Validation of flow around buildings

Simulations are always a trade off between accuracy and computational cost; at the same time it is extremely important that models accurately predict the actual flow situation for a given situation. Here is a quick run down of our model selection and validation for accurate urban wind simulation.

To ensure this accuracy in simulation results we validate our models using published experimental data. In general we prefer to use models like LES with less underlying assumptions about the flow characteristics because there is less risk of ending up with bad results when one or more of these underlying assumptions are not (completely) true for a given case. While such models come at the cost of increased computational time it is our policy at Rheologic to err on the side of quality. (We do use all kinds of models including very fast ones, depending on the task at hand and with a solid understanding of the model's internal assumptions and limitations.)

For flows with complex, turbulent flow structures and separation - which are typically encountered in the built environment - we have excellent experience using LES (Large Eddy Simulation). The image below shows the comparison of an actual, measured wind tunnel experiment and an LES performed by Rheologic. The solid black line is the stagnation line found in the experiment. The solid turquoise line are our simulation results. Inside the stagnation surface the flow reverses its direction (recirculates) compared to the main flow direction. Correct prediction of the recirculation zone is a non trivial task and can make a huge difference in the quality of the simulation results if - for example - there is danger of exhaust gases from ventilation being pulled back towards a building.

The picture below shows our LES case on the left and a standard k-Epsilon simulation on the right side. What looks like two complete different CFD simulations is actually just one correct solution (our LES) and a completely wrong one on the right side. This case clearly demonstrates how important it is to select the correct model depending on the case... and how exact CFD simulation can be if it is done right.

A very tangible representation of such simulations is the way leaves (and snow) tend to accumulate in places with little air movement - exactly like the simulation predicts, as the image below shows.

Details about the experiment that we used to validate our simulation - set-up of the wind tunnel, boundary layers and other flow conditions as well as the model geometry are published in E. Paterna: "Wind tunnel investigation of turbulent flow in urban configurations: a time-resolved piv analysis", Diss. ETH No. 22609, (2015)

Transient LES with time averaging shows excellent results for flows around buildings.