Abstract This study presents numerical simulations, back propagation artificial neural networks and genetic algorithms for optimizing laminar convective heat transfer of oil-in-water nanoemulsion fluids having non-Fourier heat conduction characteristics. Firstly, a numerical study has been conducted on laminar flow and forced convective heat transfer of oil-in-water nanoemulsion fluids in toroidal ducts using Eulerian-Lagrangian two-phase approach. New correlations of drag coefficient, effective thermal conductivity and effective viscosity were adopted to improve the accuracy of simulation. Numerical results show that convective heat transfer can be enhanced by oil nanodroplets with thermal conductivity lower than that of the base fluid. Then regression models and artificial neural network models were developed based on simulation results for predicting convective heat transfer performances of nanoemulsions, considering effects of cross-sectional aspect ratio, Reynolds number, oil nanodroplet diameter and concentration. Artificial neural network models can predict mean Nusselt number and pressure drop better than the regression model. Finally, genetic algorithms was used to optimize convective heat transfer of nanoemulsions considering droplet migration. It can be found that low cross-sectional aspect ratio of width to height is beneficial for thermal performance factor. For single-objective optimization, mean Nusselt number reaches the maximum 32.3 at aspect ratio of 0.9677 and thermal performance factor reaches the maximum 1.305 at aspect ratio of 0.3935 under certain conditions. Pareto optimal set was obtained for two-objective optimization. This study would be useful for the optimal design of convective heat transfer of emulsions in toroidal ducts.

中文翻译：

---

水包油纳米乳液在环形管道中层流对流传热的优化

摘要 本研究介绍了数值模拟、反向传播人工神经网络和遗传算法，用于优化具有非傅立叶热传导特性的水包油纳米乳液流体的层流对流换热。首先，使用欧拉-拉格朗日两相方法对环形管道中水包油纳米乳液流体的层流和强制对流传热进行了数值研究。采用新的阻力系数、有效热导率和有效粘度的相关性来提高模拟的准确性。数值结果表明，导热系数低于基液的纳米油滴可以增强对流传热。然后，基于模拟结果开发回归模型和人工神经网络模型，用于预测纳米乳液的对流传热性能，考虑横截面纵横比、雷诺数、油纳米液滴直径和浓度的影响。人工神经网络模型可以比回归模型更好地预测平均努塞尔数和压降。最后，考虑到液滴迁移，遗传算法被用于优化纳米乳液的对流传热。可以发现，低横截面宽高比有利于热性能因素。对于单目标优化，平均努塞尔数在 0.9677 纵横比时达到最大值 32.3，热性能因子在纵横比为 0 时达到最大值 1.305。3935 在某些条件下。两目标优化得到帕累托最优集。这项研究将有助于优化设计环形管道中乳液的对流传热。