Entropy and back-propagation analysis of Ag−MgO/water hybrid nanofluid flow over a radially stretching disk: Response optimization and sensitivity analysis

Abstract

We are investigating the heat transfer and other flow profiles for a hybrid nanofluid flow over a stationary stretching disk in Bodewadt flow. Sensitivity analysis was performed for the Nusselt number as a response parameter along with response optimization and back-propagation using neural networks. The present study helps in predicting the behaviour of hybrid nanofluids for such flows and optimizing the design of systems for improved thermal performances finding applications in cooling devices, centrifuges and turbomachineries. Levenberg-Marquardt algorithm was implemented on Nusselt number values for stretching parameter C, nanoparticle volume fraction of primary nanoparticle (ϕ1) and secondary nanoparticle ϕ2.Spectral quasilinearisation method (SQLM) was used for solving the dimensionless ordinary differential equations and the numerical data for Nusselt number was used in response surface methodology (RSM) and artificial neural network back-propagation algorithm (ANN-BPA) for evaluating sensitivity analysis and interpolating data respectively. The radial velocity F increased with increasing C values and nanoparticle volume fraction values. The axial velocity reduced by 39.56 % when C increased from 3.0 to 5.0 while it decreased by 6.57 % for an increase of ϕ2=0.00 to ϕ2=0.005. The pressure term decreased by 23.63 % for increasing ϕ2=0.00 to ϕ2=0.005. The mean square error for Nusselt number prediction modelling was 6.6016E-11 attained at 1000 epochs with gradient value of 1.2035E-07. Also, Nusselt number was found to be most sensitive to stretching parameter. The optimal response was obtained with 100 % desirability with maximum Nusselt number response recorded as 6.10878 for C=4.99195,ϕ1=0.00989566&ϕ2=0.00990605.

Keywords

• Bodewadt flow
• Response optimization
• Levenberg-Marquardt algorithm
• Hybrid nanofluid
• Sensitivity analysis