Conjugate heat transfer analysis of a graphics card by utilizing fins and cross-flow in ANSYS Fluent
INTRODUCTION:
Processors and graphics cards are widely used in computers today. As the functionality of the computers keeps on increasing, more and more power is required to keep up with the computational powers. A graphics card consists of three major components - the processor, the processor base and the cooling mechanism. In this project, we will analyze the cooling mechanism by adding a set of fins on top of the processor to absorb the heat by conduction as well as subjecting the graphics card setup to a cross-flow of air to absorb the heat by convection. The addition of the air-flow increases the cooling of the graphics card as convection causes additional heat transfer from the graphics card to the surrounding and away from the heat source.
COMPUTATIONAL GEOMETRY AND MESH:
The geometry of the graphics card without and with the fins base are shown below:
A fluid enclosure is generated around the graphics card to create a fluid domain for the air. Share topology is enabled for the geometry in order to generate conformal mesh as the intersecting surfaces. The conformal mesh is essential for generating good and continuous results. We use two different types of mesh - a baseline mesh and a refined mesh. The baseline mesh is generated from the default size of `10.289` `mm` calculated by ANSYS based on the geometry.
Baseline Mesh
Refined Mesh
As we can see in the baseline mesh, the processor is divided into only `4` elements, which will not return very accurate results. Also, the mesh is very coarse on the base and the fins. Hence a refined mesh with specified sizes on each of the components is generated. The sizes are:
Default: `4` `mm`
Processor: `0.5` `mm`
Base: `1` `mm`
Fins: `1` `mm`
SETUP OF THE FLOW PHYSICS:
The simulation is performed using a steady-state pressure-based solver. The energy equation solver is enabled to evaluate the heat transfer. The standard `k`-`epsilon` turbulence model with standard wall functions is implemented to capture the turbulence in the flow. The processor, fins and the base are assigned relevant solid materials. The fluid used for the convective cooling is air.
Processor - Gold
Density: `rho = 19320` `kg"/"m^3`
Specific Heat: `C_p = 129.81` `J"/"kg-K`
Thermal Conductivity: `k = 297.73` `W"/"m-K`
Base - Copper
Density: `rho = 8978` `kg"/"m^3`
Specific Heat: `C_p = 381` `J"/"kg-K`
Thermal Conductivity: `k = 387.6` `W"/"m-K`
Fins - Aluminum
Density: `rho = 2719` `kg"/"m^3`
Specific Heat: `C_p = 871` `J"/"kg-K`
Thermal Conductivity: `k = 202.4` `W"/"m-K`
Fluid - Air
Density: `rho = 1.225` `kg"/"m^3`
Specific Heat: `C_p = 1006.43` `J"/"kg-K`
Thermal Conductivity: `k = 0.0242` `W"/"m-K`
Dynamic Viscosity: `mu = 1.7894 times 10^{-5}` `kg"/"m-s`
The left side of the fluid enclosure is assigned a velocity-inlet boundary condition with a velocity magnitude of `10` `m"/"s` and a temperature of `300` `K`. The right side of the enclosure is assigned a pressure-outlet boundary condition. The wall near the base is assigned a no-slip wall boundary condition, while the remaining side walls are assigned a symmetry boundary condition so that the flow does not interfere in the nearby region of the processor. The processor is is given a source term for a constant heat-generation rate of `78125000` `W"/"m^3` which rates the processor at `5` `W` power output. The simulation is initialized using hybrid initialization and run for a total of `300` iterations.
RESULTS AND DISCUSSION:
Baseline Mesh
Residuals
Temperature
Heat Transfer Coefficient
Refined Mesh
Residuals
Temperature
Heat Transfer Coefficient
From the contour plots, we can see that the refined mesh provides more accurate and discrete results. The difference in the values obtained from both the simulations are compiled in a table below:
`{:(ul{text{Mesh}},,ul{text{Max. Processor Temp}},,ul{text{Max. Heat Transfer Coefficient}}),(text{Baseline},,320.46 K,,1053.34 W"/"m^2-K),(text{Refined},,314.72 K,,1199.76 W"/"m^2-K):}`
The potential hot spots on the model are the areas near the processor, particularly the fins since they are absorbing most of the heat generated by the processor through conduction. Hence the coolant must be able to absorb the heat from the fins into the surroundings.