Abstract:The importance of studying double-diffusive fluid flows along with the significance of non-Newtonian fluids have been well recognized in the fluid dynamics field for scientific and practical purposes. However, and given the ever-rising complexity of such non-linear flows, the conventional simulation methods are confronted with problems related to accuracy and required computation time and resources. As a result, the Machine Learning approach qualifies as a promising solution to said limitations. The present study investigates the application of Machine Learning models to study double-diffusive natural convection in rectangular cavities filled with non-Newtonian fluids. The flow is governed by four dimensionless parameters, namely: thermal Rayleigh number RaT , Lewis number Le , buoyancy ratio N , and power-law behavior index n , employed as input features. To precisely model fluid flow, three characteristics are predicted: flow intensity |ψc| , average Nusselt number Nu ̄ , and average Sherwood number Sh ̄ using four machine learning models: Artificial Neural Networks (ANN), Random Forests (RF), Gradient Boosted Decision Trees (GBDT), and Extreme Gradient Boosting (XGBoost). The analysis of investigated models' fine-tuned architectures using various tools confirms the non-linear complexity of the problem and allows to explore and discuss in details the inner-workings of each model. The ANN predicts the test data with R2=0.9999,R2=0.9996, and R2=0.9999 followed by XGBoost with R2=0.9979,R2=0.9802, and R2=0.9979 , and that for |ψc| , Nu ̄ , and Sh ̄ , respectively. The study shows the strong and correlated effects of RaT and n on the flow characteristics. The models' generalization is further examined using fluids with power-law behavior indexes outside the learning range. The present paper validates the choice of Machine Learning approach as a promising solution to model non-Newtonian double-diffusive fluid flows and encourages future works in this direction.

Machine learning based approach to fluid dynamics

Physics-Based Deep Learning for Flow Problems

Machine learning accelerated computational fluid dynamics

Predicting Pressure Fields from Incomplete Velocity Fields Based on Deep Convolutional Neural Network

Learning Incompressible Fluid Dynamics from Scratch -- Towards Fast, Differentiable Fluid Models that Generalize

Accelerating Numerical Simulations of Supercritical Fluid Flows Using Deep Neural Networks

Machine Learning for Fluid Mechanics

Active Learning over DNN: Automated Engineering Design Optimization for Fluid Dynamics Based on Self-Simulated Dataset

Predicting unsteady incompressible fluid dynamics with finite volume informed neural network

Machine Learning Approaches for the Solution of the Riemann Problem in Fluid Dynamics: a Case Study

Generalization techniques of neural networks for fluid flow estimation

Deep Learning Closure of the Navier--Stokes Equations for Transitional Flows

A Hybrid Data-driven Deep Learning Technique for Fluid-Structure Interaction

Solving Inverse Problems in Steady State Navier-Stokes Equations using Deep Neural Networks

Forecasting small scale dynamics of fluid turbulence using deep neural networks

Deep Structured Neural Networks for Turbulence Closure Modelling

Machine-learning fluid flow

Neural Network with Local Converging Input (NNLCI) for Supersonic Flow Problems with Unstructured Grids

A finite element-inspired hypergraph neural network: Application to fluid dynamics simulations

Machine Learning based algorithms for modeling natural convection fluid flow and heat and mass transfer in rectangular cavities filled with non-Newtonian fluids

Artificial Neural Networks in Fluid Dynamics: A Novel Approach to the Navier-Stokes Equations