A well-trained artificial neural network for predicting the optimum conditions of MWCNT–ZnO (10:90)/ SAE 40 nano-lubricant at different shear rates, temperatures, and concentration of nanoparticles

• Published in: Arabian journal of chemistry Vol. 16; no. 2; p. 104508
• Main Authors: Hemmat Esfe, Mohammad, Ali Eftekhari, S., Alizadeh, As'ad, Aminian, Saman, Hekmatifar, Maboud, Toghraie, Davood
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.02.2023; Elsevier
• Subjects: Artificial neural network (ANN); Least square method; Multi-walled carbon nanotubes (MWCNT); SAE 40-nano lubricant; ZnO; ANN; LM; Least square method; Artificial neural network (ANN); tansig; Multi-walled carbon nanotubes (MWCNT); MLP; logsig; ZnO; MWCNT; SAE 40-nano lubricant; MSE; SR
• ISSN: 1878-5352; 1878-5379
• DOI: 10.1016/j.arabjc.2022.104508

•ANN model has high accuracy for estimating the μnf of MWCNT–ZnO/ SAE 40 nano-lubricant.•μnfof MWCNT –ZnO (10:90)/ SAE 40 nano-lubricant was predicted using an ANN.•This ANN achieves 0.9995 and 0.00048 values for R2 and MSE.•The temperature and SR have a significant influence on the output.•φ has a direct but negligible effect on μnf. Fluid limitation in various industries due to their poor thermal conductivity has led to the improvement of the properties of the base fluid as a new method. With the development of nanofluid research, nanofluids are produced by adding metal nanoparticles and multi-walled carbon nanotubes (MWCNT). Due to the inability of theoretical models to predict the viscosity of nano-lubricants (μnf), mathematical models, especially artificial neural networks (ANNs), investigate the effect of various parameters on the properties of nanofluids and have replaced most of the usual statistical methods. This study investigates the effect of temperature, shear rate (SR), and volume fraction of nanoparticles (φ) on μnf of MWCNT –ZnO (10:90)/ SAE 40 nano-lubricant. Also, a nonlinear polynomial with terms up to power 3 is fitted in the experimental data, and its accuracy is compared to that of ANN in MATLAB. It was proposed that the ANN model has high accuracy (slightly better concerning nonlinear polynomial) for estimating the present study, the μnf of MWCNT–ZnO (10:90)/SAE 40 nano-lubricant. This ANN achieves 0.9995 and 0.00048 values for R2 and MSE, while the nonlinear polynomial showed 0.9983 and 4.0223 values, respectively, for the same parameters, which shows the good training status of the ANN. According to obtained results, the temperature and SR significantly influence the output. The experimental results showed that by increasing the temperature from 25 to 50 ˚C, the μnf of the nano-lubricant decreased from 397.5 to 90.5 cP (atφ = 1 % and SR = 400 rpm). So, the results show that with increasing temperature to 50 ˚C, the viscosity of the nano-lubricant decreases by about 77 %. ​ With increasing SR from 400 to 1000 rpm(at T = 50 ˚C and φ = 1 %), the viscosity of the nano-lubricant decreases from 90.5 to 85.3 cP. On the other hand, φ has a direct but negligible effect on μnf. In other words, the nanoparticle fraction change from 0 to 1 %, changes the μnf from 150 cP to around 200 cP. This model can be used as a design tool in future research or as an objective function in optimization problems.