Scholarlink

<p>The non-biodegradability and inherent toxicity of petroleum-based lubricants have created a strong need for environment friendly bio-based alternatives. Biolubricants enriched with nano-additives serve as a highly effective and sustainable alternative to conventional mineral oils. In this work, a set of bio-oil based mono and hybrid nanolubricants were prepared. Nanoparticles of CuO were used to develop mono nanolubricant and CuO along with MWCNT was utilized for synthesizing hybrid nanolubricant. Nanoparticles were added in a range of 0.02%–0.25% volume fraction in canola oil. Viscosity was experimentally measured over a temperature range of 25°–80&#xa0;°C. A total of 116 experimental data points was generated. The experimental data were used to develop different machine learning models, namely, random forest, linear and polynomial regression models. With a higher <i>R</i><sup><i>2</i></sup> (≥ 0.98) and prediction error less than 2% for unseen data, random forest model showcased its superior predictive capability.</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

Experimental Investigation and Machine Learning Based Prediction of Viscosity of Bio-Based Mono and Hybrid Nanolubricants

  • 期刊论文

  • 发表日期：

    2026年5月19日

查看全文
  • Santwana Mishra,
  • Soumya Mathur,
  • Vradhi Aggarwal,
  • Srishti Gupta,
  • Sneha Goswami,
  • Shipra Aggarwal

摘要

The non-biodegradability and inherent toxicity of petroleum-based lubricants have created a strong need for environment friendly bio-based alternatives. Biolubricants enriched with nano-additives serve as a highly effective and sustainable alternative to conventional mineral oils. In this work, a set of bio-oil based mono and hybrid nanolubricants were prepared. Nanoparticles of CuO were used to develop mono nanolubricant and CuO along with MWCNT was utilized for synthesizing hybrid nanolubricant. Nanoparticles were added in a range of 0.02%–0.25% volume fraction in canola oil. Viscosity was experimentally measured over a temperature range of 25°–80 °C. A total of 116 experimental data points was generated. The experimental data were used to develop different machine learning models, namely, random forest, linear and polynomial regression models. With a higher R2 (≥ 0.98) and prediction error less than 2% for unseen data, random forest model showcased its superior predictive capability.