Scholarlink

<p>This study presents the predicted thermal properties of ternary nanofluids for optimising their performance in heat transfer applications. While single- and hybrid nanofluids have received considerable attention, ternary nanofluids present unique compositional complexities that require advanced predictive models. This study developed a new ensemble machine learning approach to estimate the thermal conductivity of ZnO-TiO₂-SiC/DW-EG ternary nanofluids. Using a two-step method, the study prepared ternary nanofluids with nanoparticle mixing ratios of 0.1:0.6:0.3 in a 40:60 distilled water-ethylene glycol base at volume concentrations of 0.01–0.13 vol%. Nanofluid stability was enhanced by mechanical stirring, ultrasonication, and the addition of 1% SDBS surfactant. Thermal conductivity was measured over a temperature range of 30&#xa0;°C to 70&#xa0;°C. Significantly, thermal conductivity increased by up to 117.93% with temperature and 46.48% with concentration. The addition of SDBS surfactant enhanced thermal conductivity by up to 14.28%, demonstrating the crucial role of surfactants in nanofluid performance. The ANN model was developed using a multi-layer perceptron with a feed-forward back-propagation mechanism, employing the Levenberg-Marquardt training algorithm. The optimal architecture featured 20 neurons in a single hidden layer for thermal conductivity prediction. A bagging ensemble method was applied by training 50 individual ANN models with different bootstrap samples and aggregating their predictions. The proposed ensemble ANN model (r² = 0.9955, MSE = 3.07 × 10⁻⁵) surpassed both correlation equations and ensemble SVM approaches with very small sensitivity limits (-5.651% to 7.225%). This result represents a major step forward in accurately predicting the thermal properties of complex nanofluids, particularly for cutting operations, heat exchangers, and thermal management systems, where ideal design requires these properties to be accurately estimated.</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 thermal conductivity prediction accuracy in ternary nanofluids through ensemble machine learning methods

  • 期刊论文

  • 发表日期：

    2026年6月19日

查看全文
  • Md Mustafizur Rahman,
  • K. Karidgama,
  • M. Samykano,
  • M. M. Noor,
  • D. Ramasamy

摘要

This study presents the predicted thermal properties of ternary nanofluids for optimising their performance in heat transfer applications. While single- and hybrid nanofluids have received considerable attention, ternary nanofluids present unique compositional complexities that require advanced predictive models. This study developed a new ensemble machine learning approach to estimate the thermal conductivity of ZnO-TiO₂-SiC/DW-EG ternary nanofluids. Using a two-step method, the study prepared ternary nanofluids with nanoparticle mixing ratios of 0.1:0.6:0.3 in a 40:60 distilled water-ethylene glycol base at volume concentrations of 0.01–0.13 vol%. Nanofluid stability was enhanced by mechanical stirring, ultrasonication, and the addition of 1% SDBS surfactant. Thermal conductivity was measured over a temperature range of 30 °C to 70 °C. Significantly, thermal conductivity increased by up to 117.93% with temperature and 46.48% with concentration. The addition of SDBS surfactant enhanced thermal conductivity by up to 14.28%, demonstrating the crucial role of surfactants in nanofluid performance. The ANN model was developed using a multi-layer perceptron with a feed-forward back-propagation mechanism, employing the Levenberg-Marquardt training algorithm. The optimal architecture featured 20 neurons in a single hidden layer for thermal conductivity prediction. A bagging ensemble method was applied by training 50 individual ANN models with different bootstrap samples and aggregating their predictions. The proposed ensemble ANN model (r² = 0.9955, MSE = 3.07 × 10⁻⁵) surpassed both correlation equations and ensemble SVM approaches with very small sensitivity limits (-5.651% to 7.225%). This result represents a major step forward in accurately predicting the thermal properties of complex nanofluids, particularly for cutting operations, heat exchangers, and thermal management systems, where ideal design requires these properties to be accurately estimated.