Week 6 - CHT Analysis on a Graphics card
AIM -
To perform Conjugate Heat Transfer analysis on a graphics card.
OBJECTIVES -
- To select appropiate materials of the various parts of the graphic card for the simulation.
- To create a combination of coarse and fine mesh in different regions
- To perform the simulation for varying valocity of 1 m/s to 5 m/s for at least 3 velocities.
- To find out the maximum temperature and heat transfer coefficeint attained by the processor
INTRODUCTION -
We have a model similar to the graphics card on which we will be performing conjugate heat transfer analysis for obtain the maximum temperature and heat transfer coefficient on the processor of the graphic card . that processor is the major source of heat generation with the graphic card system as it is used to perform calculations and the rest of the components like the base and the fins are used for the connections and cooling respectively. The fins are placed on the processor such that most of the heat generated can be cconducted to the fins and as the air passes over the fins there is convection that helps to release the heat into the air. for this purpose the fins are made from high conductivity material like copper. At the same time , we do not want the heat generate in the processor to flow through the base and reach other electronic components, so the material of the base is made up of silicon or wood.
so according to this we have use the folllowing materials of the various components as shown in the table.
| Component | Material |
| Fins | Copper |
| Processor | Aluminium |
| Base | Wood |
| Enclosure | Air |
We have created an enclosure around the graphic card body to simulate external flows.
The steps to simulate the CHT analysis are
- load the geometry into spaceclaim and create a share topology throught the geometry
- meshing the geometry with required sizing
- loading the mesh into the fluent setup and setting up the physics
- setting up the solver setting and calculate the solution
- postporcessing the solution to plot various graphs and contours.
BASELINE MODEL -
for the base line setup we will be using a very coarse mesh. But first we will have to create porper geometry that can be used for meshing so for that import the geometry file in spaceclaim.
Next we will go to the workbench menu from the ribbon and select share to share the topology of entire graphic card body, it will turn all the shared faces and edge to red , we also get a message shown the number of faces and edge that are shares , now click on the green tick mark to complete share topology.
we can now see that the edges of the shares boundaries are turned blue and in the tree or structure when we look for the properties of the graphic card it shows share for the share topology. This means our meah is conforman and can be used for Conjugate heat transfer analysis. Now our geometry is ready for the meshing.
Moving to the mesh part , we will first try to create the default mesh the mesh can generate and look for the quality of this mesh.
now adding named selections for the inlet and outlet
naming the enclosure wall
naming the walls of the processor
form the above images, we can see the mesh formed by the default settings and it has very coarse sizing also the image from the section plane shows that the elements are conformal due to shared topology.
the mesh metrics shows that the element quality for few elements is less than 0.10 but as this is just our baseline setup it can be used.
alos here we have about 85000 elements so there is a lot more scope to refine this mesh further.
we will use this mesh to simulate the flow over the graphic and perform CHT.
To do that we will be loading up this mesh into flunet setup. Moving to fluent we will use a double precision solver with 4 paralllel procesors.
We will use a presure based solver with steady state condition as we are just interested in the final solution rather than what happens during the simulation.
to set up the physics we will turn on energy equation and use standard k-epsilon tubulance model
Now we will add materials wood and copper from the fluent database in the edit and create dialouge box by copying them .
now we have aliminium, copper and wood as the solids and air as the fluid
now we will set the cell zone with respective materials as aluminium for processor , copper for fins , wood for base and air for the enclosure.
also we will be adding the heat source in the porcessr of 20 W so when we input the value of the heat source in W/mm^3 the value turn out to be 312500000 W/mm^3
Now we set up the boundary conditions as 5 m/s for the inlet vlocity and we have a pressure outlet.
now lastly we will add a report definition for the volume average temperature for the processor
now we will initialize the solution and calculate for about 150 iterations but it converges earlier at about 120 iterations.
Results for flow velocity = 5 m/s :
Residuals:
Volume Average temperature plot for the processor
from the above plots we can see that the residuals and the vol average temperature have stabilized so we can conclude that the solution is converged.
the temperature contour of the graphic card is shown below
also we plot the temperature contour on the plane the cuts the graphic card in half
the surface heat transfer coefficient pot
the wall heat transfer coefficient
now we will change the velocity of flow as 3 m/s and agian run the simulation
now we will again run the simulation and see the results
Results for flow velocity = 3 m/s :
Residuals:
Volume averaged temperature of the processor
we can see that the volume averaged temp is stabilised and there is very less fluctuation in the residuals so we can conclude that the solution is converged
the temperature contour for the graphic card body is shown below
also we plot the temperature contour on the plane the cuts the graphic card in half
the surface heat transfer coefficient pot
the wall heat transfer coefficient
now we will change the velocity of flow as 1 m/s and agian run the simulation
now we will again run the simulation and see the results
Results for flow velocity = 1 m/s :
Residuals:
Volume averaged temperature of the processor
we can see that the volume averaged temp is stabilised and there is very less fluctuation in the residuals so we can conclude that the solution is converged
the temperature contour for the graphic card body is shown below
also we plot the temperature contour on the plane the cuts the graphic card in half
the surface heat transfer coefficient pot
the wall heat transfer coefficient
Following table shows the comparison of various properties with change in the flow velocity
| FLow velocity | Volume averaged temperature of the processor | Surface heat transfer coefficient of the processor | Wall heat transfer coefficeint of the processor |
| 5 m/s | 569[K] | 1.06e+3[W/(m^2_K)] | 3.51e+6[W/(m^2_K)] |
| 3 m/s | 636[K] | 8.59e+2[W/(m^2_K)] | 3.50e+6[W/(m^2_K)] |
| 1 m/s | 836[K] | 5.37e+2[W/(m^2_K)] | 3.50e+6[W/(m^2_K)] |
From the table we can see as the flow velocity decreases the average temp of the porcessor increases. we see reduction in the surface heat transfer coefficient of the processor as the flow velocity decreases.
Now we will refine our mesh to capture better and accurate results
REFINEMENT MODEL 1 -
we will be using the same geometry that we used for the earlier case and load it to the meshing ans add sizing to various components
first we add body sizing to the processor of 0.25 mm
next add body sizing to the fins component of 1mm.
finally adding a body sizing of 4 mm to the external enclosure
then generate mesh with this refinements. the mesh looks like as shown below
form the section view we can see different sizings added.
the mesh has better element quality than the baseline model , here is it above 0.10 that means our mesh is good
the mesh has about 0.25 million elements
now we will use this mesh for the simulation and compare the results of the different flow velocities.
Results for flow velocity = 5 m/s :
Reisduals -
Volume averaged temperature for the porcessor
from the above contours we can see that the volume averaged temp is stabalized and also the residuals are fluctuating in a very small range so we can conclude that the solution is converged.
Temp contours
temp contour for the section plane
Surface heat transfer coefficent for the porcessor
Wall heat transfer coefficeint for the processor
Results for flow velocity = 3 m/s :
Reisduals -
Volume averaged temperature for the porcessor
from the above contours we can see that the volume averaged temp is stabalized and also the residuals are fluctuating in a very small range so we can conclude that the solution is converged.
Temp contours
temp contour for the section plane
Surface heat transfer coefficent for the porcessor
Wall heat transfer coefficeint for the processor
Results for flow velocity = 1 m/s :
Reisduals -
Volume averaged temperature for the porcessor
from the above contours we can see that the volume averaged temp is stabalized and also the residuals are fluctuating in a very small range so we can conclude that the solution is converged.
Temp contours
temp contour for the section plane
Surface heat transfer coefficent for the porcessor
Wall heat transfer coefficeint for the processor
following table shows the comparison for the change in flow velocity
| Flow velocity | Volume averaged temperature for the processor | Surface heat transfer coefficient for the processor | Wall heat transfer coefficient for the processor |
| 5 m/s | 505[K] | 3.20e+3[W/(m^2_K)] | 1.17e+7[W/(m^2_K)] |
| 3 m/s | 563[K] | 2.58e+3[W/(m^2_K)] | 1.17e+7[W/(m^2_K)] |
| 1 m/s | 781[K] | 1.44e+3[W/(m^2_K)] | 1.17e+7[W/(m^2_K)] |
we can see as the flow velocity decreases the volume average temperatur for the processor increase and the surface heat transfer coefficient decreases.
REFINEMENT MODEL 2 -
Now in this refinement we have introduce another enclosure that is smaller in size and it is situated inside the enclosure that we have used earlier.
this inner enclosure has a top, left and right face 10 mm away , the front anf back faces are 5 mm away and the bottom face is at a distance of 1mm from the graphic card body.
usign the interference option in the prepare tab we have removed the common volume occupied by both the enclosures after that we have shared the topology from the workbench tab.
the final geometry looks as shown below
we have load this geometry into the meshing to mesh it.
here we have used various refinements shown below
the enclosure on the outside is given a body sizing of 5 mm
the inner enclosur is given a body sizing of 1.5 mm
the processor is given body sizing of 0.25 mm
the fins are given the body sizing of 1 mm.
now the mesh is generated ith usign this refinements
here we can see that the mesh is coarse on the outside and it is fine near the processor.
the metrics for the element quality is shown below
the lowest quality is 0.0988 at only a few element has that quality so we can say that our mesh is fairly good for the simulation.
also the number of elements has increased in this meshing
there are about 0.45 million elements that is good and now we use this mesh for the simulation for varying flow velocities.
The results are discussed below
Results for flow velocity = 5 m/s :
Reisduals -
Volume averaged temperature for the porcessor
from the above contours we can see that the volume averaged temp is stabalized and also the residuals are fluctuating in a very small range so we can conclude that the solution is converged.
Temp contours
temp contour for the section plane
Surface heat transfer coefficent for the porcessor
Wall heat transfer coefficeint for the processor
Results for flow velocity = 3 m/s :
Reisduals -
Volume averaged temperature for the porcessor
from the above contours we can see that the volume averaged temp is stabalized and also the residuals are fluctuating in a very small range so we can conclude that the solution is converged.
Temp contours
temp contour for the section plane
Surface heat transfer coefficent for the porcessor
Wall heat transfer coefficeint for the processor
Results for flow velocity = 1 m/s :
Reisduals -
Volume averaged temperature for the porcessor
from the above contours we can see that the volume averaged temp is stabalized and also the residuals are fluctuating in a very small range so we can conclude that the solution is converged.
Temp contours
temp contour for the section plane
Surface heat transfer coefficent for the porcessor
Wall heat transfer coefficeint for the processor
following table shows the comparison for the change in flow velocity
| Flow velocity | Volume averaged temperature for the processor | Surface heat transfer coefficient for the processor | Wall heat transfer coefficient for the processor |
| 5 m/s | 503[K] | 3.260+3[W/(m^2_K)] | 1.19e+7[W/(m^2_K)] |
| 3 m/s | 561[K] | 2.58e+3[W/(m^2_K)] | 1.18e+7[W/(m^2_K)] |
| 1 m/s | 781[K] | 1.44e+3[W/(m^2_K)] | 1.19e+7[W/(m^2_K)] |
we can see as the flow velocity decreases the volume average temperatur for the processor increase and the surface heat transfer coefficient decreases.
Comparison for various model at flow velocity = 5[m/s]
| MODEL | Volume averaged temp for the processor | Surface heat transfer coeff for the porcessor |
| BASELINE | 569[K] | 1.06e+3[W/(m^2_K)] |
| REFINEMENT 1 | 505[K] | 3.20e+3[W/(m^2_K)] |
| REFINEMENT 2 | 503[K] | 3.26e+3[W/(m^2_K)] |
CONCLUSIONS -
- With increasing the flow velocity the temperature of the processor reduces.
- With increasing the flow velocity the surface heat transfer coefficient of the porcessor increases.
- The wall heat transfer remains neearly the same when the flow velocity is changed
- We should have a high flow velocity inorder to keep the porcessor cool