Accurately simulating turbulence requires a computational model capable of capturing the complexities of external fluid dynamics. Governed by the Navier-Stokes equations, two commonly known models used today are RANS and LES. Although both are used for predicting and analysing fluid flow, RANS and LES offer two very different approaches when it comes to modelling turbulence:
RANS (Reynolds-Averaged Navier-Stokes equations) are low-fidelity simulations that solve its equations for turbulence by averaging the the details of the flow, such as oscillating wakes and individual eddies.
For external flow simulations, a RANS-based model is simply not capable of capturing all the complexities of wind behaviour. Why? As stated in its name, RANS is an averaged description of the flow. It is a steady-state flow analysis where the flow field is averaged over space and time– the data supplies the user with a general overview of the result. In some cases, RANS provides sufficient information, but in the majority of design scenarios where specific details in the data is required to make modifications and optimisations, users run the risk of losing key analytical information.
A comparative study conducted earlier this year evaluated RANS and LES-based simulations to determine which computational model most accurately predicts wind flow and mean surface pressure for buildings with balconies.
Not only did LES outperform RANS for wind directions at 90° and 180°, the study also finds:
”Because RANS systematically underestimates the absolute values of both Cp and mean wind speed on the balconies, it is suggested that building design based on RANS might result in excessive ventilation and in too high wind nuisance level.”- Zheng, Montazeri, & Blocken / CFD simulations of wind flow and mean surface pressure for buildings with balconies: Comparison of RANS and LES
In addition, the findings highlight underlying causes for RANS' performance deficits and inability to accurately calculate surface pressure coefficients along a building’s façades:
“This is mainly attributed to the well-known deficiencies of steady RANS in accurately reproducing flow separation, recirculation and reattachment, and its inability to capture vortex shedding in the wake.”- Zheng, Montazeri, & Blocken / CFD simulations of wind flow and mean surface pressure for buildings with balconies: Comparison of RANS and LES
RANS simulations are often viewed as an attractive option based on its low computational cost compared to other Computational Fluid Dynamics solutions. But as the saying goes, “you get what you pay for.” The trade-offs when choosing a low-fidelity tool can compromise accuracy and informative content, leaving the user with misleading or insufficient data.
LES (Large Eddy Simulation) is a time-dependent solution, and a preferred approach for urban wind simulations because it provides a more comprehensive analysis of complex flow behaviour.
Ingrid Cloud’s solver uses an implicit LES framework for turbulence modelling.
LES resolves a range of scales at high Reynold’s numbers, leaving only the smallest scales to be modelled. Because larger scales contain the most turbulent energy and transfer momentum (while small-scale eddies have more homogenous properties when compared to one another), an LES method is the more accurate and reliable model indicative of turbulent behaviour, as both time and spatial elements are considered within the numerical simulation.
A simulation with the ability to deliver conclusive and detailed results becomes especially relevant when dealing with large geometries such as a city, for example, or complex phenomena like vortex shedding.
”LES can provide accurate descriptions of the mean and instantaneous flow field around isolated buildings and in complex urban areas and the wind induced aerodynamic loads on building surfaces.”
- Zheng, Montazeri, & Blocken / CFD simulations of wind flow and mean surface pressure for buildings with balconies: Comparison of RANS and LES
With the scalability and advancements in high-performance computing over the years, the computational power and resources currently available are unparalleled, and one of the championing reasons why LES-based simulations are in fact very practical, cost-effective and one of the most accurate solutions available on the market today.
A representative comparison for understanding RANS & LES:
On the left, the image is taken with a high shutter speed. It shows the waterfall in a precise moment in time. Capturing the water's flow in detail at one exact instant is representative of LES’s ability to simulate the more intricate characteristics of fluid flow. The right side is a long-exposure image, taken with a slow shutter speed. It shows a smooth and blurred water flow, where the waterfall was captured over an extended period of time, and the final result is a compilation of pixels containing significantly less data in comparison to the left image.
To bridge the step from RANS to LES, various approaches have been developed, including URANS (unsteady RANS) where a time-dependence is introduced, but still only resolving the coarsest scales in the flow, and hybrid methods where LES is used only away from solid walls where the flow is modelled by RANS.
A comparison between RANS and LES is necessary for understanding two well-known predictive modelling methods for turbulent flow because:
1. RANS is a commonly known and used CFD method among the AEC Industry.
2. Emerging software-as-a-service solutions like Ingrid Cloud are further democratising CFD and LES-based methods, offering a more efficient and reliable approach to modelling turbulence in an urban wind analysis. This will become the new industry standard.
According to NASA’s vision for the future of Computational Fluid Dynamics, commercial codes need to evolve: “More effective discretizations and solvers designed specifically for LES type problems must be sought,” CFD Vision 2030 Study: A Path to Revolutionary Computational Aerosciences, NASA.
Whatever the method or CFD approach taken, safeguarding the design process from inefficiencies and error is a top priority for any skilled CFD engineer or professional designer. Users should always ensure their tools are reliable, benchmarked and scientifically validated.
*Click here to view Ingrid Cloud’s scientifically validated benchmarks.