Conjugate Heat transfer Analysis on Graphic card of a Computer with two different mesh count

Conjugate Heat transfer Analysis on Graphic card of a Computer with two different mesh count

Introduction:

A graphic card is an electronic sub component which can be seen in PC’s, Laptops and play station. It emits huge amount of heat during its working condition. During its continuous working for longer duration it may result in thermal failure if is not ventilated properly. Nowadays the graphic card and the similar electronic components are equipped with heat sink which enhances the heat dissipation from the component with the help of cooling fan. Heat sink is nothing but the group of fins with base plate made up of aluminium or aluminium alloy.

Fig 1: Geometrical model of graphic card – heat sink assembly with fluid domain

In this study conjugate heat transfer analysis is performed on a simple graphic card – heat sink assembly which is shown in Fig 1. The reason for CHT analysis is because the heat source is considered to be in the form of volumetric heat generation in the graphic card volume and the generated heat is dissipated in conduction mode at first through graphic card, heat sink and base plate material after that from those surfaces in convection mode heat is dissipated to forced air from the cooling fan.

The complication in this problem is multiple volumes in the region of study and different materials for the volumes. The fin thickness is very small and the aspect ratio in all the perspective will be poor. So the meshing strategy should be given high importance for capturing the significant properties over the heat sink.

Pre Processing:

In pre-processing stage a fluid volume is generated around the graphic card assembly with the specified dimension using enclosure option. Without removing the duplicates topology is shared to get the multiple volumes in the domains.

In I model, meshing is done as a base line setup where mesh is coarse over the surface of the heat sink – graphic card assembly and the edges on it is poorly meshed by one cell covering its thickness. It is shown in the Fig 2 given below. The mesh count is around 10 lakh for this base line model. 90 % of the meshes are constructed by tetrahedral shape only. The quality of the mesh is up to a satisfiable range only.

Fig 2: Mesh model I of the Graphic card – heat sink assembly

In II Model, meshing is very concentrated and fine meshes were created over the surface of the assembly and the edges well covered and meshed using edge sizing and face sizing options. Quality of the mesh also a good one for this model. The total mesh count reached around 30 lakh. In Fig 3, II model of mesh figure is shown, in that we can notice the edges and surfaces covered with fine meshes.

Fig 3: Mesh model II of the Graphic card – heat sink assembly

Meshing model I & II is verified whether the solid and fluid volumes were interconnected with conformal meshing or not. Conformal meshing exists between each and every volumes in the domain.

Solving:

For solving the above two models we assign the boundary conditions, solvers and models. Steady state pressure based absolute velocity formulation solver with energy and standard K-e turbulence models are enabled to solve our problem.

• Boundary condition:

Velocity inlet applying to the inlet wall with velocity = 1 m/s (the value is got from online sources for the air velocity by cooling fan of electronic components)

Volumetric heat generation from major source of our study which is the graphic card is the assigned a value of 1.25 x 10^8 W/m³. This value also got from online source by referring for heat dissipation by electronic components, we got an estimated value of 80 W by the component and then from that value we computed the volumetric heat generated by that graphic card volume. Likewise other small heat source elements are present in that assembly. Volumetric heat generated by those elements also estimated

HS1, HS2, HS3, HS4 = 83333333.33 W/m³

HS5, HS6 = 5319148.9 W/m³

Using hybrid initialization the mesh model I setup is initialized and the iteration for solution is initiated.

Fig 4: Scaled residuals plot of the Mesh model I	Fig 5: Scaled residuals plot of the Mesh model II

After crossing several iterations the scaled residuals failed to converge but yields a steady pattern in the variation of scaled residuals, so we decided to consider this steady pattern as manually converged solution. Likewise mesh model II also analyzed with same boundary condition, solvers and models. That model also behaves in same manner and converged manually around 2000 iterations as like mesh model I. The scaled residuals plot of mesh model I & II can be seen in the above Fig 4 & 5.

Results:

Fig 6: Temperature contour on Graphic card – heat sink assembly of Mesh model I.

The above Fig 6 shows the temperature contour on Graphic card – heat sink assembly of Mesh model I, in that we can notice the temperature range betweeten 393 K to 375 K. The fin plate and graphic card surface experiences the maximum temperature. Due to the flow of cooling air the first row fins have reduced temperature that too at sharp corners of the fins have minimum temperature spots. This minimum temperature spot on the corners due to the increase in velocity of air near those corners because of hydrodynamic acceleration.

Fig 7: Surface heat transfer coefficient contour on Graphic card – heat sink assembly of Mesh model I.

From the Fig 7, we could spot the surface heat transfer coefficient at the fin corner sharp edges were higher. The below Fig 8 shows the contour plot of surface nusselt number, on examining that figure there also we could spot the maximum Nusselt no of 10129.76 is reached at the same fin sharp corners of first row fins. So the convective heat transfer is predominant in those spots. It is evident that the intensity of heat transferred is higher in that spot.

Fig 8: Nusselt number contour on Graphic card – heat sink assembly of Mesh model I.

Fig 9: Temperature contour on  heat sink fin plate of Mesh model I.

The above Fig 9 shows the temperature distribution contour over the fin plate of mesh model I, where we could notice the reduced temperature region is spotted near first row of fins only and the hot region is spotted in the adjacent region of graphic card processor. Due to the poor quality of mesh we could notice the distended contour plots in the edges. Edges are not well defined, this may give deviation in the actual results.

Fig 10: Temperature contour on Graphic card – heat sink assembly of Mesh model II.

The temperature distribution over graphic card – heat sink assembly of mesh model II is shown in the Fig 10. The temperature ranges between 376 K to 395 K, when we compare it with temperature range of mesh model I there is a 2 temperature unit increment in the range due to the mesh refinement in the model II.

Comparison:

Fig 11: Temperature contour on  heat sink plate of Mesh model I.

Fig 12: Temperature contour on heat sink plate of Mesh model II.

Fig 11 & 12 shows the temperature distribution over the fin plate of the mesh model I & II respectively. In this localized temperature plot, the distribution in both cases can be seen in detail. In both the cases pattern seems similar but the temperature range alone varies.

Fig 13: Temperature contour on Graphic card of Mesh model I.

Fig 14: Temperature contour on Graphic card of Mesh model II.

Fig 13 & 14 shows the temperature distribution over the graphic card surface alone of the mesh model I & II respectively. In this localized temperature plot the distribution in both cases can be seen in detail. In both the cases pattern seems similar but the temperature range alone varies.

Fig 15: Temperature contour on mid plane of Mesh model I.

Fig 16: Temperature contour on mid plane of Mesh model II.

Fig 15 & 16 shows the temperature distribution in the mid plane of the domain for mesh model I & II respectively. This temperature contour visualizes the heat take away from the assembly by the incoming cooling air. In both the cases pattern seems similar but the peak temperature alone varies. This contours shows us the thermal boundary layers in both the cases. The thickness of this thermal boundary layer depends on the heat source and the fluid flow velocity.

Fig 17: Temperature contour on outlet of Mesh model I.

Fig 18: Temperature contour on outlet of Mesh model II.

Fig 17 & 18 shows the temperature distribution in the outlet plane of mesh model I & II respectively. This temperature contour is taken in to consideration to get rough estimate of heat out flow from our domain. In both the cases pattern and temperature ranges looks much similar. we cannot come to a conclusion that both the cases yield the same heat out flow, this similarity shows that mesh model II also should be refined and another mesh model III has to be developed to evidence this minute variations. Mesh model II also not convincingly providing the results. On increasing the mesh counts further processing time and cost increases exponentially. For time being standard deviation on the temperature distribution in the outlet surface is taken .

Table 1: Outlet temeperature table.

The above Table 1 shows the temperature outputs in outlet for both the mesh model. Looking on the standard deviation on temperature at outlet surface showing an minute variations, likewise area weighted average on temperature at outlet surface also yielding a minute difference. From this we can say that heat outflow is not the same when we compare mesh model I & II and also the refinement of mesh model I is needed further and grid independent study has to be done to validate the result.

Fig 19: Temperature contour on Graphic card - Heat sink assembly of Mesh model I Showing Hot spots I.

Fig 20: Clip Ranged Temperature contour on Graphic card - Heat sink assembly of Mesh model I Showing Hot spots I .

Fig 21: Clip Ranged Temperature contour on Graphic card - Heat sink assembly of Mesh model I Showing Hot spots II.

To find out the hot spots we need some more plots to evidently spot the region, the above Fig 19, 18 & 21 is taken in that perspective. Fig 19 shows the auto ranged temperature distribution over the Graphic card - heat sink assembly in this obviously we can say that hot spot is near the heat generation domain(graphic card).

To view the hot spot so intensely temperature range is limited with fixing a minimum temperature value say 389 K, the Fig 20 shows the clip ranged temperature contour plot and we can clearly witness the hot spot near the mid of the 2nd row fins on the heat sink plate.

The Fig 21 shows the interior surface showing the hot spot there. The ultimate higher temperature is spotted over the surface of graphic card. The adjacent surface to the garphic card in the base plate also experiances the higher temperature.

Fig 22: Temperature contour on Graphic card - Heat sink assembly of Mesh model II Showing Hot spots I.

Fig 23: Clip Ranged Temperature contour on Graphic card - Heat sink assembly of Mesh model II Showing Hot spots I .

Fig 24: Clip Ranged Temperature contour on Graphic card - Heat sink assembly of Mesh model II Showing Hot spots II.

The Fig 22, 23 & 24 showing the spots in variuos regions in the Graphic card - Heat sink assembly. The main difference between Mesh Model I & II can be clearly noticed when we examined the two sets of figures. The temperature contour plot is well defined in the Mesh model II than in the mesh model I . The distribution of temperature can be clearly noticed over the surfaces of graphic card - heat sink assembly.

The Hot spots were at the same region as like in mesh model I . There is no misplaced hot spots in mesh model II when compared to mesh model I. Fig 23 & 24 clearly spots the hot region with a well defined manner. Additionally the higher temperature spots also found near side surfaces of the graphic card.

Conclusion:

In this study two mesh models were developed by considering the mesh refinement as a main factor. Temperature range variation between mesh model I & II is accounted to be 2 temperature units. The Hot spot regions found out near the mid of 2nd row fin arranged in the fin plate, mere center region in the upper, lower and side surface of Graphic card. The adjacent surface to graphic card in the base plate also experiance the hot spots. The results provided by mesh model II is not a validated result. The model II can be refined further with increased mesh counts to undergo grid independent study for validating the results.