Development of a CFD Tool Based

on SnappyHexMesh/OpenFOAM for the Axial

Fan Performance

Van Long Le1 , Lu Tien Truong1 , and Khanh Hieu Ngo1,2(B)

1 Ho Chi Minh City University of Technology, VNU-HCM, Ho Chi Minh, Vietnam
[email protected]
2 VNU-HCM Key Laboratory of Internal Combustion Engine, Ho Chi Minh, Vietnam

Abstract. Nowadays, axial fans are widely used in industry to solve the problem
of ventilation and cooling. This type of fan converts electrical energy into kinetic
energy of the airflow to disperse heat radiation. However, at present, the manu-
facturing process of these devices in Vietnam is still simple and has not achieved
high performance. In this paper, the authors concentrate to build an automatic
meshing tool, named BKIAF, for the performance simulation of the axial fans in
normal air conditions. This automatic meshing tool is based on snappyHexMesh
and blockMesh which are tools of the open source software OpenFOAM. Addi-
tionally, the Multiple Rotating Frame (MRF) method is applied in combination
with the SIMPLE algorithm and Reynolds Averaged Navier-Stoke (RANS) tur-
bulence method. On that basis, the use of BKIAF to improve the performance
of a composite axial fan size 48 in. of Huynh Thao Fans Ltd., indicated that the
fan’s efficiency could be increased to at least 20% by modifying the shape of
fan inlet. In comparison with other CFD tools, the BKIAF tool, based on snappy-
HexMesh/OpenFOAM, would offer significant advantages to a design and analysis
process for the axial ventilating system.

Keywords: blockMesh · Reynolds Averaged Navier-Stokes · snappyHexMesh

1 Introduction
Vietnam has a territory located entirely in the tropics, so in daily life as well as in indus-
trial production activities, temperature is one of the priority issues that need to be solved
to ensure a good quality of work and quality of life for people. Excessively high tempera-
tures affect human health, making equipment and machinery reduce operating efficiency
and tool life. Currently, the most popular and widely applied solution in our country is
to use axial fans to generate cooling airflow and air exchange. Fan applications include
local ventilation, cooling towers for air and wind machines, humidifiers in textile mills,
ventilation, and exhaust in the mining industry, and dynamic cooling electromechanical
and generator. However, the design and manufacture of fans in Vietnam currently still
have many limitations, leading to the low efficiency of the fan, as well as the loss of
electrical energy.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022
A.-T. Le et al. (Eds.): RCTEMME 2021, LNME, pp. 91–103, 2022.
https://doi.org/10.1007/978-981-19-1968-8_9
92 V. L. Le et al.

Nowadays, there are many research papers/projects on axial fans in many different
aspects, to optimize the performance of this type as well as reduce energy costs. In the
study [1], D.C. Panigrahi and D.P. Mishra conducted a study on the blade design of
the general type of ventilation under the mines, by simulating 6 types of blade profiles
commonly used in air conditioning turbulent flow condition with the number Re = 3 ×
106 with the variation of the angle of attack to find the angle of attack at which the fan
has the highest aerodynamic efficiency. Panigrahi used Ansys Fluent tool and Gambit
software to create simulation mesh. In his master’s thesis at Stellenbosch University
[2], O.P.H. Augustyn used Ansys Fluent tool with multi-reference region method, built
numerical simulation cases to calculate and predict performance characteristics by the
mass flow rate of several a different type of fans. The obtained results show the high
reliability of the CFD model. Similarly, H. Kumawat [3] used Ansys CFX to study the
behavior of the fluid passing through the axial fan, thereby surveying the model with
different number of blades, and giving the numerical design propeller suitable for the
environment. However, the application of the multi-reference region method has not been
found in this publication. In the study [4], D. Dwivedi et al. conducted an experiment of
the axial fan model with the propeller slanting forward and backward, in order to compare
aerodynamic parameters such as static pressure, flow, coefficient of fluid movement
between the two models to find the most effective model, and at the same time compare
with the experimental model to check the actual errors. In two previous research projects
of these authors [5, 6], Khanh Hieu Ngo and his partners in VNU-HCM Key Laboratory
of Internal Combustion Engine have evaluated the simulation with LBV150 ROV thruster
and axial flow pump model, respectively. The authors concentrate to build the CFD
numerical model which use the Multiple Rotating Frame (MRF) method combined with
the SIMPLE algorithm in OpenFOAM - the open source software. This has proved the
suitability of the selection algorithm and the MRF method that it is completely suitable
for the rotating de-vices with 1-phase condition by reliable estimates between simulation
and experiment. In the International Conference of Fluid Machinery and Automation
System 2018 [7], Ngo Khanh Hieu and his colleagues has continued to introduce the
BKASM which is an automatic meshing software for propeller simulation in open-water
condition. At the National Technical Mechanics Conference in 2019 [8], these authors
from the Key Laboratory of Internal Combustion Engines at Ho Chi Minh University of
Technology pointed out the evaluation of the characteristic for axial - flow household
fan using numerical simulation combined with experimental method. Also in this paper,
the authors proved the accuracy of the axial fan simulation using the snappyHexMesh
and blockMesh tools of OpenFOAM software to generate the suitable mesh to apply
with k–w SST turbulence model, MRF method and the SIMPLE algorithm.
 Development of a CFD Tool Based on SnappyHexMesh/OpenFOAM 93

In this paper, the authors concentrate to introduce the BKIAF tool which is devel-
oped based on snappyHexMesh and blockMesh by the authors to provide a solution for
industrial ducted fan simulation. The Computational Fluid Dynamics (CFD) technique
is used to simulate the performance of an industrial ducted axial fans of Huynh Thao
Fan Co., Ltd., with the aim to enhance this fan performance as a study case in order to
verify the accuracy of this software. Firstly, a CFD model of the original ducted axial
fan is validated by its experimental data. Then, the improved design of this ducted axial
fan is processed by the CFD model. Finally, the performance testing of the im-proved
ducted axial fan is taken aiming to evaluate reliability of CFD technique to enhanced
industrial axial fans performance. The performance testing of the Huynh Thao ducted
axial fan is set up in the VNU-HCM Key Lab. for Internal Combustion Engine in the
context of a research, funded by VNU-HCM under grant number C2020-20-43.

2 BKIAF – the CFD Techniques for Ducted Axial Fans Simulation

In the BKIAF, the authors combine all the six steps to generate the mesh and simulate
the axial ducted fan. This tool has adjusted about 70 parameters to fit the axial ducted fan
meshing and provided the standard solver in OpenFOAM software. BKIAF is suitable
for this problem in static condition, that means in a laboratory environment and an inlet
velocity is equal to zero approximately. Users only import the geometry file; this tool
will automatically perform the following steps:

• Set up the compute domain.

• Determine the fineness of each domain and sub-partition fan surface.
• Customize the boundary layer.
• Some dynamic parameters of the fluid have been established to suit the static condition.
However, users can recalibrate to better suit specific conditions.
• The boundary conditions for this problem have also determined.
• The turbulence model and OpenFOAM solver are also recommended in this tool.

2.1 Geometry of Ducted Fan

To achieve reliable results, the digital geometry must be as close to the real model as
possible. For this purpose, a composite fan of size 48 in. of Huynh Thao Fan Co., Ltd.
is used as a case study. Figure 1 shows the 3D model of Huynh Thao composite fan of
size 48 in. (the dimensions of this fan are listed in Table 1). In this article, the authors
use this ducted fan, which is widely used in the factories in Vietnam, to verify the results
obtained from BKIAF tool.
94 V. L. Le et al.

Fig. 1. 3D model of the duct (top image), the fan (left center image), the Huynh Thao compos-
ite fan of size 48 in. (right center image) and the rounded inlet fan (bottom image - enhanced
performance geometry)

Table 1. Dimensions of the Huynh Thao composite fan size 48 in.

Details Dimensions (mm)

Input length 1472
Input width 1431
Output diameter 1500
Diameter of position close to impeller outlet 1440
Fan diameter 1210

2.2 Automatic Meshing with BKIAF

For the accuracy of simulation results, the computational domain must be large enough,
However, it is necessary to partition the volume mesh to minimize the number of mesh
cells. This paper as well as the BKIAF tool proposes a computational domain with
dimensions presented in Table 2, Table 3, and Fig. 2.
 Development of a CFD Tool Based on SnappyHexMesh/OpenFOAM 95

Fig. 2. Domain size for CFD simulation of ducted axial fans performance

Table 2. Size of blocks in the computational domain (where D is the fan diameter)

Size of blocks Cell size [mm]

9D diameter, 12D length (Block 1) (150 150 150)
4D diameter, 5D length (Block 2) (37.5 37.5 37.5)
2D diameter, 3D length (Block 3) (18.75 18.75 18.75)
1.02D diameter, 0.33D length (Block (4) (18.75 18.75 18.75)

The difficulty in our case is the gap between the duct and the fan is about 1.15 cm,
which is smaller than the diameter of the fan. Therefore, the adjustment to create bound-
ary layers and the number of boundary layers greatly affects the results of the problem.
This paper selects two boundary layers the fan and the duct. The average boundary
layer thickness of the fan is 0.3 mm. Similarly, the average boundary layer thickness
corresponds to a duct is also 0.3 mm. In addition to having four boundary layers on fan
tip surface and duct surface at the tip of fan, the minimum clearance of fan and duct is
always guaranteed to have at least 4 grid cells (4 layers) (Fig. 3).
Although the number of boundary layer are only two for both fan and duct, it is
suitable for the clearance between them which is 1.15 cm in thickness. For the mesh
elements close to the final boundary layers, they are from 1.5 to 2 times the thickness of
the outermost boundary layer depending on the location. Therefore, the streamline near
the wall boundary is relatively properly calculated in this case.
96 V. L. Le et al.

Fig. 3. Mesh distribution on the fan surface (top image), duct surface (left image) and the tip
clearance between the duct and fan (right image)

Table 3. Dimensions of the surface of the calculated domain, fan, and duct (where D is the fan
diameter)

Surfaces Surface element size [D]

Surface of the outermost layer of the computational domain (Block Min: (0.125, 0.125)
1) Max: (0.125, 0.125)
Rotation domain surface Min: (0.0078, 0.0078)
(Block 4) Max: (0.0625, 0.0625)
Fan surface Min: (0.002, 0.002)
Max: (0.0078, 0.0078)
Duct surface Min: (0.002, 0.002)
Max: (0.0078, 0.0078)

This paper applies the blockMesh meshing tool to create background mesh elements.
Subsequently, snappyHexMesh specifies the surfaces and volume areas to be finely
meshed, and manages the construction of the boundary layer on the geometric surfaces.
The mesh quality of our simulation problem is listed in Table 4. Note that, the k–ω SST
is used in combination with wall function which is pre-developed with OpenFOAM
software. Therefore, y+ is suggested in this case to be greater than 30.
 Development of a CFD Tool Based on SnappyHexMesh/OpenFOAM 97

Table 4. Mesh quality using snappyHexMesh and blockMesh tools

Criteria Value
Number of grid elements 17 788 914 cells
Max-skewness 3.9862
Max AR 21.4978
Max non-orthogonality 66.5164
Average non-orthogonality 7.0022
Number of impeller and duct boundary layers 2
Fan first boundary layer size 0.3 mm
Fan last boundary layer size 0.3 mm
Maximum boundary layer development 1.2
Average y+ (calculated after running simulation) 32.16

2.3 Turbulence Model

Three basic approaches can be utilized to calculate a turbulent flow: Direct Numerical
Simulation (DNS), Large Eddy Simulation (LES) and Reynolds Averaged Navier-Stokes
Simulation (RANS). Compared to other turbulent models, the RANS modeling method
is among the most well–developed and popularly used due to its suitability providing
acceptable accuracy and consuming less computing resources [9]. Hence, RANS is the
obvious choice to model the turbulent flow surrounding the object in this research. There
are some different types of turbulence model in the RANS methodology, some of them
are well – known to industrial applications such as the one–equation Spalart–Allmaras
or the two–equation k–ε and k–ω and their variations, and the other less well–known
models. Based on the features of each model, the authors decide to examine the k–ω
SST, which also is the most suitable for turbomachinery applications.
The first variable is turbulence kinetic energy k. It determines the turbulent energy
of the fluid flow. This variable is shown by the following equation [10]:
  
∂ ∂   ∂ui ∂ ρk ∂k
(ρk) + ρkuj = τij − β ∗ ρωk + μ + σk
∂t ∂x j ∂x j ∂x j ω ∂xj

The second variable is specific turbulence dissipation ω. This is the rate at which
turbulence kinetic energy is converted into thermal internal energy per unit volume and
time. It is shown by the following equation [10]:
  
∂ ∂ ω ∂ui ∂ ρk ∂ω σd ∂k ∂ω
(ρω) + (ρuj ω) = α τij − βρω2 + μ + σω +ρ
∂t ∂xj k ∂xj ∂xj ω ∂xj ω ∂xj ∂xj
5 3 9 1 1
α= ,β = , β∗ = , σk = , σω =
9 40 100 2 2
98 V. L. Le et al.

2.4 SimpleFoam Solver

When discretizing the governing equations for incompressible flow over a control vol-
ume, the equations will be computed without respect to the pressure in the node. The
equation is solved but the pressure is highly oscillating. This problem is called the
checker-board pressure field. Because there is no equation for the pressure, the mass
conservative equation can be modified to give an indirect equation for the pressure. Dis-
cretizing the continuity equation creates the pressure correction equation and it contains
the old velocity term and one correction term plus a term for the continuity error. Differ-
ent algorithms exist to solve the modified equations such as the Semi-Implicit Method for
Pressure-Linked Equation (SIMPLE), the Pressure Implicit with Split Operator (PISO),
the SIMPLE Consistent (SIMPLEC) and PIMPLE (merged PISOSIMPLE) algorithm
[11]. Solve the momentum equation for the velocity field. This velocity field does not
satisfy the continuity equation.

MU = −∇p

where M is a matrix of coefficients that are calculated by discretizing the velocity terms
in the incompressible Navier–Stokes equation. These coefficients are all known. Solve
the Poisson equation for the pressure field.
 
∇ • A−1 ∇p = ∇ • A−1 H

We also use the pressure field to correct the velocity field so that it satisfies the
continuity equation.

U = A−1 H − A−1 ∇p

The velocity field now does not satisfy the momentum equations. They repeat the
cycle. The turbulence scalar (k, ε, ω) and species transport equations are solved within
the loop, after the volume flux corrector: Mk k = Sk and Mω ω = Sω .

2.5 Boundary Conditions

Boundary conditions for the quantities to be calculated in the fan simulation prob-
lem have been standardized by the research team to be similar to the actual conditions
listed in the table below. Furthermore, with the application of Multiple Reference Frame
(MRF) model, the specified angular velocity of rotation domain is uniform throughout
the simulation (Table 5).
 Development of a CFD Tool Based on SnappyHexMesh/OpenFOAM 99

Table 5. The boundary conditions for the axial ducted fan simulation [8]. (f = fixedValue,
i = inletOutlet, s = slip, z = zeroGradient, c = calculated, k = kqRWall-Function, o =
omegaWallFunction

Domain U p Nut k ω
Inlet f z c f f
Outlet i f c i i
outerCylinder s s c s s
Fan and ducted f z c k o

3 Results of the Case Study

In this paper, the authors use the 10 cores CPU and 64 GB RAM computer. Each case
takes 20 h to run simulation approximately.
To evaluate the performance improvement, the numerical simulation CFD results for
both original geometry and improved geometry are presented in this. The results will
be verified unequivocally with the experiment before providing a reliable solution to
the manufacturer. The problem is to simulate the axial fan operating in the laboratory
atmosphere with the following characteristic parameters:

• Air density at laboratory temperature (30 °C degrees): 1.16 kg/m3 .

• Kinematic viscosity of air at laboratory temperature: 1.516 m/s2 .
• The inlet velocity of the air stream is 0 m/s.
• Rotation speed is 600 rpm.
• Turbulence intensity 5%.

The characteristic parameters of fan are calculated by the following formulas:

• Required power: Preq = 2 × π × n × τ.

• Useful power: Puseful = (pi × Qi ).
• Fan efficiency: ηblade = Puseful /Preq.
• Electric motor efficiency:
ηelectric_mechanic = Preq /Pelectric.

where τ is a torque which is determined by numerical simulation, n is the number of

revolutions in 1 s. pi = ρ × Vi 2 /2 is kinetic pressure of the fluid flow and Qi is the volume
rate at ith annular surface position (see Fig. 4 and Fig. 5).
100 V. L. Le et al.

3.1 Grid Independent Evaluation

Table 7 compare the differences between the proposed mesh (Table 3) and the finer mesh
(Table 6). Based on that, the authors evaluate the dependency of mesh and demonstrate
that it is appropriate to use the mesh configuration as shown in the Table 3. This mesh
quality ensures accuracy of the results and saves computational resources.

Table 6. Dimensions of the surface of the finer calculated domain, fan, and duct (where D is the
fan diameter)

Surfaces Surface element size [D]

Surface of the outermost layer of the computational domain (Block Min: (0.0625, 0.0625)
1) Max: (0.0625, 0.0625)
Rotation domain surface Min: (0.005, 0.005)
(Block 4) Max: (0.0312, 0.0312)
Fan surface Min: (0.002, 0.002)
Max: (0.005, 0.005)
Duct surface Min: (0.002, 0.002)
Max: (0.005, 0.005)

The size of duct and fan surface (finer mesh) are similar to the grid size introduced
in the Table 3 in order to ensure the average y+ is equal to 30 approximately. While the
outer computational domains have the grid element size reduced by nearly half the size.

Table 7. The comparison between proposed mesh and finer mesh of the original design

Parameter Proposed mesh Finer mesh Error (%)

Mass flow (m3 /h) 36900 37251 0.95
Torque (M/m) 10.09 10.17 0.79
Required Power (W) 633.65 641.25 0.79
Useful power (W) 415.49 427.46 2.88
Average speed at the inlet position (m/s) 6.30 6.36 0.95
Efficiency (%) 65.57 66.66 1.09

Overall, the quality of finer mesh is still guaranteed to satisfy the mesh quality
evaluation criteria of proposed mesh as in Table 4. The average y+ and the number of
boundary layers of duct/fan are 27.7 and 2, respectively. The number of cells is 47 521
332. However, the difference in the results of two meshes are lower than 3%. Therefore,
in the next section of this paper, the proposed mesh as shown on the Table 3 is selected
for our ducted axial fan simulation. It consumes less computational resources but still
ensures the accuracy of the results.
 Development of a CFD Tool Based on SnappyHexMesh/OpenFOAM 101

3.2 Simulation Results of the Original Huynh Thao Composite Fan

Table 8 describes the simulation results and experimental results of the original Huynh
Thao composite fan of size 48 in. Figure 4 shows the velocity distribution on the inlet
surface of the experiment and numerical simulation (CFD), where the difference in flow
rate is about 3%, useful power is 9%, and fan efficiency is 7.9%.

Fig. 4. Velocity distribution at inlet surface of the original Huynh Thao composite fan

Table 8. The comparison between simulation and experiment of the original design.

Parameter CFD result Exp result Error (%)

Mass flow (m3 /h) 36900 38052 3.06
Torque (M/m) 10.09 10.21 1.21
Required Power (W) 633.65 641.25 1.21
Useful power (W) 415.49 455.53 9.03
Average speed at the inlet position (m/s) 6.30 3.49 3.06
Efficiency (%) 65.57 71.04 7.70

3.3 Simulation Results of the Enhanced Huynh Thao Composite Fan

Table 9 compares the simulation results of original ducted fan design and the simulation
results of the improved ducted fan design (rounded inlet). In general, with this new
design, mass flow is improved 11.90%. Besides, the fan efficiency is also increased from
65.57% to 86.18% (see Fig. 5).
As we can see in the Fig. 6, at the corners of the original design inlet, it occurs
vortex phenomena. These vortices convert kinetic energy into pressure causing energy
loss. These problems at the rounded inlet are better solved as well as the streamline at
the rounded duct inlet is straighter. Therefore, by changing the shape of a duct from
square to round, the fan efficiency could be considerably enhanced with the same power
consumption.
102 V. L. Le et al.

Fig. 5. Velocity distribution at inlet surface of the enhanced Huynh Thao composite fan.

Fig. 6. Pressure distribution in the original design (left) and the improved design (right)

Table 9. The comparison between simulation of original design and the rounded inlet design

Parameter CFD rounded inlet CFD original Difference (%)

Mass flow (m3 /h) 41292 36900 11.90
Torque (M/m) 10.09 10.09 0.00
Required Power (W) 633.65 633.65 0.00
Useful power (W) 546.07 415.49 31.42
Average speed at the inlet position(m/s) 6.75 6.30 11.90
Efficiency (%) 86.18 65.57 31.42

4 Conclusion

In this research, the BKIAF tool is used to generate the mesh and simulate the perfor-
mance of an industrial axial fan of Huynh Thao Fan Co., Ltd., with the aim to enhance its
performance in constraints of cost-effectiveness and in-house technology. This tool acts
as a powerful support tool for manufacturers and businesses to predict fan characteristics
before setting up a test model. In industry, this tool supports the analysis of aerodynamic
characteristics as well as virtualize the distribution the velocity in space. Because of this,
the manufacturer not only evaluates the features of the fan, but also helps to consider the
installation of fans in factories effectively. In this article, the authors use the ducted fan
 Development of a CFD Tool Based on SnappyHexMesh/OpenFOAM 103

developed by Huynh Thao Co., Ltd. as a study case to prove the reliability of BKIAF tool
as well as improving the manufacturer’s design to improve performance. The enhanced
design of the Huynh Thao axial fan rounded inlet could increase the fan efficiency up
to at least 20% with the power consumption almost unchanged. In comparison with
other CFD tools, the BKIAF tool, based on snappyHexMesh/OpenFOAM, would offer
significant advantages to a design and analysis process for the axial ventilating system.

Acknowledgments. We sincerely thank the Engineering Mechanics Lab at HCMUT for the great
support on of the facility and tools to our research. This research is funded by VNU-HCM under
grant number C2020-20-43. We acknowledge the support of time and facilities from Ho Chi Minh
City University of Technology (HCMUT), VNU-HCM for this study.

References
1. Panigrahi, D.C., Mishra, D.P.: CFD simulations for the selection of an appropriate blade
profile for improving energy efficiency in axial flow mine ventilation fans. J. Sustain. Min.
13(1), 15–21 (2014)
2. Augustyn, O.P.H.: Experimental and Numerical Analysis of Axial Flow Fans. M.Sc Thesis,
University of Stellenbosch (2013)
3. Kumawat, H.: Modeling and Simulation of Axial Fan Using CFD. Int. J. Aerospace Mech.
Eng. 8(11), (2014)
4. Dwivedi, D., Dandotiya, D.S.: CFD analysis of axial flow fans with skewed blades. Int. J.
Emerging Technol. Adv. Eng. 3(10), 741–752 (2013)
5. Hieu, N.K., Thien, P.Q., Nghia, N.H.: Numerical analysis of lbv150 ROV thruster performance
under open water test condition. In: Duy, V.H., Dao, T.T., Zelinka, I., Kim, S.B., Phuong, T.T.
(eds.) AETA 2017 - Recent Advances in Electrical Engineering and Related Sciences: Theory
and Application, pp. 1037–1046. Springer International Publishing, Cham (2018). https://doi.
org/10.1007/978-3-319-69814-4_99
6. Hieu, N.K., Long, L.V., Huy, L.D.A.: Numerical simulation of an industrial centrifugal fan
5.5 kW with OpenFOAM. Sci. Technol. Develop. J. Eng. Technol. VNU-HCM Press 3(4),
508–522 (2020)
7. Long, L.V., Luan, M.N., Hieu, N.K.: Propeller simulation in open-water condition with snap-
pyHexMesh/OpenFOAM mesh generator. In: International Conference of Fluid Machinery
and Automation Systems, pp. 51–56 (2018)
8. Luan, M.N., Long, L.V., Hieu, N.K.: Evaluate the characteristic of axial fan by numer-ical sim-
ulation combined with the experiment method. In: National Technical Mechanics Conference
(2020)
9. ANSYS: Introduction to ANSYS FLUENT, Lecture 6: Turbulence Modeling (2010)
10. Wilcox, D.C.: Formulation of the k–ω turbulence model revisited. AIAA J. 46(11), 2823–2838
(2008)
11. Klasson, O.K.: A Validation, Comparison and Automation of Different Computational Tools
for Propeller Open Water Predictions, Göteborg, Sweden (2011)