Vous êtes sur la page 1sur 14

Available online at www.sciencedirect.

com
ScienceDirect

Mathematics and Computers in Simulation 157 (2019) 1–14


www.elsevier.com/locate/matcom

Original articles

Simulation of the temperature of barley during its storage in


cylindrical silos
Francisco Novoa-Muñoz
Departamento de Estadística, Universidad del Bío-Bío, Chile

Received 3 November 2016; received in revised form 21 June 2018; accepted 3 September 2018
Available online 20 September 2018

Highlights

• A mathematical model was developed to simulate the temperature of stored barley.


• Predicted malting barley temperature was in close agreement with the measured ones.
• The stability was certified, regardless the coefficient of convective heat transfer.
• The model predicted barley temperature in a 6.3 [m] diameter and 23.5 [m] height silo.
• The model created allows to study the behavior of temperature in any stored grain.

Abstract

A two-dimensional mathematical model, based on the finite-difference method of calculating heat transfer, has been developed
to simulate the temperature in a cylindrical storage silo containing malting barley. To validate the model, a real silo approximately
23.5 [m] high and 6.3 [m] in diameter was used, unlike the other investigations related to the subject, which have used tower
silos with at most 3.76 [m] in diameter and 5.5 [m] in height. These two scenarios are quite different, since it is known that
climatological variables are very different as height increases. The model has been programmed in Visual Basic for applications
in Excel, compatible with microcomputers, which enables a good display for computer simulation of temperature and an excellent
report of results, through linked graphs. The study allows to conclude that the model predicted the temperature with a standard
error of approximately between 6.6 and 12.4%, depending on the depth. In addition, the simulation of temperatures can be done
without the need to consider the weather station, because statistically, the seasons of the year turned out to be not significant for
the temperature of the stored barley.
⃝c 2018 International Association for Mathematics and Computers in Simulation (IMACS). Published by Elsevier B.V. All rights
reserved.

Keywords: Malting barley; Mathematical model; Simulation; Heat transfer; Cylindrical silo

E-mail address: fnovoa@ubiobio.cl.

https://doi.org/10.1016/j.matcom.2018.09.004
0378-4754/⃝ c 2018 International Association for Mathematics and Computers in Simulation (IMACS). Published by Elsevier B.V. All rights
reserved.
2 F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14

1. Introduction

The storage of grain is an ecological system made by man, in which the damage of it results from the interaction
among multiple variables as for example physical, chemical and biological ones. The abiotical environment of the
grain has physical (temperature), chemical (carbon dioxide and oxygen) and physio-chemical (moisture) variables.
Apart from the grain, the main biotical variables in the storage are microorganisms (fungi, actinomycetal and
bacterias), arthropods (insects and mites) and vertebrates (rodents and birds) [1,10]. The reproduction and growth
rate of these organisms are primarily determined by the temperature and the moisture content [2,1,10,17].
Of the different types of silos that are used to store grain, those that appear with greater frequency in the current
investigations are the silo-bags and the tower silos.
Silo-bags require little capital investment and can be used as a temporary measure when harvest or growth condi-
tions demand more space. Research shows that its dimensions reach 1.7 m high and 67 m long, capable of storing up to
200 tons of grain. Among the objectives that motivate to study the grain stored in bags are: determine the change in the
concentration of CO2 [3], quantify the effect of the moisture content, determine the correlation between the concen-
tration of CO2 and the moisture content [4], or identify the main causes of changes in the quality characteristics [8,12]
or the evolution of the most important quality parameters [23] and modeling of temperature and moisture [9].
On the other hand, the main reason to investigate the grain in tower silos is to model its temperature – the only
one that can be regulated causing a minimum damage to its quality – is the most important factor that regulates the
amount of insects. Furthermore, it is the only non-biological variable that can be easily measured and simulated
mathematically. Physical factors affecting the temperature of stored grain can be studied by measuring different
parameters in storage, but using a simulation model is a faster and cheaper method, which allows to study various
combinations of factors.
The regular interaction of the grain with its environment makes it exchange heat and moisture. Because of this, the
temperature and the moisture content vary in time and space. On one hand, in the warmer areas, the moisture content
must be “low” to achieve a safe storage. On the other hand, in the colder areas, the moisture can be “high” [2,1].
Given that all the living organisms stay alive and they just bloom in some limit of temperatures; the atmospheric,
grain and intergranular air temperatures are considered crucial variables to guarantee and prolong the storage of the
grain [1]. Therefore, it is essential to know the temperature distribution. One way to know it is by collecting the
temperature measurements at different points of the grain deposit, over a period of time, but it is an imprecise method,
it requires a lot of time, cost and work [1], especially if the container used is large. According to what was declared
by Abbouda et al. in [1], another way is to design a mathematical model, which based on physical principles can
accurately predict the distribution of grain temperature, in addition, it allows to study the effect of the characteristics
of the deposit (size, type of material and geographical location) on the temperature of the stored grain.
Different researchers have developed mathematical models of heat transfer to predict the temperature of the stored
grain. They have arrived to one-dimensional configurations [9,10,20,22,21], two-dimensional [1,3,4,7,14,13,16,17]
and three-dimensional [18]. These models have been simulated by using analytical methods [7] of finite differ-
ences [2,1,14,13,19] and finite elements [4,15–18,22,21]. Enabled to simulate ecosystems with and without aeration.
In these investigations, to validate the designed models we have used tower silos with the following characteristics:

(a) of different dimensions, ranging from 0.63 [m] in diameter and 1 [m] in height, up to 5.5 [m] in diameter and
8.3 [m] in height, but with grain only up to 6.5 [m],
(b) made of different materials: concrete [18,24], metal [1,10,14,13,19], corrugated steel [2,22,21], and galvanized
steel [15,17],
(c) filled with various grains, like corn [1,22,21], wheat [2,10,16–18,24], rice [14,13,15] and sorghum [7,19],
(d) with capacities that do not exceed 44 [m3 ].

But so far there are no investigations in tower silos that have stored barley beer or exceeded 6.5 [m] high with
grain.
The objective of this research was to design and validate a two-dimensional mathematical model of heat transfer,
based on the finite difference method, to predict storage temperatures of barley in cylindrical tower silos of
approximately 23.5 [m] height and 6.3 [m] in diameter, filled with about 383 tons of malting barley.
F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14 3

Fig. 1. (a) Planta C.C.U., (b) Physical characteristics of silos.

2. Materials and methods


2.1. Material and equipment

For the development of this research, a cylindrical tower silo of Planta de Silos Chillán de la Compañı́a Cervecerı́as
Unidas (C.C.U.) (see Fig. 1) was used. Moreover, we used measurement of the temperature of the barley (they are
obtained from the temperature measurement, Rolfes, MPU-85 model, in which every wire is interconnected by a
multiplexor unit and this one to a PC), air temperature, speed of wind and ventilation. Environmental measurements
were obtained from meteorological data given by Departamento de Riego y Drenaje de la Facultad de Ingenierı́a
Agrı́cola de la Universidad de Concepción, Sede Chillán. This information was gathered from January, 12th to July,
29th, 1998. Part of the data used is presented in Appendix A (see Tables 4–6).

2.2. Methodology: theory and development of the model

Simulation of the temperature


The main objective of the thermal conduction analysis is to define the profile of the temperature, which represents
the variation of temperature in any position of the studied object. Once this is known, the Law of Fourier is used to
calculate the conduction of heat flow [5]. The solution of the Partial Differential Equation given by the Conservation
of Energy Law to the stored barley is the distribution of temperature T (r, y, φ), shown in cylindrical coordinates. The
general form of the heat equation, which gives the distribution of temperature T (r, y, φ) depending on time [5], is
given by
∂T ∂T
1 ∂(k r ∂∂rT ) 1 ∂(k ∂φ ) ∂(k ∂ y ) ∂T
+ 2 + + q̇ = ρ c p .
r ∂r r ∂φ ∂y ∂t
To determine the equation of the heat flow through the barley, we will be assume that:

• The properties of the barley are constant and uniform through the silo.
• The heat conduction in the circumference direction remains constant.
4 F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14

• The heat flow is symmetric around the vertical central axis of the silo.
• The production of internal heat (respiration) inside of the silo is not considered.
Following the previous assumptions, the differential general heat flow equation of two-dimensions in a system of
cylindrical coordinates is as follows:
∂T ∂ T 1 ∂T ∂2T
( 2 )
kg
= + + , (1)
∂t ρ c p ∂r 2 r ∂r ∂ y2
where k g , ρ, c p and T are, respectively, the thermal conductivity, density, specific heat and temperature of the barley;
r and y are correspondingly the radial and vertical coordinates of the silo, and t is the time.
The initial and border conditions for Eq. (1) given in [17,18] are as follows:
T (r, y, t) = Ti (r, y), for t = 0,

∂T ∂T
( )
kg + + h c (T − Ta ) = 0, for t > 0.
∂r ∂y
where Ti is the initial temperature of the grain (◦ C), h c is the coefficient of transference of convective heat
(W m−2 ◦ C−1 ), and Ta is the air temperature (◦ C).
Since the initial and border conditions depend on the variation of the air temperature; it is hard to develop an
analytic solution for Eq. (1). As an alternative, a numerical method can be applied.
Finite difference method
On Eq. (1), the temperature T is a function of the following variables: radial length r from the center of the silo,
vertical distance y from the floor of the silo and time t of the storage.
On the r y−plane, any coordinate point (r, y) is represented by r = m ∆r , y = n ∆y (∆r = ∂r , ∆y = ∂ y).
p
Besides, t = p ∆t (∆t = ∂t) indicates a dependency on T related to time. In this way, Tm,n shows the node
temperature (m, n) in any p time; where m, n and p are integers.
If these discretizations of the equation are replaced in Eq. (1) and assuming that ∆y = ∆r , the explicit form for
finite difference to the interior node (m, n) of the silo is obtained and it can be expressed as follows:
{( ) }
p+1 1 p p p p
Tm,n = F0 1+ Tm+1,n + Tm−1,n + Tm,n+1 + Tm,n−1
m
{ ( ) }
1
+ 1− 4+ F0 Tm,np
, (2)
m
α ∆t
where F0 = (∆r )2
is the form of finite difference of the Fourier number.
Eq. (1) is not validated in any central point of the silo, since the quantity r1 ∂∂rT becomes undetermined. Applying L′
Hôpital’s rule, the finite difference form to the center of the deposit, it is obtained; where r = 0 (m = 0), in the time
( p + 1) is as follows:
p+1 ( p p p p p
T0,n = F0 2 T1,n + 2 T−1,n + T0,n+1 + T0,n−1 + (1 − 6 F0 ) T0,n ,
)
(3)
The energy balance method is applied for the surface points exposed to convective conditions [5]. Thus, the finite
difference form obtained by Eq. (1) is given by
( p )
p+1 p p 1 p Bi Ta kw
Tm,n = 2F0 Tm−1,n + Tm,n−1 + Tm,n+1 +
2 k + hc L w
( w )
2F0 Bi kw
+ 1 − 4F0 − p
Tm,n , (4)
kw + h c L w
h ∆r
where Bi = ckg is the finite difference form of the Biot number.
p
The terms kw and L w are the heat conductivity and thickness of the wall of the silo, respectively. Ta is the
environmental air temperature in p time.
Because of the initial conditions, the temperature of every node is known at t = 0 ( p = 0), calculation starts in
t = ∆t( p = 1), where Eqs. (2), (3) and (4) are, respectively, applied in every interior, central and superficial node of
the silo to determine its temperature.
F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14 5

Table 1
Physical and thermal properties of concrete and barley.
Property Unit of measurement Concrete Barley
Thickness m 0.15 –
Density kg m−3 1900 670
Specific heat J kg−1 K−1 880 1560
Thermal conductivity W m−1 K−1 1.37 0.15

Solution of the numerical method


The flux of heat of the stored grain depends on the time variable. Therefore, at the beginning of the simulation, the
temperature of barley in different places of the silo were given by measured values.
The temperatures of barley in different places of the silo and in different time intervals are given by simultaneous
solutions of Eqs. (2), (3) and (4). Air temperature and wind velocity are required in every step of the process. This
iterative process is repeated during the whole storage period.
p
It is essential that the Tm,n coefficient is non negative, in order to ensure the stability criteria in all the nodes [5],
since Bi ≥ 0. So that from Eq. (4), the stability criteria for all the nodes is given by
( )
Bi kw 1
F0 2 + ≤ , (5)
kw + h c L w 2
The inequation (5) allows us to choose the maximum value of F0 so that the ∆t and ∆r values can be defined and
applied in the simulation process.
Validation of the two-dimensional model
Of the 23 storage silos located as in Fig. 1, the silos from number 1 to number 8 are directly exposed to the north,
the other silos (9 to 23) are protected. The silos number 1 and number 8 have different shapes, as shown in Fig. 1(b).
The silos from number 3 to number 6 that were in direct contact with the environmental conditions were the most
appropriate to validate the model. A single silo was chosen because of the amount of time involved in processing all
the information and the limited budget that was available. Silo number 5 was used to validate the model because the
registration of its data was more complete. This silo was loaded with barley for two days with approximately 383
tons of barley at different temperatures at different levels. The other results are presented in Appendix B. The level is
associated to the height where the thermocouples are located, equidistant to the central vertical axis of the silo.
The floor, roof and wall of the silo are built in concrete. The roof is a horizontal platform with a gap on the center
for the ventilation of the barley.
Computing implementation
To implement the model, which will simulate the temperature of the barley, it will be necessary to consider an
amount of nodes that meet the criteria of inequation (5). Here, F0 < 14 . Replacing F0 for its original components
∆t
and considering the data of barley (see Table 1), it is obtained that (∆r )2
< 483.9. Given the values to ∆t, the ∆r
increase is achieved. If we examine, it is found that ∆r = 0.5 [m] satisfies ∆t ≤ 120 [h]. Therefore, we can choose
∆r = ∆y = 0.5 [m] and ∆t = 1 [h].
The stored barley reached a height of 17.5 [m] and since the vertical distance is y = n ∆y, thus n = 35. In addition,
the radial length is 0 ≤ r = m ∆r then m = 6 because the radio of the silo is 3 [m]. Thus, the number of nodes to be
considered is 35 × 7 = 245 nodes, because of m = 0, 1, . . . , 6.
The location, form, size, and symmetry of the nodes in the silo are in Fig. 2.
Depending on the location, the nodes get a characteristic form that can be classified as follows:

• Central nodes: the ones located in the central axis of the silo.
• Interior nodes: located in the inner part, superior or floor of the silo.
• Surface nodes: the ones attached to the wall of the silo.
6 F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14

Fig. 2. View of the nodes in silo: complete and sectional.

p p
In Eq. (3) : T−1,n = T1,n (by symmetry). Eq. (4) will be re-sequenced and it will be applied to the nodes attached
to the wall, when m = 6. Furthermore grain ventilation will be included (a = 0: without air, a = 1: with air). Thus,
the equations to apply numerically are as follows:
p+1 ( p p p p p Central nodes
T0,n = F0 4T1,n + T0,n+1 + T0,n−1 + (1 − 6F0 ) T0,n + a T0,n−1 ,
)
n = 0, 1, . . . , 34.
( )
p+1 m+1 p p p p
Tm,n = F0 Tm+1,n + Tm−1,n + Tm,n+1 + Tm,n−1
m
( ) Interior nodes
4m + 1 p
+ 1− F0 Tm,n p
+ a Tm,n−1 , m = 1, 2, . . . , 5
m n = 0, 1, . . . , 34.
p p
( )
p+1 p p p 1 p Bi (Ta − T6,n )
T6,n = 2F0 T5,n + T6,n−1 − 2T6,n + T6,n+1 +
2 1 + h ck L w w

p p Surface nodes
+ T6,n + a T6,n−1 ,
n = 0, 1, . . . , 34.
p
Observe that, Tm,−1 = 0, for m = 0, 1, . . . , 6.
But, the value of h c (convective heat transfer coefficient) relies on several factors such as air velocity, property of
air, air and the wall of the silo temperatures and dimensions of the silo. The value of h c is determined by

• Natural convection: without wind.


• Forced convection: with wind.

For natural convection, Holman [11] established


hc L Ta + Tw
Nu = = E (Gr f Pr f )k , Tf = , (6)
kf 2
where N u is the Nusselt number, L is the height of the silo, k f is the thermal conductivity of the air at a film
temperature (W m−1 K−1 ), T f is the film temperature (◦ C), Ta is the temperature of the air (◦ C), Tw is the wall
temperature of the silo (◦ C), Gr f is the Grashof number at film temperature, Pr f is the Prandtl number at film
temperature, and E and k are constants, which depend on the numbers of Grashof and Prandtl.
F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14 7

g β ∆T L 3 νf
The numbers of Grashof and Prandtl are Gr f = ν 2f
and Pr f = αf
, where g is the gravity acceleration (m
s ), β is the volumetric thermal expansion coefficient ( K), ∆T is the difference between the temperature of the air
−2 ◦

and the wall of the silo (◦ C), ν f is the air kinematic viscosity at film temperature (m2 s−1 ), and α f is the thermal
diffusivity air at film temperature (m2 s−1 ).
To find E and k it will be necessary to obtain the variation of the product
g ∆T L 3
Gr f Pr f = ,
Ta ν f α f
where Ta is the temperature of the air in absolute degrees.
During the storage, the variation of the temperature of the air was
1 1 1
− 1.693 [◦ C] ≤ Ta ≤ 34.710 [◦ C] ⇒ ◦
≤ ≤ .
307.860 [ K] Ta 271.457 [◦ K]
The values for ν f and α f were obtained from the tabulations in Holman [11]. By interpolation, it is achieved
106 1 106
12.62213 × 10−6 ≤ ν f ≤ 17.44446 × 10−6 ⇒ ≤ ≤ ,
17.44446 νf 12.62213

104 1 104
0.16996 × 10−4 ≤ α f ≤ 0.23343 × 10−4 ⇒ ≤ ≤ .
0.23343 αf 0.16996
Besides, the rank of variation of the wall temperature of the silo and air temperature verifies 1 ≤ ∆T ≤ 10.
The constant acceleration of gravity and the resulted height of the accumulation of barley in the silo are g = 9.8
[m s−2 ] and L = 17.5 [m], respectively. So the product of Grashof and Prandtls verifies 109 ≤ Gr f Pr f ≤ 1013 .
Hence (see Holman [11]) E = 0.1 and k = 13 and Eq. (6) is reduced to
( )1
g β ∆T 3
h c = 0.1 k f Pr f .
ν 2f
All the terms of the previous equation depend on the film temperature except g. The values of the film temperature
are tabulated for certain temperatures. Using the second degree Lagrange polynomial interpolation, the following is
obtained:
1
P2 (x) = ((x − 27)(x − 77)y1 − 2(x + 23)(x − 77)y2 + (x + 23)(x − 27)y3 ) ,
5000
where x is the film temperature (◦ C), y1 , y2 and y3 are the tabulated values, whether is k f , ν f or Pr f to
−23, 27, 77 [◦ C], respectively, and P2 (x) is the interpolation, whether is k f , ν f or Pr f to temperature x.
For forced convection, Welty et al. [25] establish that
hc D v∞ D
Nu = = 0.227 Re0.633 and Re = ,
kf νf
where Re is the Reynolds number, D is the diameter of the silo (m) and v∞ is the speed of the wind (m s−1 ).
Solving the necessary algebraic clearing, it is obtained that h c = D0.367f ( vν∞f )0.633 .
0.227k

3. Results and discussion


In order to decide whether the studied model predicted the temperature of barley in the silo in an accurate way,
variations of measured and simulated temperatures were analyzed. Table 2 shows the average temperatures of the
first days of the simulation process, with their respective standard deviations. Fig. 3 shows the daily averages of the
temperatures of the thermocouples and their simulated temperatures together with the air temperatures of the entire
simulated period. The plots of each of the levels are presented in Appendix B.
To validate the simulation model, the temperatures of barley given were compared with the ones obtained by
the thermocouples through a dispersion plot, which presented a linear tendency. Thus, a calibration technique
8 F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14

Table 2
Average daily temperatures (with standard deviation) of thermocouples (Therm) and simulated (Sim) in 1998.
Day Therm Sim Day Therm Sim Day Therm Sim
01/14 29.7 ± 2.9 29.8 ± 3.0 01/15 29.8 ± 2.9 29.8 ± 2.9 01/16 30.2 ± 3.2 29.7 ± 2.9
01/17 30.1 ± 3.4 29.7 ± 2.9 01/18 30.0 ± 3.4 29.7 ± 2.8 01/19 30.1 ± 3.4 29.6 ± 2.8
01/20 30.0 ± 3.1 29.3 ± 3.1 01/21 30.1 ± 3.1 28.3 ± 5.0 01/22 30.1 ± 3.1 28.5 ± 4.4
01/23 28.7 ± 4.0 28.2 ± 5.1 01/24 26.4 ± 4.4 27.5 ± 5.5 01/25 26.3 ± 4.4 27.5 ± 5.4
01/26 26.4 ± 4.6 26.9 ± 5.1 01/27 23.6 ± 4.1 25.7 ± 5.6 01/28 22.0 ± 3.1 24.0 ± 6.1
01/29 21.0 ± 2.4 23.0 ± 5.7 01/30 20.2 ± 2.0 22.0 ± 5.2 01/31 20.3 ± 1.9 22.0 ± 5.1
02/01 20.3 ± 1.9 22.0 ± 5.1 02/02 20.5 ± 1.8 22.0 ± 5.1 02/03 20.0 ± 1.8 22.0 ± 5.1
02/04 19.8 ± 1.7 22.0 ± 5.1 02/05 19.5 ± 1.5 21.9 ± 5.1 02/06 19.2 ± 1.4 21.9 ± 5.0
02/07 19.2 ± 1.4 21.9 ± 5.0 02/08 19.2 ± 1.4 21.9 ± 5–0 02/09 19.4 ± 1.3 21.9 ± 4.9
02/10 19.5 ± 1.5 21.9 ± 4.9 02/11 19.5 ± 1.5 21.9 ± 4.9 02/12 19.4 ± 1.3 21.9 ± 4.8
02/13 19.3 ± 1.2 21.9 ± 4.8 02/14 19.6 ± 1.3 21.9 ± 4.7 02/15 19.6 ± 1.3 21.9 ± 4.7
02/16 19.6 ± 1.3 21.9 ± 4.7 02/17 20.0 ± 1.6 21.8 ± 4.7 02/18 20.1 ± 1.4 21.8 ± 4.7
02/19 20.0 ± 1.6 21.8 ± 4.6 02/20 20.3 ± 1.3 21.8 ± 4.6 02/21 20.3 ± 1.2 21.8 ± 4.6
02/22 20.3 ± 1.2 21.8 ± 4.5 02/23 20.3 ± 1.2 21.8 ± 4.5 02/24 20.6 ± 0.8 21.7 ± 4.5
02/25 20.5 ± 0.8 21.7 ± 4.4 02/26 20.5 ± 0.8 21.7 ± 4.4 02/27 20.9 ± 0.6 21.7 ± 4.4
02/28 20.9 ± 0.6 21.7 ± 4.3 03/01 21.0 ± 0.5 21.7 ± 4.3 03/02 21.0 ± 0.5 21.7 ± 4.3
03/02 20.9 ± 0.3 21.7 ± 4.2 03/03 21.4 ± 0.5 21.7 ± 4.2 03/04 21.4 ± 0.5 21.7 ± 4.2
03/05 21.4 ± 0.5 21.6 ± 4.2 03/06 20.9 ± 1.1 21.6 ± 4.2 03/07 20.9 ± 1.1 21.6 ± 4.1
03/08 20.9 ± 1.1 21.6 ± 4.1 03/09 20.6 ± 1.2 21.5 ± 4.1 03/10 20.5 ± 1.3 21.5 ± 4.0
03/11 20.5 ± 1.3 21.5 ± 4.0 03/12 20.2 ± 1.5 21.5 ± 4.0 03/13 20.0 ± 1.6 21.5 ± 3.9
03/14 19.9 ± 1.7 21.5 ± 3.9 03/15 19.9 ± 1.7 21.5 ± 3.9 03/16 19.6 ± 1.5 21.4 ± 3.8
03/17 19.2 ± 1.7 21.4 ± 3.8 03/18 19.2 ± 1.5 21.4 ± 3.8 03/19 19.4 ± 1.3 21.4 ± 3.8
03/20 19.4 ± 1.3 21.4 ± 3.7 03/21 19.5 ± 1.3 21.3 ± 3.7 03/22 19.2 ± 1.1 21.3 ± 3.7
03/23 19.2 ± 1.1 21.3 ± 3.6 03/24 19.1 ± 1.3 21.3 ± 3.6 03/25 19.1 ± 1.3 21.3 ± 3.6
03/26 19.0 ± 1.2 21.2 ± 3.6 03/27 19.1 ± 1.0 21.2 ± 3.5 03/28 19.1 ± 0.7 21.2 ± 3.5
03/29 19.1 ± 0.7 21.2 ± 3.5 03/30 19.1 ± 0.7 21.2 ± 3.4 04/01 19.1 ± 0.7 21.2 ± 3.4
04/02 19.1 ± 0.7 21.2 ± 3.4 04/03 19.1 ± 0.7 21.1 ± 3.4 04/04 19.1 ± 0.7 21.1 ± 3.3
04/05 19.1 ± 0.7 21.1 ± 3.3 04/06 19.1 ± 0.7 21.1 ± 3.3 04/07 19.1 ± 0.7 21.0 ± 3.3
04/08 19.1 ± 0.7 21.0 ± 3.3 04/09 19.1 ± 0.7 21.0 ± 3.2 04/10 19.1 ± 0.7 21.0 ± 3.2
04/11 19.2 ± 0.4 20.9 ± 3.2 04/12 19.0 ± 0.7 20.9 ± 3.2 04/13 18.9 ± 0.7 20.9 ± 3.1
04/14 19.2 ± 0.9 20.9 ± 3.1 04/15 19.2 ± 0.9 20.8 ± 3.1 04/16 19.1 ± 1.0 20.8 ± 3.1
04/17 19.1 ± 1.0 20.8 ± 3.1 04/18 19.1 ± 1.0 20.8 ± 3.0 04/19 19.2 ± 1.1 20.7 ± 3.0
04/20 19.1 ± 1.4 20.7 ± 3.0 04/21 18.9 ± 1.4 20.7 ± 3.0 04/22 18.9 ± 1.4 20.6 ± 3.0
04/23 18.9 ± 1.4 20.6 ± 3.0 04/24 18.4 ± 2.0 20.5 ± 3.0 04/25 18.0 ± 1.8 20.5 ± 3.0
04/26 18.0 ± 1.8 20.5 ± 3.0 04/27 17.5 ± 1.6 20.5 ± 2.9 04/28 17.3 ± 1.5 20.4 ± 2.9
04/29 17.1 ± 1.6 20.4 ± 2.9 04/30 16.8 ± 1.6 20.4 ± 2.9 05/01 16.4 ± 1.6 20.3 ± 2.9

Fig. 3. Average daily temperatures of air, thermocouples and simulated.


F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14 9

Table 3
Coefficients and statistical estimators of regressions.
Height b c Correlation t-value SEE(%)
Level 1 2.99 0.76 0.94 39.6 6.6
Level 2 9.38 0.47 0.76 16.5 10.5
Level 3 8.57 0.55 0.74 15.3 10.2
Level 4 7.57 0.64 0.82 20.3 9.3
Level 5 8.19 0.64 0.86 23.1 9.1
Level 6 9.10 0.67 0.91 30.5 11.8
Level 7 10.62 0.65 0.94 39.3 12.3
Level 8 10.73 0.73 0.91 31.5 12.5
Level 9 12.08 0.68 0.80 18.5 12.8
Level 10 −1.65 1.19 0.84 21.9 12.1

Fig. 4. Scatter plot between real and simulated temperatures.

was applied [6] through a linear simple regression between those temperatures, where the results given by the
thermocouples are placed in the X axis (independent variable) and the results given by the simulation are located
in the Y axis (dependent variable).
So, the lineal model to adjust was E(Y/ X ) = B + C X , which was obtained through a linear regression given by
Ŷ = b + cX , where X is the temperature of barley given by the thermocouple, and Y is the temperature of barley
given by the simulation.
The results are shown in Table 3, where the parameters b and c represent the coefficient of position and the slope
of the regression line, respectively.
To decide whether the correlation coefficient r was indeed significant, a contrast of hypothesis was used:

H0 : ρ = 0, (there is not a correlation between X and Y ),

against the alternative

H1 : ρ ̸= 0 (there is a correlation between X and Y ).



The statistic of the test applied was tm = |r√| n−2
, which has a t-student distribution, where r is the correlation
1−r 2
coefficient sample, and n is the number of pair data (n = 197).
The p-value for the previous contrast was near 0, so we must conclude that the correlation was highly significant.
Therefore the relation between the variables is linear to each one of the depths.
The Standard Error of the Estimate (SEE) in relation to the mean is very low. Therefore we can conclude that the
estimated temperatures of barley by the simulation model are closer to the measured values by the thermocouples
along the vertical central axis of the silo. Fig. 4 shows the dispersion between the average temperatures of the
10 F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14

Table 4
Air temperature (◦ C).
Hours Date (year 1998)
01/14 01/15 01/16 01/17 01/18 01/19 01/20 01/21 01/22 01/23 01/24 01/25 01/26
0:00 14.65 17.17 15.38 15.41 17.82 14.29 13.64 16.07 14.42 16.26 17.96 19.84 12.42
1:00 14.69 16.11 16.38 14.20 16.10 13.72 10.38 15.45 13.05 15.17 16.40 19.05 11.53
2:00 12.02 14.10 16.21 13.08 15.18 12.86 8.81 15.49 11.73 14.18 14.99 18.03 9.58
3:00 11.62 12.70 15.38 12.78 14.38 11.40 8.57 13.67 10.67 13.35 13.93 16.47 11.00
4:00 10.97 12.05 15.28 12.08 13.51 11.41 8.42 12.29 9.75 12.74 13.07 15.04 10.37
5:00 10.75 11.43 14.83 11.50 12.76 11.88 8.94 11.48 9.15 12.19 12.50 14.59 11.15
6:00 9.23 10.51 15.29 10.67 12.47 11.06 8.88 10.90 8.89 12.17 11.80 13.20 11.76
7:00 12.40 13.40 15.36 11.51 13.21 11.96 11.17 12.19 9.84 13.22 12.91 14.18 12.72
8:00 16.40 17.29 16.34 13.74 15.76 15.88 15.83 14.19 11.98 15.29 14.88 16.06 13.18
9:00 17.53 18.84 18.67 16.99 18.37 17.55 19.04 15.80 15.16 17.64 17.44 18.69 13.95
10:00 19.44 21.02 21.44 20.22 20.00 19.07 20.23 17.30 17.74 19.81 20.03 21.41 14.98
11:00 21.53 22.30 22.61 22.50 21.95 20.36 22.13 19.11 20.01 21.92 22.40 23.84 15.15
12:00 23.12 23.59 24.94 24.44 24.01 23.00 23.15 20.12 21.84 23.72 24.89 25.89 16.15
13:00 24.81 24.61 26.19 26.48 26.23 24.94 23.91 21.19 23.65 24.85 27.01 27.48 17.70
14:00 25.71 25.76 27.35 27.55 25.84 26.22 24.76 21.82 25.23 26.00 29.08 28.41 17.94
15:00 26.67 26.98 26.31 28.70 26.41 27.21 24.04 22.50 26.51 26.53 30.39 29.12 19.68
16:00 27.01 27.37 26.37 29.80 24.75 27.95 24.27 22.92 26.70 26.85 31.94 29.86 19.92
17:00 27.14 27.12 25.99 29.86 25.41 26.64 24.20 22.92 27.08 26.77 31.77 28.69 20.31
18:00 25.40 26.58 25.08 29.59 25.11 25.40 23.29 22.59 26.75 26.70 31.83 26.69 20.33
19:00 23.78 24.28 23.49 28.61 24.67 22.87 21.65 21.64 25.92 26.00 29.34 24.68 20.31
20:00 21.79 21.42 22.03 25.98 21.63 19.49 19.80 19.88 23.86 24.09 26.64 21.62 17.81
21:00 20.03 19.36 20.44 22.95 20.15 16.76 18.09 18.27 21.07 22.33 24.12 18.40 16.25
22:00 18.80 18.38 18.06 20.96 17.86 15.35 17.23 16.79 19.04 20.98 20.44 16.93 15.92
23:00 18.14 17.46 16.51 19.00 16.08 14.35 16.84 15.65 17.72 19.69 20.68 14.89 15.62

Table 5
Thermocouple temperature (◦ C).
# Thermocouple Date (year 1998)
01/14 01/15 01/16 01/17 01/18 01/19 01/20 01/21 01/22 01/23 01/24 01/25 01/26
1 29 29 29 29 29 29 28 26 26 22 21 21 21
2 31 31 31 30 29 29 29 27 27 24 20 20 20
3 29 29 29 29 29 28 28 27 27 26 23 22 22
4 31 31 31 30 30 30 30 29 29 27 24 24 24
5 31 31 32 32 32 32 32 30 30 29 26 26 26
6 29 29 30 30 30 30 30 30 30 29 27 27 27
7 36 36 37 38 38 38 37 33 33 31 29 29 29
8 29 30 31 31 31 32 32 33 33 31 30 30 30
9 27 27 27 27 27 27 28 36 36 34 32 32 32
10 25 25 25 25 25 26 26 30 30 34 32 32 33

thermocouples and the simulated ones, with their respective correlation coefficient (r ). Plots like these, for each
of the levels are presented in Appendix C.
The best prediction was obtained near the floor (thermocouple 1), in which, the error was just of 6.6%; the largest
error was of 12.8% near the roof (thermocouple 9).
A good estimation in the proximities of the center of the silo was also achieved (thermocouple 5), with an error of
9.1%. In Fig. 4 this is graphically observed.
Even though there are errors in the estimation, these are suitable for the analysis of the stored grain ecosystem.
The simulation model applied on it does not include the creation of internal heat (because of the respiration of the
F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14 11

Table 6
Simulated temperature (◦ C).
# Level Date (year 1998)
01/14 01/15 01/16 01/17 01/18 01/19 01/20 01/21 01/22 01/23 01/24 01/25 01/26
1 28.64 28.33 28.07 27.85 27.61 27.33 27.02 26.75 26.53 26.41 26.37 26.32 25.98
2 29.00 29.01 29.02 29.03 29.04 29.05 29.07 29.08 29.09 29.11 29.12 29.13 29.14
3 29.00 29.01 29.03 29.04 29.06 29.08 29.11 29.13 29.16 29.18 29.21 29.23 29.26
4 31.00 30.99 30.98 30.97 30.96 30.95 30.93 30.91 30.90 30.88 30.87 30.85 30.83
5 31.00 31.00 31.00 31.00 31.00 30.99 30.99 30.99 30.99 30.98 30.98 30.97 30.96
6 29.01 29.04 29.08 29.13 29.18 29.25 29.31 29.38 29.45 29.52 29.59 29.66 29.73
7 35.69 35.41 35.16 34.95 34.76 34.60 34.46 34.33 34.21 34.11 34.02 33.93 33.85
8 32.96 32.92 32.88 32.85 32.82 32.80 32.78 32.76 32.74 32.72 32.71 32.69 32.68
9 27.09 27.17 27.24 27.30 27.35 27.39 27.43 27.46 27.49 27.52 27.54 27.56 27.57
10 24.99 24.98 24.97 24.95 24.94 24.91 24.89 24.85 24.82 24.78 24.74 24.72 24.68
The “Level” is associated with the height where the thermocouples are located.

barley, insects and microorganisms), not even the variation of thermal properties of the barley (because of the moisture
content and strange materials in the silo), which are factors that influence the temperature of the barley.
In order to find out whether there are significant differences between the correlation coefficient for different depths
(ten levels) by weather seasons (summer, fall and winter), it was considered a Randomized Block Design, blocking
by seasons. The results demonstrated that there is no meaningful difference, concluding that the average percentage
of variation (of the simulated temperatures are explained by the measures) is equal in different depths and seasons.

4. Conclusions

A two-dimensional finite-difference model was developed to predict malting barley temperature in cylindrical
tower silos at any location. The model including air temperature requires few parameters that are easy to obtain, and
this model can be applied with ease to any cereal grain or cylindrical silo size and diverse environmental conditions.
This model with an axial and radial node spacing network of 0.5 [m] was able to predict grain temperature in a 6.3
[m] diameter and 23.5 [m] high silo containing malting barley. Validation was performed by comparing predicted and
measured grain temperature.
Predicted grain temperature was in close agreement with the measured ones with standard error of 1.09 to 2.67 [◦ C].
The model and parameters used in the model are applicable for temperature prediction purposes.
The stability obtained by the finite difference method was independently guaranteed by the value of the convective
heat transfer coefficient, even though; this value is involved in the calculation of stability criteria. This is a great
advantage because it is very hard to obtain this coefficient.
The simulation of the temperature for the malting barley in a cylindrical silo can be made without considering the
weather season; so this variation is not significant.

Acknowledgments

The author would like to thank the Universidad del Bı́o-Bı́o Research department and the support of project DIUBB
172409 GI/C at Universidad del Bı́o-Bı́o, Chile. He also thanks the anonymous reviewers and the editor of this
magazine for their valuable time and their careful comments and suggestions with which the quality of this paper
has been improved. Thanks to Dr. Fernando Miranda and Ma. Florencia Osorio for their valuable comments in the
revision of the manuscript.

Appendix A

See Tables 4–6.


12 F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14

Appendix B

Average daily temperatures of air, thermocouples and simulated, in the ten height “levels” where the thermocouples
are located.
F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14 13

Appendix C

Scatter plots between real and simulated temperatures, in the ten height levels where the thermocouples are located.
14 F. Novoa-Muñoz / Mathematics and Computers in Simulation 157 (2019) 1–14

References
[1] S.K. Abbouda, A.A.E. Ayoub, S.G.A. Dafaalla, Application of heat transfer model for prediction of temperature distribution in stored corn, J.
Sci. Eng. Res. 4 (2017) 187–193.
[2] A.M.S. Al-Amri, S.K. Abbouda, A heat transfer model to predict temperature distribution in stored wheat, Agric. Sci. 5 (2000) 85–90.
[3] A. Arias Barreto, R. Abalone, A. Gastón, R. Bartosik, Analysis of storage conditions of a wheat silo-bag for different weather conditions by
computer simulation, Biosyst. Eng. 116 (2013) 497–508.
[4] R. Bartosik, D. Ochandio, L. Cardoso, D. de la Torre, Storage of malting barley with different moisture contents in hermetic silo-bags, in:
Proceedings del 9th International Conference Controlled Atmospheres and Fumigation of Stored Products, 15 al 19 de Octubre de 2012,
Antalya, Turquía, pp. 549–554.
[5] T.L. Bergman, A.S. Lavine, F.P. Incropera, D.P. Dewitt, Fundamentals of Heat and Mass Transfer, seventh ed., John Wiley & Sons Inc., USA,
2011.
[6] J.P. Buonaccorsi, Measurement Error: Models. Methods, and Applications, Chapman & Hall/CRC Interdisciplinary Statistics, 2010.
[7] M. Carrera-Rodríguez, G.M. Martínez-González, J.L. Navarrete-Bolaños, J.E. Botello-Álvarez, R. Rico-Martínez, H. Jiménez-Islas, Transient
numerical study of the effect of ambient temperature on 2-D cereal grain storage in cylindrical silos, J. Stored Prod. Res. 47 (2011) 106–122.
[8] R. Fourar-Belaifa, F. Fleurat-Lessard, Z. Bouznad, A systemic approach to qualitative changes in the stored-wheat ecosystem: Prediction of
deterioration risks in unsafe storage conditions in relation to relative humidity level, infestation by Sitophilus oryzae (L.), and wheat variety,
J. Stored Prod. Res. 47 (2011) 48–61.
[9] A. Gastón, R. Abalone, A. Cassinera, M.A. Lara, Modelización de la distribución de la temperatura y humedad en granos almacenados en
silos-bolsa, Mec. Comput. XXVI (2006) 3547–3561.
[10] F. Hammami, S. Ben Mabrouk, A. Mami, Modelling and simulation of heat exchange and moisture content in a cereal storage silo, Math.
Comput. Model. Dyn. Syst. 22 (3) (2016) 207–220.
[11] J.P. Holman, Heat Transfer, McGraw-Hill, New York, 2010.
[12] C. Idler, A. Wagner, U. Weber, T. Hoffmann, Effect of short-term storage on quality of wheat stored in large polyethylene bags, Agric. Eng.
Int. CIGR J. 14 (1) (2012) 149–156.
[13] A. Iguaz, C. Arroqui, A. Esnoz, P. Vírseda, Modelling and simulation of heat transfer in stored rough rice with aeration, Biosyst. Eng. 89 (1)
(2004) 69–77.
[14] A. Iguaz, C. Arroqui, A. Esnoz, P. Vírseda, Modelling and validation of heat transfer in stored rough rice without aeration, Biosyst. Eng.
88 (4) (2004) 429–439.
[15] C. Jia, D.W. Sun, C. Cao, Finite element prediction of transient temperature distribution in a grain storage bin, J. Agric. Eng. Res. 76 (2000)
323–330.
[16] C. Jia, D.W. Sun, C. Cao, Mathematical simulation of temperature fields in a stored grain bin due to internal heat generation, J. Food Eng. 43
(2000) 227–233.
[17] C. Jia, D.W. Sun, C. Cao, Computer simulation of temperature changes in a wheat storage bin, J. Stored Prod. Res. 37 (2001) 165–177.
[18] F. Jian, D.S. Jayas, N.D.G. White, K. Alagusundaram, A three-dimensional, asymmetric, and transient model to predict grain temperatures in
grain storage bins, Trans. ASAE 48 (1) (2005) 263–271.
[19] H. Jiménez-Islas, J.L. Navarrete-Bolaños, E. Botello-Álvarez, Estudio numérico de la convección natural de calor y masa 2-D en granos
almacenados en silos cilíndricos, Agrociencia 38 (2004) 325–342.
[20] D.C. Lopes, J.H. Martins, E.C. Melo, P.M.B. Monteiro, Aeration simulation of stored grain under variable air ambient conditions, Postharvest
Biol. Technol. 42 (2006) 115–120.
[21] M.D. Montross, D.E. Maier, K. Haghighi, Development of a finite-element stored grain ecosystem model, Trans. ASAE 45 (5) (2002) 1455–
1464.
[22] M.D. Montross, D.E. Maier, K. Haghighi, Validation of a finite-element stored grain ecosystem model, Trans. ASAE 45 (5) (2002) 1465–1474.
[23] D. Ochandio, L. Cardoso, R. Bartosik, D. de la Torre, J. Rodríguez, J. Massigoge, Storage of quality malting barley in hermetic plastic bags,
in: Proc. of the 10th international working conference on stored product protection, 2010, pp. 331–337.
[24] Y. Wang, H. Duan, H. Zhang, Z. Fang, Modeling on heat and mass transfer in stored wheat during forced cooling ventilation, J. Therm. Sci.
19 (2) (2010) 167–172.
[25] J.R. Welty, G.L. Rorrer, C.E. Wicks, R.E. Wilson, Fundamentals of Momentum, Heat and Mass Transfer, fifth ed., Wiley, Hoboken, NJ, 2008.

Vous aimerez peut-être aussi