The present study investigates the feasibility of predicting the thermal properties of MgO/water nanofluid using artificial neural networks (ANNs). To design the ANN, Reynolds number and volume fraction of nanoparticles ($\phi$) were considered as ANN inputs and in principle independent parameters and on the other hand, the parameters of relative pressure drop and relative heat transfer coefficient were considered as the outputs of this ANN. One of the important innovations in the present study is the attempt to simultaneously predict the relative pressure drop and the relative heat transfer coefficient, the first of which is known as an undesirable parameter and the second as a desirable parameter. The designed ANN using the radial basis function (RBF) was able to predict the laboratory parameters of relative pressure drop and relative heat transfer coefficient (RHTC) with 99.3% and 99.5% accuracy, respectively. It can drastically reduce the time and financial costs of laboratory methods and minimize the need for them. Based on the obtained results in $\phi$ = 0.125%, the unfavorable parameter (relative pressure drop) has a significant advantage over the optimal parameter (RHTC) and therefore the $\phi$is not suitable for use in thermal cycles. But on the other hand, with $\phi$ = 0.5%, the optimal parameter (RHTC) has a significant advantage over the undesirable parameter (relative pressure drop), and therefore for MgO/water nanofluid, $\phi$>0.5% are suitable and good efficiency will have. Also, $\phi$ = 0.5%, the best returns were reported for Re = 5000 to 14000.

本研究调查了使用人工神经网络 (ANN) 预测 MgO/水纳米流体的热性能的可行性。为了设计人工神经网络，纳米粒子的雷诺数和体积分数（$\phi$) 被认为是 ANN 的输入，原则上是独立的参数，另一方面，相对压降和相对传热系数的参数被认为是该 ANN 的输出。本研究的一个重要创新是尝试同时预测相对压降和相对传热系数，其中第一个被称为不希望的参数，第二个被称为理想的参数。使用径向基函数 (RBF) 设计的 ANN 能够分别以 99.3% 和 99.5% 的准确度预测相对压降和相对传热系数 (RHTC) 的实验室参数。它可以大大减少实验室方法的时间和财务成本，并最大限度地减少对它们的需求。根据获得的结果$\phi$ = 0.125%，不利参数（相对压降）比最佳参数（RHTC）有显着优势，因此 $\phi$不适用于热循环。但另一方面，随着$\phi$ = 0.5%，最佳参数 (RHTC) 比不希望的参数（相对压降）具有显着优势，因此对于 MgO/水纳米流体， $\phi$> 0.5% 是合适的，并且会有良好的效率。还，$\phi$ = 0.5%，报告了 Re = 5000 到 14000 的最佳回报。