Analysis of different dimensions of a virtual wind tunnel and different wall conditions

Growing interest in improving the performance of vehicles is a subject that allows for continuous research. One such branch of these researchers is energy and consumption performance. These performances describe, on a global level, the need for lower fuel consumption but also a higher energy and power performance. Therefore, to create this balance, direct control of polluting emissions, especially C02 emissions, is necessary. Creating the best mix between low fuel consumption, maximum power of internal combustion engine and low pollutant emissions is influenced by many factors. One such factor, of prime importance, is represented by the aerodynamics of vehicle, in particular the aerodynamic performance, which needs to be tested to be confirmed, so numerical values are needed. Therefore, different ways of testing in the physical or virtual environment are necessary. To facilitate the optimization of a vehicle from an aerodynamic point of view, the most practical way is the virtual wind tunnel, given the possibility to test, at the same time, more configurations. But the CFD (Computational Fluid Dynamics) simulations, sometimes, have the disadvantage of very long compilation times. So, in the presented paper the construction of an optimal virtual wind tunnel for trucks will be presented, focusing on the test area (chamber) and several configurations will be presented such as: different proximity of the tunnel walls to the tested object so a different geometry of the test chamber and condition for the type of floor with different roughness than base case to simulate a more realistic road surface. Following the CFD simulations, it was observed the influence that increasing the dimensions of the test chamber has on the results obtained, so that moving the walls away from the test model generates much more realistic results. In addition, another thing that has an impact is the roughness of the floor, which, in this particular case, generate a 10% difference between the values obtained. Therefore, it will be possible to optimize the tunnel as best as possible to have results as close as possible to reality and in a reduced time.


Introduction
Nowadays, the demand for the transportation of goods from a place to another is constantly growing.As a result of this demand, there is also a significant diversity in the ranks of vehicles that transport goods and valuables.Transport must be done quickly, easily and safely, in one word: efficient.This efficiency is given, among other things, by the dynamic capacity of the vehicle; in this paper the emphasis is placed on the aerodynamic performance.This performance is dictated by the aerodynamic coefficient (Cd) of the vehicles which can also be translated through a reduced fuel consumption.Cd is a dimensionless numerical value that is difficult to estimate but can be determined through various physical tests and/or in the virtual environment.The physical tests take place in large wind tunnels, tests used both for optimizing the bodywork and/or its various auxiliary elements or used for 1303 (2024) 012043 IOP Publishing doi:10.1088/1757-899X/1303/1/012043 2 homologation.A physical test requires the presence of a model of the vehicle and can be at scale 2:5, 1:1 or any other scale.This model must be in very good condition and be built mainly from body parts or clay or 3D printed elements.In addition, it must have a very solid construction to withstand the high wind speed as well as possible.A physical test in a wind tunnel has advantage, especially in the case of 1:1 scale model, of precise results and very close to the real exploitation of the vehicle, but given the relatively high dimensions and costs, they are produced in a reduced number.Most of the time, different car manufacturers test their products in the same wind tunnel.This involves rigorous and though organisation to meet the all-testing requirements.This describes a time limitation in terms of testing multiple configurations of the same model to optimize it.That's why, to streamline and speed up the development process of the vehicles it switched in parallel, to testing in the virtual environment, of the CFD type, with wind tunnel.This type of testing has the advantage that it is not necessary to model the entire wind tunnel, but only the test chamber (the area of interaction between the model and the air flow).Besides the gain from point of view of taking space (because it takes place in the virtual environment), it has the advantage that several tests (simulations) can be started at the same time; in other words, a bigger image on several models in a much shorter time.However, there is the disadvantage of not having results close to reality; these being dictated by discretization method of the models, which will be discussed in the following lines.Next, different types of wind tunnels will be presented.These are used in automotive industry, aviation, construction and others [1], [2].Based on Figure 1 can be distinguishing two different wind tunnels: open-return in where the air that reach in the test section is gathered from the room where the wind tunnel it is placed and close-return where the air is conducted from the exit of the test section back to the fan by a series of turning vanes or blades.The close-return wind tunnel has a superior quality of the flow in the test section due to the fact that it is relatively uniform in comparison with the open-return where it has a poor flow quality.[3], [4], [5].For this paper, the idea is to have a close-return type wind tunnel from which only the test chamber (test section) will be modelled.This will be configured in different dimensions to be able to present which of them have the best potential to be used for tests in order to determine an aerodynamic performance.Based on this paper, the purpose of the importance of the good definition an selection of appropriate dimensions for the virtual test chamber will be concluded and the necessity of adding different parameters, such as the roughness of the floor, to obtain the most conclusive results will be presented.

Virtual wind tunnel
For this paper, the construction of an aerodynamic tunnel for trucks will be presented and the emphasis will be placed only on the test section.In the CFD virtual environment (we have the possibility to simulate only an area of the entire tunnel, the rest can be entered as parameters.For example, to simulate the fan we will give a speed of the air as an input parameter and in addition it is not necessary to specify which type of tunnel will be used (open-return or close-return).Having to do 1303 (2024) 012043 IOP Publishing doi:10.1088/1757-899X/1303/1/0120433 with a numerical simulation, a very important thing in the accuracy of the results is represented by discretization.The finer the mesh, the better and more realistic the results will be.In the actual paper, for truck with a scale of 1:1, it is necessary that the discretization to be as fine as possible.However, certain details that can complicate the calculation process can be removed, such as: design details of door handles, of the mirrors, elements that meet the air flow such as those inside the passenger compartment.In this way, the time to obtain the results is minimized.Therefore, in the numerical simulation program, a wind tunnel will be modelled with a rectangular test chamber and the minimal starting dimensions will have an average, approximate value, following the verification of several wind tunnels after, it will be used with various other configurations.In Figure 2, it can be observed a test chamber for a truck assembly (tractor unit and semi-trailer) and, at the same time, it can be seen the large dimensions of this type of structure can have.In this case, the open jet test section it is part of a test hall of 50m x 30m x 20 [3].It can be said that these give the chance of having the best and most conclusive results, eliminating the possibility of measurements errors as a result of unwanted air interference with the tunnel walls.The physical aerodynamic tunnels for vehicles have rotation lanes for each wheel, thus simulating the operation in reality as best as possible.In addition, the wheels are a significant generator of vortices that influence the overall aerodynamic performance of the vehicle.In the virtual environment, this rotation band/lane can be simulated through a rotation domain, in other words: each wheel will be given the condition to rotate with a certain angular speed (the same for all wheels).For this paper, a wheel rotation domain was not considered given the simplistic model used for the test.
In the virtual environment, as was mentioned in the rows above, there is a much greater possibility of having large dimension wind tunnels, but the waiting time for solving the results will also be longer.At the same time, in addition to large dimensions, discretization is another important factor.In addition, there is the possibility that many virtual tunnels are over-dimensioned and the values on the 3 axes (X, Y, Z) are far above the maximum limit at which relevant results can be obtained.In this case, sometimes the excessive size of dimensions leads to very long waiting times with, mainly the same results as smaller tunnels.That is why it is important to analyse several configurations to observe from which point the increase in dimensions is no longer necessary [7], [8].

Set-up
To observe the possibility of having measurements influenced by the dimensions of the wind tunnel, several simulations were made at different dimensions and different conditions for the walls.Also, a simplified model was used for these simulations.By analysing the results, it will be possible to approximate, the point from which increasing the dimensions of the tunnel is not necessary, thus shortening the simulation period.

Geometrical model
The geometric model was created to simulate a basic shape of a truck, so that an air flow similar to that of a normal operation or a real truck can be generated.As can be seen in Figure 3, the test model has similar dimensions to a real truck and does not have complex elements that could complicate the calculation process.In addition to this aspect, as mentioned previously, for the truck, the condition of the field rotation for the wheels was considered, therefore, these are a common (solid) body with the base geometry of the truck.Auxiliary structural elements such as: mirrors, air tanks, storage boxes, signalling elements were, also, removed to simplify the model so that the processing time of the simulation is as low as possible.

Mesh
Discretization is one of the most important factors in terms of delivering the best possible results.As a consequence of the fact that a test model with a 1:1 scale was used for this paper this lead, as a result of low computing power (12 GB RAM, 2.50 GHz CPU, 2GB GPU), to use a relatively quality criteria sufficient to satisfactorily cover all the geometry; it is good to remember that the discretization it is made both: for the geometric model and for the test chamber with the same value.As can be seen in the Figure 4, discretization of the wind tunnel and the geometrical model are unified as a single block due to the meshing of the solid (truck) and the area where the airflow should be (test chamber).From meshing point of view, four dimensions were used for the elements: 60 mm, 160 mm, 260 mm and 360 mm.For these, the following results of the forces on the X axis were obtained: 407297 N (for 60 mm), 397974 N (for 160 mm), 416721 N (for 260 mm) and 412081 N (for 360mm).It can be seen that there is a difference between the forces and the reason why it was chosen that the discretization of all the models in this paper should be done with elements that have size of 60 mm is given by the fact that, in general, a finer meshing determines better results.However, for the value of 160 mm, the force on X axis is lower than that at 60 mm, 260 mm and 360 mm.This particular case describes way in which the value of the elements and the type of discretization must be carefully chosen because improper values for these parameters can generate final results that are not definitively conclusive.In addition meshing with a value of 60 mm for element size sums up 2854386 nodes and 16438941 elements.

Preparation
Given the fact that for these CFD simulations (performed in Ansys software) only the test chamber will be used, the airflow will be given as a parameter to simulate the other areas of a real wind tunnel.Therefore, for all simulations, an air speed of 150 km/h will be used for the entrance in the chamber and a pressure of 1 atm (atmospheric pressure) will be used as a data for the exit of the air flow from the test chamber, also air temperature it is considered at 298,15 K and density at 1,185 kg/m 3 .The value of velocity for these tests was chosen as a result of the fact that as a vehicle's forward speed increases, its forward resistance also increases.For trucks, although the speed is much lower, the purpose of this paper is to present in an informative way the influence of the dimension of the test chamber on the model.In the case of the analysis of the specific aerodynamic performance of a truck or its various specific equipment, after the correct dimensioning of the test chamber, the velocity can have values of 80 km/h or 90km/h, so specific to this kind of vehicle.In addition, at this speed the air flow can be observed more easily and more clearly.Other parameters such as wall conditions will be presented for each individual configuration.Regarding the values for the distance in X between the air inlet in the test chamber and the geometrical model for the 11 configurations they are (in this order): 100 mm, 500 mm, 1000 mm, 1500 mm, 2000 mm, 2500 mm, 3000 mm, 4000 mm, 5000 mm and 700 mm for the last two.

Configurations
As previously mentioned, for this paper, different dimensions will be checked for a virtual wind tunnel to estimate the point from which additional dimensioning is no longer necessary.For this, it was chosen to use as a starting point a base simulation in which to successively increase the dimensions on the 3 axes.The base simulation, in terms pf dimensions, was chosen to be the most constraining model.In addition, for few simulations, different conditions were set for the walls to observe the influential possibilities they could have on the results obtained.Table 1 shows the dimensions used for the test chambers and also the boundary conditions for the walls.

Rest of wall with free slip condition
As can be seen in the Table 1, Simulation 1 will be considered as a base starting point simulation and from this one, dimensions were increased up to Simulation 9.For all this nine-simulation condition for all walls it is to be free slip so, no friction between fluid and wall.Another two iterations are Simulation 10 and Simulation 11 where was tested a condition of rough wall for the floor but it is kept the free slip condition for the other three walls.Dimensions for simulations 10 and 11 were chosen randomly.For an easy identification of the impact generated by each of the configurations in Table 1, it was chosen to present the force on the X axis for each of the 11 iterations.Through this method, having a numerical value, it is possible to determine the simulation that has the best solution, respectively finding the point from which dimensioning is no longer necessary.As can be seen in Table 2 forces on X axis for the 11 th simulations are presented.Focusing on the iterations from 1 to 9, it can be observed that as the dimensions of the test chamber increase from one simulation to another, the value of the force on the X axis decreases, which means that the resistance force is reduced.In parallel, for simulations 10 and 11, the impact of the rough wall condition of the floor can be observed, which significantly influences the results.In terms of airflow, impact can be seen in Figure 5 where is presented a comparison between Simulation 1one with the smallest dimensions of test chamber and Simulation 9one with the largest dimensions of the test chamber.It is very easily visible that for the Simulation 9, especially in the rear part of the vehicle, the airflow allows the generation of a more natural wake compared to Simulation 1 in which the wake is almost non-existent, the airflow being forced to end up suddenly.At the same time, in the front part for Simulation 9, the air has the advantage of covering the body (cabin) of the vehicle much better so that the flow is as natural as possible.It can also be seen on the sides that IOP Publishing doi:10.1088/1757-899X/1303/1/0120438 in Simulation 1 the air is very compressed between the semi-trailer and the walls, which does not allow the simulation of a normal, natural flow.

Conclusion
As it was presented in the lines above, by means of this paper, an attempt was made to present how the dimensions of a wind tunnel, in this case of a test chamber, influence the obtained results.Based on the numerical values achieved by measuring the force on the X axis, a decrease of this index is clearly observed, which concludes that a larger chamber test size also leads to good results, close to reality.The difference between the 1 st simulation and the 9 th one shows the impact that the proximity of the walls has to the tested model and describes the fact that due to an inadequate dimensioning the results obtained can be considered wrong.Although the simulation was stopped for a test chamber value of 29300 mm x 12300 mm x 8400 mm, the values show that an extra increase in the values of dimensions will lead to even better results.Based on values from Table 2 and images from Figure 5, it can be said, in a general way, that the dimensions of the tunnel for a geometric model must be large enough to avoid pressure zones between the walls and the geometry and also to have the possibility to allow the generation of good and natural wake.Regarding the boundary conditions, the influence of the floor wall with a roughness of 8 mm was observed.Compared to Simulation 10, Simulation 11 had about 10% higher value in terms of force on X axis.Therefore, this describes the need to integrate the road conditions (roughness) into the simulations.As a general conclusion, it is important to consider the dimensions of the geometry that will be tested and to choose the appropriate values so that there is no high pressure and friction between walls and model.In addition, the insertion of condition for the runway has the potential to generate better results.

Figure 3 .
Figure 3. Geometrical model for test (values in mm)

Figure 5 .
Figure 5.Comparison of streamlines between Simulation 1 and Simulation 9

Table 1 .
Dimensioning of the test chamber and boundary conditions.

Table 2 .
Forces on X axis.