This work presents a robust artificial neural network (ANN)-driven computational framework to investigate mixed convection flow over a rotating cone with variable viscosity and variable thermal conductivity, incorporating the coupled influences of Soret and Dufour effects. The nonlinear coupled ordinary differential equations, derived via similarity transformations, are approximated using a multilayer perceptron network. The trial solutions include adjustable network settings (weights and biases) that are improved using the adaptive moment estimation algorithm to meet the physical laws involved. A constant wall temperature condition is applied at the cone surface to simulate realistic thermal boundaries. The proposed ANN strategy exhibits excellent accuracy when benchmarked against traditional numerical methods while offering superior flexibility and generalization in handling complex transport phenomena. Parametric studies visually demonstrate the sensitivity of flow, thermal, and concentration fields to variations in key fluid and geometric parameters, affirming the strength of neural networks as a viable alternative for modeling sophisticated fluid dynamics systems.
Rajender et al. (Tue,) studied this question.