Scholarlink

<p>Compared to conventional solid weirs, porous/gabion weirs offer advantages in ecological effects, hydraulic efficiency, and structural integrity, with scientists and engineers paying more attention to them. However, the possibility of sediment buildup upstream, which could result in partial blockages, is a major problem with these designs. This study used a machine learning model of CatBoost to estimate the discharge coefficient for partially blocked porous broad-crested weirs. The Catboost model's effectiveness was evaluated compared to standard nonlinear regression (SNR) techniques and gene-expression programming (GEP). 140 experimental data points from the literature were used in this investigation. With root mean square errors of 0.002 and 0.003 and coefficients of determination of 0.999 and 0.996 for the training and testing datasets, the CatBoost model outperformed the alternatives, according to the findings. This model accurately predicted all data, with an error margin of less than 3%. The other models were ranked in order of accuracy: GEP and SNR.</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

Multi-approaches Evaluation for Prediction of Discharge Coefficient of Porous Broad-Crested Weirs Under Upstream Partial Blockage

  • 期刊论文

  • 发表日期：

    2025年2月7日

查看全文
  • Sanaz Hasanian Shirvan,
  • Bahareh Pirzadeh,
  • Seyed Hosein Rajaei,
  • Mahmood Shafai Bejestan

摘要

Compared to conventional solid weirs, porous/gabion weirs offer advantages in ecological effects, hydraulic efficiency, and structural integrity, with scientists and engineers paying more attention to them. However, the possibility of sediment buildup upstream, which could result in partial blockages, is a major problem with these designs. This study used a machine learning model of CatBoost to estimate the discharge coefficient for partially blocked porous broad-crested weirs. The Catboost model's effectiveness was evaluated compared to standard nonlinear regression (SNR) techniques and gene-expression programming (GEP). 140 experimental data points from the literature were used in this investigation. With root mean square errors of 0.002 and 0.003 and coefficients of determination of 0.999 and 0.996 for the training and testing datasets, the CatBoost model outperformed the alternatives, according to the findings. This model accurately predicted all data, with an error margin of less than 3%. The other models were ranked in order of accuracy: GEP and SNR.