function (RBF) kernel emerged as the most effective predictor of <i>C</i><sub><i>d</i></sub>, surpassing Linear, Polynomial, and Sigmoid kernel functions. Similarly, in the ANN, the Multilayer Perceptron (MLP) network exhibited greater precision compared to the RBF. Also, the findings highlighted MARS model’s equation as a significant contributor to high prediction accuracy, although its accuracy is low compared to other models. Following the investigations, the ANN-MLP model emerged as the most promising candidate, demonstrating best results compared to other models. Notably, the ANN-MLP (test results) model achieved notable results with an <i>R</i><sup>2</sup> value of 0.983, an RMSE of 0.028, MRE% of 2.62%, and KGE of 0.977.</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

Soft Computing and Predicting Labyrinth Gates Discharge Coefficient

  • 期刊论文

  • 发表日期：

    2025年7月19日

查看全文
  • Hamidreza Abbaszadeh,
  • Reza Tarinejad

摘要

Abstract

This research is concentrated on analyzing the discharge coefficient (Cd) of labyrinth gates with varying cycles and internal angles, utilizing soft computing techniques (SVM, ANN and MARS models). Through the analysis, significant variables, called cycle numbers (N), upstream depth to the gate opening (H/G), and internal angle (θ) were determined to predict the Cd. Data partitioning (210) involved allocating 70% for training and the rest for testing across all models. Results revealed that within the SVM model, the Radial Basis Function (RBF) kernel emerged as the most effective predictor of Cd, surpassing Linear, Polynomial, and Sigmoid kernel functions. Similarly, in the ANN, the Multilayer Perceptron (MLP) network exhibited greater precision compared to the RBF. Also, the findings highlighted MARS model’s equation as a significant contributor to high prediction accuracy, although its accuracy is low compared to other models. Following the investigations, the ANN-MLP model emerged as the most promising candidate, demonstrating best results compared to other models. Notably, the ANN-MLP (test results) model achieved notable results with an R2 value of 0.983, an RMSE of 0.028, MRE% of 2.62%, and KGE of 0.977.