Akademik Veri Yönetim Sistemi
18th International Conference on Agents and Artificial Intelligence, ICAART 2026, Marbella, İspanya, 5 - 08 Mart 2026, cilt.3, ss.2640-2647, (Tam Metin Bildiri)
The present study develops an artificial neural network (ANN) model to predict bioconvective flow generated by magnetotactic bacteria in a square cavity with a rounded upper corner under an external magnetic field. The induced magnetic field is directly incorporated into the governing equations of bioconvection. A dataset is constructed using radial basis function (RBF) method by varying key physical parameters, including the Rayleigh, bioconvective Rayleigh, Peclet, Lewis, Hartmann and magnetic Reynolds numbers, and radius of the rounded corner. The ANN is trained and tested using multiple architectures, activation functions, and partition ratios to evaluate performance. Results indicate that a trilayer ANN with ReLU activation and an 80:20 training-to-testing split achieves the lowest mean squared error across all target outputs. Unlike conventional numerical solvers, the proposed ANN acts as a fast model capable of accurately predicting heat, mass, and bioconvective transport indicators across a high-dimensional parameter space.