Conjugate Heat Transfer (CHT) Analysis on a desktop graphics card using ANSYS FLUENT.

Objective

1. Find out the maximum temperature attained by the processor.

2. Prove that the simulation has achieved convergence with appropriate images and plots.

3. Find out the heat transfer coefficient in appropriate areas of the model.

4. Identify potential hotspots on the model.

Conjugate Heat Transfer (CHT)

Conjugate heat transfer (CHT) is a process that involves the variation of temperature within solid and fluid due to thermal interaction between them. CHT analysis allows the simulation of the heat transfer between solid and fluid domains by exchanging thermal energy at the interfaces between them.

CHT analysis is advantageous in accurately simulating the surface temperature and heat exchanging capabilities of various devices and engineering machines. With the help of CHT analysis, we can optimize the thermal properties during the product design phase and save millions of dollars before actually producing the prototype or the final machine.

Applications:

CHT analysis helps to understand the effectiveness of various thermal systems involving solid and fluid medium. It helps us to simulate and understand thermal systems like heat sinks, coolers, heat exchangers, exhaust ports of vehicles, Internal combustion engines, boilers, HVAC, Electronic components cooling, etc.

Graphics card

A graphics card is an expansion card which generates a feed of output images to a display device. At the core of graphics card is the graphics processing unit (GPU), which is the main part that does the actual computations. The graphics processor can perform additional processing, removing some load from the central processing unit of the computer. The graphics card from NVIDIA and AMD can use the computing capabilities of the graphics processor to solve non-graphics tasks.

Case 1: CHT analysis on baseline mesh

1. Geometry

Figure 1: Enclosure

Figure 2: Graphics card

• To perform CHT analysis on graphics card an enclosure has been created around the graphics card. Air will be used as cooling fluid.
• The graphics card contains three main components. The orange part is the base and the part above it is the fins. And the processor is sandwiched in between them.
• The base and the fins will conduct the heat away from the processor. And that heat will be carried away by cooling air through forced convection.
• To generate a conforming mesh we have to enable the shared topology.

2. Meshing

• Before generating the mesh we have to create the named selections. The required named selections are inlet, outlet, symmetry wall for the enclosure walls, base wall, fin wall, and processor wall.
• For case 1 we will create a baseline mesh. Mesh details are given below.
• Element size for enclosure: 0.005 m.
• Element size for all components of the graphics card: 0.001 m.
• Nodes: 38,701.
• Total Elements: 213,540.

Figure 3: Meshed Enclosure

Figure 4: Meshed Graphics card

Figure 5: Meshed Base and processor

Figure 6: Section plane

Figure 7: Section plane zoomed view

Figure 8: Mesh metrics

3. Setup

• Steady-state analysis.
• Solver: pressure based.
• Velocity formulation: absolute.
• Solution method used: SIMPLE.
• Viscous model: k-omega.
• Solid material: Aluminium, Copper, Silicon, Fluid: air.
• Solution method: SIMPLE.
• Initialization method: Hybrid.
• Number of iterations: 500.

Cell zone conditions:

• Enclosure: Fluid-Air.
• Graphics card Base: Solid-Copper.
• Graphics card fins: Solid-Aluminium.
• Graphics card processor: Solid-Silicon.
• We have to create silicon, as it is not available in Ansys fluent material database.
• Processor heat generation rate: 2.534e008 W m^-3. (Maximum heat generation rate of NVIDIA GTX 960M graphics card).

Boundary conditions:

• Inlet: velocity = 5 m/s, Temperature = 300 K.
• Outlet: pressure = 0 pascals (Guage Pressure).
• Lateral walls of the enclosure: Symmetry.

4. Solution

Figure 9: Residuals

• The above plot is of residuals, here we can see that the solution has converged at 200 iterations.

Figure 10: Temperature contour on base and fins

Figure 11: Temperature contour on base and processor

• In the above two images, we can see that the heat is spreading outward from the center of the graphics card.
• The maximum temperature reached by the processor is 379.49 K (106.34 'C).

Figure 12: Temperature contour on a plane

• Here we can see that the results are a bit coarse near the graphics card walls due to the coarse mesh. We need to refine the mesh to remove this inaccuracy in the results.

Figure 13: Velocity contour on a plane

Figure 14: Wall heat transfer coefficient

• In the above image, we can see the wall heat transfer coefficient, the maximum heat transfer coefficient reached in case 1 is 1150.87 W m^-2 k^-1.

5. Potential hot spots

To see the potential hotspots we have to change the temperature range so the temperature contour will better visualize the hotspots. The lower temperature range is increased to 367.26 K.

Figure 15: Hot spots at the processor and the base

Figure 16: Hot spots at fins

Figure 17: Hot spots at fins zoomed view

• The hotspots are seen in the center at the processor as expected. The heat is spreading out from the processor. The other hotspot is seen at the center of the fin.

Case 2: CHT analysis on a refined mesh

1. Geometry

The geometry is same as used in case 1, in case 2 we will refine the mesh to check the results obtained in case 1.

2. Meshing

In case 2 we have to refine the mesh to better simulate the CHT analysis on the graphics card. In the graphics card, the processor is the only source of heat generation. So we need to refine the processor and mesh it with very fine elements. The other region for mesh refinement is base and the fin. The heat generated by the processor is conducted away by the base and fins. That why it is essential to refine the mesh around the base and fin. Details of mesh refinement are given below.

• Before generating the mesh we have to create the named selections. The required named selections are inlet, outlet, symmetry wall for the enclosure walls, base wall, fin wall, and processor wall.
• For case 1 we will create a baseline mesh. Mesh details are given below.
• Element size for enclosure: 0.0025 m.
• Body sizing for processor: 0.00015 m.
• Body sizing for the graphics card base: 0.0005 m.
• Body sizing for the graphics card fins: 0.0005 m.
• Nodes: 201,449.
• Total Elements: 1,066,383.

Figure 18: Meshed Enclosure

Figure 19: Meshed Graphics card

Figure 20: Meshed Base and processor

Figure 21: Meshed Processor

Figure 22: Section plane

Figure 23: Section plane zoomed view

Figure 24: Mesh metrics

3. Setup

• Steady-state analysis.
• Solver: pressure based.
• Velocity formulation: absolute.
• Solution method used: SIMPLE.
• Viscous model: k-omega.
• Solid material: Aluminium, Copper, Silicon, Fluid: air.
• Solution method: SIMPLE.
• Initialization method: Hybrid.
• Number of iterations: 500.

Cell zone conditions:

• Enclosure: Fluid-Air.
• Graphics card Base: Solid-Copper.
• Graphics card fins: Solid-Aluminium.
• Graphics card processor: Solid-Silicon.
• We have to create silicon, as it is not available in Ansys fluent material database.
• Processor heat generation rate: 2.534e008 W m^-3. (Maximum heat generation rate of NVIDIA GTX 960M graphics card).

Boundary conditions:

• Inlet: velocity = 5 m/s, Temperature = 300 K.
• Outlet: pressure = 0 pascals (Guage Pressure).
• Lateral walls of the enclosure: Symmetry.

4. Solution

Figure 25: Residuals

• The above plot is of residuals, here we can see that the solution has converged at 250 iterations.

Figure 26: Temperature contour on base and fins

Figure 27: Temperature contour on base and processor

• The heat is spreading out from the center to outward of the graphics card. Here in case 2, we can see that the maximum temperature reached by the processor is 377.23 K (104.08 'C).

Figure 28: Temperature contour on a plane

Figure 29: Velocity contour on a plane

• Temperature contour and velocity contour is not coarse as compared to case 1, mesh refinement has led to the much more refined and proper contours near the walls of the graphics card.

Figure 30: Wall heat transfer coefficient

• In the above image, we can see the wall heat transfer coefficient, the maximum heat transfer coefficient reached in case 1 is 2035.56 W m^-2 k^-1.

5. Potential hot spots

To see the potential hotspots we have to change the temperature range so the temperature contour will better visualize the hotspots. The lower temperature range is increased to 367.26 K.

Figure 31: Hot spots at the processor and the base

Figure 32: Hot spots at fins

Figure 33: Hot spots at fins zoomed view

• The hotspots are seen in the center at the processor as expected. The heat is spreading out from the processor. The other hotspot is seen at the center of the fin.

Conclusion

• Conjugate heat transfer analysis was performed on the graphics card in case 1 with a baseline coarse mesh and in case 2 with much finer mesh.
• The finer mesh led to a change in temperature contour, velocity contour, and heat transfer coefficient.
• The comparison of the two cases is given below.

Comparison: