Erdoğan Beytullah, Güneş Abdulsamed, Kılıç İrfan, Yaman Orhan
Department of Mechanical Engineering, Zonguldak Bülent Ecevit University, Zonguldak 67100, Turkey.
Department of Electric and Energy, Firat University, Elazığ 23119, Turkey.
Micromachines (Basel). 2025 Apr 26;16(5):504. doi: 10.3390/mi16050504.
Machining processes often face challenges such as elevated temperatures and wear, which traditional cutting fluids are insufficient to address. As a result, solutions involving nanoparticle additives are being explored to enhance cooling and lubrication performance. This study investigates the effect of thermal conductivity, an important property influenced by the densities of mono and hybrid nanofluids. To this end, various nanofluids were prepared by incorporating hexagonal boron nitride (hBN), zinc oxide (ZnO), multi-walled carbon nanotubes (MWCNTs), titanium dioxide (TiO), and aluminum oxide (AlO) nanoparticles into sunflower oil as the base fluid. Hybrid nanofluids were created by combining two nanoparticles, including ZnO + MWCNT, hBN + MWCNT, hBN + ZnO, hBN + TiO, hBN + AlO, and TiO + AlO. A dataset consisting of 180 data points was generated by measuring the thermal conductivity and density of the prepared nanofluids at various temperatures (30-70 °C) in a laboratory setting. Conducting thermal conductivity measurements across different temperature ranges presents significant challenges, requiring considerable time and resources, and often resulting in high costs and potential inaccuracies. To address these issues, a feedforward artificial neural network (FFANN) method was proposed to predict thermal conductivity. Our multilayer FFANN model takes as input the temperature of the experimental environment where the measurement is made, the measured thermal conductivity of the relevant nanoparticle, and the relative density of the nanoparticle. The FFANN model predicts the thermal conductivity value linearly as output. The model demonstrated high predictive accuracy, with a reliability of R = 0.99628 and a coefficient of determination (R) of 0.9999. The average mean absolute error (MAE) for all hybrid nanofluids was 0.001, and the mean squared error (MSE) was 1.76 × 10. The proposed FFANN model provides a State-of-the-Art approach for predicting thermal conductivity, offering valuable insights into selecting optimal hybrid nanofluids based on thermal conductivity values and nanoparticle density.
加工过程常常面临诸如温度升高和磨损等挑战，而传统切削液不足以应对这些挑战。因此，人们正在探索涉及纳米颗粒添加剂的解决方案，以提高冷却和润滑性能。本研究调查了热导率的影响，热导率是一种受单组分和混合纳米流体密度影响的重要特性。为此，通过将六方氮化硼（hBN）、氧化锌（ZnO）、多壁碳纳米管（MWCNTs）、二氧化钛（TiO）和氧化铝（AlO）纳米颗粒加入葵花籽油作为基础流体，制备了各种纳米流体。混合纳米流体是通过将两种纳米颗粒组合而成的，包括ZnO + MWCNT、hBN + MWCNT、hBN + ZnO、hBN + TiO、hBN + AlO和TiO + AlO。通过在实验室环境中测量制备的纳米流体在不同温度（30 - 70°C）下的热导率和密度，生成了一个由180个数据点组成的数据集。在不同温度范围内进行热导率测量面临重大挑战，需要大量时间和资源，并且常常导致高成本和潜在的不准确。为了解决这些问题，提出了一种前馈人工神经网络（FFANN）方法来预测热导率。我们的多层FFANN模型将进行测量的实验环境温度、相关纳米颗粒的实测热导率以及纳米颗粒的相对密度作为输入。FFANN模型以线性方式预测热导率值作为输出。该模型显示出高预测准确性，可靠性R = 0.99628，决定系数（R）为0.9999。所有混合纳米流体的平均平均绝对误差（MAE）为0.001，均方误差（MSE）为1.76×10。所提出的FFANN模型为预测热导率提供了一种先进方法，为基于热导率值和纳米颗粒密度选择最佳混合纳米流体提供了有价值的见解。
