Neural networking analysis on heat transfer in Casson fluid with mixed convection equipped in staggered cavity with anti-parallel moving boundary

Abstract

It is an important fact that the Casson liquid suspension staying in staggered domains with moving boundaries claims daily life engineering standpoints. In this direction, anti-parallel drive of upper and lower walls of cavity results in a complicated formulation and hence it remains a challenging task for researchers to identify flow field properties of the liquid suspension. Motivated from this fact, we consider cavity with non-Newtonian Casson fluid. An inclined magnetic field and natural convection are applied. To be more precise, we considered three different cavity aspect ratios (AR = 0.4 ≤ 0.6 ≤ 0.8) of the staggered cavity. The walls are subject to no-slip conditions, except for the top and bottom walls, which have antiparallel boundaries. The right wall is kept at a cold temperature, while the left wall is uniformly heated. The remaining walls are exposed to adiabatic conditions. The flow equations are solved via finite element method (FEM). Velocity and temperature are all presented using contour plots with an inclined magnetic field. The value of kinetic energy is noted against the magnetic, Casson parameters, and Rayleigh number in tabular form. It is observed that increasing the aspect ratios across various physical parameters leads to a reduction in the size of vortices. Moreover, neural networking model is constructed to enhance the accuracy of predicting kinetic energy values. These models consist of three inputs, ten hidden layers, and one output. The training of this network utilizes the Levenberg-Marquardt algorithm. The best average mean squared error (MSE) value is identified as 6.29807E-05 in the case of ANN model-I. The comparison of numerical values and ANN predicted values are displayed through graphs, which are in good agreement.