b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/issn/15375110

Research Paper

CFD modeling of air flow distribution in rice bin

storage system with different grain mass
configurations

Gbenga Olatunde a, Griffiths G. Atungulu a,*, Sammy Sadaka b

a
Department of Food Science, University of Arkansas Division of Agriculture, 2650 N Young Avenue, Fayetteville,
AR 72704, USA
b
Department of Biological and Agricultural Engineering, University of Arkansas Division of Agriculture, 2301 S.
University Avenue, Little Rock, AR 72204, USA

article info
Poor airflow distribution in grain mass during in-bin aeration, drying and storage may lead to
Article history: moisture content variations that could be detrimental to grain quality. The effects of grain
Received 18 March 2016 mass configuration and porosity on airflow distribution inside a rice bin were investigated
Received in revised form using three-dimensional computational fluid dynamics simulations and experiments. A finite
13 September 2016 volume method with porous media formulation was used to simulate air flow characteristics
Accepted 13 September 2016 in peaked, inverted, and levelled grain mass configurations for long-grain rough rice with a
Published online 10 October 2016 porosity of 0.55, and mean particle size distribution of 2.94 mm. The airflows through the
rough rice masses were simulated for airflow rates of 0.55, 0.825 and 1.1 m3 min
1[air] t
1 [rice].
Keywords: The model was validated using a bench scale pressure drop system and an actual long-grain
Rough rice rice in-bin storage with peaked grain mass configuration having a capacity of 700 Mt. The
Drying and storage results showed that long-grain rice has viscous and inertial resistance coefficients of 9.72Eþ06
Aeration and storage and 36,185, respectively. Non-uniform airflow distribution dominated peaked and inverted
Airflow grain mass configurations with peaked configuration having the highest restriction to airflow.
CFD modelling Airflow at peak positions in the bed were significantly (p < 0.05) lower compared to other parts.
The average non-uniformity coefficient (NUF) measured directly from the bin was 34% and
those obtained from the model using constant and variable porosities were 19% and 71%,
respectively. For inverted scenario, a maximum of 50 t of rice is needed to be removed from the
rice storage bin to ensure an airflow distribution with an NUF <50%.
© 2016 IAgrE. Published by Elsevier Ltd. All rights reserved.

storage. To prevent spoilage and maintain grain quality, grain

1. Introduction condition must be stabilised as soon as possible. Aeration is a
process of forcing natural air from the bottom of the bin at
The temperature and moisture content (MC) of freshly- relatively low airflow (1e2 s
1 [air] m
3 [grain]) through the
harvested rough rice varies considerably as it comes into bulk grain for the purposes of cooling, ventilation and

* Corresponding author.
E-mail address: atungulu@uark.edu (G.G. Atungulu).
http://dx.doi.org/10.1016/j.biosystemseng.2016.09.007
1537-5110/© 2016 IAgrE. Published by Elsevier Ltd. All rights reserved.
 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7 287

storage system, grain samples at the bin core tend to be

Nomenclature associated with low porosity while that at the periphery (close
to the wall) high porosity (Bartosik & Maier, 2006; Lawrence &
CFD Computational Fluid Dynamics
Maier, 2011). Hence, during aeration lower specific air velocity
NUF Non-uniformity coefficient (%)
passes through the core compared with the periphery of the
MC (s) Moisture content (% wet basis)
grain mass. Airflow resistance as the result of the concentra-
UDF Users Defined Function
tion of fines was investigated by Siebenmorgen and Jindal
MRD Mean relative deviation (%)
(1987) using a long-grain rough rice cultivar at varying MCs
DDP Dimensionless Dynamic Pressure
(12%, 18% and 24%) and fine material concentrations (0%, 5%,
DV Dimensionless Velocity
10%, 15%, 20%, 25% and 30%) at airflow velocities ranging from
SIMPLE Semi-Implicit Method for Pressure Linked
0.0135 to 0.387 m s
1. They found a 1% increase in fine con-
Equation
centrations resulted in 0.87% increase in airflow resistance. A
Symbols similar result was obtained by Chung, Maghirang, Kim, and
S Source term for the momentum equation Kim (2001) who also investigated the effect of fine content
D Normal viscous resistance (m
2) on rough rice at different MCs (12%, 13%, 15%, 16%, and 18%)
C Inertial resistant coefficient (m
1) and four levels of fine materials (0%, 1%, 3%, and 5%) at ve-
3 Porosity locities ranging from 0.05 to 0.38 m3 m
2 s
1.
r Density (kg m
3) Loading of the bin from the centre of the roof can result
u Velocity (m s
1) in peak/cone shape grain mass configuration. As the grain
P Pressure (Pa) drops into the middle of the bin, and spreads radially out-
DP Pressure drop (Pa) ward toward the bin wall, the grain follows the natural angle
L Grain column depth (m) of repose (typically 30 ). The formation of a peak configu-
m Viscosity (Pa s) ration results in beds with non-uniform surfaces and un-
k Turbulent energy (m2 s
2) equal bed depth with surface grain close to the periphery
g Acceleration due to gravity (m s
2) being a few metres to several metres below the top of the
S Mean modulus of the rate of strain tensor peak. The extra grain depth in the centre of the bin increases
s Turbulent based on Prandtl number air flow resistance. In particular, the region at the peak of
Q Airflow (m
3 s [air] m
2) the cone usually has low exposure to airflow which could
cause grain at those locations to remain at high MCs for
Subscript
extended periods of time; grain quality reduction, mould
bulk Bulk density (kg m
3)
development, and mycotoxin contamination may ensue
particle Particle density (kg m
3)
(Bartosik & Maier, 2006). To reduce problems associated with
i, j Prescribed matrices
peak grain mass configuration, producers use coring and
n Normal component
levelling of the grain bed surface to improve airflow distri-
t Tangential component
bution within the grain. The coring process involves
3 Turbulent dissipation rate constant
partially drawing out grain at the bin core. The coring pro-
cess results in an inverted cone mass configuration with
bottom of the inversion at several metres below the surface
improving grain storability (Calderon, 1972; Foster, 1979; of the grain at the periphery.
Navarro & Noyes, 2001). However, grain conditions can make A few experimental and modelling studies on airflow dis-
the aeration process less effective by impeding uniform flow tribution in rice bins are reported in literature. Calderwood,
of air throughout the bed. Presence of fines and/or dirt in the Cogburn, Webb, and Marchetti (1984) studied the difference
grain (dockage), grain mass configuration, and the physical in viability reduction in long-grain rice stored in aerated and
properties of the grain can produce less exposure to aeration non-aerated bin for 30 months storage. They found that non
in sections of the bin; areas of the grain mass subjected to high aerated grains had zero viability. Using a corn bin, Bartosik
exposure to aeration could over-dry (Flinn, Hagstrum, & Muir, and Maier (2006) found a reduction from 89% to 36% in non-
1997). As a consequence, poor aeration is associated with uniformity factor (NUF) when a peaked grain mass con-
problems of grain moisture build up, mould growth, insect figuration was changed to flatbed grain mass configuration.
attack and, spoilage (Bartosik & Maier, 2006; Khatchatourian & NUF is defined as (peripheral air velocity
central air
Binelo, 2008). There is need to understand airflow distribution velocity)  100/(peripheral air velocity þ central air velocity).
and pattern at different bed conditions of grain for successful De Ville and Smith (1996) presented an analytical solution for
aeration to prevent the quality of rice deteriorating, myco- solving nonlinear air flow distribution through a bed of rape-
toxin contamination occurring, and to reduce economic seed, wheat, barley and corn. Lai (1980) presented three-
losses. dimension nonlinear partial differential equations that
Atungulu et al. (2013) investigated dockage in freshly- describe the axisymmetric airflow through a porous media
harvested rice of different varieties (M104, M202, M205, and using the Ergun equation. The core and the periphery of the
M206) and harvest MCs that ranged from 18% to 27% wet basis grain mass were modelled using two different porosities (0.4
and found that the dockage varied between 0.2% and 2.0% but and 0.6). Similarly, Lawrence and Maier (2011) used the NUF to
the amounts could be more depending on harvester and validate the work of Bartosik and Maier (2006). The author
weather conditions. When the grain is transferred to bin used Fluent computational fluid dynamics software with
288 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7

porous media theory to solve a 3D non-uniform airflow model coefficient and subscript n and t are normal and tangential
for peaked, levelled and, inverted grain mass configurations in components, respectively. Turbulence was incorporated
a maize silo. They estimated the porosity using linear inter- using the standard k-3 model according to the following
polating function with porosity of 0.34 at the core of the bin equations:
and 0.38 at the wall. k equation
The objectives for this research are to (1) investigate airflow   
vðkÞ v m vk pffiffiffi
distribution in a rice bed for various airflow conditions in ¼ mþ t þ mt S2
rε; S ¼ 2Sij Sij (4)
vt vxj sk dxj
peaked, inverted and flatbed grain mass configurations, and
(2) investigate the effects of variable porosity on airflow dis- 3 equation
tribution in the bin.   
vðεÞ v m vε ε

¼ mþ t þ C1ε mt S2
rC2ε ε (5)
vt vxj sε dxj k

2. Materials and methods k2

2.1. Numerical algorithm
A porous media formulation was used to model three-
dimensional models for peaked, inverted and flatbed grain
mass configurations consisting of a long-grain rough rice. A
finite volume approximation method was used to solve the
governing equations of the numerical model using CFD
software FLUENT (ANSYS, version 16.2, Canonburg, PA,
USA). The governing equations are solved using pressure-
based solver in unsteady formulation and iterated to ach-
ieve a converged solution. The SIMPLE (Semi-Implicit
Method for Pressure Linked Equation) algorithm was used
for pressure and velocity coupling. The presto and second
order upwind differencing scheme was used for the spatial
discretisation of pressure and momentum equation and first
order implicit scheme was used for discretisation of the
transient terms.
2.1.1. The governing equations
For the porous-flow model, continuity (Eq. (1)) and mo-
mentum (Eq. (2)) equations were solved in three dimensions
for the superficial gas velocities and pressure distribution with
two additional momentum sink terms to simulate flow
through porous media where the standard conservation
equations for turbulence quantities (Eqs. (4) and (5)) in the
porous medium were also solved.
Continuity equation:
mt ¼ C m r (6)
ε
where, Ci3 , C3 2, sk and s3 are constants, S is the mean modulus
of the rate of strain tensor; sk is turbulent based on Prandtl
number; and mt is the mean turbulent viscosity.
2.1.2. Porous media resistance coefficients
The experimental data of pressure drop against velocity
through a bed of rice was extrapolated to determine the
coefficients for the porous media following the method
described in the Fluent (2016) manual. A pressure drop
testing device (Fig. 1) was constructed consisting of a
transparent acrylic cylindrical pipe with internal diameter of
0.1 m and a height of 1.0 m. The lower end of the bed
domain was fitted with a distributor plate made of wire
mesh that served as a support to the samples. Air was
supplied by industrial compressor (Model 5120, Quincy,
Minette, AL, USA) powered by three-phase electric motor.
The air flow was regulated by valve (d) attached to the
pressure test kit and measured by air flow meter (c) (Model
3000, Kestrel Birmingham, MI, USA). Pressure drop across
the bed was measured by of digital manometer (b) (Series
4753, Michigan City, IN, USA) connected to pressure taps
located at the top (i.e. close to the outlet) of the pipe and
0.01 m below the distributor (a).

vr vðrui Þ
þ ¼0 (1)
vt vxi
Momentum equation:

  
vðrui Þ v rui uj vP v vui vuj
þ ¼
þ rgi þ ðm þ mt Þ þ þ Si (2)
vt vxi vxi vxj vxj vxi

where, r is the density (kg m
3), t is time (s), ui and uj are the
average superficial velocity component respectively (m s
1),
P ¼ pþ (2/3)rk, where p is the pressure (Pa), k is the turbulent
energy (m2 s
2), gi is the acceleration due to gravity (m s
2), m is
the viscosity (Pa s); i, j ¼ 1, 2, 3 (x, y, z) and Si is the source term
for the momentum equation as shown in Eq. (3).
 
1
si ¼
Dij mu þ Cij rjuju (3)
2
Fig. 1 e Pressure drop test device. (a) Lower pressure
where, Dij and Cij are prescribed matrices of the material, D is tapping; (b) Digital manometer; (c) Anemometer; (d) Airflow
the normal viscous resistance, C is inertial resistance valve.
 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7 289

The following procedure was carried out for pressure drop volume of the material in the sample cell. The particle density
measurement. A sample of rice weighing 4840 g was loaded was computed from the ratio of the sample mass to the
into the pressure test chamber. The valve was then opened measured particle volume. The porosity was then calculated
gradually to increase the rate at which air flowed through the using Eq. (7):
bed. At each air velocity, at least 60 s were allowed for stabi-
rbulk
lisation before recording the pressure drop across the cham- ε¼1
(7)
rparticle
ber; the corresponding air velocity through the bed was
recorded. A polynomial fit of pressure drop against superficial where 3 is the porosity, rbulk is bulk density (kg m
3) and rparticle
gas velocity was executed in a Microsoft Excel spreadsheet. is particle density (kg m
3).
The test was repeated three times and the average result of In order to determine the variable porosity, the rough rice
the viscous (1/a) and inertial (C2) losses were estimated from was cleaned using a dockage equipment (MCi Kicker Dockage
the plot to be 9.72  106 (m
2) and 36,185 (m
1), respectively. Tester, Mid-Continent Industries Inc., Newton, KS, USA). The
foreign materials obtained were collected and marked sepa-
2.1.3. Geometry, mesh grids and boundary conditions rately. The samples were then remixed for variable porosity
Geometry for peaked, inverted, and levelled bin configurations test. The first mixed sample was fifty to fifty (50-50) percent
was developed using the ANSYS workbench. For the levelled proportion by weight and the second was seventy-five to
grain mass configuration design, the bin had a diameter and twenty-five (75-25) percent proportion by weight. Porosity of
bed height of 15.63 m and 4 m, respectively. However, the clean, 50-50, 75-25 and unclean samples were determined
peaked grain mass configuration had a base height of 4 m and using the same procedure as described previously and found
base angle of 24 ; which is the natural angle of repose of long- to be 0.55 ± 0.10 and 0.45 ± 0.10, 0.52 ± 0.01 and 0.55 ± 0.01
grain rough rice determined by experiment. The inverted cone respectively.
was constructed from the peaked grain mass configuration.
This was done by removing volumes based on an inverted 2.3. Simulation procedure
angle of repose. The inverted bin configurations were formed
using scenarios where 1 t, 50 t, and 100 t of the grain mass was In order to perform the simulation, porous zone was first
removed. activated from the cell zone conditions and the values of
The meshes for the peaked, inverted and the flatbed grain porosity, viscous and inertial resistance were added to the
mass configurations were generated using a tetrahedral mesh. appropriate categories. Turbulence scheme was activated
Minimum and maximum orthogonal skew were used for and the model constant for the standard kee turbulence
determining the optimal mesh type and grid size considering model with standard wall functions Cmu ¼ 0.09, C1 ¼ 1.44,
mesh quality, total number of cells and computational accu- C2 ¼ 1.92, sk ¼ 1.0 were set. The airflow velocity selection
racy. The coordinate system for the numerical model was a 3D was based on the velocities used in rough rice aeration
Cartesian coordinate system. The origin of coordinates was process in Arkansas region which varies between 0.55 and
set to the centre of the bin. In the coordinate system, z was 1.1 m3 min
1 [air] t
1 [rice]. In this simulation, 0.55, 0.825 and
positive towards the flow direction, x was perpendicular to the 1.1 airflows m3 min
1 [air] t
1 [rice] were investigated.
flow direction in the horizontal plane. The base of the bin was However, since the bed configurations resulted in different
described by the velocity-inlet boundary condition, while the bin volumes, the airflow was adjusted based on the bin
top was described by the pressure outlet boundary condition. volume (Table 1).
The solid side surface was modelled using the wall boundary The variable porosity was modelled by implementing a
condition specified shear stress set at zero for x, y and z- User-Defined Function (UDF) in the Fluent software written in
components. the C-programming language. For variable porosity simula-
tions, the porosity of the core was chosen to be 0.45 while
2.2. Physical properties estimation 0.55 was used for the region closer to the wall. An interme-
diate region is the bed region between the bin core and its
2.2.1. Determination of bulk and particle density periphery. The porosity of the intermediate region was
Bulk density was determined using standard bulk density determined using linear interpolation. Similarly, variable
measuring apparatus. The procedure involved allowing the porosity was also implemented for the various degrees of
sample to freely fall through a funnel onto a standard cup inversion. A porosity of 0.45 implies the worst case scenario
(volume of 118 mm3). The material was then levelled across
the surface of the container and weighed using a digital bal-
ance (Model SP6001, Ohaus Corp, Pinebrook, NJ, USA) with Table 1 e Airflow velocity at various grain weights.
0.1 g precision. Bulk density was then estimated as the ratio of Peaked Flat Inverteda
the mass of the sample in the container to the volume of the 700 640 1-t 50-t 100-t
container. A gas pycnometer (Accupyc 1330, Micromeritics 3
1
1
1
Airflow (m min -t ) Airflow (m s )
Instrument Corp., Norcross, GA, USA) was used for particle
0.55 0.033 0.027 0.033 0.031 0.029
density measurement. The pycnometer measures the pres-
0.83 0.050 0.040 0.049 0.046 0.043
sure difference between a reference cell and a cell containing
1.10 0.067 0.054 0.066 0.061 0.057
the measured sample. The pressure was generated by allow- a
Mass removed from a bed with peaked grain mass configuration
ing helium to flow from the reference cell into the cell con-
(700 t).
taining the sample. The pycnometer then calculates the
290 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7

where the rough rice has a high fine dirt content of more
than 50%.

2.4. Validation of model

2.4.1. Laboratory validation

The average result of the viscous (1/a) and inertial (C2)
losses that was estimated to be 9.72  106 (1 m
2) and
36,185 (m
1), respectively were validated by predicting the
pressure drop across the porous bed consisting of long-
grain rough rice at different air flow velocities. The bed
(0.1 diameter and 1.0 m bed height) was developed using
3-D Cartesian coordinate system with porous media
formulation of 0.55 constant porosity. The pressure based
solver with transient time simulation were activated. The
inlet boundary has uniform air velocity inlet (0.2e2 m s
1)
while the outlet boundary is the pressure boundary con-
dition set at zero-gauge pressure to model the system
open to the atmosphere. For the wall, boundary conditions
were set at specified shear set to zero for x, y and z
components. For all simulations, the time step, maximum
iteration time and reporting interval was set at 10 s, 20 s,
and 1 s, respectively. Statistical comparisons were made
between the result of the pressure drop obtained at
different velocity for experimental, simulated and ASABE
airflow resistance curve (ASABE, 2011) calculated using Eq.
(8), which took into account rice depth and a correction
factor for the cleanness of the rice:

DP aQ 2
¼ C (8)
L loge ð1 þ bQÞ

where, DP is pressure drop (Pa); L is grain column depth Fig. 2 e Funnel and digital anemometer used for measuring
(m); Q is airflow (m
3 s [air] m
2); a and b are grain specific airspeed for a cone shape grain mass configuration.
constants. For rough rice, a ¼ 2.24  104 (Pa s2 m
3) and
b ¼ 13.2 (m2 s m
3) when Q ε (0.0056, 0.152). The term C is
the correction factor which depends on the cleanliness of
3. Results and discussion
the rice (1 represents clean kernels only, and 1.5 represents
rice with maximum amount of foreign materials) (Arana,
3.1. Laboratory validation
2012).

Figure 3 shows the pressure against the superficial velocity for

2.4.2. Field validation
experimental, simulated and ASABE data. Generally, the
Validation of the model was carried out using a field stationed
pressure drop increased with increase in superficial velocity.
bin with long-grain rice. The bin which was located at Bur-
As velocity increased the pressure drop at different airflows
dette, AR, USA, had rice filled in a cone-shape grain mass
configuration and was subjected to natural air drying. The bin
was 15.63 m in diameter and 7.32 m in height. The airflow was 7000
generated by a 29.4 kW centrifugal fan operating at 1740 RPM.
6000
Eq. (8), and the fan curve provided by the fan manufacture Simulation
5000
Pressure (Pa)

(The GSI Group, Assumption, IL, USA) were used to estimate Experimental
the airflow of the fan. 4000
ASABE
A wind vane anemometer (Omega HHF91, Omega En- 3000
gineering, Inc., Stanford, CT, USA) and a custom-built 2000
funnel with small and large diameters of 0.41 m and
1000
0.08 m respectively (Fig. 2) were used to measure the ve-
0
locity of the air because the low air velocity found could
0 0.1 0.2 0.3 0.4 0.5 0.6
not be measured directly by the flowmeter when it was
Velocity (m s-1)
placed directly at the grain surface. The concept of conti-
nuity was then used to calculate the airflow at a particular Fig. 3 e Validation of viscous and inertial resistance on
position. predicted pressure drop at different airflow velocity.
 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7 291

Fig. 4 e Predicted airflow velocity profile for peaked shape grain mass configuration using the constant porosity (0.55) at
0.067 ms¡1.

predicted from ASABE equation under predicted the experi- 3.1.1. Simulation of the effect of grain mass configuration on
mental values. The difference between predicted and experi- airflow through the bed at constant porosity
mental data was evaluated by determining the mean relative Figure 4 shows the airflow distribution pattern in a cone shape
deviation (MRD) (Nemec & Levec, 2005; Olatunde et al., 2016). grain mass configuration for a bin containing long-grain rough
At lower airflow velocities (<0.3 ms
1), the MRD between the rice with airflow of 0.067 m s
1 (1.1 m3 min
1 [air] t
1 [rice]). It
experimental and ASABE calculations and between the can be seen that airflow varied with location in the bed with
experimental and simulated values were <8%. However, as the peak of the bed having a velocity that was <0.012 m s
1
airflow velocity increased, the MRD between experimental while the base of the cone had the highest air flow (0.07 m s
1).
and ASABE increased to 17%. Statistical analysis showed that This represents more than five-fold difference in airflow be-
there was no significant difference between experimental and tween the peak and the base of the cone. Reduction in airflow
simulated results while the ASABE vs experimental results as the bed height increased was also reported by Bartosik and
were significantly different (p < 0.05). Since the MRD is below Maier (2006) and Lawrence and Maier (2011). The authors
10%, then the viscous and inertial resistance used in the explained this by the fact that higher grain depth increases
simulated study was adjudged to predict the experimental the resistance to airflow due to increase in the residence time
data, hence validating the laboratory setup. that the air required to navigate tortuously through the grain
void before reaching the surface. Since air follows the path of
least resistance, in case of the peaked grain mass configura-
1.0 tion, the air moves toward the shallow depth at the periphery
Flat bed of the mass. Hence the pressure per unit length decreases as
Dimensionless dynamic pressure

0.8 Airflow (m3 min-1-t-1) the air approaches the peak of the grain mass configuration.
0.55 0.55 The airflow velocity profile for flatbed grain mass configu-
0.6 ration was uniform. This condition of bed ensures uniform
0.83 0.83
(DDP)

Cone shaped bed drying and prevents moisture build up in certain areas of the
0.4 1.10 1.10
bed. This is the best configuration for optimal airflow distri-
bution in the bin but it is not widely used because of additional
0.2
instrumentation installation costs (spreader) and/or the use of
0.0
manual labour which adds to operating costs. Also, for the
0 2 4 6 8 10 same bin size, the flatbed grain mass configuration holds less
Bed height (m)
volume of grain than the peaked shape grain mass configu-
Fig. 5 e Dimensionless dynamic pressure (DDP) for flat and ration. Results from the flatbed grain mass configuration
peaked shape grain mass configuration at the centre further showed that increases in bed height coupled with non-
position of the bin and at various velocity. DDP is the ratio uniformity of bed surface are responsible for the non-
of Pi to Pmaximum; Pi is the local dynamic pressure (Pa) and uniformity in air flow in the bed.
Pmaximum is the highest dynamic pressure (Pa) at a Further, Fig. 5 shows the plot of the dimensionless dy-
particular velocity. namic pressure (DDP) of air (fluid kinetic energy) as bed height
292 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7

Fig. 6 e (a): Predicted airflow velocity profile for inverted grain mass configuration in which 1-t grain mass was removed
with constant porosity (0.55) and at 0.06 m s¡1 airflow. (b): Predicted airflow velocity profile for inverted grain mass
configuration in which 50 t grain mass was removed with constant porosity (0.55) and 0.057 m s¡1 airflow.

increased. The DDP was calculated by dividing the dynamic
pressure data with the maximum dynamic pressure at a
particular airflow velocity. Generally, DDP reduced with in-
crease in bed height in peaked grain mass configuration while
DDP values were uniform for flatbed grain mass configuration.
This further showed that the additional layer of non-uniform
grain surface/level impacts the airflow penetration through
the bed. When the airflow was increased from 0.033 m s
1
(0.55 m3 min
1 [air] t
1 [rice]) to 0.067 ms
1 (1.1 m3 min
1 [air]
t
1 [rice]), the DDP varied marginally between 1 and 5 m bed
thickness area (peaked grain mass configuration). This could
be attributed to difference in turbulence experienced within
the bed as airflow increased (turbulence intensity increased to
0.005% from 0.001% as velocity was increased). Major practical
problems could be the increased energy required to run the
fan to achieve the high airspeeds and its associated cost and
overexposure of grains to high airspeeds causing over-drying
of grains in that are close to the air inlet zone. Furthermore,
if new bin technology uses cables for sensors it is possible that
airflow turbulence could distort the accuracy of data and
affect the timeliness of the control system as drying airflow
increased.
The air velocity patterns for the inverted grain mass
configuration for 1-t and 50-t removed configurations with
constant porosity (0.55) and 1.1 m3 min
1 [air] t
1 [rice] airflow
are shown in Fig. 6. The airflow velocity within the bed varied
 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7
Fig. 7 e Predicted airflow velocity profile for inverted grain mass configuration in which 100-t grain mass was removed with
constant porosity (0.55) and at 0.057 m s¡1 airflow.
1.6
Dimensionless airflow
1.2
0.8
1-t
0.4
50-t
100-t
0 removed configuration increased due to reduction in bed
-8 -6 -4 -2 0 2 4 6
Radial coordinate (m)
Fig. 8 e Dimensionless velocity (DV) for inverted grain
mass configuration for 1-t, 50-t and 100-t removed
configurations at bed height of 5 m. DV is the ratio of Ui to
Uapplied; Pi is the local velocity (m s¡1) and Papplied is the
velocity at the plenum (m s¡1).
between 0.061 and 0.070 m s
1 and 0.068e0.072 m s
1 for 1-t
and 50-t removed configuration, respectively. Also, the air
flow velocity at the cleavage (area where grain mass was
removed) varied between 0.0460 and 0.0537 m s
1 and
0.135e0.144 m s
1 for 1-t and 50-t removed configuration,
respectively. The peak of the bed for both cases has less than
0.01 m s
1 air flow. Increased airflow at the centre of the invert
was also observed when 100 t of grain was removed (Fig. 7).
The airflow values at the centre of the invert were in the range
0.170e0.181 m s
1, while airflow at the top of the periphery
(Fig. 8) was reduced substantially to less than 0.01 m s
1. High
airflow within the bed (0.053e0.064 ms
1). Increase in airflow
at the cleavage could be attributed to reduce bed height as a
result of coring (Lawrence & Maier, 2011).
293
In order to compare the airflow profile of the inverted grain
mass configurations, Fig. 8 shows the plot of the dimension-
less airflow (DV) measured at bed height of 5 m. The DV was
calculated by dividing the local velocity data with the velocity
at the plenum. Generally, the grain mass with 50-t removed
configuration has the highest airflow toward the periphery
(DV > 1) while grain mass with 100-t resulted in the lowest
airflow toward the periphery (DV < 0.8). This showed the
impact of additional layer of grain at the core resulted in air
flow diversion from the centre (DV < 0.8) to the periphery.
However, at the core of the bin, the airflow of 50-t and 100-t
8
height. The result also revealed that some section of the bed
may have low airflow for instance, 1-t removed configuration
have lower airflow at the core (DV < 0.8), 50-t removed
configuration have lower airflow between 4 and 6 m radial
distances and lower air flow was observed at the periphery for
100-t removed configuration. These are area of the bed with
potential for poor aeration and processor must pay a closer
attention.
3.1.2. Simulation of the effect of grain mass configuration on
air flow through the bed at variable porosity
Figure 9 shows the peaked grain mass configuration with
variable porosity at 0.067 m s
1. The region with high resis-
tance to airflow increased tremendously compared to Fig. 4
and extends downward to the inlet region at the core of the
bed. The air velocities within the low porosity section ranged
between 0.0503 and 0.059 m s
1 while the airflow at the pe-
riphery ranged between 0.078 and 0.101 m s
1 as shown on the
scale. Poor airflow at the middle section and peak of the bed
could be the principal reasons why peaked grain mass
configuration is susceptible to mould development because
high moisture grains microclimates create a condition for
mould growth, discolouration of kernels, and in worst case
scenarios, mycotoxin contamination. The air velocities and
294 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7

Fig. 9 e Predicted airflow velocity profile for peaked shape grain mass configuration for the variable porosity of 0.55 at the
peripheral to 0.45 at the core of the bin and 0.067 ms¡1.

Table 2 e Predicted air velocities at the central and peripheral surface locations of the grain mass for different grain mass
configurations and variable porosity.
Type Measurement location Mass removed (t) Airflow velocity (m s
1)
Peripheral Centre NUFc (%)
a
Cone shape North 0.106 0.01 71.09
West 0.106 0.01 71.09
South 0.106 0.01 71.09
East 0.106 0.01 71.09
Invertedb North 1 0.099 0.016 71.88
West 0.099 0.016 71.88
South 0.099 0.016 71.88
East 0.099 0.016 71.88
North 50 0.093 0.029 52.93
West 0.093 0.029 52.93
South 0.093 0.029 52.93
East 0.093 0.029 52.93
North 100 0.024 0.14
71.36
West 0.024 0.14
71.36
South 0.024 0.14
71.36
East 0.024 0.14
71.36
a
The velocity was measured at the base (periphery) and at the tip of the peaked.
b
The velocity was measured at the eye of the inverted cone and top of the inversion (periphery).
c
NUF symbolizes non-uniformity coefficient and was calculated from (peripheral air velocity
central air velocity)  100/(peripheral air
velocity þ central air velocity).

NUF for variable porosity at the peaked grain mass configu-
ration predicted at different bed positions are shown in Table
2. For the peaked grain mass configuration, the predicted
average NUF was 71%.
The inverted grain mass configuration with variable po-
rosities at different angles of inversions corresponding to 1 t,
50 t and 100 t of rice removed from the grain mass at airflow
velocity of 0.066, 0.061 and 0.057 ms
1 are shown in Figs.
10e12, respectively. For the three scenarios, the airflow at
the centre of inversion increased. When 1 t, 50 t and 100 t of
rice were removed, the airflows at the centre of inversion were
0.05 ms
1, 0.135 ms
1, and 0.151 ms
1, respectively. Reduction
in bed height at the core resulted in reduction in airflow re-
striction and hence more airflow penetration. Also, as the
angle of inversion increase from 1 t to 50 t of grain mass was
removed, the airflow resistance at the peripheral reduced
from 0.099 ms
1 to 0.015 ms
1, respectively and the NUF
decreased from 70% to 50%, but a further increase in mass
removed to 100 t caused an increase in NUF to 70%. A negative
NUF means that the airflow at the periphery is lower than the
airflow at the centre of the grain mass. This result showed that
not more than 50 t of bed should be removed in order to
 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7 295

Fig. 10 e Effect of degree of inversion in terms of grain mass removed and variable porosity on airflow resistance through a
rice bed in which 1 t of rough rice was removed.

Fig. 11 e Effect of degree of inversion in terms of grain mass removed and variable porosity on airflow resistance through a
rice bed in which 50 t of rough rice was removed.

maximise bin volume and optimise airflow distribution, even
with the presence of dockage.
3.2. Validation of simulation and airflow measurement
The results of the air velocities and NUF of both experimental
and simulated data using the peaked shape grain mass
configuration are shown in Table 3. The NUF determined for
the experimental bin varied between 29.9% and 34.4% with
respect to different locations in the bin. The west and the
southern locations of the rice bed had the lowest and highest
NUF of 29.8% and 34.4%, respectively. Generally, the velocity
increased radially outward. For instance, the velocity at the
periphery for the measurement done at the North location was
found to be 0.076 ms
1 while the corresponding velocity to-
ward the centre was found to be 0.047 ms
1. This further
confirmed that the airflow was not uniform within the bin and
that the region closer to the wall received more air compared
to the centre of the bin. The experimentally estimated airflow
of 1.1 m3 min
1 [air] t
1 [rice] (0.067 ms
1) with a constant
porosity of 0.55 was used for the validation and the result is
shown in Table 3. The percent difference between experi-
mental and predicted airflow at the peripheral was les that 2%.
However, the simulation over predicted the velocity at the
centre of the bin by 10%. The disparity in the prediction could
be due to porosity values used at the core because in reality the
296 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7

Fig. 12 e Effect of degree of inversion in terms of grain mass removed and variable porosity on airflow resistance through a
rice bed in which 100 t of rough rice was removed.

Table 3 e Air velocity measured at the centre and 4. Conclusion

periphery (close to the bin wall) of the peaked shape grain
mass configuration.
The airflow distribution in peaked, inverted and flatbed grain
Data Measurement Airflow velocity (m s
1) mass configurations during natural air drying process of
type location
Periphery Centre NUFb (%) rough rice in bin systems were modelled using three-
Experiment North 0.076 0.047 33.58 dimension finite volume method with porous media formu-
West 0.087 0.047 29.85 lation. The model was validated with data obtained from a bin
South 0.086 0.042 34.38 containing long-grain rice in a peaked shape grain mass
East 0.088 0.045 33.52 configuration. Also, the effect of variable porosity and various
Constant North 0.076 0.052 18.75
angles of inversion were investigated and the following were
porositya West 0.076 0.052 18.75
concluded:
South 0.076 0.052 18.75
East 0.076 0.052 18.75
(1) Average NUF of the experimental data was found to be
The velocity was measured at the base (periphery) and at the tip of
34% while that of constant and variable porosity simu-
the peak.
a
The velocity was measured at the eye of the inverted cone and lations were found to be 18% and 71%, respectively.
top of the inversion (periphery). (2) The NUF of simulated inverted grain mass configuration
b
NUF symbolises non-uniformity coefficient and was calculated was found to be reduced from 50% to 71% as the mass of
from (peripheral air velocity
central air velocity)  100/(periph- grain removed from the bin increased from 1 to 50 t.
eral air velocity þ central air velocity). (3) For inverted scenario, simulations indicate that a
maximum of 50 t of grain can be removed to ensure
uniform airflow distribution in the bin.
bin is not clean. However, grain producers typically approxi-
(4) For peaked shape scenario, simulations indicate that
mate the amount of the grain stored in their bins. They typi-
there are low airflows at the peak position in the bed
cally have rough estimate and hence difference between
this could create a conditions for rapid microbial
predicted and the experimental data may be marginal in terms
growth.
of field practice. However, the result of NUF with constant
porosity for peaked grain mass configuration is shown in Table
3 to be 19%. The NUF between predicted and experimental
were about 50% different. This implies that the grains in field
bins were not clean and the assumption of constant porosity Acknowledgement
may not apply. However, the NUF obtained for variable
porosity was found to have an NUF of 71% (Table 2). This shows The authors greatly appreciate Rice Research and Promotion
that even, with clean rough rice, peaked grain mass configu- Board for financially supporting this study, The University
ration still have non-uniformity in airflow through the bed. of Arkansas Division of Agriculture, Grain Processing
 b i o s y s t e m s e n g i n e e r i n g 1 5 1 ( 2 0 1 6 ) 2 8 6 e2 9 7
Engineering Program at the University of Arkansas Division of
Agriculture for collaborations and providing facilities for
research. Rice producers in the Arkansas Delta Region espe-
cially Mr. Mike Sullivan in Burdette and Paul McKinney in
Newport, Arkansas.
references
Arana, I. (2012). Physical properties of foods: Novel measurement
techniques and applications (1 ed.). London: CRC Press, Taylor &
Francis Group.
ASABE. (2011). Resistance to airflow of grains, seeds, other agricultural
products, and perforated metal sheets (Vol. D272.3). St. Joseph,
Mich.: American Society of Agricultural and Biological
Engineering.
Atungulu, G., Prakash, B., Wang, X., Wang, T., Fu, R., Khir, R., et al.
(2013). Determination of dockage for accurate rough rice
quality assessment. Applied Engineering in Agriculture, 29.2,
253e261.
Bartosik, R., & Maier, D. (2006). Effect of airflow distribution on the
performance of NA/LT in-bin drying of corn. Transactions of the
ASABE, 49(4), 1095e1104.
Calderon, M. (1972). Aeration of grain e Benefits and limitations.
EPPO Bulletin, 2(6), 83e94. http://dx.doi.org/10.1111/j.1365-
2338.1972.tb02137.x.
Calderwood, D., Cogburn, R., Webb, B., & Marchetti, M. (1984).
Aeration of rough rice in long-term storage. Transactions of the
ASAE, 27(5), 1579e1585.
Chung, D., Maghirang, R., Kim, Y., & Kim, M. (2001). Effects
of moisture and fine material on static pressure drops in a bed
of grain sorghum and rough rice. Transactions of the ASAE,
44(2), 331.
297
De Ville, A., & Smith, E. A. (1996). Airflow through beds of cereal
grains. Applied Mathematical Modelling, 20(4), 283e289. http://
dx.doi.org/10.1016/0307-904X(95)00121-Y.
Flinn, P., Hagstrum, D., & Muir, W. (1997). Effects of time of
aeration, bin size, and latitude on insect populations in stored
wheat: A simulation study. Journal of Economic Entomology,
90(2), 646e651.
Fluent, A. (2016). 16.2 User's guide (Version 16.2). Canonsburg, PA
15317: Ansys.
Foster, G. (1979). Managing grain for year-round storage. West
Lafayette IN U.S.A: Purdue University. Retrieved from https://
www.extension.purdue.edu/extmedia/ae/ae-90.html.
Khatchatourian, O. A., & Binelo, M. O. (2008). Simulation of three-
dimensional airflow in grain storage bins. Biosystems
Engineering, 101(2), 225e238. http://dx.doi.org/10.1016/j.
biosystemseng.2008.06.001.
Lai, F. (1980). Three-dimensional flow of air through nonuniform
grain beds. Transactions of the ASAE, 23(3), 729e0734.
Lawrence, J., & Maier, D. E. (2011). Three-dimensional airflow
distribution in a maize silo with peaked, levelled and cored
grain mass configurations. Biosystems engineering, 110(3),
321e329.
Navarro, S., & Noyes, R. T. (2001). The mechanics and physics of
modern grain aeration management (1 ed.). Florida U.S.A.: CRC
Press, Taylor & Francis Group.
Nemec, D., & Levec, J. (2005). Flow through packed bed reactors: 1.
Single-phase flow. Chemical Engineering Science, 60(24),
6947e6957.
Olatunde, G., Fasina, O., Adhikari, S., McDonald, T. P., &
Duke, S. R. (2016). Size measurement method for Loblolly Pine
grinds and influence on predictability of fluidization. Canadian
Biosystems Engineering, 58(1), 4.1e4.10.
Siebenmorgen, T. J., & Jindal, V. K. (1987). Airflow resistance of
rough rice as affected by moisture content, fines
concentration and bulk density. Paper/ASAE, 86, 3036.