Harnessing Lorentz force interactions and gyrotactic microorganisms: A novel design of explicit feature interaction-aware graph neural network for advanced bio-convection modeling in transverse fluid dynamics

The term “bio convection” refers to the complex convection patterns that result from the movement of microorganisms in a fluid under the influence of magnetic fields and gravity. The main objective of the work is to integrate Gyrotactic microorganisms and Lorentz force interactions to improve bio-convection modeling. The proposed EFIAGNN technique is used to predict the bioconvective flow in a fluid. This method addresses both physical and biological effects on fluid dynamics with the goal of increasing the accuracy and efficiency of transverse fluid transportation. The impact of Gyrotactic microorganisms in relation to bioconvection constant σ, bioconvection Lewis number LB, Peclet numberPe, and Prandtl number Pr are analyzed. The performance metrics like Efficiency, Mean Squared Error (MSE) and Coefficient of Determination (R²) are analyzed and compared with existing methods like Multi-Layer Perceptron Network (MLPN), Artificial Neural Networks (ANN) and Levenberg-Marquardt Back propagation Neural Networks (LMBNN). The errors obtained by the existing methods (MLPN, ANN, and LMBNN) are 1.7%, 2.8%, and 3.8%, respectively. Furthermore, the error of proposed EFIAGNN method obtains 0.5%. The efficiency value of the proposed technique is 99.8%, while the efficiency values of the MLPN, ANN, and LMBNN methods are 80.4%, 70.8%, 60.2%.