Int. J. Food Eng.

2015; 11(6): 715–729

David Mosna and Giuseppe Vignali*

Three-Dimensional CFD Simulation

of a “Steam Water Spray” Retort Process
for Food Vegetable Products
DOI 10.1515/ijfe-2015-0062 however, it adversely affects the nutritional quality of
foods such as vitamins and essential fatty acids particu-
Abstract: This work aims at simulating a retort steriliza-
larly for products processed in metal or glass containers.
tion process of flexible packaging, performed in auto-
The basic method for retorting a food product is pla-
clave. ANSYS CFD© software has been used in this
cing the food inside a can or a retort pouch (flexible
study to simulate the heat transfer inside the retort cham-
packaging) sealing the container, and then placing the
ber. Flexible packaging filled with a pumpkin soup has
container in a retort, which is a large industrial-scale
been modelled. Three-dimensional (3D) geometrical con-
pressure cooker. The retort is sealed and heated to a
figuration of the equipment has been designed and then
specific temperature, generally exceeding the boiling
simulated to evaluate the sterilization level achieved by
point of water [1]. The food is then heated for a specific
each packaging inside the retort chamber. In particular,
period of time depending on the nature of the food and
the internal temperature of the product situated in the
several other factors. The processing time and temperature
cold spot of the chamber has been monitored during
must be sufficient to allow the commercially sterile of the
5 min of the process (after reaching 85°C inside the
product. After cooking, the container is cooled to room
product) in order to monitor the time/temperature trend.
temperature. The severe heat treatment of retort proces-
Experimental tests have been finally performed to vali-
sing can also destroy essential nutrients and negatively
date the simulation model of heat exchange. The sterili-
affect sensory properties. The goal when choosing a pro-
zation process takes place in a retort chamber by means
cessing temperature and time for a retorted food product is
of “overheated water sprayed” at 403 K and a pressure of
to maximize the quality retention and prevent overcook-
5 atm. Results show a good agreement between the simu-
ing, while maintaining the safety of the product [2].
lated results and the real retort process and confirm the
Until today various types of retorts have been devel-
potential value of the simulation model to evaluate the
oped to handle different packaging and manufacturing
process performance.
requirements. They can be classified based on three main
Keywords: CFD, food packaging, retort process, vegetable aspects: method of heating, batch versus continuous and
soup mode of agitation. The different methods of heating
include saturated steam, water immersion, steam–air
retorts and steam water spray [3].
1 Introduction Saturated steam retorts are typically used for proces-
sing cans. They consume a lot of energy, but are less
Thermal sterilization is one of the most efficient methods expensive compared with other types of retort. Since any
of food preservation practiced worldwide. The main air inside the retort could insulate the containers being
objective of this thermal processing is to achieve long- heated, the retort has to be saturated with steam while air
term shelf-stability. escapes through vent valves. Often, overpressure air can
The retort sterilization is one of the most adopted be applied during cooling to prevent the containers defor-
thermal processing in order to reduce the microbial load mation. The use of overpressure means that it can be
of packaged food and achieve commercially sterile food processed in many types of containers, not just metal
products. In achieving the needed sterilization effect, cans. Fragile containers, such as glass, and flexible con-
tainers, such as pouches, generally require some over-
*Corresponding author: Giuseppe Vignali, Department of Industrial
pressure. Water immersion retorts are commonly used
Engineering, University of Parma – Parco Area delle Scienze, 181/A,
to process containers using an overpressure process.
43100 Parma, Italy, E-mail: [email protected]
David Mosna, Department of Industrial Engineering, University of Steam–air retorts can also provide an overpressure pro-
Parma – Parco Area delle Scienze, 181/A, 43100 Parma, Italy cess. In these retorts, a large fan is used to mix steam

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
716 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process

with the air to prevent any cold spots inside the retort [4]. along a defined time span. The temperature values are then
Steam water spray retorts are less expensive compared compared with those obtained from experimental tests and
with other types of retort. They differ from saturated used to define the sterilization efficiency of the process.
steam retorts in that air can be present in the retort In summary, the objectives of this work can be
during processing, meaning it can be an overpressure described as follows:
process. Nowadays retort treatment using pressured – To carry out the simulation of an industrial “steam
“steam water spray” is used especially for vegetables or water spray” retort process of a pumpkin soup pack-
soups, which can hardly be sterilized by means of con- aged in carton flexible container, to assess the tem-
tinuous processes, due to the possible presence of some perature trend as a function of time
solid pieces inside a homogeneous liquid. Metallic con- – To validate the simulation results experimentally
tainers, jars or heat-resistant beverage cartons, such as
Tetra Recart® or Combisafe®, could be used in this pro-
cess. Dividing retorts based on the presence and mode of
agitation, three categories can be defined: static or still,
2 Materials and methods
rotary and oscillating [3].
Two types of heat transfer occur during this process: 2.1 Equipment
convection and conduction. The convection rate can be
considerably increased by inducing a forced convection “Steam water spray” retorts from Allpax is used in this
by means of mechanical agitation of the trays. For this study (Figure 1); it utilizes a high-volume pump with an
reason many retort chambers are designed to provide array of spray nozzles strategically located to provide
longitudinal motion or reciprocating agitation of the pro- complete load coverage and create even temperature dis-
duct [5–13]. By means of these movements, it is possible tribution in both heating and cooling. The system uses a
to reduce the treatment time and to obtain higher quality host computer that stores product-configurable recipes.
products [5]. These recipes define the process parameters to be fol-
Literature widely explored the issue of retort process lowed, which can be automatically corrected to avoid
for vegetables and other food products [14, 15], with the process deviations. The retort system can contain until 6
aim to develop better understanding of this technology [16]. baskets, each composed of 16 trays, which contain 70
Some authors tried to approach the problem experimen- cartons per tray. This system is then able to sterilize up
tally [17], but the high costs and time consumption asso- to 6,720 bricks having a capacity of 500 ml or 8,640
ciated with experimentation have often precluded containers with a capacity of 390 ml per cycle.
producing efficient in-depth results [18]. In recent years The thermal cycle of the pumpkin soup lasts approxi-
the development of the numerical simulation helped to mately 2 h and it is composed of the following steps:
overpass this problem and to identify the best thermal – Preheating phase: the internal temperature of the
setting of the process [19]. Among the different numerical sterilizer rises up to 50°C at 2 bars (7 min)
approaches, computational fluid dynamics (CFD) appeared – Heating phase: the internal temperature of the steri-
as the most suitable in order to understand the evolution of lizer rises up to 110°C with an increased pressure of
the temperature inside the product during a retort process up to 4 bars (20 min)
[20–25]. All of these latter studies simulated a three-dimen- – Sterilization phase: inside the sterilizer the tempera-
sional (3D) process considering the real configuration of the ture reaches 130°C and the overheated water is
packaging, but none of them aims at analysing the flow of inserted with a pressure of about 4 bars; the product
steam water moving inside a retort chamber. Similar stu- is handled in these conditions for a defined steriliza-
dies have been performed for other food products, like tion time (45 min)
pasta [26, 27] or refrigerated chamber [28], showing how – Precooling phase: the sterilizer temperature drops to
is important the knowledge of the air/vapour flows for food 105°C with a decreased pressure to 2 bars (10 min)
processes. – Cooling phase: the whole system achieves an ambient
In this work, by means of the support of a numeric temperature and a pressure close to 1.5 bars (30 min)
solver (ANSYS CFX version 14.5), CFD simulations have – Discharge phase: when the system reaches the atmo-
been carried out to evaluate the temperatures reached spheric pressure, the retort system can be opened to
inside the brick during a “steam water spray” retort process, discharge the products (5 min)

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process 717

Figure 1: Retort system used in this study.

At the end of the sterilization, all the cartons are discharged well known and can be easily retrieved from the soft-
on conveyor belts and they are sent to the secondary ware library. Conversely, for the food product, it was
packaging operation. necessary to find (or compute) the correct values for
From a technological point of view, the most critical each physical property. In particular, for the pumpkin
phase is the product heating phase, especially in the 5 min soup, the following properties were considered: den-
after the reaching of 85°C inside the product, due to sity, dynamic viscosity, thermal conductivity and heat
the difficulties in temperature control (usually after 1,680 capacity. The values of these properties were provided
s of treatment, including the preheating phase). This is the by a supporting company which, during experimental
time range in which the simulations were performed tests, has measured the corresponding data. Thermal
(1,680–1,980 s of treatment). The analysis was focused on conductivity was measured using the method
the heating phase, instead of the sterilization one, because described by Bellet et al. [29], dynamic viscosity was
it is the most critical phase in terms of product overcooking, determined by using Cannon-Fenske glass capillary
because the time and temperature are not controlled in this viscometer and density was determined by pycno-
phase instead they are fixed for the sterilization. metric and hydrostatic methods. Heat conductivity
was measured using the differential scanning calori-
metry (DSC) method.
To perform the simulation, packaging materials have
2.2 Materials also been considered. In fact thermal exchange happened
between the overheated water and the food liquid
The product used in this study is pumpkin soup, pack- through a thin multilayer food package. In this type of
aged in TetraRecart® brick having a volume of 500 ml packaging, paper is the material with the lowest thermal
with a width of 84 mm, maximum height of 46 mm and conductivity. For this reason only the paper layer was
length of 136 mm. The food product is filled by means of considered. All the properties of each material are
a piston filler, which ensures a good accuracy and a reported in Table 1.
limited head space (on average 13 ml). The sealed con- Before the simulation, preliminary tests have been
tainer was treated in the retort system (Figure 1). performed in order to identify the cold spot of the
In order to perform CFD simulations, three factors process. Six points located inside the chamber have
were considered: overheated water, pumpkin soup and been monitored in order to find the most disadvantaged
packaging materials. Properties of the first one are position in reaching the sterilization effect after the

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
718 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process

Table 1: Properties of overheated water, pumpkin soup and whole process. This point has been identified as located
packaging paper used in the study. in the centre of the third layer starting from the bottom
of the vessels (Figure 5).
Overheated water

Thermodynamic state Mixture liquid/gas in equilibrium

2.3 Mathematical modelling
Molar mass . kg/mol
Density  kg/m
2.3.1 Simulation setting: geometry and mesh
Reference pressure  bars
Reference temperature °C
Heat capacity , J/(kg K) Based on the large size of the retort sterilizer, the simula-
Thermal conductivity . W/(m K) tions were carried out considering the central section of
Dynamic viscosity . Cp the whole system. Considering that the retort has an
Pumpkin soup
overall length of 8 m, the section chosen is the one
between 3.850 and 4.150 m.
Thermodynamic state Liquid Due to the complexity of the domain, it has been
decided to approach the issue considering both the pre-
Density , kg/m
Heat capacity , J/(kg K) sence of the trays under the carton bricks and a domain
Thermal conductivity . W/(m K) without trays (Figure 2(A) and 2(B)).
Dynamic viscosity . Pa s For each configuration, 320 bricks were loaded, con-
sidering however only 1 brick for the thermal analysis. As
Food packaging paper
shown in Figure 2(C), the considered brick is located in
Thermodynamic state Solid the centre of the third layer starting from the bottom of
the vessels. This is selected based on preliminary experi-
Density  kg/m
Heat capacity , J/(kg K)
mental tests described previously.
Thermal conductivity . W/(m K) The grid generation for the obtained fluid domains was
performed using ANSYS ICEM CFD meshing software, the

Figure 2: Configuration with (A) and without trays (B), and focus on the analysed brick (C).

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process 719

modeller associated with ANSYS CFX software. By means of Table 2: Values of the surface mesh.
this action the fluid volumes are divided into a finite num-
ber of cells, on which the analysis is carried out. This Part Size (mm) Height (mm) Height ratio Tetra ratio
operation has been performed for both the configurations Wall   . .
(with trays and without trays). Open   . .
The meshes were created following a gradient, which Inlet   . .
respects the Courant number. This number is crucial for Brick   . .
Trays   . .
transient flows. For a one-dimensional grid, it is defined as
Pipe   . .

Δt
Courant ¼ u ð1Þ
Δx
reports the setting used to generate the surface mesh for
where u is the fluid speed, Δt is the time step and Δx is each part.
the mesh size. The Courant number calculated in ANSYS The volume mesh was initially generated by creating
CFX is a multidimensional generalization of this expres- a uniform subdivision and then thickened in the critical
sion where the velocity and length scale are based on the areas of the fluid volume. In particular, a finer mesh was
mass flow into the control volume and the dimension of used near the outlet section of the nozzles, because in
the control volume [30]. To decrease oscillations, improve these areas high shear rates are expected.
accuracy and decrease numerical dispersion, the Courant Figure 4 shows the generated mesh used for the
number must be kept sufficiently small (always < 10), calculation: an unstructured tetrahedral meshing scheme
which explains that fluid particles move from one cell was used for each configuration.
to another within one time step. The final meshes were determined when the increase
The number of cells used in the simulations was in quality of the mesh did not provide any significant
determined starting from a coarse mesh and gradually improvements in the results. The overall number of the
refined evaluating the changes in the results. The exter- cells created for the first mesh (Figure 4(A)) is about
nal surface mesh (Figure 3) was first generated as sug- 12,500,000, while, for the second one (Figure 4(B)), the
gested in ANSYS solver modelling guide [31]. Table 2 overall number of cells created is about 11,200,000.

2.3.2 Simulation setting: domain equations

and boundary conditions

Three-dimensional, multiphase, two-fluid model simula-

tions were developed to investigate the temperature trend
inside the carton package. ANSYS CFX 14.5 software was
used to solve the governing continuity, momentum and
energy equations for the defined geometry and associated
boundary conditions. The solved generalized transport
equations are as follows:
The continuity equation:

@ρ
þ Ñ  ðρU Þ ¼ 0 ð2Þ
@t
The momentum equation:

@ ðρU Þ
þ Ñ  ðρU # UÞ ¼  Ñp þ Ñτ þ SM ð3Þ
@t
where the stress tensor, τ, is related to the strain rate by
this following equation:
 
T 2
τ ¼ μ ÑU þ ðÑUÞ  δÑ  U ð4Þ
Figure 3: External mesh surface.
3

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
720 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process

Figure 4: Volume mesh of the configuration with trays (A) and without trays (B).

In this work, according to the adopted materials, two fluid soup, which is contained inside the brick, includes nat-
domains were created: “autoclave” and “pumpkin soup”. ural convection, and the Boussinesq approximation was
ANSYS CFX uses the concept of domains to define the assumed.
properties of regions where a fluid, porous or solid material Inside the autoclave domain the “thermal energy”
is located. In these domains, the equations of fluid flow or model was set in order to consider the heat exchange in
heat transfer are then solved. This implies the selection of this region.
3D bounding regions and the choice of appropriate physical Being also the water flowing in the “autoclave”
models for each domain. The interface between the “auto- domain at high velocity, a turbulent model has been
clave” domain and the “pumpkin soup” domain has been adopted. One of the main problems in turbulence model-
modelled as “thin material interface” meaning that this ling is the accurate prediction of flow separation from a
surface will be considered by the solver as a thin layer of smooth surface. For this reason, the model adopted was
solid material. In this case, paper has been set up as solid the shear stress transport (SST). The SST model, proposed
since it is the main material and also that with the lowest by Menter and Ekaterinaris [32], is an eddy-viscosity
thermal conductivity in Tetra Recart packages. A thickness model which is a combination of k–ω and k–ε models.
of 0.5 mm has been chosen for this solid layer. For the The first is used in the inner boundary, while the second
physical properties of packaging paper, see Table 1. in the outer region and outside of the boundary layer.
The following “thermal energy” model is used to pre- The SST model has been used in order to overcome the
dict the temperature inside the “pumpkin soup” domain: problems of both the methods. These features make the
SST model more accurate and reliable for a wider class of
@ ðρhtot Þ
þ Ñ  ðρUhtot Þ ¼ Ñ  ðkÑT Þ þ Ñ  ðU  τ Þ ð5Þ flows than the standard k–ω and k–ε models [32].
@t
The boundary conditions (Figure 5) are referred to 10
where htot is total enthalpy, which can be expressed as a inlet holes and 2 opening sections; in particular, a uni-
function of the static enthalpy h (T, p) as follows: form orthogonal velocity for the inlets and a relative
pressure for the opening are set. The external wall was
1
htot ¼ U 2 ð6Þ considered as adiabatic, while the other bricks (which are
2
not considered for the internal thermal process) were
The term Ñ  ðU  τÞ in eq. (5) is due to viscous stresses kept at 85°C. Table 3 resumes the boundary and initial
and it is known as the viscous work term. conditions for the autoclave domain.
Instead, as far as the flow of the food product inside All the simulations were carried out in the transition
the “pumpkin soup” domain is concerned, the laminar state, to evaluate the trend of the sterilization temperature
flow model has been chosen. The heat transfer in the inside the brick as a function of time. In particular for both

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process 721

Figure 5: Boundary condition in CFX-pre-solver: (A) brick boundary, (B) inlet boundary, (C) opening boundary and (D) wall boundary.

the configurations, a total time of 5 min with a 0.5 s time step 1. Preliminary analysis of the product (density, dynamic
was adopted for each simulation. The simulation models viscosity, thermal conductivity and heat capacity)
were run on an Intel® Xeon® E3-1270 Quad Core (3.4 GHz) 2. Collecting of the samples
PC, with 16 GB RAM memory, under Windows 7 professional 3. Positioning of the temperature probes inside the
platform. With this configuration, the computational time for package
each simulation was about 4 days and 6 h. 4. Partial loading of the retort
5. Controlling and analysing the samples by opening
the cartons
2.4 Experimental method

The experimental tests have been performed to validate the According to the preliminary test (as reported in Section
simulations and to ensure a heat treatment able to give a 2.2) the tests were carried out using temperature probes
healthy and safe product to the consumers. The tests have located in the two bricks, which are previously considered
been developed according to the following procedure: as positioned in the cold spot inside of the retort.

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
722 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process

Table 3: Boundary and initial condition. has been placed in the mentioned position is that in some
cases [20] the cold spot inside the container was located in
Boundary conditions the lower part of it.
Inlet  m/s After the probes connection, the retort has been
Opening  relative bars loaded with only a few hundred cartons, disposed in
Wall No slip wall and adiabatic two sections as supposed in the simulation model.
thermal condition
Brick No slip wall and initial
temperature at °C

Initial conditions (t ¼ ) 3 Simulation results

Pressure  relative bars
The CFX-Post Solver was used to analyse velocity, tem-
Velocity  m/s
“Autoclave” initial temperature °C perature and pressure fields inside the retort sterilizer
“Packaging paper” initial °C and pumpkin soup domains. Different section planes
temperature were used to view the results of the process.
“Pumpkin” initial temperature °C Furthermore, the cold spot was identified to monitor the
temperature during the thermal cycle.
For both simulations, the autoclave domain
The thermocouple probes were thus inserted inside the two (Figure 6(A)) shows a constant temperature of 130°C.
containers and connected to an ELLAB data logger. For each As regards the steam velocity, Figure 6(B), 6(C) and
brick two probes were used. The first one was located in the 6(D) shows the flows in a range between 0 and 5 m/s.
geometrical centre of the brick, while the second one was The section planes cut the spray nozzles.
located above the brick base, at a distance of approximately For each simulation, the temperatures inside the
30% of the brick height. The reason why the second probe bricks were reported in the following sections.

Figure 6: Temperature inside the sterilizer (A); velocity inside the sterilizer at three planes: (B), (C) and (D).

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process 723

3.1 Simulation results of the configuration 3.2 Simulation results of the configuration
with trays without trays

Analysing the temperature of the product in the configura- A similar increase of the temperature, with respect to the
tion with trays, it was observed that the product heats up previous simulation, is shown by the food product with
with the increase of the time. Figure 7 shows the increase of the configuration without trays. Figure 9 shows the tem-
temperature inside the brick during the sterilization phase perature inside the brick during the sterilization phase
at different time intervals (between 0 and 5 min) in a range at different time intervals (between 0 and 5 min) in a
between 85°C and 130°C. The first image on the left in range between 85°C and 130°C. In the case without
Figure 7 shows the temperature in the brick at the beginning trays, the temperatures of the external layer are higher
of the simulation, the second one shows the temperature than before (Figure 9). The first image on the left in
after 2.5 min and the last one at the end of the simulation. Figure 9 shows the temperature in the brick at the
The temperatures were also analysed as a function of beginning of the simulation, the second one shows the
time, in the previously defined point inside the product. temperature after 2.5 min and the last one at the end of
Starting by 85°C, the cold spot of product reaches about the simulation.
92.5°C after 5 min. The temperature trend in the cold spot As before, the temperatures were analysed as a func-
is described in Figure 8. tion of time in the previously defined point inside the
The simulation results were processed by means of product. Starting by 85°C, considered as uniformly dis-
Microsoft Excel to achieve all the simulated temperatures tributed inside the products, the cold spot of the product
in the brick. With those data, we were able to analyse the reaches about 94.5°C after 5 min. Figure 10 describes this
temperature trend in the product volume and not only in temperature trend.
a specific point. The simulation results were processed by means of
The percentage of volume with temperature values Microsoft Excel to achieve all the simulated temperatures
higher than some reference temperatures was reported in in the brick. As reported in Table 5, an increase of the
Table 4. percentage volume with higher temperature than the
Analysing Figure 7 and Table 4 it is possible to study reference ones was observed with respect to the case
the evolution of the temperature inside the brick during with trays. As before values equal to 100% mean that
the sterilization. At the beginning of the process there is a the entire volume is located at a temperature higher or
high temperature gradient between the brick’s surface equal to the reference temperature set.
(that reaches 108°C), and the lower part whose tempera- Even in the configuration without tray, a high tem-
ture is 85°C. perature gradient is present at the beginning of the
This difference decreases during the process but with sterilization (Figure 10), but it decreases in a more con-
a slow trend, because the heating exchange is due to a siderable way compared to the previous simulation,
conduction phenomenon and because in the mathemati- because of the heating exchange between the lower
cal model the heating exchange between the paper brick’s surface and the overheated water in the autoclave
packaging and the trays is equal to zero. (Table 5).

Figure 7: Temperature inside the brick at t ¼ 0 min, t ¼ 2.5 min, t ¼ 5 min.

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
724 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process

Time–temperature trend (with trays CFD simulation)

100.0
T inside brick
97.5

95.0

92.5
(°C)

90.0

87.5

85.0

82.5

80.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Figure 8: Time–temperature trend in the pre-
(min) viously defined point inside the brick.

Table 4: Percentage of volume with T higher or equal to which has been evaluated by the simulation during all
Tref (at t ¼ 2.5 min and t ¼ 5 min). the process.
Figure 12 shows instead the trend of the temperatures
Tref (t ¼ . min) of the pumpkin soup and the retort chamber in the refer-
Tref °C K %Volume
ence time (after 1,680 s of treatment).
During the reference time (1,680–1,980 s), the tem-
T  .  perature values in the retort sterilizer and inside the brick
T  . .
at a step of 0.5 min are reported in Table 6. The product
T  . .
T  . .
temperature has been taken in the same point before
analysing in the CFD simulation.
Tref (t ¼  min)

Tref °C K %Volume
4.2 Comparison with simulation results
T  . 
T  . . Table 7 compares the experimental results with them
T  . . provided by the software simulations. The following
T  . . values refer to those measured and calculated at the
bottom of the product (cold spot).
Figure 13 shows the comparison between time–tem-
4 Experimental validation perature data obtained experimentally and by simula-
tions. From the figure, it is possible to observe a slight
4.2 Experimental tests deviation between the experimental results and those
calculated by the software for the configuration with
Following the procedure reported in Section 2.4, an trays. In fact the trays are modelled as an adiabatic wall
experimental test has been performed. Figure 11 reports and thus do not exchange heat with the overheated water
the temperature trend as a function of time for the steri- so they cannot transfer thermal energy to the base sur-
lizer chamber and inside the brick in the same point, face of the bricks. This limits the heat transfer surface for

Figure 9: Temperature inside the brick at t ¼ 0 min, t ¼ 2.5 min, t ¼ 5 min.

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process 725

Time–temperature trend (no trays CFD simulation)

100.0
T inside brick
97.5
95.0
92.5
(°C)

90.0
87.5
85.0
82.5
80.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Figure 10: Time–temperature trend in the local
(min) point inside the brick.

Table 5: Percentage of volume with T higher or equal to Tref

(at t ¼ 2.5 min and at t ¼ 5 min).

Tref (t ¼ . min)

Tref °C K %Volume

T  . 
T  . .
T  . .
T  . .

Tref (t ¼  min)
Figure 12: Temperature trend of the retort chamber and inside the
Tref °C K %Volume brick in the reference time (1,680–1,980 s).

T  . 
T  . 
T  . . Table 6: Experimental temperature inside the retort chamber and
T  . . inside the brick (step of 0.5 min).

Time (min) T retort sterilizer (°C) T inside brick (°C)

. . .

. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .

Figure 11: Temperature trend of the retort chamber and inside the
brick. whole package surface allowing a very good agreement
between the experimental and the simulated data.

each package and consequently the increasing of tem-

perature follows a different trend. In the real case trays 5 Discussion
are made of stainless steel and thus a high conduction
heat transfer is present with the bricks’ base surface. By One of the main goals of this work was to show, through CFD
removing the trays the heat exchange happened on the simulation, how a complex internal geometry of a loaded

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
726 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process

Table 7: Comparison between experimental data and those of the simulations with trays and
without trays.

Time Experimental CFD simulations ΔError CFD simulations ΔError

(min) data (°C) with trays (°C) (°C) without trays (°C) (°C)

 . . . . .

 . . . . .
 . . −. . −.
 . . . . −.
 . . . . −.
 . . . . .

Figure 13: Comparison between the

experimental data and simulation
results.

retort chamber can affect the heating process of the slowest face”, with paper chosen as solid material (thickness 0.5
heating zone (SHZ) inside the brick located in the “autoclave” mm). The properties of these materials are the same as the
cold spot. To achieve this goal, the results presented in the ones shown in Table 1 and the initial conditions are the same
previous paragraphs are compared with those of a new CFD as the previous simulations. Also in this case, the simulated
simulation. In this new simulation only one brick, containing time is of 300 s.
pumpkin soup, is present and is surrounded by overheated The differences in the temperature distribution inside
water. Also in this case the interface between the soup and the packaging at the end of the simulation are shown in
the overheated water is modelled as a “thin material inter- Figure 14.

Figure 14: Temperature inside the brick at 5 min in the case of (A) simulation without trays and (B) “one brick” simulation.

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process 727

Figure 15: Comparison between experimental data and simulation without trays and one brick.

Table 8: Comparison between experimental data and those of the simulations without trays and “one brick”.

Time Experimental CFD simulations ΔError CFD simulations ΔError

(min) data (°C) without trays (°C) (°C) “one brick” (°C) (°C)

 . . .  −.

 . . . . .
 . . −. . .
 . . −. . .
 . . −. . −.
 . . . . −.

It can be noticed that the temperature in the “one brick” 6 Conclusions

case reaches higher values and that there is a lower
thermal gradient. This work aimed at simulating a retort sterilization pro-
Figure 15 and Table 8 show instead the differences cess of pumpkin soup packaged in flexible packaging.
in time–temperature trend in the brick’s cold spot for ANSYS CFD© software has been used in this study to
the two simulations compared with the experimental simulate the heat transfer inside the retort chamber.
tests. Three-dimensional geometrical configuration of the
It can be noticed that, at the end of the simulation equipment has been modelled in order to evaluate the
time, the brick’s cold spot reaches a temperature of sterilization level reached by each flexible packaging
96.6°C in the “one brick” case, which is 2.3° higher inside the retort chamber. Until now, no studies aim at
than the temperature reached in the retort simulation. considering the flow of overheated water reaching the
The results highlighted in Figure 15 show the impor- packaging inside a complex geometry like those of an
tance of simulating the effects of whole retort section autoclave. In this study, in particular, the internal tem-
instead of simulating only one brick. perature of cold spot has been monitored during 5 min
It is important to remark that the new simulation of the process in order to control the time/temperature
results have been compared with the results of the simula- trend. The simulations have been performed consi-
tion without trays because, as explained in Section 4.2, dering two configurations: one with trays and another
this CFD analysis fits in a better way to experimental data. without them.
In the literature there are no papers concerning the The experimental validation demonstrated that the
simulation of the whole section of a complex retort cham- better results were obtained by the simulation without
ber. Previous works dealt with the simulation of a single trays, having it the ability to better understand the beha-
package for thermal sterilization process, focusing mainly viour of the heat exchange during the considered pro-
on the shape and the position of the SHZ [20, 33, 34] or on cess. The comparison between the simulation of retort
bacteria deactivation kinetics [35]. chamber and the “one brick” simulation shows that the

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
728 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process

chamber’s internal geometry affects the heath exchange fluids (CMC) during reciprocation agitation thermal processing.
process since there are evident differences in the time– J Food Eng 2015;165:1–12.
10. Singh AP, Singh A, Ramaswamy HS. Modification of a static
temperature trend of the brick’s cold spot.
steam retort for evaluating heat transfer under reciprocation
Future research will then be addressed to better agitation thermal processing. J Food Eng 2015;153:63–72.
simulate the process considering also the heat exchange 11. Singh AP, Ramaswamy HS. Effect of can orientation on heat
between the stainless steel trays and the paper packa- transfer coefficients associated with liquid particulate mix-
ging, in order to make the simulation results still more tures during reciprocation agitation thermal processing. Food
Bioprocess Technol 2015;8:1405–18.
adherent to the experimental ones.
12. Singh A, Singh AP, Ramaswamy HS. A refined methodology for
evaluation of heat transfer coefficients in canned particulate
fluids under rapid heating conditions. IChemE Food Bioprod
Nomenclature Process 2015;94:169–79.
13. Walden R, Emanuel J. Developments in in-container retort
Symbol Property Unit technology: the Zinetec Shaka® process. In: Doona CJ, Kustin
K, Feeherry FE, editors. Case studies in novel food processing
ρ Density kg/m3 technologies: innovations in processing, packaging, and pre-
t Time s dictive modeling. Cambridge, UK: Woodhead Publishing Ltd,
U Vector of velocity m/s 2010: 389–406.
p Static pressure kg/(m s2) 14. Durance TD. Improving canned food quality with variable
τ Molecular stress tensor kg/(m s2) retort temperature processes. Trends Food Sci Technol
SM Momentum source kg/(m2 s2) 1997;8:113–18.
δ Kronecker delta function 15. Teixera AA, Tucker GS. On-line retort control in thermal ster-
μ Dynamic viscosity kg/(m s) ilization of canned foods. Food Control 1997;8:13–20.
T Static temperature K 16. Simpson R, Teixeira A, Almonacid S. Advances with intelligent
htot Specific total enthalpy m2/s2 on-line retort control and automation in thermal processing of
k Thermal conductivity W/(m K) canned foods. Food Control 2007;18:821–33.
17. Dwivedi M, Ramaswamy HS. An empirical methodology for
evaluating the fluid to particle heat transfer coefficient in bi-
axially rotating cans using liquid temperature data. Food
Bioprocess Technol 2010;3:716–31.
References 18. Norton T, Sun DW. Computational fluid dynamics (CFD) – an
effective and efficient design and analysis tool for the food
1. Singh RP, Heldman DR. Introduction to food engineering, 4th industry: a review. Trends Food Sci Technol 2006;17:600–20.
ed. Burlington, MA: Elsevier Inc., 2009:841. 19. Miri T, Tsoukalas A, Bakalis S, Pistikoupoulos EN, Rustem B,
2. Ansar Ali A, Sudhir B, Srinivasa Gopal TK. Effect of rotation on Fryer PJ. Global optimization of process conditions in batch
the heat penetration characteristics of thermally processed thermal sterilization of food. J Food Eng 2008;87:485–94.
tuna in oil in retort pouches. Int J Food Sci Tech 2006;41: 20. Abdul Ghani AG, Farid MM, Chen XD, Richards P. Thermal
215–19. sterilization of canned food in a 3-D pouch using computa-
3. Hardt-English PK. Commercial sterilization systems. In: tional fluid dynamics. J Food Eng 2001;48:147–56.
Heldman DR, editor. Encyclopedia of agricultural, food, and 21. Abdul Ghani AG, Farid MM, Zarrouk SJ. The effect of can rota-
biological engineering. New York: Marcel Dekker Inc, tion on sterilization of liquid food using computational fluid
2003:139–42. dynamics. J Food Eng 2003;57:9–16.
4. Williams JR, Steffe JF, Black JR. Economic comparison of can- 22. Augusto PE, Cristianini M. Numerical evaluation of liquid food
ning and retort pouch systems. J Food Sci 1982;47:284–90. heat sterilization in a brick-shaped package. Proc Food Sci
5. Dwivedi M, Ramaswamy HS. Comparison of heat transfer rates 2011;1:1290–4.
during thermal processing under end-over-end and axial 23. Kiziltas S, Erdogdu F, Palazoglu TK. Simulation of heat transfer
modes of rotation. LWT – Food Sci Technol 2010;43:350–60. for solid–liquid food mixtures in cans and model validation
6. Hassan FH, Ramaswamy HS. Bio-validation of bi-axial rotary under pasteurization conditions. J Food Eng 2010;97:449–56.
thermal processing. Food Sci Technol 2013;53:418–25. 24. Lee GM, Yoon BW. Developing an effective method to deter-
7. Ramaswamy HS, Dwivedi M. Effect of process variables on mine the deviation of F value upon the location of a still can
heat-transfer rates to canned particulate Newtonian fluids during convection heating using CFD and subzones. J Food
during free bi-axial rotary processing. Food Bioprocess Process Eng 2014;37:493–505.
Technol 2011;4:61–78. 25. Singh A, Singh AP, Ramaswamy HS. Computational techniques
8. Rattan SN, Ramaswamy HS. Comparison of free/bi-axial, fixed used in heat transfer studies on canned liquid-particulate
axial, end-over-end and static thermal processing effects on mixtures. Trends Food Sci Technol 2015;43:83–103.
process lethality and quality changes in canned potatoes. 26. Armenzoni M, Solari F, Marchini D, Montanari R, Bottani E,
Food Sci Technol 2014;58:150–7. Vignali G, et al. Advanced design of the pasta drying process
9. Singh A, Ramaswamy HS. Effect of product related parameters with simulation tools. Int J Simul Process Modell 2013;8:
on heat-transfer rates to canned particulate non-Newtonian 104–17.

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM
 D. Mosna and G. Vignali: Three-Dimensional CFD Simulation of a Retort Process 729

27. Bottani E, Ferretti G, Manfredi M, Vignali G. Modeling and 32. Menter FR, Ekaterinaris JA. Computation of oscillating air foil
thermo-fluid dynamic simulation of a fresh pasta pasteuriza- flows with one- and two-equation turbulence models. AIAA J
tion process. Int J Food Eng 2013;9:327–39. 1994;32:2359–65.
28. Ambaw A, Verboven P, Delele MA, Defraeye T, Tijskens E, 33. Abdul Ghani AG, Farid MM, Chen XD, Richards P.
Schenk A, et al. CFD-based analysis of 1-MCP distribution in Numerical simulation of natural convection heating of canned
commercial cool store rooms: porous medium model applica- food by computational fluid dynamics. J Food Eng 1999;41:
tion. Food Bioprocess Technol 2014;7:1903–16. 55–64.
29. Bellet D, Sangelin M, Thirriot C. Determination des proprietes 34. Abdul Ghani AG, Farid MM, Chen XD. Numerical simulation of
themophysiques de liquides non-newtniens a l’áide d’une cellule a transient temperature and velocity profiles in a horizontal can
cylindres coaxiaux. Int J Heat Mass Transfer 1975;18:1177–87. during sterilization using computational fluid dynamics. J Food
30. Löhner R. An adaptive finite element scheme for transient Eng 2002;51:77–83.
problems in CFD. Comput Method Appl Mech Eng 35. Abdul Ghani AG, Farid MM, Chen XD, Richards P. Investigation
1987;61:323–38. of deactivation of bacteria in a canned liquid food during
31. ANSYS. CFX-solver modelling guide, release 14.0. Southpointe sterilization using computational fluid dynamics (CFD). J Food
275 Technology Drive, Canonsburg, PA: ANSYS, Inc., 2011. Eng 1999;42:207–14.

Brought to you by | University of California

Authenticated
Download Date | 12/11/15 11:10 AM