A review on the applications of intelligence methods in predicting thermal conductivity of nanofluids

Intelligence methods, including Artificial Neural Networks (ANNs) and Support Vector Machine, are among the popular approaches for modeling the engineering systems with high accuracy. Nanofluid's thermal conductivity depends on several factors such as the dimensions of nanoparticles, their concentration, synthesis method and temperature. Intelligence methods are appropriate tools to precisely estimate nanofluids' thermal conductivity. Different methods and structures are used for the modeling of this property. In the present article, the related studies, using intelligence methods in thermal conductivity estimation, are comprehensively reviewed. According to the literature review, the accuracy of the predictive models has an association with their structure, utilized functions, selected input variables and employed algorithm. For instance, compared with mathematical correlations, obtained by curve fitting, ANNs are more accurate. Moreover, it is concluded that the structure of the NN, including numbers of hidden layer and neurons, can noticeably influence their performance. In the reviewed articles, trial and error are performed to distinguish the most favorable structure of ANNs. Due to the dependency of the models on the input variable, considering all the factors affecting the nanofluid's thermal conductivity results in higher precision of the models.