Abstract

This study aimed to investigate the accuracy of the artificial neural network in estimating thermal conductivity (k) of ferrofluid-based nanofluids. The parameters of ${\mathrm{k}}_{{\mathrm{ND}+\mathrm{Fe}}_2{\mathrm{O}}_3/\mathrm{EG}-\mathrm{water}}$ and ${\mathrm{k}}_{\mathrm{EG}-\mathrm{water}}$ have been measured at 20–60$°\mathrm{C},$ 0.05, 0.1, and 0.2 vol.% and the results showed that ${\mathrm{k}}_{{\mathrm{Fe}}_2{\mathrm{O}}_3/\mathrm{EG}-\mathrm{water}}$ was greater than ${\mathrm{k}}_{\mathrm{EG}-\mathrm{water}}$ by 89%, which is obtained at 60$°\mathrm{C}$ and 0.2 vol.%. To estimate ${\mathrm{k}}_{{\mathrm{ND}+\mathrm{}\mathrm{Fe}}_3{\mathrm{O}}_4/\mathrm{EG}-\mathrm{water}}$ a three-layer ANN was developed that contained two, three, and one neurons, respectively. This neural network was able to estimate ${\mathrm{k}}_{{\mathrm{ND}+\mathrm{}\mathrm{Fe}}_3{\mathrm{O}}_4/\mathrm{EG}-\mathrm{water}}$ with less than 0.8% error considering of ${\mathrm{R}}^2=0.996.$ The response surface methodology was also implemented, and it was observed that cubic polynomials, taking to account of ${\mathrm{R}}^2=0.994,$ will figure out the best results so that ${\mathrm{k}}_{{\mathrm{ND}+\mathrm{}\mathrm{Fe}}_3{\mathrm{O}}_4/\mathrm{EG}-\mathrm{water}}$ can be predicted with an error of less than 0.5%.

摘要

本研究旨在研究人工神经网络在估算基于铁磁流体的纳米流体的热导率 (k) 时的准确性。的参数${\mathrm{k}}_{{\mathrm{无损检测}+\mathrm{铁}}_{\mathrm{2个}}{\mathrm{欧}}_{\mathrm{3个}}/\mathrm{例如}-\mathrm{水}}$和${\mathrm{k}}_{\mathrm{例如}-\mathrm{水}}$已被测量为 20–60$°\mathrm{C},$0.05、0.1 和 0.2 vol.%，结果表明${\mathrm{k}}_{{\mathrm{铁}}_{\mathrm{2个}}{\mathrm{欧}}_{\mathrm{3个}}/\mathrm{例如}-\mathrm{水}}$大于${\mathrm{k}}_{\mathrm{例如}-\mathrm{水}}$89%，这是在 60 时获得的$°\mathrm{C}$和 0.2 体积％。估计${\mathrm{k}}_{{\mathrm{无损检测}+\mathrm{}\mathrm{铁}}_{\mathrm{3个}}{\mathrm{欧}}_{\mathrm{4个}}/\mathrm{例如}-\mathrm{水}}$开发了一个三层 ANN，分别包含两个、三个和一个神经元。这个神经网络能够估计${\mathrm{k}}_{{\mathrm{无损检测}+\mathrm{}\mathrm{铁}}_{\mathrm{3个}}{\mathrm{欧}}_{\mathrm{4个}}/\mathrm{例如}-\mathrm{水}}$考虑到误差小于 0.8%${\mathrm{R}}^{\mathrm{2个}}=0.996.$还实施了响应面方法，观察到三次多项式，考虑到${\mathrm{R}}^{\mathrm{2个}}=0.994,$会找出最好的结果，这样${\mathrm{k}}_{{\mathrm{无损检测}+\mathrm{}\mathrm{铁}}_{\mathrm{3个}}{\mathrm{欧}}_{\mathrm{4个}}/\mathrm{例如}-\mathrm{水}}$可以预测误差小于0.5%。