Hybrid artificial neural network-particle swarm optimization approach to investigate the blood flow through an artery affected by aneurysm

Abstract Weakening of the arterial wall leads to the formation of a sac-like structure called an aneurysm. Shear stress at the arterial walls can contribute to bulge initiation, growth, expansive arterial remodelling, and an increased risk of bulge rupture. This study aims to explore the variations in wall shear stress (WSS) and entropy generation in the context of irreversible blood flow through an artery affected by an aneurysm under magnetohydrodynamic conditions utilizing machine learning techniques. The Casson fluid model is employed to depict the non-Newtonian behaviour of blood. A model describing the above-considered blood flow problem with mild aneurysms is developed, and the obtained partial differential equations, along with the suitable conditions, have been numerically treated using an explicit finite difference scheme. In this study, it has been observed that an intensified heart pulse frequency results in lower wall shear stress on the aneurysm-affected arterial wall, while the wall shear stress is significantly reduced on the outer layer of the arterial wall affected by an aneurysm. Entropy generation and WSS are optimized using a machine learning approach, employing the Particle Swarm Optimization (PSO) algorithm along with an Artificial Neural Network (ANN). The network is trained using the Levenberg–Marquardt training procedure. Through the application of this machine learning approach, specific values for the Hartmann number, dimensionless temperature difference, heart pulse frequency, and amplitude of the pulsatile component of pressure gradient have been identified to achieve the minimum entropy generation.