A reliable model to estimate the effective thermal conductivity of nanofluids

Abstract

The thermal conductivity is a key parameter to study the applicability of nanofluids for heat transfer enhancement of flowing liquids. This paper is an effort on implementing various methods to model the effective thermal conductivity of 26 nanofluids under different situations and evaluate the authenticity of the reported experimental data in the open literature. The most influential physical properties of nanofluids, such as the nanoparticle volume fraction, nanoparticle diameter, thermal conductivity of base fluid, temperature, and thermal conductivity of solid particle are considered as the input variables. With the purpose of introducing a comprehensive and pragmatic model with desired accuracy, a Multilayer Perceptron-Artificial Neural Network (MLP-ANN) approach is constructed and tested using data generated from 993 experiments. To appraise the creditability of the MLP-ANN model, a comparison with other 10 alternative techniques is carried out. The predictions made by the MLP-ANN yield excellent match with the experimentally generated samples against those of the other approaches. The coefficient of determination and relative root mean squared error are found to be 0.994 and 1.534%, respectively. Likewise, the results of the data analysis and the outlier detection method have proved that some of the data samples are significantly inconsistent with the remainder of the data set.