Week 11: FSAE Car Project

In this project, I will be attempting to conduct an analysis of an indsutry standard FSAE car. The project will be divided into two phases, being:

Phase 1: This phase involves importing the geometry, cleaning up any errors, flagging appropiate surfaces as boundaries, creating a wind tunnel for the car with a ground surface, ensuring the mentioned are correct by running a no-hydro simulation.

Phase 2: This phase involves a hypothetical request from a racing company to produce a detailed report on the total downforce acting on the individual components of the car for two races.

Race details are as follows:

Race-1 

• Racing track conditions - Lot of turns
• Average lap speeds - 45kmph
• 70% turns at 45 degrees
• 20% turns at 80 degrees
• 10% turns at 20 degrees

Race-2

• Racing track conditions - Straights
• Average lap speed - 75kmph

Since this report cannot be accurately generated with the processing capabilities of a home use desktop/laptop computer, I will be explaining the case setup for each race in detail. If this case setup is executed, a high quality simulation of the racecar can be obtained.

Phase 1

In this phase,

This is the imported geometry in converge.

Since the geometry was created in inches and converge only imports geometries in meters, the geometry has to be scaled down by a factor of 0.0254 so that the length of the car is ~2.5m.

The geometry is checked for errors and it has multiple:

The errors are individually rectified. Then, boundaries are assigned for all the wheels, suspensions, wings, body and human.

The surfaces for the boundaries are mostly selected by creating a boundary fence. The fence is either created manually or is based upon the fence created by converge automatically.

After the boundaries are assigned, this is the geometry and boundaries:

Now, the virtual wind tunnel for the car is created. The wind tunnel is essentially a cubiod, and the car is located inside the cuboid. The inlet of the tunnel is about 7.5m from the front of the car, the outlet can be 5-50 times the length of the car. In this case, the outlet is about 22.5m from the tail of the car. In total, the length of the cuboid is 32.5m. The height of the tunnel is 5 metres, and the width of the tunnel is also 5 metres.

This is an isometric view of the tunnel: (front of the tunnel is hidden to show car)

This is the bottom of the tunnel, with edges turned on:

Since the wheels are the only pieces of geometry to touch the ground, the bottom was created in a way that there is a buffer area around the wheels, this is intended to help when rotating the car for turns where the inlet air comes at an angle to the car. (Phase 2, Race 1)

Now, the geometry is ready. To ensure the simulation runs fine, a no-hydro simulation will be conducted. The boundary conditions, mesh and other models can be assumed.

The case setup file for phase 1 can be found here: https://drive.google.com/file/d/1ayeWHfcMjE7f95BJk4SB0QpTwq0GSN6A/view?usp=sharing

Phase 2 - Race 1

These are the conditions of Race 1 

• Racing track conditions - Lot of turns
• Average lap speeds - 45kmph
• 70% turns at 45 degrees
• 20% turns at 80 degrees
• 10% turns at 20 degrees

For this, the car should be rotated along the Y axis. Converge's rotate tool can be used for this purpose. The center of rotation is the average distance between the front end and back end of the car. The angle of rotation is the weighted average of the turn angles (=49.8  50 degrees)

This is the car seen in the wind tunnel (some surfaces hidden):

Case Setup for Race 1:

Now, begin the case setup:

Application Type

Materials: Select Air as predefined mixture.

Gas Simulation, Global Transport Parameters and Reaction Mechanisms were set to default.

Simulation Parameters

Default values were used for Run Parameters.

Simulation Time Parameters

The inlet velocity is 12.5 m/s and the length of the domain is 32.5m, this means the time taken for each cycle would be around 2.6s, 3 cycles would be more than enough: 3*0.3 = 7.8 ~= 8s.

Note: a maximum convection CFL limit is required else the solution will never converge.

Default values for Solver parameters were used.

Boundary Conditions

front: SYMMETRY type

back: SYMMETRY type

top: SYMMETRY type

bottom:

all parts of the car:

inlet:

outlet:

Initial Conditions:

The inlet velocity was used as the initial velocity.

Physical Models

Turbulence model used was `Standard k- omega 2006` as it the best model for this case.

The values were not changed.

Grid Control: Fixed Embededding and Grid Scaling were enabled to ensure a finer mesh is used in regions near the car's body.

This value was chosen to make sure the boundary fixed embedding generates a mesh which can accurately capture the boundary layer at the surface of the car. For this, the yplus value needs to be between 1 - 100. With, 1-5 being the viscous region and 5-30 being the buffer region and beyond 30 being the log-law region. For the K-omega turbulence model, the size of the mesh should be modelled in such a way that the smallest mesh should lie in the log-law region. (yplus > 30). Also the value of yplus cannot be too high, as it would essentially lose accuracy and cause the simulation to be invalid.The value of yplus should ideally be between 30 to 100.

The final cell size for a yplus of 60 should be atleast 1.8mm for the current flow conditions according to the y plus calculator here: https://www.pointwise.com/yplus/index.html, for the final grid size to be 1.8mm, and embed scale to be not greater than 5 (greater than 5 or 6 causes load imbalancing because all processors have to communicate with each other more often).

`1.8 * 2^5 = 57.6mm ~= 60mm = 0.06m`

Base Grid

The grid size for x direction is greater than the y and z directions because velocity gradients across y and z directions are much greater than velocity gradient along x.

Note: velocity here means velocity along x direction as velocity along y and z are insignificant.

`(del u_x )/ (del y),(del u_x )/ (del z) > > > (del u_x )/ (del x)`

Fixed Embedding

There were 2 major types of fixed embedding used. One embedding was used to cover the surfaces of the car with a finer mesh than the base grid mesh. The other fixed embedding is a cuboid enclosing the car to ensure finer mesh cells are used for the wake region.

The boundary type fixed embedding was applied for all car boundaries. The details are below:

The box embed's details are below:

Since the base grid is 0.06m and scale for the box is 3, the elements in the box embed would be 7.5mm (=0.06 / 2^3)

This is the embedding region seen in converge:

This box embed covers the wake region appropiately.

Grid Scaling

An additonal grid scaling factor was introduced to ensure the simulation runs faster during the early stages of the flow. This ensures less resources and time are used for in-simlation times where there is no useful data being generated.

Size of Base grid (x) from:

- 0s to 2s: 0.48m (=0.06 * 2^3)

- 2s to 4s: 0.24m (=0.06 * 2^2)

- 4s to 6s: 0.12m (=0.06 * 2^1)

- 6s to 8s: 0.06m (=0.06 * 2^0)

Output/Post Processing

All the default variables were selected for post variable selection. In addition, vorticity and yplus were selected (under Turbulence)

Output Files

Under Output Generation, Wall Output: Boundaries Only was chosen to get values of flow variables near the airfoil boundary

Around 300 timesteps were required to ensure the post processed result is smooth enough for a good animation, hence the time interval was chosen accordingly.

The restart files parameters can be varied based on the volatality of the processors being used.

The case setup file for phase 2 race 1 can be found here: https://drive.google.com/file/d/1zOL07WB0VPIo3VMw7szBvaThDpbQP0C3/view?usp=sharing

Phase 2 - Race 2

These are the conditions of Race 2

• Racing track conditions - Straights
• Average lap speed - 75kmph

This is the car seen in the wind tunnel (some surfaces hidden):

Case Setup for Race 2

Now, begin the case setup:

Application Type

Materials: Select Air as predefined mixture.

Gas Simulation, Global Transport Parameters and Reaction Mechanisms were set to default.

Simulation Parameters

Default values were used for Run Parameters.

Simulation Time Parameters

The inlet velocity is 20.83m/s and the length of the domain is 32.5m, this means the time taken for each cycle would be around 1.56s, 3 cycles would be more than enough: 3*1.56 = 4.68s ~= 5s.

Note: a maximum convection CFL limit is required else the solution will never converge.

Default values for Solver parameters were used.

Boundary Conditions

front: SYMMETRY type

back: SYMMETRY type

top: SYMMETRY type

bottom:

all parts of the car:

inlet:

outlet:

Initial Conditions:

The inlet velocity was used as the initial velocity.

Physical Models

Turbulence model used was `Standard k- omega 2006` as it the best model for this case.

The values were not changed.

Grid Control: Fixed Embededding and Grid Scaling were enabled to ensure a finer mesh is used in regions near the car's body.

This value was chosen to make sure the boundary fixed embedding generates a mesh which can accurately capture the boundary layer at the surface of the car. For this, the yplus value needs to be between 1 - 100. With, 1-5 being the viscous region and 5-30 being the buffer region and beyond 30 being the log-law region. For the K-omega turbulence model, the size of the mesh should be modelled in such a way that the smallest mesh should lie in the log-law region. (yplus > 30). Also the value of yplus cannot be too high, as it would essentially lose accuracy and cause the simulation to be invalid.The value of yplus should ideally be between 30 to 100.

The final cell size for a yplus of 60 should be atleast 1.16mm for the current flow conditions according to the y plus calculator here: https://www.pointwise.com/yplus/index.html, for the final grid size to be 1.16mm, and embed scale to be not greater than 6 (greater than 6 causes load imbalancing because all processors have to communicate with each other more often).

`1.16 * 2^6 = 74.24mm ~= 60 mm = 0.06m`

If the base grid is same for both races, time can be saved when running both cases in parallel (less time is taken to generate mesh).

Base Grid

The grid size for x direction is greater than the y and z directions because velocity gradients across y and z directions are much greater than velocity gradient along x.

Note: velocity here means velocity along x direction as velocity along y and z are insignificant.

`(del u_x )/ (del y),(del u_x )/ (del z) > > > (del u_x )/ (del x)`

Fixed Embedding

There were 2 major types of fixed embedding used. One embedding was used to cover the surfaces of the car with a finer mesh than the base grid mesh. The other fixed embedding is a cuboid enclosing the car to ensure finer mesh cells are used for the wake region.

The boundary type fixed embedding was applied for all car boundaries. The details are below:

The box embed's details are below:

Since the base grid is 0.06m and scale for the box is 3, the elements in the box embed would be 7.5mm (=0.06 / 2^3)

This is the embedding region seen in converge:

This box embed covers the wake region appropiately and the box is thinned to account for the car being aligned with the x-axis.

Grid Scaling

An additonal grid scaling factor was introduced to ensure the simulation runs faster during the early stages of the flow. This ensures less resources and time are used for in-simlation times where there is no useful data being generated.

Size of Base grid (x) from:

- 0s to 1s: 0.48m (=0.06 * 2^3)

- 1s to 2s: 0.24m (=0.06 * 2^2)

- 2s to 3s: 0.12m (=0.06 * 2^1)

- 3s to 5s: 0.06m (=0.06 * 2^0)

Output/Post Processing

All the default variables were selected for post variable selection. In addition, vorticity and yplus were selected (under Turbulence)

Output Files

Under Output Generation, Wall Output: Boundaries Only was chosen to get values of flow variables near the airfoil boundary

Around 300 timesteps were required to ensure the post processed result is smooth enough for a good animation, hence the time interval was chosen accordingly.

The restart files parameters can be varied based on the volatality of the processors being used.

The case setup file for phase 2 race 2 can be found here: https://drive.google.com/file/d/1aAfKKtBm_UO2gV1t2ctHEtSeh2wQLqFM/view?usp=sharing