Thermal conductivity prediction of WO 3 -CuO-Ag (35:40:25)/water hybrid ternary nanofluid with Artificial Neural Network and back-propagation algorithm
Chunlei Lin,Junhui Zhou,Qianqian Lu,Mohamad Khaje Khabaz,Amirreza Karimi Andani,Mortatha Al-Yasiri,Guangyong Pan
DOI: https://doi.org/10.1016/j.mtcomm.2023.106807
IF: 3.8
2023-08-07
Materials Today Communications
Abstract:One way to enhance the thermal conduction process in heating systems is to attach substances with high thermal conductivity (TC) to the base fluids. The interesting properties of NF (nanofluids) and the significant capability for growing TC have caused this group of fluids to be of great interest to researchers in recent years. Hence, in this study, the thermal conductivity of WO 3 -CuO-Ag (35:40:25) /water nanofluid (NF) with affected Artificial Neural Network (ANN) and back-propagation algorithm is investigated and predicted. The objective of this research is to develop a sophisticated data analysis model that can evaluate intricate relationships that impact the thermal conductivity of HNF in terms of temperature and solid volume fraction (SVF). This study is performed at 6 temperatures (T) of 25, 30, 35, 40, 45, and 50 °C and 4 vol fractions (φ) of 1%, 0.2%, 0.3%, and 0.4%. The regression plot for the datasets a remarkable resemblance to unity, indicating a robust correlation between the predicted values from ANN and the actual values. Notably, the predicted values in ANN are indistinguishable from 0.998, indicating a near equivalence between the projected and actual values. The results show that increasing the φ of NP and T has led to a rise in the TC of studied NF, and at a high-T level, the maximum TC value is reached when the φ is extended to the maximum value. Also, the results indicate that T sensitivity plays a more significant role than φ in increasing the TC of WO 3 -CuO-Ag (35:40:25) /water NF. The maximum and minimum values of the sensitivity analysis for T equal 177.4501–431.489098 in five ANNs, and φ equals 16.2539901–37.2332646, respectively. Eight neurons are the best number in the hidden layer (HL) based on different trained ANNs.
materials science, multidisciplinary