Purpose

Convection is one of the main heat transfer mechanisms in both high to low temperature media. The accurate convection heat transfer coefficient (HTC) value is required for exact prediction of heat transfer. As convection HTC depends on many variables including fluid properties, flow hydrodynamics, surface geometry and operating and boundary conditions, among others, its accurate estimation is often too hard. Homogeneous dispersion of nanoparticles in a base fluid (nanofluids) that found high popularities during the past two decades has also increased the level of this complexity. Therefore, this study aims to show the application of least-square support vector machines (LS-SVM) for prediction of convection heat transfer coefficient of nanofluids through circular pipes as an accurate alternative way and draw a clear path for future researches in the field.

Design/methodology/approach

The proposed LS-SVM model is developed using a relatively huge databank, including 253 experimental data sets. The predictive performance of this intelligent approach is validated using both experimental data and empirical correlations in the literature.

Findings

The results show that the LS-SVM paradigm with a radial basis kernel outperforms all other considered approaches. It presents an absolute average relative deviation of 2.47% and the regression coefficient (R²) of 0.99935 for the estimation of the experimental databank. The proposed smart paradigm expedites the procedure of estimation of convection HTC of nanofluid flow inside circular pipes.

Originality/value

Therefore, the focus of the current study is concentrated on the estimation of convection HTC of nanofluid flow through circular pipes using the LS-SVM. Indeed, this estimation is done using operating conditions and some simply measured characteristics of nanoparticle, base fluid and nanofluid.

中文翻译：

---

支持向量机在纳米流体通过圆管对流换热系数精确预测中的应用

目的

对流是高温到低温介质中的主要传热机制之一。准确预测传热需要准确的对流传热系数 (HTC) 值。由于对流 HTC 取决于许多变量，包括流体特性、流动流体动力学、表面几何形状以及操作和边界条件等，因此其准确估计通常太难了。在过去二十年中广受欢迎的基液（纳米流体）中纳米粒子的均匀分散也增加了这种复杂性。所以，

设计/方法/方法

所提出的 LS-SVM 模型是使用相对庞大的数据库开发的，包括 253 个实验数据集。使用实验数据和文献中的经验相关性验证了这种智能方法的预测性能。

发现

结果表明，具有径向基核的 LS-SVM 范式优于所有其他考虑的方法。它提供了 2.47% 的绝对平均相对偏差和0.99935的回归系数 ( R ² )，用于估计实验数据库。所提出的智能范式加快了圆形管道内纳米流体流动的对流 HTC 的估计过程。

原创性/价值

因此，当前研究的重点集中在使用 LS-SVM 估计通过圆形管道的纳米流体流动的对流 HTC 上。事实上，这种估计是使用操作条件和一些简单测量的纳米颗粒、基液和纳米流体的特性来完成的。