Artificial neural network analysis for radiative bioconvective Jeffrey nanofluid flow with Soret and Dufour impacts

Presently, the concept of artificial neural networks represents an innovative and transformative approach with wide-ranging applications across industrial, mechanical, pharmaceutical, and automotive domains. This article employs numerical computing approach for chemically reactive flow of an incompressible Jeffrey nanofluid between coaxial cylinders. Backpropagated neural networks (BNNs) are utilized. Outer cylinder remains stationary while the inner cylinder is stretched. Presence of gyrotactic microorganism is ensured. Besides this the salient features of thermophoresis, radiation and Brownian movement. Convection conditions of heat and mass transfer in presence of Soret and Dufour features are explored. The relevant problems are computed through BVP4c for reference data used in view of training Levenberg–Marquardt algorithm based upon (LMA-BNNs). The neural networks are trained, validated and tested to obtain optimal level of generalization and learning. The Levenberg–Marquardt backpropagation algorithm is utilized to minimize mean square error (MSE) in order to predict accurate solution under different parametric conditions. Comparative analyses (using statistical measures such as correlation coefficient, MSE, regression plot and error histogram) confirm the strength and validity of BNN treatment. Finally, the physical quantities of interest for influential variables are examined. Main results are concluded. Both temperature and Nusselt number against thermal Biot number have same response qualitatively.