Machine learning-based approaches for modeling thermophysical properties of hybrid nanofluids: A comprehensive review

Abstract

Abstract Thermophysical properties of hybrid nanofluids remarkably affect their behavior in engineering systems. Among these properties, dynamic viscosity and thermal conductivity are more crucial in thermal sciences. In recent years, several models have been introduced based on intelligence methods for predicting these properties of the hybrid nanofluids. Confidence and accuracy of these models are influenced by the modeling algorithm used, the data implemented to train the model, the input parameters that are considered, etc. In the present review article, models created by several different machine learning approaches are comprehensively reviewed. According to the studies conducted in this field so far, it is concluded that artificial neural network is a very attractive approach for modeling both dynamic viscosity and thermal conductivity. The performance of these ANN-based methods can be modified by applying appropriate optimization approaches in order to find their optimum architecture design which minimize error margins. In addition to available correlations and implementation of ANNs, other intelligent approaches such as support vector machine and adaptive neuro fuzzy interface system are also applicable for accurate modeling of rheological properties of hybrid nanofluids.