1. The application of an atmospheric boundary layer to evaluate truck aerodynamics in CFD “A solution for a real-world engineering problem”
Ir. Niek van Dijk DAF Trucks N.V.

2. CONTENTS Scope & Background Theory: the atmospheric boundary layer
Cause: the earth’s roughness
Turbulence in the atmospheric boundary layer
Modeling of an atmospheric boundary layer in STAR-CCM+
Problem analysis
Numerical effects on the atmospheric boundary layer
Improving numerical settings
Conclusions & Recommendations

3. INTRODUCTION
Scope
Results shown are part of a graduation project performed at DAF Trucks NV
Project focus: Effects of the atmospheric boundary layer on truck aerodynamics
Presentation focus: Highlights of the numerical aspects
Background
Significant drag differences are observed between on-road testing methods and computational fluid dynamics (CFD) simulations
The atmospheric boundary layer is not present in CFD simulations
Vtruck
Vwind 3

4. THE ATMOSPHERIC BOUNDARY LAYER
Aerodynamic roughness
What is the atmospheric boundary layer (ABL)?
The atmospheric boundary layer describes the wind profile over the earth's surface as varying over height due to a certain roughness
Cause for an ABL is aerodynamic roughness ( 𝑧 0 )
Defined by terrain type, rougher terrain results in a higher aerodynamic roughness
𝑉 𝑤,𝑙𝑜𝑔 𝑧 = 𝑉 𝑟𝑒𝑓𝑙𝑜𝑔 ⋅ ln 𝑧+ 𝑧 0 𝑧 ln 𝑧 𝑟𝑒𝑓𝑙𝑜𝑔 + 𝑧 0 𝑧 0 4

5. THE ATMOSPHERIC BOUNDARY LAYER
Aerodynamic roughness
Velocity variation can be approximated with a logarithmic profile
Variation depends on aerodynamic roughness
𝑉 𝑤𝑖𝑛𝑑 = 𝑉 𝑙𝑜𝑔 (𝑧, 𝑧 0 )
3 𝑚/𝑠 at 𝑧=2
“Low roughness” profile (simulating highway conditions) used in numerical investigations 5

6. THE ATMOSPHERIC BOUNDARY LAYER
Turbulence in the atmospheric boundary layer
Turbulent intensities and length scales in the ABL are typically higher than in wind tunnels and the traditional CFD approach
Turbulent intensity (TI)
Turbulent length (TL)
High turbulence levels used in simulations
TI= 5% and TL = 5 m 6

7. MODELLING OF AN ABL IN STAR-CCM+
Initial analysis
The truck will not experience the profile specified at the domain boundaries without surface roughness and mesh modifications due to unwanted development of the profile
In the first part this unwanted development is analysed for an empty domain
Inlet boundary conditions:
3 𝑚/𝑠 at 𝑧=2 𝑚
Wind angle 𝜙=0°
TI = 5% and TL = 5 m (constant over height) 7

8. MODELLING OF AN ABL IN STAR-CCM+
Initial analysis: empty domain
Velocity inlet
Pressure outlet
Symmetry planes
No-slip floor with 𝑽=𝟐𝟓 𝒎 𝒔 without surface roughness
𝟑𝟎 𝐦
𝟏𝟔𝟎 𝐦
Red plane used to monitor the ABL development
𝟕𝟓 𝐦 8

9. MODELLING OF AN ABL IN STAR-CCM+
Initial analysis
Solver settings
All simulations are performed with a steady state RANS modelling approach in combination with the K-Omega SST turbulence model
Velocity magnitude profile is not constant through the domain
Front of truck
Truck height 9

10. MODELLING OF AN ABL IN STAR-CCM+
Initial analysis
The velocity profile development with “standard” CFD settings at position of the truck 1 2
Due to the symmetry plane as a top boundary, a zero normal velocity is “forced”
Aerodynamic roughness is not included 10

11. MODELLING OF AN ABL IN STAR-CCM+
Initial analysis
Friction velocity shows a rapid change in the first few meters of the domain
𝚫𝟏𝟔𝟎 𝐦
Floor (top view)
Near-ground velocity profile: changes rapidly in 3 m from inlet
𝚫𝟎.𝟏𝟓 𝐦
𝚫𝟑 𝐦 11

12. MODELLING OF AN ABL IN STAR-CCM+
Numerical effects on the ABL: top part (1)
The observed unwanted development is caused by absence of aerodynamic roughness
Goal: to make flow profile at position of the truck equal to the input
To avoid change of the ABL profile in the top part of the domain
 Change top boundary to a velocity inlet with only a parallel velocity component 12

13. MODELLING OF AN ABL IN STAR-CCM+
Numerical effects on the ABL: lower part (2)
Preventing development in the lower part of the profile
Literature:
Match the aerodynamic roughness with the surface roughness of the no-slip floor Surface roughness: 𝑘 𝑠 =30 𝑧 0 → first cell layer >2 𝑘 𝑠
Y+ values > 30, to ensure wall functions are applied (which can be modified)
Literature & STAR-CCM+ user guide:
Specific combinations of numerical settings in order to prevent development of the profile
Result: an error of more than 10% close to the ground w.r.t. to the input
Unknown effects on the flow characteristics
Wind profile develops into a constant (over height) wind profile 13

14. MODELLING OF AN ABL IN STAR-CCM+
Numerical effects on the ABL: alternative approach
Alternative approach: modifying the floor boundary condition upstream of the truck
Two alternatives: “slip floor” or a “velocity plane”
This prevents adaption of the ABL profile to the near wall region and the creation of a boundary layer on micro scale, which cannot be neglected downstream of the truck
1. Slip floor 2. Velocity plane
No-slip floor
Truck position 14

15. MODELLING OF AN ABL IN STAR-CCM+
Numerical effects on the ABL: alternative approach
Slip floor
The air profile close to the ground can be influenced since the flow is forced to have a zero normal velocity
This allows control via mesh settings and the use of the “undershoot”
Velocity plane (velocity inlet with only a parallel velocity component)
Both velocity and turbulence quantities can be specified
Less sensitive to mesh settings
ABL input
≪𝟏 𝐦
Slip floor 15

16. MODELLING OF AN ABL IN STAR-CCM+
Numerical effects on the ABL: results
Both the slip and velocity plane floor give very good fits of the ABL profile
Errors within 0.5% for both methods 16

17. MODELLING OF AN ABL IN STAR-CCM+
Numerical effects on the ABL: mesh details
Differences in mesh of both alternatives
“Volumetric cells”: 0.5 𝑚
Slip floor
First layer = 1 ⋅10 −5 𝑚
25 prism layers
Larger mesh required for the “slip floor”
Velocity plane
First layer = 1 ⋅10 −2 𝑚
15 prism layers 17

18. MODELLING OF AN ABL IN STAR-CCM+
Numerical effects on the ABL: turbulence
Effects of turbulent intensity and turbulent length scale profiles
Initial conditions: profiles through the domain with constant TI and TL specified at the inlet
Turbulent intensity (TI)
Turbulent length (TL) 18

19. MODELLING OF AN ABL IN STAR-CCM+
Numerical effects on the ABL: turbulence
Various tests performed to match the turbulent intensity and turbulent length scale profiles to the profiles found in literature
Both a “realistic” profile for the TI and TL increase the development of the ABL profile (compared to the input) at the position of the truck
The Realizable K-Epsilon model did increase the development even more
A decrease in development is obtained with modifying turbulence model parameters
Undesired: since the effects on truck aerodynamics are unknown
With initial settings: turbulence quantities at truck position are still within the range found in literature without turbulence profiles
Conclusion: TI and TL are kept constant over height (5% and 5m) 19

20. MODELLING OF AN ABL IN STAR-CCM+
Optimal numerical settings
Optimal numerical settings: achieved with the “slip floor”
Based on wind angle variation
“Velocity plane” seemed to be very sensitive to mesh refinement while the “slip floor” achieved the same accuracy without changing the mesh
Slip floor
No-slip floor
Vair
ABL
Vtruck
Yaw angle 20

21. CONCLUSIONS
Standard CFD settings result in an unwanted development of the ABL in an empty domain
As a result, the standard CFD settings are not useful to evaluate the effects of the ABL on truck aerodynamics
Appropriate CFD settings have been found to keep the development of the ABL to a minimum
Proposals shown in literature proved to be insufficient
Defining a slip floor in front of the truck in combination with a very fine mesh showed best results
Numerical settings shown only apply to a specific ABL
A different aerodynamic roughness requires, most likely, other mesh settings 21

22. RECOMMENDATIONS
The ABL approach might be useful as a alternative of the traditional CFD approach to match real world conditions
Significant differences in the flow field analyses were found for all yaw angles
The definition of the drag coefficient is ambiguous because the air velocity is not constant
An averaging method can be applied to calculate a single unique reference velocity and yaw angle
When specifying the ABL profile via a table at the boundaries, a perfect match should be ensured with the cell centroids of the mesh to avoid unnecessary interpolation 22

23. Ir. Niek van Dijk A-Niek.van.Dijk@DAFTRUCKS.com

24. SUPPLEMENTARY SLIDES

25. CALCULATION OF THE DRAG COEFFICIENT
Drag force in driving direction
𝐹 𝑥 = 1 2 𝜌 𝑉 𝑟𝑒𝑓 2 𝐴 𝑟𝑒𝑓 𝐶 𝑥 𝐶 𝑥 =𝑓(𝛽)
How to formulate the reference velocity to calculate the drag coefficient for the atmospheric boundary layer
Both air velocity and yaw angle vary over height

26. CALCULATION OF THE DRAG COEFFICIENT
Reference velocity for the atmospheric boundary layer
Average air velocity over truck height
This averaging method is one of many methods
Most accurate method
Corresponding yaw angle
Average yaw angle over truck height