Estimation of the viscosity of nine nanofluids using a hybrid GMDH-type neural network system

• A hybrid system of group method of data handling and polynomial neural network (GMDH–PNN) is applied to calculate the viscosity of nine common nanofluids.
• The total percentage of absolute relative deviation (AARD%) was obtained 2.14% with a high regression coefficient of R = 0.9978.
• Compared to traditional neural networks, the proposed model is superior due to its simplicity.

The introduction of nanoparticles into the fluids traditionally used in heat transfer processes, such as water, ethylene glycol and propylene glycol, has led to the advent of nanofluids which have become widely applicable due to their improved heat transfer properties. Dispersion of nanoparticles in base fluid affects the viscosity of system to a noticeable degree. In this regard, we developed a hybrid self-organizing polynomial neural network on the basis of group method of data handling (GMDH) to study the viscosity of nine nanofluids based on water, ethylene glycol and propylene glycol. The results show that the hybrid GMDH model can accurately predict the viscosity of nanofluids. The percentage of average absolute relative deviation (AARD%) for all systems was 2.14% with a high regression coefficient of R = 0.9978. The results estimated by the hybrid GMDH model, when compared to those of various theoretical models and an empirical equation, exhibit a higher accuracy.

Figure optionsDownload as PowerPoint slide