Session R05: CFD: General II

Adaptive mesh refinement in complex flow simulations

Adaptive mesh refinement (AMR) involves dynamically adjusting the computational grid resolution during a numerical simulation based on the local properties of the solution, allowing the simulations to concentrate computational resources in regions of high complexity or rapid changes. As a result, AMR has made turbulent flow simulations at very high Reynolds numbers more practical. However, AMR strategy may also adversely affect the solution accuracy if not done carefully. For instance, AMR strategy generally requires a user-defined cut-off value on a flow parameter like vorticity, based on a priori knowledge of the flow, to select all the appropriate grid zones for further refinement. Sometimes, this strategy may also ignore regions with important flow physics, e.g., high strain-rate regions. To avoid these issues, we propose an AMR strategy where the grid zones are selected for refinement by optimizing a sub-modular set function. This function is representative of the captured complexity in flow physics or numerical accuracy of the simulation and does not require a priori knowledge of the flow. This optimization strategy is implemented in Python and integrated with a c++ based CFD toolbox, OpenFOAM. The initial simulations of canonical flows with this strategy have shown promising results, and further details of the approach with simulations of complex flows will be presented at the meeting.   

Investigation on population-based optimization of complex flows

Optimization of complex high-speed flows involving multi-component and multiphase flow phenomena is often challenging due to their unsteady and multi-scale nature. In this study, we present a computational strategy for model-free, population-based optimization of such flows. In the method, to optimize a set of flow parameters for targeted objectives, groups of simulation samples are distributed and iteratively search the parameter space until convergence. We evaluate the efficiency and accuracy of this method using canonical model problems. The method is implemented in an in-house compressible flow code to demonstrate optimization of three-dimensional high-speed reacting flows on multiple GPUs in a scalable fashion.   

Boussinesq Convection in Spherical Shells using a Hybrid Discrete Exterior Calculus and Finite Difference Method

We present a new hybrid discrete exterior calculus (DEC) and finite difference (FD) method to simulate Boussinesq convection in spherical shells. We discretize the surface spherical operators using DEC, taking advantage of its unique features including coordinate system independence to preserve the spherical geometry, while we discretize the radial operators using FD method. The grid employed for this novel method is free of problems like the coordinate singularity, grid non-convergence near the poles, and the overlap regions. We have developed a parallel in-house code using the PETSc framework to verify the hybrid DEC-FD formulation and demonstrate convergence. We have performed a series of numerical tests which include quantification of Nusselt and Reynolds numbers for basally heated spherical shells, quantification of the critical Rayleigh numbers for spherical shells characterized by aspect ratios ranging from 0.2 to 0.8, and the simulation of robust convective patterns in addition to stationary giant spiral roll covering all the spherical surface in moderately thin shells near the weakly nonlinear regime.   

A CFD study on the effects of air cleaner and mechanical ventilation on the particulate matter concentration at a semi-outdoor space

One of the approaches to mitigate the outdoor particulate matter (PM) concentration is the use of air cleaners. In the present study, the effects of air cleaners as well as mechanical ventilation units on the PM reduction of a semi-outdoor bus terminal are examined by using computational fluid dynamics (CFD) simulations. The governing equations for the fluid flow and the PM concentration are the steady incompressible three-dimensional Reynolds-averaged Navier-Stokes equations (RANS) with the realizable k-ε model and a passive scalar transport equation with the standard gradient-diffusion hypothesis, respectively. As per the numerical results, the air cleaners alone significantly reduce the PM concentration of the bus terminal. The effect of mechanical ventilation units on the PM reduction is positive without air cleaners but negative with air cleaners. To understand the role of ventilation on the PM reduction, a simple theoretical model is derived based on the mass conservation of the PM. This model considers the effects of traffic-related PM emission, air cleaners, and ventilation. The PM concentrations predicted by the proposed model show a good correlation with those from the CFD simulations. From the model, we suggest that the ratio of the mass emission rate of PM to the clean air delivery rate of air cleaners is an important value in determining the effect of ventilation.   

Physical, mathematical, and numerical modeling of a gas flow in pipeline networks with low Mach number expansion

This study aims to investigate gas flows at low velocities through pipeline arrangements. We consider one-dimensional Navier-Stokes equations

Insights into the particle-free and particle-laden turbulent flow statistics in sharply bent channels

Turbulence in canonical wall-bounded and particle-laden flows shows a wide range of regimes due to considerable interaction between scales. Reynolds number is primarily used to characterize the fluid dynamics and particle-laden fluid corresponding to single phase and particle-laden channel flows. Nonetheless, different flow behaviour exists in curved channels even at a fixed Reynolds number, as delineated by Geert Brethouwer (J. Fluid Mech., vol. 931, 2022, pp. A21). In our study, we show the behavior of wall-bounded turbulent flows with and without particles in sharply bent channels by exploring the time averaged velocity profiles at various channel cross-section. The well-known logarithmic behaviour of the time averaged normalized velocity profile is found to be retained, where the von Kármán and the additive constants assume altered values depending on the bend angle. An enhanced near-wall fluctuations at the bend is obtained owing to the diffusion of counter rotating vortices that led to increased turbulent activity in case of particle-free as well as in particle-laden turbulent channel flows. Turbulent kinetic energy (TKE) budget for particle-free and particle-laden cases are elaborated for various bend inclinations at different sections of the channel, which establish that TKE is modulated at the bend with an overall attenuation on loading the channel with particles.   

Numerical and experimental study of the flow converging at or emerging from the apex of a conical surface for near zero Re values.

The results of a numerical and experimental study of the steady state flow in the vicinity of a conical nozzle in the injection and withdrawal processes for near zero Re values are presented. The flow was numerically solved for nozzles with aperture angles of 90, 105, 120 and 135 degrees at Re values of 0.01, 0.1, 1.0 and 10.0. Numerical solutions were obtained by applying a sharp interface immersed boundary method to a streamfunction-vorticity formulation in cylindrical coordinates discretized in finite differences. The velocity distribution was experimentally measured using the PIV technique for aperture angles of 90 and 135 degrees and for Re 0.01. Results are presented in the form of velocity fields, streamline plots, contour levels of the streamfunction and plots of the streamline separating the main and surrounding flows. A comparison of the results of both studies, when the geometry and flow regime are equivalent, is presented.   

WaterLily: A fast differentiable CPU/GPU flow simulator in Julia

Integrating computational fluid dynamics (CFD) software into optimization and machine-learning frameworks is hampered by the rigidity of classic computational languages and the slow performance of more flexible high-level languages. WaterLily is an open-source incompressible viscous flow solver written in the Julia language. The small code base is multi-dimensional, multi-platform and backend-agnostic (serial CPU, multi-threaded, & GPU execution). The simulator is differentiable and uses automatic-differentiation internally to immerse solid geometries and optimize the pressure solver. The computational time per time step scales linearly with the number of degrees of freedom on CPUs, and we see up to a 182x speed-up using CUDA kernels. This leads to comparable performance with Fortran solvers on many research-scale problems opening up exciting possible future applications on the cutting edge of fluid mechancis and machine-learning research.