An Artificial Intelligence Approach to Predict the Thermophysical Properties of MWCNT Nanofluids

Experimental data of thermal conductivity, thermal stability, specific heat capacity, viscosity, UV-vis (light transmittance) and FTIR (light absorption) of Multiwalled Carbon Nanotubes (MWCNTs) dispersed in glycols, alcohols and water with the addition of sodium dodecylbenzene sulfonate (SDBS) surfactant for 0.5 wt % concentration along a temperature range of 25 degrees C to 200 degrees C were verified using Artificial Neural Networks (ANNs). In this research, an ANN approach was proposed using experimental datasets to predict the relative thermophysical properties of the tested nanofluids in the available literature. Throughout the designed network, 65% and 25% of data points were comprehended in the training and testing set while the other 10% was utilized as a validation set. The parameters such as temperature, concentration, size and time were considered as inputs while the thermophysical properties were considered as outputs to develop ANN models of further predictions with unseen datasets. The results found to be satisfactory as the (coefficient of determination) R-2 values are close to 1.0. The predicted results of the nanofluids' thermophysical properties were then validated with experimental dataset values. The validation plots of all individual samples for all properties were graphically generated. A comparison study was conducted for the robustness of the proposed approach. This work may help to reduce the experimental time and cost in the future.