Accelerated convergence for city-scale flow fields using immersed boundaries and coupled multigrid

Abstract

Fast and accurate prediction of Chemical, Biological and Radiological (CBR) dispersion patterns and toxic extents is important for public safety as conventional and non-conventional airborne threats released into the built environment can pose serious dangers to human health ranging from difficulty breathing and eye irritation to incapacitation and death. Computational fluid dynamics (CFD) is routinely used for gas dispersion modeling when increased accuracy over traditional Gaussian-based methods is required, although CFD requires greater modelling efforts. The combination of structured Cartesian meshes and immersed boundary methods reduces this effort and has been resurging in popularity over the past decade.Due to a lack of related literature, a simple and robust methodology was developed by the authors to treat immersed boundaries at multigrid levels, and this integration focused on a segregated multigrid approach using structured Cartesian meshes with adaptive mesh refinement. In the current work, a coupled multigrid treatment is adapted to the integrated immersed boundary methodology and adaptive mesh approach and compared to the segregated technique in terms of convergence rates. To accelerate convergence, the coupled approach selectively applies grid refinement at coarse grid levels, and by design, neither methodology detects the immersed boundaries on coarse grid levels.The two multigrid approaches are compared at three flow scales: wind tunnel, mock-urban and full-urban. At the wind tunnel scale, the coupled multigrid solver shows 39.3 times speed up over single grid and 10.4 times improvement over the baseline segregated approach. At the full urban scale, the coupled approach is shown to converge wind fields with a 6.3 times speed improvement over the previous segregated approach and therefore is suitable for CFD-based CBR assessment tools.