Intelligent numerical computing paradigm for heat transfer effects in a Bodewadt flow

The current research article is intended to examine the nonlinear input-output fitting with two-layer feed forward neural network for numerical treatment of heat transfer effects in Bodewadt flow over a permeable disk. Estimation parameters in Bodewadt flow model include wall suction parameter (1.7 <= A <= 6.7), magnetic field (0.1 <= M <= 1.0), Brandt' number (0.2 <= Pr <= 5.0) and variable viscosity (2.0 <= theta(e) <= 20.0). The influences of magnetic field, Joule heating, internal heat generation/absorption, wall suction, viscous dissipation along with variable fluid properties are also contemplated as well. The energy equation in Bodewadt flow for permeable disk governing heat transfer and fluid motion are transformed into self-similar dimensionless differential equation by using Von-Karman variables. The temperature and velocity of the fluid about the disk by taking various values of the physical parameter are solved by Adams Bashforth method to determine the reference computational results of Bodewadt flow model. The values of skin friction co-efficient and Nusselt number are also calculated and physically interpreted for the assorted parameters. Further, the obtained experimental dataset of the system is used to authenticate the artificial neural network modeling with optimization of Levenberg-Marquardt backpropagation. Fitting data precision is examined on mean squared error based cost function for the system and the outputs of intelligent networks are demonstrated using performance parameter, error histograms, regression and fitting plots. Least mean square error trained at decreasing gradient with optimized weights having strong correlation R = 1 between target and network output and a consistent convergence further certified the worth of methodology.