Research Projects

This research proposal introduces a novel deep learning framework based on Physics-Informed Neural Networks (PINNs) for modeling the complex behavior and predicting both linear and nonlinear stability features of hybrid nanofluid flow within a porous channel, with a focus on analyzing convective onset and detecting subcritical zones. The convective instability mechanism of flow inside an infinitely long channel immersed in a porous region is considered one of the most significant processes due to its relevance to practical applications in science and engineering. The stability boundaries of isothermal channels are notoriously difficult to capture using classical linear and nonlinear analyses, which rely on complex numerical solutions of differential equations that are both computationally intensive and sensitive to discretization. Physics-Informed Neural Networks (PINNs), a recent advancement in machine learning, offer an efficient and innovative approach by explicitly incorporating fundamental physical laws into the learning process, enabling the reconstruction of velocity, pressure, and temperature fields. While limited experimental, theoretical, and numerical studies exist for non-isothermal channels, the stability characteristics remain unclear when the channel is filled with a porous medium containing hybrid nanofluids. Traditional heat transfer fluids such as water, ethylene glycol, and oil possess inherently low thermal conductivity, limiting their performance in heat exchangers. The development of nanofluids, which are base fluids uniformly dispersed with metallic or metal oxide nanoparticles, marked a significant advancement in enhancing heat transfer efficiency and overall thermal performance. More recently, hybrid nanofluids, which combine different types of nanoparticles within a single base fluid, have exhibited superior thermophysical properties, including enhanced thermal conductivity, improved viscosity control, and increased stability. Under thermal gradients, these fluids significantly influence the onset of convective motion by modifying key parameters such as buoyancy forces and thermal diffusivity. By tuning nanoparticle type, size, and concentration, it becomes possible to manipulate the critical Rayleigh number and optimize heat transfer for advanced applications in nuclear reactors, solar collectors, chemical processing units, and high-performance computing systems. In this context, the proposed research seeks to bridge the gap between classical theory and modern data-driven methods by integrating physics-based modelling with deep learning techniques. The objective is to perform both linear and nonlinear stability analyses of hybrid nanofluid flows through porous media and to predict critical parameters such as the Rayleigh number using PINNs, thereby supporting the design and optimization of next-generation thermal systems.