Predicting the density of carbon-based nanomaterials in diesel oil through computational intelligence methods

The density of nanofluid is a crucial property in heat transfer applications, and it is important in the determination of various heat transfer parameters such as the Reynolds number, Nusselt number, the friction factor, and the pressure loss. Unlike thermal conductivity and viscosity of nanofluids, estimating the density of nanofluids has received very little attention. Since accurate models can speed up the design of thermal devices, therefore, modeling the density of nanofluids is highly important. This study modeled the density of carbon-based nanomaterials, including graphene nanoparticles, multiwall carbon nanotubes, and hybrid of all suspended in diesel oil. This modeling was done using an artificial neural network (ANN) and Bayesian support vector regression (BSVR). The model was developed using the temperature and mass fraction of the nanoparticles as the model inputs. The temperature considered ranges from 5 to 100 degrees C while the mass fraction concentration ranges from 0.05 to 0.5%. During model training, both the SVR and the ANN models achieved a very high correlation coefficient of 99.63% and 99.88%, respectively. Finally, the accuracy of the models was validated on the 22 new experimental datasets (testing dataset), and the root means square error of the Pak and Cho, BSVR, and ANN models are 4.2E-3, 1.7E-3, and 1.3E-3 (g/cm(3)). The ANN model achieved a 3 -fold improvement in accuracy than the existing Pak and Cho model for the nanofluids. This study provides a simple and accurate approach for modeling the density of nanofluids.

Automated summary

This study modeled the density of nanofluids using an artificial neural network and Bayesian support vector regression. The ANN model showed a 3-fold improvement in accuracy compared to existing models.