Conjugate Heat Transfer Analysis on a graphics card.
AIM
Conjugate Heat transfer analysis on a graphic card model.
OBJECTIVES
1) To Perform a steady-state conjugate heat transfer analysis on a model of a graphics card.
2) To understand the mesh effect in the different regions using a coarse and fine mesh.
3) To understand the effect of velocity on the graphic card model using different velocities.
4) Find out the maximum temperature and heat transfer coefficient attained by the processor.
INTRODUCTION
Conjugate heat transfer refers to a heat transfer process involving the interaction of conduction within a
solid body and the convection from the solid surface to fluid moving over the solid surface. Therefore, a realistic analysis of conjugate heat transfer problems necessitates the coupling of the conduction in the solid and the convection in the fluid. This is generally accomplished by satisfying the conditions of continuity in temperature and heat flux at the fluid-solid interface.
In the graphic card, There are 3 main parts,
1) Fins (the extended surface which uses to increase the heat transfer rate)
2) Base (Pcb)
3) Processor (run using the electrical power and generate some heat)
A short fan is provided for the cooling purpose near the graphic card for reducing the heating problems. In our study air comes from the inlet with some velocity and takes some amount of heat from the fins and gives the cooling effect.
Assumptions
1) The processor is the only heat source in the entire graphic card. (Vol= 8x8x1= 64 mm^3)
2) Power consumption by the processor 9.6 watts.
3) The heat generated with no loss = 9.6/10^-9
=15*10^7 W/m^3
Methodology
1. Modeling of Graphic Card
2. Meshing
3. Pre-processing
4. Setting-up physics and solving
5. Post-processing
1) Modeling
3D model of graphic card is generated using any cad software then import into the ANSYS Spaceclaim.
The enclosure is created to see the effect of velocity on the graphic card model and enable the shared topology to transfer the information among them.
2) Meshing
In CFD meshing we should care about the mesh size because large mesh size will not predict the correct solution near the exact solution and we will not choose the small mesh size with proper justification because it takes more computation time. So we start with the Baseline mesh and set up the solver for the simulation. Once the simulation is run then reduce the element size to get the result close to exact.
Start with name defining process.
CASE 1: BASELINE MESH
Start with the baseline mesh first to check the solver.
Default Element size 10.289 mm
Nodes: 15765
Elements: 83786
3. Pre-processing
Check the mesh first after that defines the solver.
Simulation Type: Steady-State Simulation
Solver: Pressure Based
Use the Energy Equation and select the K-epsilon turbulence model with standard wall function.
there are basically two solid and fluid cell zones created here so provide the material accordingly.
Fins: Copper (solid)
Base (PCB): Aluminium (solid)
Processor: Gold (solid)
we already take an assumption that the processor is the only heat source in the entire graphic card so provide a source term (15x10^7 w/m^3)
Enclosure: Air (fluid)
4. Setting-up physics and solving
Providing Boundary conditions.
Inlet Velocity: Take 2 m/s for the baseline mesh case.
Inlet Temperature: 300k
Outlet: Pressure 0
rest all boundary conditions take fluent itself accordingly.
The current set up is initialized using hybrid initialization and after running for the 500 iterations
a) Residual Plot
b) Area Weighted Avg Temp
c) Temperature plot on the plane
d) Velocity plot on the plane
e) Temperature Contour on the graphic card
f) Wall Heat transfer Coefficient plot on plane
g) Hotspot Regions (Fins, Base, Processor)
h) Wall heat transfer Coefficient (fin, Base, Processore)
Since we run the set-up for a baseline mesh we get the results for 2 m/s Inlet velocity as shown above and the solution is converged near 170 iterations. The temperature plot shows the variation of total temperature and maximum temperature obtained after the heat transfer from the processor is 390k. We also get the velocity and wall heat transfer coefficient plot on the plate. But we need a mesh refinement near the contact region to obtain the result close to the exact solution. The Graphic card model is complex geometry so we can't create the inflation layers and also we don't create a very fine mesh refinement without proper justification because it takes too much computational time. So start the refinement using body sizing for the FINS, BASE, and Processor.
CASE 2: REFINED MESH
In this case, we refined the mesh using body sizing of different parts. Here I am using the academic version of Ansys so I can't go below .5 million cells. Also here we checked the mesh quality and it should not go below 5 % otherwise it shows bad results.
Default Element size 2.5 mm
a) Fin: Body sizing 0.5 mm
b) Processor: body sizing 0.5 mm
c) Base: body sizing 1.0 mm
Nodes: 85292
Elements: 480099
Fig: Refined Mesh (Enclosure with graphic card)
Fig: Refined Mesh cut plane view
Fig: Refined Meshon Graphic card
Fig: Mesh Quality
Pre-processing
Check the mesh first after that defines the solver.
Simulation Type: Steady-State Simulation
Solver: Pressure Based
Use the Energy Equation and select the K-epsilon turbulence model with standard wall function.
there are basically two solid and fluid cell zones created here so provide the material accordingly.
Fins: Copper (solid)
Base (PCB): Aluminium (solid)
Processor: Gold (solid)
we already take an assumption that the processor is the only heat source in the entire graphic card so provide a source term (15x10^7 w/m^3)
Enclosure: Air (fluid)
4. Setting-up physics and solving
Providing Boundary conditions.
Here we run the case for three Velocities (1 m/s, 2m/s and 5 m/s)
i) Velocity:. 1 m/s
Inlet Velocity: Take 1 m/s for the baseline mesh case.
Inlet Temperature: 300k
Outlet: Pressure 0
rest all boundary conditions take fluent itself accordingly.
The current set up is initialized using hybrid initialization and after running for the 250 iterations
a) Residual Plot
b) Area Weighted Avg Temp
c) Temperature plot on the plane
d) Velocity plot on the plane
e) Temperature Contour on the graphic card
f) Wall Heat transfer Coefficient plot on plane
g) Hotspot Regions (Fins, Base, Processor)
h) Wall Heat Transfer Coefficient (Fins, Base, Processor)
ii) Velocity:. 2 m/s
Inlet Velocity: Take 2 m/s for the baseline mesh case.
Inlet Temperature: 300k
Outlet: Pressure 0
rest all boundary conditions take fluent itself accordingly.
The current set up is initialized using hybrid initialization and after running for the 500 iterations
a) Residual Plot
b) Area Weighted Avg Temp
c) Temperature plot on the plane
d) Velocity plot on the plane
e) Temperature Contour on the graphic card
f) Wall Heat transfer Coefficient plot on plane
g) Hotspot Regions (Fins, Base, Processor)
h) Wall Heat Transfer Coefficient (Fins, Base, Processor)
iii) Velocity:. 5 m/s
Inlet Velocity: Take 5 m/s for the baseline mesh case.
Inlet Temperature: 300k
Outlet: Pressure 0
rest all boundary conditions take fluent itself accordingly.
The current set up is initialized using hybrid initialization and after running for the 250 iterations
a) Residual Plot
b) Area Weighted Avg Temp
c) Temperature plot on the plane
d) Velocity plot on the plane
e) Temperature Contour on the graphic card
f) Wall Heat transfer Coefficient plot on plane
g) Hotspot Regions (Fins, Base, Processor)
h) Wall Heat Transfer Coefficient (Fins, Base, Processor)
Results and Observation:
From the above simulation, we get the result for baseline and refined mesh for different velocities and described by the table
| Baseline Mesh |
Refined Mesh Case (i) |
Refined Mesh Case (i) |
Refined Mesh Case (i) |
|
| Velocity (m/s) | 2 | 1 | 2 | 5 |
| Max Total Temperature of entire Graphic Card Model | 390.8 | 430.1 | 378 | 343.9 |
| Max Area Weighted Avg total Temp | 333 | 420 | 369 | 333 |
| Avg of Area weighed total Avg Temp | 313.41901 | 325.74128 | 314.1284 | 306.85773 |
| Area Avg Wall Heat Transfer Coefficient of Fin | 63.6736 | 256.067 | 257.735 | 261.367 |
| Area avg Wall Heat Transfer Coefficient of Base | 85.9601 | 163.417 | 165.267 | 169.202 |
| Area Avg Wall Heat Transfer Coefficient of Processor | 33.4622 | 402.926 | 406.243 | 417.543 |
| Max Temp of Processor | 390 | 430 | 377 | 343 |
We run the simulation for all three velocities and get a converged solution as shown from the residual plots.
We start with the baseline mesh and after refinement we get the solution close to exact and on the based of these we can say that Baseline mesh is not good to capture the temperature distribution on the interface zone. It can also be observed by the temperature contours.
Fins are used to increase the heat transfer rate and in this model, it is situated above the processor and base placed below the processor. When we increase the velocity from 1 to 5 m/s we can see that there will be a small change in the heat transfer coefficient of fin and base while on the other side they both connected with the processor so it will not cause a major change. But, processor-fin heat transfer increasing which transfers more heat from the processor to surrounding due to an increase in velocity and temperature of the processor start decreasing that increases the heat transfer rate of fin-air and base-air.
Conclusions
1) Velocity directly affects the heat transfer rate and increases the heat transfer coefficient.
2) Mesh plays important roles for capture the heat transfer in complex geometry.
4) Inflation layers are not generated in the very complex geometry so we choose body sizing to refine the mesh and get close to exact simulation results.
3) Fins play an important role because it is used to increase the heat transfer coefficient. so it should be chosen accordingly.