Professional Documents
Culture Documents
Available at www.sciencedirect.com
article info
A mathematical model and software were developed for the three-dimensional simulation of
Article history:
airflow through high capacity grain storage bins by considering the non-uniformity of the seed
mass. To validate the proposed model, empirical relationships between air velocity and static
pressure drop were obtained for compacted layers of several storage depths for soya bean,
29 May 2008
maize, rice and wheat mass. The software was written in ANSI C which is transferable to
a variety of platforms. For the construction of 3D geometry and the generation of meshes
free-of-charge software was used. The solver software generated a system of linear algebraic
equations using the finite -element method. Three iterative processes were carried out: (1) calculation of a local permeability coefficient, using the pressure distribution in the immediately
previous iteration step, (2) search for the system design point, located in the performance curve
of the aerator fan, and (3) adaptation to refine the mesh. A local criterion to estimate the efficiency of complex aeration system in storage bins was proposed. The simulations showed
good performance. It was considered that the method could be applied to optimise the performance of existing grain stores and lower the engineering costs of new grain stores.
2008 IAgrE. Published by Elsevier Ltd. All rights reserved.
1.
Introduction
* Corresponding author. Department of Physics, Statistics and Mathematics, Regional University of the Northwest, Rio Grande do Sul,
R. Sao Francisco, 501, 98700-000 IJUI, RS, Brazil.
E-mail addresses: olegkha@unijui.edu.br (O.A. Khatchatourian), mbinelo@yahoo.com.br (M.O. Binelo).
1537-5110/$ see front matter 2008 IAgrE. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.biosystemseng.2008.06.001
226
Nomenclature
A
a
b
C
c
G
H
i
k
L
LX
M
m
n
n
P
Pe
Q
q
R
S
surface area, m2
product-dependent constant
product-dependent constant
compaction function, dimensionless
product-dependent constant
product-dependent constant
bed depth, m
order number of corresponding inlet
permeability coefficient, m3 kg1 s
bed depth, m
full length of a trajectory, m
total number of experimental points
grain mass, kg
product-dependent constant; inlet number
unit vector normal
pressure, Pa
air entrance or exit pressure in Pa;
global airflow rate, m3 s1 kg1
local specific airflow rate, m3 s1 kg1
product-dependent constant
empirical coefficient
To simulate the aeration of grain, with any type of air distribution systems, it is necessary to develop software to predict the distribution of the parameters, because obtaining
empirical data is very difficult and costly. Most research
on airflow simulation in grain stores is related to onedimensional, two-dimensional or axisymmetric cases; although the flow is usually three-dimensional. Even when
the grain mass distribution is two-dimensional or axisymmetric, the airflow inlets do not satisfy these conditions.
Also, the aeration of large grain stores is frequently carried
out separately in different segments.
The principal objectives of the present work were as follows:
(a) to create a mathematical model, algorithm, and software,
to calculate the static pressure, streamlines, and airflow
velocity distribution in three-dimensions under nonhomogeneous conditions;
(b) to determine the variation in compaction factor for several
depths of grain;
(c) to study the relationship between the air velocity and the
pressure gradient as a function of the compaction factor;
(d) to develop and incorporate into the software a criterion
for system performance based on estimating threedimensional air distribution in grain storage bins; and
(e) to carry out numerical simulations of real and hypothetical
grain stores with aeration to detect areas of operational risk.
2.
Mathematical model
The problem of incompressible viscous isothermal flow is described by the system of equations of continuity [Eq. (1)] and of
NavierStokes [Eq. (2)]:
t
U
V
V
Xi
x
y
z
3
r
DP
m
time, s
intermediate argument
velocity vector, m s1
velocity, m s1
product-dependent constant (i 1, 2, 3)
coordinate located in floor plan, m
coordinate along airflow axis, m
coordinate located in floor plan, m
porosity factor, dimensionless
density, kg m3
pressure drop, Pa
dynamic viscosity, Pa s
Subscripts
a
air
b
bulk
e
entrance, exit
g
grain
i
order number of corresponding inlet
k
kernel
L
local
X
in point X(x, y, z)
div V 0;
DV
grad P mV2 V;
Dt
(1)
(2)
227
(3)
(4)
dP=dyfV2 0V kjdP=dyj
(5)
dlnjdP=Lj 21 bVln1 bV bV
(9)
ln1 bV
L
(8)
(10)
(7)
1 RS V
dln V
:
dlnjdP=Lj 1 2RS V
1.0
d(lnV)/d(ln|gradP|)
dln V
0:5:
dlnjdP=dyj
0.9
Laminar-flow
condition
0.8
Transient
regime
0.7
Turbulent
condition
0.6
0.5
-2
10
ln|gradP|, Pa m-1
Fig. 1 Observed and predicted variation of derivative
dln V=dlnjgradPj[fgradP for airflow through soya bean
mass (blue points and curves) and wheat mass (black points
and curves): ,, 6, Shedds (1953) data ; -, :, authors data;
predicted by Eq. (12); - - -, predicted by Eq. (9).
228
(12)
(13)
ln V ln 1 U2 2U arctanU p 3U 4a c;
where c is a constant of integration.
As will be shown, this equation, depending on three constants (a, b and c), describes well the experimental data in all
regions. In addition, unlike Eq. (7), explicit dependence of velocity on a pressure gradient in Eq. (13) essentially simplifies
its use together with the equation of continuity for problems
formulated in two-dimensional and three-dimensional.
Finally, the mathematical model of the airflow in the particular media for the three-dimensional case consists of a system of two equations:
div V 0;
k
k
0:
(18)
vx
vx
vy
vy
vz
vz
The boundary conditions for the problem considered have
the form:
P Pe
of the silo;
(15)
The scalar equation (14) is the continuity equation for incompressible fluid. The vector equation (15), which has replaced
the NavierStokes equation, shows that the velocity vector
and pressure gradient are collinear in all points of the airflow
domain and that the ratio of the absolute values of these vectors is a function of the pressure gradient. Expressing the coefficient of proportionality k by
k exp ln 1U2 2U arctanU p3U 4ac jgrad Pj;
(16)
and using Eq. (15), the velocity components u, v and w for the
three-dimensional case can be expressed in the form
vP
vP
vP
; v k ; w k ;
vx
vy
vz
20
3.
(14)
V
u k
(19)
3.1.
grad P
exp ln 1 U2 2U arctanU p
jgrad Pj
3U 4a c :
(17)
Geometry construction
3.2.
Mesh generation
3.3.
229
4.
Validation of the mathematical model for
non-homogeneous conditions in a grain mass
To validate the proposed mathematical model, the empirical
relationships between air velocity and static pressure drop
were obtained for compacted layers with several grain storage
depths. The coefficients a, b and c presented in the mathematical model were obtained experimentally for soya bean,
maize, rice and wheat grains. In large storage bins, due to
compaction, grain mass is a non-homogeneous medium and
the permeability coefficient varies as a function of the grain
layer depth as well as pressure gradient. Therefore the influence of the grain mass compaction factor on the permeability
coefficient was investigated.
Experimental equipment
4.2.
Experimental results
The experimental results, presented in Fig. 3, show the relationship between airflow velocity and static pressure drop in
the soya beans, shelled maize, rice and wheat mass. Table 1
shows the values of empirical model coefficients a, b and c,
230
0.5
Velocity, m s-1
ln(dP/dy), Pa m-1
0.4
-1
-2
-3
0.3
0.2
0.1
0.0
-4
500
1000
lnV, m s-1
Fig. 3 Relationship between air velocity (V) in m sL1 and
air pressure drop (dP/dy) in Pa mL1; ,, soya bean,
coefficient of correlation R2 [ 0.9954; 6, shelled maize
R2 [ 0.9982; B, rice, R2 [ 0.9934; >, wheat, R2 [ 0.9972; d,
predicted by Eq. (13).
obtained by minimising the residual error between experimental and simulated data. The simulations based on these
coefficients satisfactorily described the experimental data
(Fig. 3).
Experimental data in Fig. 4 show the significant influence
of the storage layer depth on the aerodynamic resistance of
the grain mass over the studied depths (from 1 m up to 50 m).
Fig. 5 presents a reduction of the measured porosity factor 3
with storage depths for soya bean, maize and rice, where 3 is
the ratio of the void volume to the total bed volume. Experimental porosity values were measured using a specially developed and adjusted pycnometer. The relationship between the
reduction in porosity and layer depth H can be presented as
H n
3
eS50 ;
30
(21)
1500
2000
2500
3000
dP/dy, Pa m-1
Fig. 4 Influence of bed depth (H ) on the air velocity (V) in
m sL1 as function of air pressure drop (dP/dy) in Pa mL1
(one-dimensional storage), shelled maize: -, H [ 1 m; ,,
H [ 10 m; C, H [ 20 m; B, H [ 30 m; :, H [ 40 m; 6,
H [ 50 m; d predicted; $, predicted by porosity reduction
for H [ 50 m; - - -, predicted by Eq. (25).
However, the greater porosity did not guarantee smaller resistance to airflow in the grain mass. For example, the rice in
the husk (or paddy) had a resistance greater than the maize although the rice porosity was greater. It is possible that free
volumes of air between husk and the grain increased porosity
but did not increase cross-sectional area for airflow.
It must be concluded that, besides the global non-uniformity defined by the alteration of the mean porosity factor
with the depth variation, there is local non-uniformity caused
by the seed form and that this does not significantly alter the
porosity factor value of the medium. Probably, the compaction
of non-spherical seeds creates local dense regions through
which airflow is hindered.
The experimental data presented in Fig. 6 show that for the
studied velocity and depth variation intervals, the relative
pressure gradient increment C (jgrad PHj jgrad P0j)/jgrad P0j
can be considered as being independent of air velocity and depends only on the storage layer depth H, where H is a distance
between the upper seed surface (free surface) and the layer
under consideration. This hypothesis was confirmed by
multi-factorial analysis of variance and by a nonparametric
association test of the Spearman rank order correlation
(Table 3).
The function C C(H ) relates to the initial pressure gradient jgrad P0j, where P0 corresponds to grain depth H 1 m,
and the pressure gradient jgrad PHj for considered depth H at
Table 1 The empirical coefficients a, b and c with 95% confidence bounds for different seeds, sum squared error (SSE),
coefficient of determination (R2) and root mean squared error (standard error) for Eq. (13)
Soya bean
Maize
Rice
Wheat
SSE
R2
RSME
0.82 0.12
0.61 0.07
0.51 0.13
0.86 0.15
3.57 0.66
2.92 0.39
3.08 0.82
5.49 0.98
2.77 0.12
2.75 0.08
2.23 0.13
2.18 0.06
0.5013
0.1304
0.3526
0.1296
0.9954
0.9982
0.9934
0.9972
0.0480
0.0289
0.0525
0.0348
231
0.5
1.00
C=(gradP - gradP0)/gradP0
0.99
0.98
0.97
0.96
0.95
0.94
0.4
0.3
0.2
0.1
0.93
0.92
0.0
0.1
0
10
20
30
40
Bed Depth, m
Fig. 5 Porosity reduction with bed depth; -, soya bean;
:, maize; d, predicted; - -, 95% confidence bounds for
prediction.
(24)
As a result, Eqs. (13) and (22) and with the intermediate argument in Eq. (24) relate the air velocity for the storage layer located in the depth H, and the necessary pressure gradient.
Fig. 4 shows close agreement between observed and predicted
data.
To take account of the grain bulk density (and in implicit
form the bed depth) the ASAE Standards 2000 recommends
using the equation obtained by Bern and Charity (1975):
0.3
0.4
0.5
2
r
rb
b
V
V2
DP
rk
rk
X
;
X1 X2
3
r 3
r 3
L
1 b
1 b
rk
rk
(25)
(26)
Navarro and Noyes (2001) recommended calculating the average value of the grain bulk density during filling a silo by
Probability > F
R2
Soya bean
Depth, m
Velocity, m s1
76.6
0.54
0.002
0.855
0.958
0.035
Variable
0.2
Velocity, m s-1
50
Seed type
30
Coefficient of
determination (R2)
Shelled corn
Depth, m
Velocity, m s1
221.6
5.05
0.001
0.103
0.954
0.031
Soya bean
Maize
Rice
0.43
0.44
0.61
0.0680
0.0736
0.0664
0.5261
0.4683
0.5134
0.9985
0.9997
0.9989
Rice
Depth, m
Velocity, m s1
147.2
4.03
0.001
0.138
0.958
0.045
232
Standard error
Coefficient of
(RSME)
determination (R2)
0.9943
0.9858
0.9872
0.0111
0.0176
0.0155
0.6
0.5
the initial pressure; and (3) the pressure and airflow rate in
an iterative process for the chosen fan and electric motor (by
estimating system design point).
5.
Numerical simulations
0.4
0.3
0.2
0.1
0.0
10
20
30
40
50
Layer Depth, m
Fig. 7 Variation of the compaction function
C [ (jgradPHj L jgradP0j)/jgradP0j with bed depth (H ) in m
for soya bean and shelled maize: C, soya bean, observed
data; 6, shelled maize, observed data; , 95% confidence
bounds for prediction; - -, non-simultaneous bounds for
observation; , predicted.
233
Fig. 11 Comparison of three simulated aeration systems: distribution of pressure (left column) and risk regions
(right column).
234
Fig. 13 Visualisation of regions with inadequate ventilation (qL < 4.5) for three air inlet systems: (1) central inlet system; (2)
central and upper lateral inlet systems; and (3) central, lower lateral and upper lateral inlet systems; Q [ 9 m3 tL1 hL1.
Qa
VA
V
mg rb HA rb H
5.1.
Criterion for describing the efficiency of aeration
system
qL X
To estimate the efficiency of an aeration system the specific
airflow rate is usually specified. This is obtained from the ratio
between the total airflow rate and the total product mass. This
criterion is suitable for simple silo designs with constant
cross-sectional area, when the air velocity is uniform throughout the storage. If variations in the cross-sectional area are
significant, or the aeration distribution system is complex
(e.g. case 3), this criterion is not suitable.
To evaluate aeration efficiency for storage bins with variable cross-sectional area and with complex air distribution
system, a local specific airflow rate is proposed. For simple
storage layouts (Fig. 12(a)) with constant cross-sectional area
the specific volume airflow rate Q is
(27)
VX
;
rXLX
(28)
Fig. 14 Visualisation of regions with the excessive intensity of ventilation (qL > 18) for three air inlet systems: (1) central
inlet system; (2) central and upper lateral inlet systems; and (3) central, lower lateral and upper lateral inlet systems;
Q [ 9 m3 tL1 hL1.
235
Fig. 15 Distribution of local specific airflow rate, Q [ 9 m3 hL1 tL1 (2.5 3 10L6 m3 sL1 kgL1): (a) central inlet and (b) lower
lateral inlet.
5.2.
To visualise risk domains in the grain storage bin, the distribution of local specific airflow rates was studied (Figs. 1319).
Fig. 13 shows the visualisation of domains with inadequate
ventilation (qL < 4.5) for three air inlet systems: (1) a central
inlet; (2) central and upper lateral inlets; and (3) central, lower
lateral and upper lateral inlets. As simulations show, the system with central, lower lateral and upper lateral inlets considerably improved the conditions of storage in regions close to
walls when compared with other inlet systems. For all cases
considered there was an area of risk in the uppermost part
of grain mass.
The regions with the raised intensity of ventilation (qL > 18)
are shown in Fig. 14. The results obtained show that the second system (central and upper lateral inlets) has a smaller volume with excessive intensity of ventilation in comparison
with others, i.e. has improved efficiency.
Fig. 16 Distribution of local specific airflow rate, Q [ 9 m3 hL1 tL1 (2.5 3 10L6 m3 sL1 kgL1): (a) upper lateral inlet and (b)
central, lower lateral and upper lateral inlets with identical initial pressures.
236
Fig. 17 Distribution of local specific airflow rate, Q [ 9 m3 hL1 tL1 (2.5 3 10L6 m3 sL1 kgL1): (a) central and lower lateral
inlets with different initial pressures and (b) central, lower lateral and upper lateral inlets with different initial pressures.
Additional more detailed comparative analyses of the efficiency of different aeration systems were made for the same
grain storage bin with the same global specific airflow rate
Q 9 m3 h1 t1 (2.5 106 m3 s1 kg1). The number of inputs
(from one up to three), their position (upper lateral, lower lateral and central inlets), and ratio of pressure between various
inputs were varied. Using the additive property of the local
specific airflow rate, estimations of ventilation system efficiency
were carried out separately using each of inlets during the
different periods of time. Relationships between the duration
Fig. 18 Distribution of resultant local specific airflow rate with separated functioning inlets, Q [ 9 m3 hL1 tL1
(2.5 3 10L6 m3 sL1 kgL1): (a) upper lateral, lower lateral and central inlets with equal application times (1:1:1) and (b) upper
lateral, lower lateral and central inlets with different application times (1:2:2).
237
Fig. 19 Visualisation of domains with the lowered intensity of ventilation (qL < 4.5; right) and with excessive intensity of
ventilation (qL > 18; left); qL was obtained by superposition of simulations for separated functioning of upper lateral, lower
lateral and central inlets with different application times (1:2:2); Q [ 9 m3 hL1 tL1 (2.5 3 10L6 m3 sL1 kgL1).
(29)
where qi is the local specific airflow rate corresponding to ventilation with only one inlet (order number i); ti is ventilation
time with only one inlet (i); n is total number of inlets; i is order
number of corresponding inlet.
For example, Fig. 18 shows the distribution of resultant local specific airflow rates with separated operation of the upper
lateral, lower lateral or central inlets. In case (a) the application time is the same for each inlet, and the resultant local
specific airflow rate at each point of the storage bin can be calculate by
1
1
1
qL q1 q2 q3 ;
3
3
3
(30)
where q1, q2 and q3 are local specific airflow rates corresponding to upper lateral, lower lateral or central inlets.
The simulations presented in Figs. 16(b) and 18(a) indicate
the significant advantage of ventilation carried out in turn by
each of inlets in comparison with the simultaneous use of all
inlets at equal pressures. This improvement is caused because
the capacity for air to penetrate to all zones under the dominant influence of each inlet results in a more uniform distribution of qL. By varying the duration of aeration for each inlet, it
6.
Conclusions
A mathematical model of three-dimensional airflow in an aerated grain storage system was developed for non-uniform
conditions of the seed mass. Experiments were conducted to
obtain the relationship between air velocity and pressure gradient and the values of the porosity factors for different seed
types and different storage layer depths. A local criterion
was proposed to estimate the efficiency of complex aeration
system in grain storage bins.
Software was developed to determine the velocity, pressure and local specific airflow rates distributions, the global
airflow rate or initial pressure head in the grain mass store
for three-dimensional cases. The aeration system efficiency
of several stored seeds was analysed to provide the airflow
distribution uniformity and the static pressure head values
that generate the appropriate airflow rate for safe storage.
It was shown that the aeration system of grain storage bin
can be essentially improved by the use of inlets system with
different initial pressures selected for each inlet. Also, it was
shown that it is possible to optimise air distribution in a grain
storage bin by operating each inlet in turn and by selecting
a suitable aeration period for each inlet.
Acknowledgements
The authors would like to thank CNPq for the financial support for this work (process No. 464380/00-6).
238
references