Viscosity and rheological behavior of Al2O3-Fe2O3/water-EG based hybrid nanofluid: A new correlation based on mixture ratio

• Main Authors: Vicki Wanatasanappan V., Kumar Kanti P., Sharma P., Husna N., Abdullah M.Z.
• Other Authors: 58093867000
• Format: Article
• Published: Elsevier B.V. 2024
• Subjects: Al₂O₃-Fe₂O₃; ANN; Bayesian optimization; Hybrid nanofluid; Machine learning; Rheological behaviour; Viscosity; Alumina; Aluminum oxide; Bayesian networks; Ethylene; Ethylene glycol; Hematite; Nanofluidics; Neural networks; Newtonian liquids; Rheology; Al2O3-fe2O3; Machine-learning; Mixture ratio; Nanofluids; New correlations; Rheological property; Viscosity properties

The present study is a pure experimental investigation of the viscosity and rheological properties of the Al2O3-Fe2O3 hybrid nanofluid and the development of a new correlation. The main purpose of the study is to evaluate the effect of the Al2O3-Fe2O3 mixture ratio on the viscosity property and develop a correlation for the viscosity prediction. The Al2O3 and Fe2O3 were first characterized using XRD diffraction and the FESEM technique. The nanofluid was prepared using a two-step method using base fluid consisting of water and ethylene glycol mixture at 60/40 ratios. Five different Al2O3-Fe2O3 nanoparticle compositions were investigated experimentally for the viscosity and rheological properties at temperatures between 0 and 100 �C. The experimental data shows that the Al2O3-Fe2O3 composition of 40/60 resulted in the highest viscosity value at all temperatures investigated, while the 60/40 composition recorded the lowest viscosity value. Besides, the increase in temperature of nanofluid shows a maximum viscosity reduction of 87.2 % as the temperature is increased from 0 to 100 �C. Also, the rheological analysis on a hybrid nanofluid for all compositions of Al2O3-Fe2O3 indicates a Newtonian fluid characteristic. The experimental research data was utilized to create an artificial neural network (ANN)-based architecture. An autoregressive method called the Bayesian approach was adopted for training hyperparameters. During model training, the autoregressive technique assisted in achieving outstanding correlation values of more than 99.99 % with minimal mean squared errors as low as 0.000036. � 2023 Elsevier B.V.