An experimental study on the agitating efficiency and power consumption for viscoelastic-based nanofluids: Elasticity, impeller effects, and artificial neural network approach

Abstract

In industrial mixing applications, the power consumption and mixing time are employed widely for engineering equipment designs containing non-Newtonian, especially viscoelastic fluids. Though readings on nanofluids are growing, the attentions on nanofluids built on three ingredients elements of viscoelastic-based nanofluids (VBN) are insufficient. Subsequently, in previous study, multi-walled carbon nanotubes (MWCNT) were functionalized chemically with carboxyl groups (to prepare f-MWCNT) nanoparticles and characterized using X-ray diffraction, Fourier transforms infrared spectroscopy, dynamic light scattering, and transmission electron microscopy analyses. In this study, three constituents, viscoelastic-based fluid have been made by using (f1) a mixture of polyacrylamide, glycerol, and water as the base fluid and (f2) synthesized f-MWCNT as the nanoparticles. Further, the power consumption and mixing time of the VBN, i. e. f1+f2, with Rushton turbine disk (RTD), 45° pitched blade turbine (PBT), and hydrofoil (HF) impellers were measured in the transition region (10 < Re < 1800). The mixing times of the RTD, PBT, and HF impellers were measured for different VBN by the thermal response method resulting in minimum mixing time for the RTD. It was shown that mixing time increases with increasing of both nanoparticle and polyacrylamide (PAA) concentrations. Also, by increasing the both PAA and f-MWCNT mass fraction (high elasticity), the power number of the impellers rises and falls in low and high Reynolds numbers, respectively. In addition, artificial neural network (ANN) modelling with two hidden layers (4:15:12:1) was developed to predict the power consumption using impeller types, rotational speed, f-MWCNT, and PAA weight fraction. The correlation coefficient (R), and root mean square (RMSE) parameters of the test dataset are 0.99 and 0.0014, respectively which confirms the high accuracy of the presented ANN relationship.