A recurrent neural network approach for magneto‐hydro‐dynamic flow of second‐grade fluid with dissipation effect

Abstract

Artificial neural networks (ANNs) with feedback loops known as recurrent neural networks (RNNs) are appropriate for handling temporal dependencies. The accuracy of the results in computational fluid dynamics (CFD) has gradually improved with the integration of artificial intelligence (AI) with CFD. This research article aims to decipher the dynamics of magneto‐hydro‐dynamic flow of second‐grade fluid with dissipation effect (MHD‐FSGF‐DE) using the Levenberg–Marquardt backpropagation (LMB) based on RNNs (LMB‐RNNs). The dataset is produced by the cutting‐edge homotopy analysis method for variation of different parameters including fluid parameter β, Marangoni parameter Ma, Hartmann number M, and Prandtl number Pr. RNNs are trained on the points of the dataset to maximize the outcome's accuracy and provide a comprehensive knowledge of the long‐term correlations between the input and output data points. The accuracy of the state‐of‐the‐art LMB‐RNNs approach is validated by performance graphs, error histograms, training estate analyses, regression plots, and input and output correlation plots. The profiles of physical properties like velocity, temperature, and concentration against β, Ma, M, and Pr are graphically shown to further highlight how these parameters affect physical properties. After 1000 iterations, the mean squared error (MSE) is 10−10 and the observed value of correlation coefficient R is 1 endorsing the worth of the LMB‐RNNs. The velocity of fluid upsurges for increasing β while declines for increasing M and Ma. The outcomes are comparable with the previously published findings. The core findings of this study have potential applications in various fields like polymer processing and cooling of electronic devices, specifically in the areas of electronic coolant system design and optimization.