MHD bioconvection flow in different curvy cornered cavity configurations: CFD and machine learning analysis

Abstract

In this study, four different curvy cornered cavity geometries are examined on Fe3[jls-end-space/]O4[jls-end-space/]-water nanofluid magnetohydrodynamics (MHD) bioconvection problem in the presence of magnetotactic bacteria. The time independent governing equations are built in terms of stream function, temperature, iron concentration, bacteria density and vorticity. The numerical simulations are performed by radial basis function (RBF) collocation method. The obtained results are presented by plots of values of average Nusselt, Sherwood and bacteria density in important problem parameters. The results demonstrate a notable dependence on corner placement. The top left corner-curved cavities have more convective heat, mass and bacteria density transfer than the bottom left corner-curved cavities. The increase in radius of curvy corner results in increase in all convective transports in case of left-top curvy cavities. Among all configurations, the left-top–right-top (ltrt) curvy cornered cavity consistently achieves the highest values of Nu¯[jls-end-space/], Sh¯[jls-end-space/], and Nm¯[jls-end-space/], identifying it as the optimal geometry for magneto-bioconvective transport under an external horizontal magnetic field. A machine learning analysis using neural networks is also utilized on the data collected from numerical results. Model stability, generalizability and feature importance analysis were also conducted. The modeling results also verified the superiority of ltrt cavity configuration. © 2026 Published by Elsevier Ltd.