Optimized neural network modeling of ternary hybrid nanofluid dynamics in double rotating disks with radiation and Cattaneo-Christov heat flux

Abstract

This study employs recurrent neural networks (RNN) with the Levenberg-Marquardt technique (RNN-LMT), based on back-propagation to find the solutions for MHD flow of a ternary hybrid nanofluid with the CattaneoChristof heat flux model (C-CHFM) over double disks. Unsupervised learning exclusively makes use of training data using input vectors. The data for this study were obtained through the ND solve method and later optimized using RNN. The RNN method is trained using 80 % of the data, with 10 % allocated for testing and the remaining 10 % reserved for validation. The network creates clusters during training by employing the input patterns to identify novel features. The fitting of the data (Ft), performance (Pf), mean square error (MSE), and training (Tr) are all evaluated using a stochastic numerical technique. The problem has been validated by error histograms (Eh) and regression (Rg) tests, as well as by robust conformance with empirical solutions, with accuracy in the range from E-3 to E-6. The influence of key variables on temperature, velocity, and nanoparticle concentration is graphically represented. The graph shows that the velocity distribution profile of the nanofluid rapidly decreases with an incremental increase in the melting parameter. An increase in the Prandtl number leads to a rise in the temperature profile. The thermal distribution profile decreases as the thermal relaxation parameter varies further. Additionally, the acquired numerical responses were cross-checked against previously released data for a few unusual cases, resulting in a workable compromise. This model can also be utilized to study organic instruments, medical approaches, nanopharmacological drug delivery schemes, and the application of nanotechnology to the treatment of diseases such as cholesterol.