Abstract Accurate estimation of viscosity, one of the most important thermo-physical properties of nanofluids, is essential in heat transfer fluid applications in many industries. In this paper, for the first time, the random forest (RF), a robust artificial intelligence method is utilized to accurately estimate the viscosity of Newtonian nanofluids. To develop the model five input parameters were used, namely the temperature, solid volume fraction, viscosity of the base fluid, nanoparticle size, and density of nanoparticle. Further, 2890 datasets were collected from 50 references representing a wide range of experimental settings. The model’s predictive performance was assessed against those of a multilayer perceptron (MLP) model, a support vector regression (SVR) and various classical and empirical models. The models’ performance were analyzed using various statistical performance indicators and graphical plots. Performance criteria assessment, using the testing dataset, showed that the RF model provided the best prediction of the viscosity of nanofluids (R ​= ​0.989, RMSE ​= ​0.139, MAPE ​= ​4.758%) in comparison to those of the MLP (R ​= ​0.915, RMSE ​= ​0.377, MPE ​= ​16.194%) and the SVR (R ​= ​0.941, RMSE ​= ​0.315, MAPE ​= ​7.895%). Moreover, a sensitivity analysis demonstrated that the volume fraction and density of nanoparticles were the most and second most significant factors affecting the viscosity of nanofluid, respectively.

中文翻译：

---

使用随机森林 (RF) 方法预测纳米流体粘度

摘要 粘度是纳米流体最重要的热物理特性之一，在许多行业的传热流体应用中，粘度的准确估计是必不可少的。在本文中，首次利用随机森林（RF）这一稳健的人工智能方法来准确估计牛顿纳米流体的粘度。为了开发该模型，使用了五个输入参数，即温度、固体体积分数、基液粘度、纳米颗粒尺寸和纳米颗粒密度。此外，从代表各种实验设置的 50 个参考文献中收集了 2890 个数据集。该模型的预测性能是根据多层感知器 (MLP) 模型、支持向量回归 (SVR) 以及各种经典和经验模型的预测性能进行评估的。使用各种统计性能指标和图形来分析模型的性能。使用测试数据集进行的性能标准评估表明，与 MLP 相比，RF 模型提供了对纳米流体粘度的最佳预测（R = 0.989, RMSE = 0.139, MAPE = 4.758%） (R = 0.915, RMSE = 0.377, MPE = 16.194%) 和 SVR (R = 0.941, RMSE = 0.315, MAPE = 7.895%)。此外，敏感性分析表明，纳米颗粒的体积分数和密度分别是影响纳米流体粘度的最重要和第二重要因素。表明与 MLP 模型相比，RF 模型对纳米流体的粘度（R = 0.989, RMSE = 0.139, MAPE = 4.758%）提供了最好的预测（R = 0.915, RMSE = 0.377, MPE = 16.194%) 和 SVR (R = 0.941, RMSE = 0.315, MAPE = 7.895%)。此外，敏感性分析表明，纳米颗粒的体积分数和密度分别是影响纳米流体粘度的最重要和第二重要因素。表明与 MLP 模型相比，RF 模型对纳米流体的粘度（R = 0.989, RMSE = 0.139, MAPE = 4.758%）提供了最好的预测（R = 0.915, RMSE = 0.377, MPE = 16.194%) 和 SVR (R = 0.941, RMSE = 0.315, MAPE = 7.895%)。此外，敏感性分析表明，纳米颗粒的体积分数和密度分别是影响纳米流体粘度的最重要和第二重要因素。