Synthesis of Fe3O4-Al2O3-ZnO/water ternary hybrid nanofluid: Investigating the effects of temperature, volume concentration and mixture ratio on Specific heat capacity, and development of Hybrid machine learning for prediction

• DOI: 10.1016/j.est.2021.102947
• Publication Date: 2021-07-21T09:52:05Z
• Authors: Humphrey Adun; Doga Kavaz; Ifeoluwa Wole-Osho; Mustafa Dagbasi
• Keywords: Zeta potential; Atmospheric temperature range; Volume fraction

Abstract In this experimental study, the specific heat capacity of water-based Fe3O4-Al2O3-ZnO was fabricated. Three mixture ratios of 1:1:1 (33.33%% Fe3O4, 33.33%% Al2O3, 33.33%% ZnO), 1:2:1 (25% Fe3O4, 50% Al2O3, 25% ZnO,) and 1:1:2 (25% Fe3O4, 25% Al2O3, 50% ZnO,) were synthesized at volume fraction of 0.5%, 0.75%, 1% and 1.25%. All the experiments were carried out at a temperature range of 25°C and 65°C. Zeta potential test and particle size analyzer were used in examining the dispersal stability of the nanofluids, and nanoparticle size, respectively. Also, a high-resolution scanning electron microscope was used in describing the morphological structure of the nanocomposites. The result of the study showed that there exists a linear effect of temperature on the specific heat capacity of the ternary hybrid nanofluid. Also, when the volume concentration increases, it causes the specific heat capacity to decrease. At the mixture ratio of 1:1:1(Fe3O4-Al2O3-ZnO), 1:2:1 (Fe3O4-Al2O3-ZnO) and 1:1:2 (Fe3O4-Al2O3-ZnO), the maximum specific heat capacity increment of 11.9399%, 14.6491% and 13.5572% respectively was recorded at 1.25% volume concentration and 25°C temperature, as compared to the base fluid. The experimental result showed a `peaking effect` in the specific heat capacity, which was measured at a mixture ratio of 1:1:1. The least specific heat capacity values recorded are for the mixture ratio of 1:2:1. Correlation and machine learning models were developed in this study, and the result showed that the most accurate prediction of the experimental data was obtained using the support vector regression model. When the support vector regression and correlation models were compared to the experimental results, the maximum deviation recorded was 0.2% and 12.467% respectively.