Scholarlink

<p>This study addresses the challenge of accurately predicting the density of liquid siloxanes, which are widely used as working fluids in systems designed for recovering waste heat from low-temperature and minor-scale origins such as solar energy, biomass, geothermal energy, and turbine exhaust. Accurate modeling of transport properties, including density, is essential for improving the performance of these systems. In this research, several machine learning algorithms were created and assessed to predict siloxane density based on temperature, pressure, boiling point, and molar mass. The methods include decision trees, ensemble methods, boosting techniques, random forest, convolutional neural networks, support vector machines, k-nearest neighbors, and multilayer perceptron artificial neural networks. A sensitivity assessment was performed to evaluate the influence of each input variable, and an outlier detection method was applied to enhance data reliability. The multilayer perceptron model outperformed all other algorithms, achieving a coefficient of determination (R<sup>2</sup>) of 0.982 and an average absolute relative error of 0.58% on the testing dataset. Furthermore, temperature exhibited a negative correlation with density, while pressure was identified as the most influential factor.</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

Accurate Estimation of Density of Liquid Siloxanes Via Robust Machine Learning Methods

  • 期刊论文

  • 发表日期：

    2025年8月15日

查看全文
  • Huan Liu,
  • Ayat Hussein Adhab,
  • G. PadmaPriya,
  • Anupam Yadav,
  • Aditya Kashyap,
  • Kattela Chennakesavulu,
  • Biswaranjan Mohanty,
  • Manish Srivastava,
  • Morug Salih Mahdi,
  • Shirin Shomurotova,
  • Aseel Salah Mansoor,
  • Usama Kadem Radi,
  • Nasr Saadoun Abd,
  • Samim Sherzod

摘要

This study addresses the challenge of accurately predicting the density of liquid siloxanes, which are widely used as working fluids in systems designed for recovering waste heat from low-temperature and minor-scale origins such as solar energy, biomass, geothermal energy, and turbine exhaust. Accurate modeling of transport properties, including density, is essential for improving the performance of these systems. In this research, several machine learning algorithms were created and assessed to predict siloxane density based on temperature, pressure, boiling point, and molar mass. The methods include decision trees, ensemble methods, boosting techniques, random forest, convolutional neural networks, support vector machines, k-nearest neighbors, and multilayer perceptron artificial neural networks. A sensitivity assessment was performed to evaluate the influence of each input variable, and an outlier detection method was applied to enhance data reliability. The multilayer perceptron model outperformed all other algorithms, achieving a coefficient of determination (R2) of 0.982 and an average absolute relative error of 0.58% on the testing dataset. Furthermore, temperature exhibited a negative correlation with density, while pressure was identified as the most influential factor.