Speaker
Description
Many engineering and scientific applications involve the simulation of incompressible flows around bluff bodies. Among these problems, flow past a circular cylinder is one of the most widely studied benchmark cases for understanding wake behavior and testing numerical methods. Although the geometry is simple, accurately resolving the flow often requires solving the Navier–Stokes equations with high numerical resolution. When such simulations must be repeated for different parameter values, the computational cost can become significant. Reduced-order modeling has therefore attracted considerable attention as a way to capture the essential flow behavior while reducing the cost of repeated simulations.
In this study, we develop a computational framework for the parametric analysis of steady incompressible flow past a circular cylinder. The high-fidelity simulations are performed by solving the two-dimensional steady incompressible Navier–Stokes equations on a structured grid using a finite-difference formulation. The steady solution is obtained through pseudo-transient iterations in which convective terms are treated with an upwind-biased scheme and viscous diffusion is handled explicitly. Incompressibility is enforced using a projection step that solves a pressure Poisson equation to correct the velocity field. Convergence of the solution is monitored through residual histories of the continuity and momentum equations.
To build a parametric dataset, simulations are carried out for a range of Reynolds numbers. Proper Orthogonal Decomposition (POD) is then applied to snapshots of velocity magnitude and pressure to extract the dominant flow structures and construct a reduced representation of the system. Surrogate models based on radial basis function interpolation (POD–RBF) and neural networks (POD–NN) are used to predict the reduced coefficients for new parameter values. The predictive performance of these models is assessed using a leave-one-out validation strategy, showing that the proposed framework can provide efficient and reliable predictions of the flow behavior.