Modelling and prediction of 3D carreau fluid behaviour using machine learning for Cattaneo Christov Double diffusion with variable conductivity
Aqsa Zafar Abbasi, Mamoon Aamir, Mariyam Sattar, Nermeen Abdullah, Tarek Salem Abdennaji, Badr M. Alshammri, Lioua Kolsi
Abstract
This paper proposes a smart numerical analysis of the Cattaneo–Christov Double Diffusion using a three-dimensional Carreau fluid flow model (CCDD-CFM) through an Intelligent Back-Propagated Neural Network combined with the Levenberg-Marquardt algorithm (IBPNN-LM). The nonlinear partial differential equations that govern the problem are reduced to ordinary differential equations via similarity transformations and solved by the Lobatto IIIA-based bvp4c method. A wide-ranging dataset is created by changing source parameters such as the Prandtl number (Pr), Schmidt number (Sc), stretching rate ratio (α), thermal conductivity parameter (γ), and local Weissenberg number (β). The created dataset is further separated into 80% training, 10% test, and 10% validation subsets for supervised learning of the IBPNN-LM model. Validation with the reference solution ensures accurate prediction with regression coefficients (R-values) around 1 and Mean Squared Error (MSE) around 10 -9 . The histogram of error and fitness curves further establish the credibility of the model. It has been identified in key findings that as γ increases, it creates a 23% increase in fluid temperature in the case of shear-thinning (n<1) and 17% increase for shear-thickening (n>1) fluids. On the other hand, increasing Pr produces a 29% decrease in the thickness of thermal boundary layer. Increases in Sc and α have the effect of reducing the concentration profiles on average by 22–28%. Gradient values declined systematically across epochs with up to 92% drop from initial levels, demonstrating model convergence. Results verify the effectiveness and dependability of IBPNN-LM in representing complicated fluid behavior in diffusion processes.