Application of computational fluid dynamics – optimal cooling of

hardware component

Faizan Ahmada, Amjad Riaza,b*

a
Process Engineering Research Group (PERG), Department of Chemical Engineering,
COMSATS Institute of Information Technology, Lahore, Pakistan

b
Process Systems Engineering Centre (PROSPECT), Department of Chemical
Engineering, Faculty of Chemical Engineering, Universiti Teknologi Malaysia, Skudai
81310, Johor Bahru (JB), Malaysia

* Corresponding author. Tel.: +60-108804973; Email address: ariaz@ciitlahore.edu.pk

(Amjad Riaz)
Application of computational fluid dynamics – optimal cooling of
hardware component

Computational Fluid dynamics (CFD) is the science of predicting fluid flow, heat
and mass transfer, chemical reactions, and related phenomena by solving
numerically the set of governing mathematical equations. The results of CFD
analyses are relevant in conceptual studies of new designs, detailed product
development, trouble shooting and redesign. CFD analysis reduces the total effort
required in the experimental design and data acquisition. This paper includes the
study of different aspects to find the best possible cooling solution for the hardware
components using simulation in software FLUENT. The geometries and grids were
generated in “GAMBIT” software. For the particular geometry, by introducing
forced convection maximum temperature at the hardware components is reduced
by approximately 10 °C, but including multiple fans have no particular effect on
maximum temperature. Changing the orientation of hardware components by 90
°C upright, the maximum temperature achieved by natural convection is 15 °C.
Thus by such a modification minimum temperature is achieved without fan.

Keywords: Optimal cooling, computational fluid dynamics, FLUENT, GAMBIT

1. Introduction

Computational Fluid Dynamics is the branch of fluid dynamics that deals with the

prediction the fluid flow, heat transfer, mass transfer and different transport phenomena

based on the computerized solution rather than a practical solution. Different types of

software are available for CFD and FLUENT is widely used among the available. It

works based on some mathematical models and approximates to the real practical

situation (Chung 2002, FLUENT v 6.1 2003). In order to work with the CFD, we have to

follow some special pattern to obtain the results close to the practical situation. For this

purpose, first of all, a pre-processing is required which involves the solution of different

set of equations related to the transport equations like mass, momentum and energy

transport on the mesh generated for a piece of equipment. Physical models, which
describe the physical situation, condition or state of the fluid like its turbulence, viscosity,

phase change or a process occurring in the fluid like vaporization, combustion or

volatilization help the mathematical model to solve the equations for a practical solution

rather than a computerized ideal situation. Solver settings and the input of boundary

conditions, which involve the inlet condition or physical state and properties of the fluid

and conditions at the outlet after a process of post processing give the end results

(Ferziger and Perić 2002 , Thévenin and Janiga 2006).

The basic techniques that enable CFD to give the practical results are as follows.

 Problem Identification and Pre-processing in which the process have to be

determined according to the requirement of the end result we are looking for and

generation of grid for the process equipment being used for the process involved.

As far as the grid generation is concerned, there are different types of grids each of

which is based on the requirement and degree of accuracy. The grid may be

quad/hex or tri/tet depending upon the end results which we want to obtain. For a

simple shape like a pipe, quad/hex mesh gives accurate results with only fewer

cells rather than tri/tet mesh, and the mesh is aligned with the flow; whereas for a

complex geometry like a distillation column or a boiler quad/hex mesh give no

numerical advantage; hence it is better to use a simple tri/tet mesh to save extra

meshing effort. The mesh was generated in GAMBIT software for grid

generation.

 Solver execution in which a numerical model which involves the physical models

such as turbulent, combustion or homogeneous or heterogeneous along with the

material properties, operating conditions and boundary conditions at each

boundary zone is setup and computation is monitored. To monitor the

 computation, we need to solve the mathematical model by iteration and usually a

lot of iterations are required and the results are best close to the practical and

desired results if the iteration results are close to each other.

 Post processing in which the results are analyzed and alterations are made in the

above two steps to achieve the required results. A very good advantage of using

FLUENT is to obtain the results graphically drawn by the software. We can

visualize the end results qualitatively by knowing the different zones where

turbulence, temperature rise, shocks or shear layers are formed as well as

quantitatively by using some numerical reporting tool which gives numerical

values for the zones of turbulence, temperature rise, shocks or hammering etc.

The model is reviewed by checking the physical model, boundary conditions and

grid generation given to the solver for execution (Thévenin and Janiga 2006,

Warsi 2006).

2. Mathematical Modelling

The goal was to find the best possible cooling solution for the hardware components by
natural convection as well as forced convection with different locations of inlets and
outlets ports. The tasks included the geometry and simulation steps which were fulfilled
by GAMBIT and FLUENT respectively. The geometry (
List of Figure Captions

Figure 1) consisted of two hardware components enclosed in a square box with openings
on both sides. The length of each side was 11cm each. The openings allow the air to flow
through the square box for cooling.

For natural convection case, two grids (coarse grid with 2,825 cells and fine grid
with 11,300 cells) were generated using GAMBIT shown in Figure 2. The properties of
air, numerical operating and boundary conditions were set for the solver. A constant heat
flux of 110W/m2.K was considered with walls considered at constant temperature for the
solver.

Moreover, two dimensional steady flow condition and standard k-epsilon

turbulence model were assumed for the given problem. In k-epsilon model, the baseline
two-transport equation model solves for k and ε and coefficients are empirically derived;
valid for fully turbulent flows only. Options to account for viscous heating, buoyancy,
and compressibility were shared with other k-ε models.

For forced convection case, again, the grids were generated and the properties of
air, numerical values for the heat flux from the hardware components, operating and
boundary conditions were given for different positions of inlet and outlet for the air.

3. Results and Discussion

In the natural convection case, the fluid medium was air with changing velocity. The
operating conditions were standard with g = -9.8 m/s2 and pressure = 101,325 Pa. The
boundary conditions were walls at constant temperature of 295 K and hardware
components at constant heat flux of 110 W/m2K. Continuity, velocities, energy, k and
epsilon residuals were set to 1e-6.

The temperature and velocity contours, obtained by iterating the solution both for
coarse as well as fine grid, are shown in Figure 3.

It can be observed from the contours that the maximum temperature achieved by
natural convection was 344.36 K that was near to the hardware components. The grids
were adapted (Figure 4) based on these contours and finally comparison graph (Figure 5)
was plotted showing the static temperature at different positions in the square box with
and without adaptation.

For the forced convection, the cases were divided into single inlet- single outlet
and single inlet-multiple outlets. In the case of forced convection with single inlet and
single output, it was observed that different temperatures were obtained by applying inlet
and outlet at different positions (Figure 6). The maximum temperature observed inside
the boxes were noted and compared for the minimum among them. The minimum
temperature, among all different cases of inlet and outlet positions, achieved was 334.74
K that was observed in the case of middle inlet and middle outlet. The same procedure
was repeated, as in natural convection, by plotting contours and adapting grid. Finally,
comparison of static temperatures vs. position for adaptation and without adaptation
cases were plotted (Figure 7).

In the case of multiple inlets and single outlet for forced convection, the inlet
stream was divided into two and three inlet streams respectively (Figure 8) and observed
for minimum temperature among all cases of different positions. The minimum
temperature achieved is 335.58 K, which was obtained by dividing inlet stream into three
streams and keeping outlet stream in the middle position.

List of Table Captions

Table 1 summarizes the results of temperatures obtained by natural convection as

well as forced convection (both single inlet- single outlet and multiple inlet- single
outlet). It can be observed from the table that forced convection with single inlet and
single outlet gives the minimum temperature. In other words, it can be said that optimal
cooling is possible by applying forced convection with single inlet and single outlet
(Tseng et al. 2007).

At this moment, it was thought to check the effect of change in the orientation of
hardware components. The geometry was revised in a way that hardware components are
places vertically in a square box instead of horizontally (Figure 9) (Kim and Kim 2008).
By adapting this approach, the temperature achieved was 330.74 K that was even less
than forced convention cases. Finally, comparison between static temperature and
position was made with and without adaptations is made (Figure 10), as in previous cases.

4. Conclusion

It can be deducted from the comparison chart (Figure 11); an approximate temperature of
72 °C is achieved by natural convection, 62 °C by forced convection with single input and
single output and 63 °C under the same mode of forced convection with multiple input
and single output. Thus, for the given geometry, by introducing force convection,
maximum temperature at the hardware component is reduced by approximately 10 °C.
Thus, it can be said that introducing multiple fan has no particular effect on maximum
temperature.

Moreover, it can also be observed from the chart changing the orientation of
hardware component by 90 °C upright, the maximum temperature achieved is 57 °C,
which is 15 °C less than that achieved by natural convention in given geometry and 5 °C
less than that by forced convection. Thus by such a modification, minimum temperature
is achieved without any fan (forced convection).
REFERENCES

Chung, T.J., 2002. Computational fluid dynamics: Cambridge University Press.

Ferziger, J.H. & Perić, M., 2002 Computational methods for fluid dynamics: Springer.

Fluent V 6.1, 2003. Introductory fluent notes. FLUENT Inc.

Kim, H.J. & Kim, C.-J., 2008. An optimal parametric design to improve chip cooling.
Applied Thermal Engineering, 28 (14-15), 1896-1905.

Thévenin, D. & Janiga, G., 2006. Lectures of cfd. Germany: Otto-Von Guericke
University Magdeburg.

Tseng, Y.-S., Fu, H.-H., Hung, T.-C. & Pei, B.-S., 2007. A numerical analysis for the
cooling module related to automobile air-conditioning system. Applied Thermal
Engineering, 27 (11-12), 1823-1831.

Warsi, Z.U., 2006. Fluid dynamics: Theoratical and computational approaches, 3rd ed.:
Taylor and Francis group.
 List of Table Captions

Table 1 Summary of results of temperatures achieved by natural convection and forced

convection
 List of Figure Captions

Figure 1 Geometry of hardware components enclosed in a square box

Figure 2 Meshing of geometry using GAMBIT (a) Coarse grid (b) Fine grid

Figure 3 Contours of geometry (a) Static Temperature (b) Velocity

Figure 4 Adaptation of grid (a) Coarse grid (b) Fine grid

Figure 5 Comparison plot b/w static temperature and position without and with
adaptation for natural convection

Figure 6 Maximum temperature achieved by using different positions of inlet and outlet

Figure 7 Comparison plot b/w static temperature and position with and without
adaptation for forced convection

Figure 8 Maximum temperature achieved by using multiple inlets and single outlet at
different positions

Figure 9 Revision of geometry by change in the orientation of hardware components

Figure 10 Comparison plot b/w static temperature and position with and without
adaptation for revised geometry

Figure 11 Comparison of temperatures achieved by applying different methods