Scholarlink

<p>Additive manufacturing (AM) enables the creation of intricate microchannel heat sink designs, overcoming the geometric constraints of traditional fabrication methods. While numerous studies have explored nanofluid cooling and novel channel configurations, experimental investigations that directly couple AM-fabricated channels with nanofluids are limited, and few have extended this integration to predictive machine learning frameworks. This study integrates these aspects by experimentally evaluating two AlSi10Mg heat sinks fabricated via direct metal laser sintering: a straight microchannel (MC-1) and a wavy microchannel with secondary passages (MC-2) designed to enhance fluid mixing. Their thermo-hydraulic behaviour was evaluated using water-ethylene glycol and TiO<sub>2</sub> nanofluids under controlled heat flux levels of 20–40 W/cm<sup>2</sup>. The wavy MC-2 configuration reduced surface temperature and achieved Nusselt numbers up to 29.5% higher than MC-1, owing to geometry-induced secondary flows. The addition of TiO<sub>2</sub> nanofluids further improved heat transfer; however, it also increased the viscosity and pressure drop, resulting in performance evaluation factors that remained below unity. To extend predictive capability, machine learning (ML) models were trained on the experimental dataset, with Gradient Boosting Regression yielding excellent accuracy (<i>R</i><sup>2</sup> &gt; 0.998) for both surface temperature and Nusselt number. The results indicate that AM-enabled geometrical modifications are the dominant factor in enhancing heat transfer, nanofluids act as supplementary enhancers with hydraulic trade-offs, and machine learning provides a reliable framework for predictive optimization. Together, this hybrid experimental-computational approach offers a validated framework for designing next-generation cooling systems.</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

Thermal performance and flow behaviour of nanofluid-cooled additively manufactured microchannels: an experimental and machine learning approach

  • 期刊论文

  • 发表日期：

    2026年2月11日

查看全文
  • K. Vijetha,
  • Dumpala Lingaraju,
  • Satish Geeri

摘要

Additive manufacturing (AM) enables the creation of intricate microchannel heat sink designs, overcoming the geometric constraints of traditional fabrication methods. While numerous studies have explored nanofluid cooling and novel channel configurations, experimental investigations that directly couple AM-fabricated channels with nanofluids are limited, and few have extended this integration to predictive machine learning frameworks. This study integrates these aspects by experimentally evaluating two AlSi10Mg heat sinks fabricated via direct metal laser sintering: a straight microchannel (MC-1) and a wavy microchannel with secondary passages (MC-2) designed to enhance fluid mixing. Their thermo-hydraulic behaviour was evaluated using water-ethylene glycol and TiO2 nanofluids under controlled heat flux levels of 20–40 W/cm2. The wavy MC-2 configuration reduced surface temperature and achieved Nusselt numbers up to 29.5% higher than MC-1, owing to geometry-induced secondary flows. The addition of TiO2 nanofluids further improved heat transfer; however, it also increased the viscosity and pressure drop, resulting in performance evaluation factors that remained below unity. To extend predictive capability, machine learning (ML) models were trained on the experimental dataset, with Gradient Boosting Regression yielding excellent accuracy (R2 > 0.998) for both surface temperature and Nusselt number. The results indicate that AM-enabled geometrical modifications are the dominant factor in enhancing heat transfer, nanofluids act as supplementary enhancers with hydraulic trade-offs, and machine learning provides a reliable framework for predictive optimization. Together, this hybrid experimental-computational approach offers a validated framework for designing next-generation cooling systems.