International Journal of Thermal Sciences 151 (2020) 106284

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

journal homepage: www.elsevier.com/locate/ijts

Numerical investigation of the fluid flow distribution for the

hydrofluidisation food freezing method
Michal Stebel ∗, Jacek Smolka, Michal Palacz, Wojciech Adamczyk, Edyta Piechnik
Institute of Thermal Technology, Silesian University of Technology, Konarskiego 22, 44-100 Gliwice, Poland

ARTICLE INFO ABSTRACT

Keywords: The food freezing method of hydrofluidisation is based on submerging small food products in an appropriate
Hydrofluidisation water-based solution, which is pumped into the tank through orifices in order to increase the heat transfer
Food freezing between the food products and the liquid by creating agitating jets. The highly turbulent refrigerating
CFD
medium flow, which can be analysed using Computational Fluid Mechanics (CFD), creates significantly higher
PIV
heat transfer coefficients in comparison to other techniques offering effective freezing, e.g. immersion or
Freezing
impingement jet methods. Therefore, rapid freezing of food products may occur. The aim of this study was to
investigate the turbulent refrigerating medium flow within a hydrofluidisation freezing unit. The liquid used for
the process was an aqueous solution of ethanol with a mass concentration of 5%. An axisymmetric CFD model
has been developed to analyse the configuration with a single orifice and a spherical-shaped artificial food
product located at different positions. The parameters being analysed were the orifice diameter (2–5 mm),
sphere diameter (5–30 mm), position of the product above the orifice (20–80 mm) and the refrigerating
medium mass flow rate (0.00267–0.15 kg⋅s−1 ). All cases covered the Reynolds number range of 1700–93,000.
The developed model employed two turbulence models: 𝑘-𝜔 SST and Reynolds stress model (RSM) which
are appropriate for modelling an impinging flow over spherical objects. Validation has been performed using
the PIV technique on a hydrofluidisation freezing unit designed for this study. The results being compared
were the angle of the boundary layer separation and velocity profiles along the impinging jet and around
the spherical object. The results of the CFD simulations confirmed by PIV measurements proved that the
refrigerating medium flow is highly agitating around the samples. In addition, the preliminary results of sphere
shape potato samples during the cooling process have been presented in this paper.

1. Introduction food freezing methods, i.e., impingement jet freezing (IJF) or hy-
drofluidisation freezing (HF), both of which are characterised by high
The food industry has been transformed significantly over the last freezing rates [3].
several decades due to customer requirements. Food products need The issue of damage to food products has been investigated using
to keep their freshness during storage and transport because of both scanning electron microscopy [4]. The authors compared fresh and
biological stability and customer preferences. The most widespread frozen strawberries subjected to differing freezing times. The structure
method used for post-harvest pre-treatment products to achieve a long of the fruits that were frozen slowly was strongly damaged, while
shelf-life is cooling and freezing, which is a basic part of the food cold the second sample was almost unaffected at the micro scale. Addi-
chain [1]. Apart from health and hygiene matters, the market demands tionally, slow food freezing results in protein damage [5] and high
have become crucial; therefore, huge significant development of this moisture loss when thawed [3]. Finally, freezing rates adversely affect
important branch of industry and technology has occurred. the taste, which was mentioned in a comparison of different food
However, during the freezing process, food structures may be af- freezing techniques [6].
fected by the formation of ice crystals within and outside cells [2]. The most common technique used currently is air-blast freezing
It was proven that rapid freezing has a favourable impact on the based on convective heat transfer in continuous or batch systems [6].
biological quality of food products [1]. The small size and uniform Its advantages are the ease of maintenance and flexibility; however, the
distribution of those crystals may be assured by employing emerging freezing times are long due to the low heat transfer coefficients relative

∗ Corresponding author.
E-mail address: [email protected] (M. Stebel).

https://doi.org/10.1016/j.ijthermalsci.2020.106284
Received 14 July 2019; Received in revised form 3 January 2020; Accepted 18 January 2020
Available online 24 January 2020
1290-0729/© 2020 Elsevier Masson SAS. All rights reserved.
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

food is directly submerged in the liquid. Thus, an uncontrolled solute

Nomenclature
uptake by the product may occur as an unwanted effect [8].
Two novel methods that have recently been investigated are air
Roman Symbols fluidisation [9] and IJF, which was investigated for real food prod-
C𝑃 Pressure coefficient ucts [10] and for objects with simpler shapes like spheres [11]. Another
𝑐𝑝 Specific heat capacity, J kg−1 K−1 emerging method is HF freezing, which is the subject of this study. In
all of these methods, the heat transfer rate is much larger due to the
𝐷 Sphere diameter, mm
forced fluid flow. Therefore, the freezing process is relatively fast and
𝑑 Orifice diameter, mm
the quality of the food products can satisfy market needs.
𝑔 Gravitational acceleration, m s−2
Freezing and cooling with the IJF method is widely used in industry
𝐻 Orifice-stagnation point distance, mm because high heat transfer rates may be reached using the highly
𝐻∗ Dimensionless orifice-stagnation point dis- turbulent flow. High velocity jets flowing through nozzles strike the
tance, – surface directly, and therefore, the turbulent flow intensifies the heat
𝑘 Turbulence kinetic energy, m2 s−2 transfer within the boundary layer. This method is characterised by the
𝑘𝑡 Thermal conductivity, W m−1 K−1 highest heat transfer rate among all single-phase arrangements [12].
𝑚̇ Mass flow rate, kg s−1 Food freezing method of hydrofluidisation, which is a subject of
𝑝 Static pressure, Pa this paper, has been patented by Fikiin [13]. Basically, this technique
𝑅 Radial coordinate measured from axis, mm brings together advantages of immersion and impingement methods.
𝑟 Radial distance for spherical coordinates, Strongly turbulent refrigerating medium, e.g. ice slurry, [1,14], an
mm aqueous solution of ethanol and sucrose or glucose [15,16] or NaCl
Re Reynolds number, (𝜌 ⋅ 𝑣𝑜 ⋅ 𝐷 ⋅ 𝜇 −1 ) water solution [17], is pumped upwards through orifices or nozzles into
𝑇 Temperature, K the refrigerating medium tank. This kind of the flow creates agitating
jets inside that tank, where small food products are immersed. During
𝑡 Time, s
the process, the group of food products behaves as a fluidised bed [6].
𝑉 Velocity value, m s−1
The highly turbulent flow over the submerged products increases the
𝑉 ∗ Dimensionless velocity value in reference
convective heat transfer coefficients, which has been confirmed in
to maximum value within considered
several studies.
range, –
Firstly, the method has been compared with other immersion-type
𝑣𝑜 Refrigerating medium mean velocity in methods for the fish chilling and freezing [18]. It was determined that
orifice, m s−1 the freezing times of different kinds of fish using this technique were
y+ Dimensionless wall-cell centre distance, – reduced by half and the heat transfer coefficients were two times or
Greek Symbols almost three times higher comparing with sprinkling and immersion
methods, respectively. The HF system design was proposed in this paper
𝛿𝑖𝑗 Kronecker delta as well. This method has shown its potential; however, there is scarce
𝛿 Relative difference, % literature describing it and currently no commercial manufacturers
𝜃 Polar angle, ◦ offer HF food freezing systems to the best of the authors’ knowledge.
𝜇 Dynamic viscosity, Pa s Further laboratory-scale studies were aimed at measuring the cool-
𝜇𝑇 Turbulent viscosity, Pa s ing process with simpler objects instead of real food products. Verboven
𝜌 Density, kg m−3 et al. performed an experimental study with aluminium spheres instead
𝜙 Azimuthal angle, ◦ of food samples [15]. To describe the heat transfer, they proposed
Nusselt number correlation based on the sphere diameter, fluid flow
𝜔 Specific turbulent dissipation rate, s−1
rate, the temperature of the fluid during the process and the level of
Subscripts fluid agitation. They suggested that the use of numerical methods could
bring a valuable contribution in terms of fluid flow investigations for
0 Initial conditions
the freezing technique. Recent studies were performed to examine the
𝑎𝑥 Axial velocity component
HF method using laboratory-scale measurements performed in similar
𝑟𝑎𝑑 Radial velocity component way together with computational fluid dynamics (CFD). Peralta et al.
Abbreviations presented the experimental work that resulted in Nusselt number cor-
relation taking into account the main geometrical dependencies in the
𝐶𝐹 𝐷 Computational Fluid Dynamics flow within the HF unit [17]. In the next work of the same group,
𝐻𝐹 Hydrofluidisation Freezing a mathematical model of the simplified system, which was in good
𝐼𝐽 𝐹 Impingement Jet Freezing agreement with measurements performed in the previous study, was
𝐼𝐹 Immersion Freezing proposed [19]. That numerical work was complemented with heat and
𝑃 𝐼𝑉 Particle Image Velocimetry mass transfer phenomena occurring within real food products [20].
𝑅𝐴𝑁𝑆 Reynolds–averaged Navier–Stokes Later studies dealt with more complex systems, namely, with numerous
𝑅𝑆𝑀 Reynolds Stress Model products and orifices [21] and an HF numerical model with movement
of the food products and mutual collisions [22]. The latter has been
complemented with a sensitivity analysis of the variables affecting fluid
flow, heat, and mass transfer in an HF unit [23].
to alternative methods [6,7]. Another method being used widely is The most recent paper describing the HF method is an experimental
study of the same system as proposed in the current paper [24]. Palacz
immersion freezing (IF). In this case, food products are directly sub-
et al. used a PIV technique to capture the refrigerating liquid flow
merged in refrigerating medium (water or aqueous solutions) [6]. This
field around the stationary spherical sample. They analysed the velocity
method has both pros and cons. The heat transfer is more intensive in distribution under different Reynolds numbers and for different values
comparison with the aforementioned air-blast freezing; however, the of the orifice-stagnation point distance to the orifice diameter ratio. It

2
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

is the only experimental work focused on the velocity distribution in the inflow plate with the orifice, a steel deflector was placed to stabilise
hydrofluidisation units. the pressure level and the liquid flow under the entire plate that occurs
In this study, a numerical model of the refrigerating liquid flow due to the impact of the fluid stream from the inflow pipe. Each of two
around the food products for the idealised case in the hydrofluidisation side walls of the tank had two rectangular outflow openings, 20 mm
unit has been presented. Using the CFD modelling supported with high and 30 mm wide, located in the top corners as presented in Fig. 1
PIV measurements allowed for a detailed and reliable assessment of at the left side. They allowed the liquid to flow back to the refrigerating
the velocity field in such a system. The motivation of this study was medium reservoir by the gravity. Their location prevents the tank front
to investigate the refrigerating liquid flow for different geometrical wall and the middle part of side walls from getting wet during the
configurations using two turbulence modelling approaches. This is continuous operation of the HF system. This is significant for the PIV
particularly valuable because local quantities have not been measured measurements performing, because the liquid film or droplets would
and validated for HF food freezing systems yet. perturb the laser beam and result in the light reflections.
Spherical food product sizes and positions above the orifice as
well as the orifice diameter are the parameterised variables. Highly 2.2. Objects within the field of the refrigerating medium flow
turbulent refrigerating medium flow has been described employing 𝑘-𝜔
SST turbulence model, which was confirmed to be appropriate for such During the investigation of parameters that affect the distribution
a flow in a similar study [19] and Reynolds stress modelling, which of the fluid flow within the HF system, processes that occur inside food
has not been used for HF system models before. The second approach products may be neglected. Therefore, for the fluid flow study, the food
products have been replaced with a single artificial object, namely,
is more relevant for complex fluid flow analyses [25].
a copper sphere with a 20 mm diameter attached to the movable
The real process of hydrofluidisation food freezing takes place in
aluminium frame. Only preliminary analyses of the cooling process
the tank with more complex setup of orifices. In addition, the ar-
have been performed using the potato sample having also the spherical
rangement of real-shape food products is arbitrary and dynamic due to
shape, which will be discussed further. Samples were hung on a plastic
the process of intensive fluid flow. However, this study is focused on
rod, which assures good thermal insulation and stability of the sphere
the investigation of some parameters identified as important in terms
position. Moreover, the rod does not disturb the refrigerating medium
of the fluid flow pattern. The numerical model validation using PIV
turbulent flow in the sensitive areas, which is important for simplifying
method could be performed for that kind of the arrangement. The
the numerical model and ensuring that it is possible to perform the PIV
results obtained from the study of the reported simplified system can
measurements. The position of the sphere hanging frame is controlled
bring some conclusions and suggestions for further works with more
by two stepper motors. Therefore, the location of samples may be fixed
complex setups, e.g. the appropriate selection of the diameter of the
at an exact place and their movements within the tank may be assured
orifices for the particular size of food products. in both the vertical and horizontal dimensions.

2. Laboratory test rig and measurements 2.3. Refrigerating medium selection

2.1. HF cooling chamber features As mentioned before, different liquids have been used in recent
studies of the HF freezing method. For the fluid flow analyses, a diluted
For this investigation of the HF method using CFD, it is necessary to solution of ethanol was prepared with a mass concentration of 5%.
validate the obtained results with measurements. A laboratory test rig The potato cooling analyses were performed with a different solution
has been designed and built at the Institute of Thermal Technology in concentration, namely 30%. The main thermal requirement for the
Gliwice, Poland. Apart from the refrigerating medium preparation loop, refrigerating medium, apart from the density, the viscosity and the heat
the HF freezing unit consists of the cooling chamber shown in Fig. 1. capacity, is that its freezing temperature be below the initial freezing
The latter was made of ASI 316L stainless steel, thermally insulated point of food products. Therefore, using different solutes is a common
and equipped with transparent windows to make the PIV measurements practice in freezing methods using liquids as a freezing medium. During
possible. The exact process of HF freezing takes place inside it within the process of food product freezing, simultaneous heat and unwanted
a cuboid-shaped tank made of a transparent material, opened from the mass transfer occurs. However, for a rapid freezing method such as HF,
top. The tank is 150 mm high, 445 mm wide and 280 mm deep. The the solute intake is expected to be significantly lower when the freezing
refrigerating medium flows into it through orifices made in a plate starts due to a decrease in the effective diffusion coefficient [20]. The
placed at the bottom of the tank. The plate, made of stainless steel, is effect of mass transfers between the liquid and food products will be
removable to perform tests with different inflow configurations. Under investigated during further stages of this research.

Fig. 1. The refrigerating medium tank inside the HF cooling chamber (middle) with a close-up of the side outflow (left) and the copper sphere and orifice arrangement (right).
The PIV observation field is shown in red frame.

3
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

2.4. Measurement instruments and equipment employed The numerical 3-D domain of the refrigerating medium tank has
been reduced to one-quarter with two vertical planes of symmetry.
A test rig has been equipped with appropriate instrumentation. The placement of boundary conditions for that domain is shown in
For mass flow rate measurements, two Coriolis type Endress+Hauser Fig. 2 on the left side. Instead of modelling the free surface of the
Promass E 300 flow meters were installed with a stated accuracy of liquid, the tank has been closed by the wall from its top, since it is
±0.1% for mass flow rates higher than 5 kg s−1 . A maximum error not affected by the refrigerating fluid flow within the region of interest.
of approximately ±2% is stated for lower flow rates close to zero. The dimensions exactly reflect the test rig described previously. For the
The pressure is measured in the refrigerating medium preparation case being considered, the cooled sphere had a 20 mm diameter and
loop and directly below the plate with orifices by the side of the was placed 50 mm above the plate. A single orifice with a diameter
inflow tank section. Measurement errors from the WIKA S-10 pressure of 5 mm was placed in the centre of the 5 mm thick plate. The orifice
sensors are below ±0.5% of its range. Temperature measurements are had a chamfered edge with an angle of 45◦ and a depth of 0.5 mm at
performed in five places in the refrigerating medium preparation loop the bottom to represent the geometry of the real plate in detail. This
using thermocouples. The location of these sensors along the whole case has been selected as a benchmark for the simplified 2-D domain
loop is as follows: (1) at the side wall of the compartment being verification.
below the refrigerating medium tank, (2) between the refrigerating Due to the tank size, the geometry of the sphere and orifice as
medium reservoir and the pump, (3) between the pump and the mass well as their mutual placement, an axisymmetric simplified domain
flow meter, (4) after the mass flow meter, (5) before the reservoir in could be a proper approach because of the significant decrease in the
the return pipe. Temperature measurements in these places allowed required computational effort. A comparison of results obtained using
to control if the fluid flow is already stabilised and homogenised.
the axisymmetric and 3-D domains is presented in further paragraphs.
Additionally, temperature measurements may be made at the centre
The axisymmetric 2-D domain has been prepared to have the same
of the cooled sphere and in 36 different spots within the refrigerating
configuration as the previously described 3-D domain. It is presented
medium tank if necessary. T-type thermocouples with a maximum error
in Fig. 2 on the right side. The radius of the refrigerating medium tank
of ±1 K were applied for these measurements.
was 200 mm. This value was determined with the assumption of a fixed
The fluid flow distribution of the refrigerating medium within the
tank volume. The diameter of the sphere and the distance between the
tank was visualised using a PIV method using the Dantec Flow Sense
inflow plate and the stagnation point at the bottom of the sphere are
4M MK4 II system equipped with a high-resolution camera and a
parameterised dimensions. Another parameterised dimension for this
twin pulse laser. The camera has a CCD sensor with a resolution of
study was the orifice diameter. The numerical model makes it possible
2048 × 2048 and a pixel size of 7.3 μm. Additionally, a Carl Zeiss APO-
to investigate those values in the range of 20–80 mm for the distance
Sonnar T* 135 mm f/2 lens with a magnification of 4 was used. The
between the inflow plate and the stagnation point of the sphere, 2–
camera was placed facing the front side of the chamber. The field of
observation is a rectangle area around the sphere, 63 mm high and 5 mm for the single orifice diameter and 5–30 mm for the sphere
wide. The laser illuminates that section from the side. It is a Quantel diameter.
Twin BSL 200 with a wavelength of 532 nm and an energy output of For the fluid flow study, the artificial food product has been ex-
200 mJ. For the PIV measurements 5 μm polyamide particles were used cluded from the domain because only the refrigerating medium be-
as a seed material. haviour has been investigated. However, including of the food product
was required for the further preliminary analyses of the cooling process.
3. Mathematical modelling and computing procedure In addition, the supporting rod that maintains the sphere position has
been excluded from the computational domain, like in the numerical
3.1. Computational domain model proposed by Peralta et al. [19], because it has no practical effect
on the freezing process and the impingement jet flow. Nonetheless,
The numerical domain includes the tank with the refrigerating the location of the rod has been marked on the scheme in Fig. 2.
medium, the inflow pipe and the section below the orifice plate with Furthermore, the authors of the mentioned paper presented the distri-
the deflector placed inside, as shown in Fig. 2, which presents the 3- bution of the heat transfer coefficient at the sphere surface. It proved,
D domain containing one-fourth of the real system and a reduced 2-D that the heat transfer occurs mainly in the area below the separation
axisymmetric domain. point. Therefore, the area of recirculation is less significant in terms

Fig. 2. Numerical domain of the 3-D variant, including one-quarter of the real geometry (left side) and a 2-D axisymmetric variant (right side) with all dimensions indicated in
mm.

4
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

of the heat transfer. In the unit presented in this study, the rod is 1.5 mm in most significant areas, i.e., within the jet flow region,
situated far above the boundary layer separation point. Hence, the around the sphere or near the orifice. Mesh independence was verified
perturbation of the flow occurs only in the area of recirculation, which comparing the velocity profiles. Further mesh refinements do not cause
was not analysed in comparison with the measurements. The results significant changes in the refrigerating medium flow distribution.
of preliminary analyses proved, that in the area being compared with Numerical meshes for 2-D axisymmetric simulations were generated
PIV measurements, flow field distribution was not affected due to automatically using the CoolFoodPL tool, which makes it possible to
supporting rod removing. The velocity has been verified in several generate the geometry and numerical mesh based on input parameters,
points around the sphere. The average differences between the velocity namely, the orifice diameter, the sphere diameter and the distance
magnitude for the cases with and without the rod were in order of between the inflow plate and the sphere stagnation point. Each mesh
1%. Maximum difference was 3% and was reported for the point in contains approximately 130,000 quadrilateral elements (depending on
the region of the flow recirculation. the geometry configuration). For that domain, the mesh was arranged
to keep the same grid structure, nodes distribution and cell size for
3.2. Computational procedure the 3-D mesh. Fig. 4 presents an axisymmetric mesh generated for a
sample case for the fluid flow field analysis with details in significant
The procedure of numerical processing and data generation has areas. A mesh independence study has been performed comparing a
been arranged using the CoolFoodPL tool developed for this research. It few axial and radial velocity profiles, for example along the jet axis.
makes it possible to perform numerous serial simulations by automating The thickness of the first cell layer around the sphere was set to keep
the process of geometry and mesh generation, running CFD simula- the 𝑦+ value far below 1.0 for each case being calculated. In this case,
tions and generating results. The scheme of the operations sequence the grid complexity was reduced and the quality of cells was improved.
is presented in Fig. 3. The commercial software used for the process is
ICEM CFD and ANSYS Fluent. The input data describing the operating 3.4. Governing equations
conditions and geometry are processed, and based on that, a numerical
mesh is generated to perform numerical simulations and finally to Fluid flow using the finite volume method was solved by a set of
generate results. equations. They form so-called governing equations used to describe
To solve the mathematical model in ANSYS Fluent software, the incompressible flow in terms of mass and momentum conservation in
COUPLED solver has been used for the pressure–velocity coupling for a steady-state or transient flows. The continuum and momentum equa-
the steady-state computations of the fluid flow field. It turned out to tions are solved using Reynolds–averaged equations. These equations,
be stable and the most efficient one. All the equations were discretised given in Eqs. (1) and (2), respectively, include accumulative terms,
using second-order upwind schemes. To guarantee the stability of each being neglected during steady-state fluid flow analyses. Moreover,
simulation, the first 100 iterations were performed with lowered values the time-dependent cooling analyses required an additional energy
of the under-relaxation factors. equation, presented in Eq. (3).
Each solution was considered to be converged when the normalised • Continuity
residuals were less than 1 ⋅ 10−5 for the continuity, both components
of the velocity and the turbulence quantities and less than 1 ⋅ 107 for
𝜕𝜌 𝜕 ( )
+ 𝜌𝑢𝑖 = 0 (1)
𝜕𝑡 𝜕𝑥𝑖
the energy equation. Moreover, the local values of the velocity were
monitored in several points selected arbitrarily located less than 10 mm • Momentum
from the sphere surface. [ ( )]
𝜕 ( ) 𝜕 ( ) 𝜕𝑝 𝜕 𝜕𝑢𝑖 𝜕𝑢𝑗 2 𝜕𝑢
Apart from the steady-state analyses of the fluid flow, which are 𝜌𝑢𝑖 + 𝜌𝑢𝑖 𝑢𝑗 = − + 𝜇 + − 𝛿𝑖𝑗 𝑘
𝜕𝑡 𝜕𝑥𝑗 𝜕𝑥𝑖 𝜕𝑥𝑗 𝜕𝑥𝑗 𝜕𝑥𝑖 3 𝜕𝑥𝑘
the main results of this study, the time-dependent simulations of the ( )
𝜕 ′ ′
cooling process were performed. For these cases, the cooling process + −𝜌𝑢𝑖 𝑢𝑗 + 𝜌𝑔𝑗 (2)
𝜕𝑥𝑗
was solved after the flow field has been established during steady-state
computations. • Energy
[ ]
𝜕𝑇 𝜕𝑇 𝜕 𝑘𝑡 𝜕𝑇
3.3. Numerical mesh 𝜌 + 𝜌𝑢𝑖 = − 𝜌𝑇 ′ 𝑢′𝑖 (3)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖 𝑐𝑝 𝜕𝑥𝑖
The numerical 3-D mesh containing one-fourth of the refrigerating where 𝜌 is the density, 𝑡 is the time, 𝑢𝑖 is the average velocity vector
medium tank consists of over 8,000,000 hexahedronal elements, which component, 𝑢′𝑖 is the fluctuating velocity component, 𝑝 is the average
leads to time-consuming calculations. This grid was prepared using pressure, 𝜇 is the dynamic viscosity, 𝛿𝑖𝑗 is the Kronecker delta and
a blocking method, which resulted in a structured topology around 𝑔𝑗 is the gravitational acceleration vector component, 𝑇 is the aver-
curved surfaces like the sphere and cylindrical orifice. Within these age temperature, 𝑇 ′ is the fluctuating temperature, 𝑘𝑡 is the thermal
areas, o-grid mesh structures were applied. Cells sizes do not exceed conductivity and 𝑐𝑝 is the specific heat capacity.

Fig. 3. Scheme of the operations performed for automatic serial computations.

5
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

Fig. 4. Details of the 2-D axisymmetric numerical mesh with nearly 130,000 elements presenting quadrilateral elements near the sphere surface (A) and within the orifice area
(B) for the case described by the geometrical parameters: 𝐷 = 20 mm, 𝐻 = 50 mm, 𝑑 = 5 mm.

The 𝑘-𝜔 SST turbulence model solves two additional equations most complex among all classical turbulence models and solves many
i.e. for the turbulence kinetic energy and the specific dissipation rate: simple and complex flows with high accuracy [25].
The first model is widely used for wall-bounded flows. This model
• Turbulence kinetic energy
[( ) ] solves two additional equations for the turbulence kinetic energy and
𝜕 ( ) 𝜕 𝜇 𝜕𝑘
𝜌𝑘𝑢𝑖 = 𝜇+ 𝑇 + 𝐺𝑘 − 𝑌𝑘 (4) the specific dissipation rate [27]. It has proven excellent performance
𝜕𝑥𝑖 𝜕𝑥𝑗 𝜎𝑘 𝜕𝑥𝑗 with the CFD model of the HF system developed by Peralta et al. [19].
• Specific dissipation rate A similar study, namely, a numerical analysis of food cooling using
[( ) ] a single jet, was performed, again using the 𝑘-𝜔 SST model [11]. In
𝜕 ( ) 𝜕 𝜇 𝜕𝜔
𝜌𝜔𝑢𝑖 = 𝜇+ 𝑇 + 𝐺𝜔 − 𝑌𝜔 + 𝐷𝜔 (5) that case, the air jet impinging the cylindrical surface was modelled
𝜕𝑥𝑖 𝜕𝑥𝑗 𝜎𝜔 𝜕𝑥𝑗 with good accuracy. Yet, more complex arrangements with multiple
where 𝑘 is turbulence kinetic energy, 𝜇𝑇 is the turbulent viscosity, 𝜎𝑘 spherical samples and inlets were analysed in a work by Olsson et al.
is the turbulent Prandtl number for 𝑘, 𝐺𝑘 is the generation of 𝑘 due to using the same turbulence model [26]. In that paper, the heat transfer
mean velocity gradients 𝑌𝑘 is dissipation of 𝑘 due to turbulence, 𝜔 is coefficient distribution was determined with good agreement with
the specific dissipation rate, 𝜎𝜔 is the turbulent Prandtl number for 𝜔, experimental data. The further development of HF numerical models
𝐺𝜔 is the generation of 𝜔, 𝑌𝜔 is the dissipation of 𝜔 due to turbulence, described previously proved that the 𝑘-𝜔 SST is a good approach for
and 𝐷𝜔 is the cross-diffusion term. more complex systems involving multiple products and jets [21] and
The effects of turbulence are represented by the term −𝜌𝑢′𝑖 𝑢′𝑗 in even sphere movement within the refrigerating medium tank [22].
Eq. (2). These are the Reynold stresses. The 𝑘-𝜔 SST turbulence model However, it also shows some drawbacks, like an overprediction of
solves Reynolds stresses using the Boussinesq approximation according the heat transfer rate [26] and a turbulence intensity in some areas
to Eq. (6). The turbulent viscosity is computed in this approach using mentioned in the paper by Peralta et al. [19].
a blending function as shown in Eq. (7).
( ) Another turbulence model employed in this study was the RSM
𝜕𝑈𝑖 𝜕𝑈𝑗 2 that is known as the most complex among all RANS approaches. It
−𝜌𝑢′𝑖 𝑢′𝑗 = 𝜇𝑇 + − 𝑘𝛿𝑖𝑗 (6)
𝜕𝑥𝑗 𝜕𝑥𝑖 3 does not assume isotropic eddy viscosity but solves additional trans-
port equations for the Reynolds stresses, with one quantity providing
𝜌𝑘 the turbulence length or time scale, e.g., dissipation [28]. The study
𝜇𝑇 = ( ) (7)
1 𝑆𝐹2 comparing several turbulence models for impingement flow proved that
𝑚𝑎𝑥 ;
𝛼 ∗ 𝛼1 𝜔
this model may give more accurate results for the jet region’s velocity
where 𝛼 ∗ is the constant for high-Reynolds formulation, 𝑆 is the strain than other models [29]. Olsson et al. confirmed that RSM predicts heat
rate magnitude, 𝐹2 is the blending function for this model and 𝑎1 is the transfer better than 𝑘-𝜔 SST within the recirculation region [30].
model constant. Constants 𝛼 ∗ = 1 and 𝛼1 = 0.31
The numerous studies mentioned above are based on turbulence
For the Reynolds Stress Model, the Boussinesq approximation is not
modelling in systems close to HF in terms of flow pattern. All those
valid because all the stresses are solved separately. As described in
studies, together with the comparison of turbulence numerical mod-
detail in Section 4.2, this model has finally been rejected due to its
elling and PIV measurements [31], indicate the necessity of inves-
accuracy comparing with 𝑘-𝜔 SST for this particular case. For that rea-
tigating this novel food freezing method using different turbulence
son, a detailed mathematical description of this turbulence modelling
approach is not provided in that paper. modelling approaches.

3.5. Turbulence modelling 3.6. Fluid physical properties

The intensified freezing rate during the HF process is a result of

turbulent fluid flow over the food product’s surface. Hence, the proper The refrigerating medium was assumed to be a Newtonian fluid. All
modelling of such a flow is fundamental for the numerical model properties required for modelling the fluid flow within the HF system
development of the HF system. Likewise, in impingement flows, Olsson were defined using the CoolPack software, wherein the water-ethanol
et al. pointed out the jet region, the stagnation zone and the wake properties are based on the VDI Heat Atlas [32]. Fluid thermo-physical
zone behind the frozen object [26]. In this study, the 𝑘-𝜔 SST and properties, namely, the density, the thermal conductivity, the specific
RSM models were used. The first one combines the advantages of heat and the dynamic viscosity, were computed using polynomial func-
Wilcox’s original 𝑘-𝜔 and the widely used 𝑘-𝜖 two-equation models, tions of temperature that are presented in Table 1 for constant ethanol
while the second, also called the second-moment-closure model, is the concentrations.

6
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

Table 1 • The inlet boundary condition is based on the liquid mass flow rate
Polynomial functions of temperature in Kelvin describing the thermo-physical properties
value. Defined quantities are the mass flow rate specified for each
for a water-ethanol solution of 5% and 30% concentration.
case, the temperature of the refrigerating medium as 0 ◦ C and
Property 5% mass concentration 30% mass concentration
the turbulent intensity, estimated at 5% for this kind of boundary
Density (kg m−3 ) −0.0076019 ⋅ 𝑇 2 + 4.1835 ⋅ −0.0026348 ⋅ 𝑇 2 + 0.95759 ⋅
condition arrangement [36].
𝑇 + 415.60 𝑇 + 899.74
Thermal conductivity 0.0016005 ⋅ 𝑇 + 0.94434 0.00063878 ⋅ 𝑇 + 0.22316 • The outlet boundary condition is defined by the static pressure
(W m−1 K−1 ) value of 105 Pa, which corresponds to the ambient pressure.
Specific heat −0.000014411 ⋅ 𝑇 2 + −0.000044944 ⋅ 𝑇 2 + • In terms of boundary conditions, the top wall was assumed as slip
(J kg−1 K−1 ) 0.012460 ⋅ 𝑇 + 6.6587 0.028190 ⋅ 𝑇 − 0.18395 wall, because it is a representation of the liquid free surface. All
Dynamic viscosity 0.0042965 ⋅ 𝑇 2 − 2.4563 ⋅ 𝑇 + 0.016267 ⋅ 𝑇 2 − 9.4377 ⋅ 𝑇 +
(10−3 kg m−1 s−1 ) 352.83 1370.9
the other walls are defined as no-slip, adiabatic walls.
• Symmetry within the HF unit was assured by symmetry planes in
the case of the 3-D domain and by an axis for the 2-D axisymmet-
Table 2 ric case.
Thermo-physical properties of unfrozen potato sample.
Property Value
4. Numerical model verification and validation
Density (kg m−3 ) 1062
Thermal conductivity (W m−1 K−1 ) 0.485 [33]
4.1. Verification
Specific heat (J kg−1 K−1 ) 3670 [34]; 3450 [35]

4.1.1. Domain dimensionality analysis

The laboratory scale HF system designed by Peralta et al. [17],
3.7. Physical properties of potato sample and the numerical model of that unit developed later [19], consists
of a cylindrical-shaped tank that can be easily reduced to its parts,
For the preliminary cooling analyses with spherical-shaped potato e.g., a 60◦ section due to the axial symmetry of that shape. This type
samples, the properties of this product also had to be evaluated. For a of tank shape also assures more effective freezing since there are no
potato with a temperature being above its initial freezing temperature, areas such as corners where a decreased agitation level of the liquid
i.e. about −1.8 ◦ C, the phase change does not occur. Therefore, the flow occurs [15].
properties can be assumed as constant. It is worth to mention that when However, PIV measurements would be difficult to perform in a
the phase change is progressing, the sample temperature is decreasing, refrigerating medium tank of such shape. Therefore, for this study,
while the ice content gradually rises. It has a crucial effect on all the a rectangle-shaped tank has been designed to make those measure-
properties. For that reason, evaluation of thermal properties of frozen ments possible. Furthermore, this configuration of the tank enables a
and unfrozen foods is a challenging task. Table 2 presents the properties numerical domain reduction into its 90◦ section.
assumed for this study. The density has been evaluated by measuring Further reductions of the rectangle-shaped tank to its axisymmetric
the mass and the volume of the group of over 20 potato samples. The
simplification may be performed only if the flow pattern is not affected
thermal conductivity and the specific heat of unfrozen potato have been
by the shape of the tank. In this study, the orifice and sphere are located
found in the literature. Because the literature presents different values
at the centre of the tank. Additionally, the field of interest includes only
of the specific heat, this paper covers the results obtained by using two
the central area of the tank. For that reason, a simplified axisymmetric
different values of this quantity.
domain has been prepared, as described before.
3.8. Boundary conditions To verify if the flow is dimension-independent, values of the re-
frigerating medium velocity were compared at several points located
Both the 3-D and 2-D axisymmetric numerical domains are de- 2 mm or 4 mm around the sphere surface. The location of these points
scribed using several boundary conditions, including the axis of sym- and their velocity values are shown in Fig. 5. The origin has been
metry, as well as the inlet, outlet and wall boundary conditions in order set as the centre point of the sphere. The polar angle 𝜙 is measured
to obtain a complete definition of the mathematical model. The inlet, starting from the stagnation point placed at the bottom of the sphere
outlet and spherical wall are indicated in Fig. 2 using captions. The and increases along the flow direction. The comparison was performed
symmetry planes are shown as grey surfaces on the 3-D domain scheme, for ten groups (P1–P10) of points within the 3-D domain differing only
while the axis is marked using a green line on the 2-D axisymmetric by the azimuthal angle 𝜃 together with a corresponding point from the
domain scheme. The remaining surfaces are the inflow pipe, the de- 2-D axisymmetric domain having the same polar angle. Due to the as-
flector, orifice edges and walls of both chambers, which are described sumption that the refrigerating medium flow is symmetrical around the
as adiabatic walls. axis, the azimuthal position around the sphere should not significantly

Fig. 5. Velocity values in m s−1 for groups of corresponding points for the 𝑘-𝜔 SST (left chart) and the RSM (right chart) models with explanations of their placement.

7
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

affect the flow, so the velocity values should be approximately the same of the orifice that were compared for model verification. Variants A–
for all points within each group. C are simplified cases with inlet boundary conditions being used with
Fig. 5 compares the velocity values within point groups P1–P10 a uniform velocity value along the boundary. The difference in these
located around the sphere for both the 𝑘-𝜔 SST and RSM turbulence cases is its location. For Variants A and B, this boundary condition was
models. As seen in the figure, the velocity values match with very good defined at the top or bottom of the inflow plate. Variant C includes an
agreement for the 𝑘-𝜔 SST model. The maximum variance of velocity additional section for developing the velocity profile. Other options that
magnitudes equals 0.04 m s−1 for group P1 for that model. The velocity were compared, namely, Variants D and E, include the fluid domain
values from the 2-D axisymmetric domain are the same as those from below the orifice. The orifice in Variant D is a simplified cylindrical
the 3-D domain. Therefore, it can be confirmed that the physical opening, while Variant E describes the geometry in detail with all
behaviour of the flow allows the domain to be reduced into a 2-D chamfers.
axisymmetric view. However, for the RSM turbulence model, variations The velocity profiles along the axis from the orifice level to the
in the values within several groups of points, especially P1 and P6, stagnation point for all described Variants are shown in Fig. 7. For each
are significantly higher and reach up to 0.54 m s−1 for the P6 group. case, the velocity in the jet core remains stable until reaching a height
This may be a result of anisotropic turbulence formulation, which is a of 25 mm. Only Variant C results in a smooth drop in the velocity,
characteristic feature for that particular numerical approach [25]. This which may be an effect of the jet profile formation below the orifice.
suggests that this model might be more adequate for 3-D analyses. Similar behaviour may be observed in experimental investigations of
fully developed impingement jets performed by Lee et al. [38] or in
a numerical study by Peralta et al. [19], where the authors assumed
4.1.2. Mesh sensitivity study
the 1/7th power velocity profile. The velocity values for Variant D are
As mentioned before, the numerical mesh sensitivity study has
significantly higher than those of the others. This is caused by the sharp
been performed comparing some velocity profiles. Fig. 6 presents the
orifice edge at the bottom. The fluid flowing through the orifice was
velocity profile along the jet axis (the left chart) and the velocity profile
stagnated near its wall, so the velocity in the jet core increases. For
in the radial direction from the sphere surface (the right chart). Five
Variant E, this was not observed. The flow had the same behaviour as
different meshes were generated for the case with the sphere diameter
for Variants A and B in that area. The velocity values are the same for
of 20 mm, orifice diameter of 5 mm and 50 mm in the distance between
those three variants until it starts to drop after reaching a distance of
the orifice and the stagnation point. They consisted of approximately
approximately 30 mm. Variant B tends to show similar behaviour as
38, 57, 83, 125 and 183 thousands of hexagonal cells. The mass flow
the one with a developed velocity profile.
rate for the grid sensitivity study was set as 0.15 kg s−1 , which was a
This comparison suggested that the proposed Variants A–D give
maximum value of a range being studied.
more or less different velocity profiles than the most detailed variant,
As seen, compared profiles obtained with the mesh having approx-
namely, E. For that reason, the numerical model of the HF system
imately 125,000 elements were roughly the same as the ones obtained
presented in this study includes Variant E.
with the finest mesh. For that reason, it has been selected as appropriate
for this numerical model. The thickness of the first cell layer around the 4.2. Validation
sphere was set to keep the 𝑦+ value far below 1.0 for every considered
case. The numerical model validation has been carried out on the labo-
ratory test rig described previously. Measurements of the inflow mass
4.1.3. Orifice arrangement flow rate and the temperature of the refrigerating medium within the
The orifice being used to create the fluid jet may be represented in loop were used to control the operating conditions. PIV recordings were
two general ways within the domain. The first way is to define it as an directly used to compare the velocity within the jet region and around
inlet boundary condition using the proper velocity profile, which was the sphere. At this stage of the experimental study, the PIV fluid flow
done in the numerical model of an HF system developed by Peralta distribution analysis was conducted at ambient temperature. Therefore,
et al. [19]. Another solution is to include the orifice together with the heat transfer during that investigation was neglected.
the entire inflow section of the domain. This decision is significant, The full description of the experimental study in the same HF
because in such a flow, the orifice design may affect the shear layer system is presented in the separate paper [24]. This study describes the
behaviour at the jet exit and the jet evolution as a result. That issue velocity distribution using PIV method in detail. Measurements were
was described in the experimental comparison of two jets, where the performed during the continuous process of the fluid flow inside the
first one was formed by a smooth contraction nozzle and the second test rig in a steady-state. Each PIV recording was executed during the
was created by the flow in a long pipe, where the flow profile was period of approximately 30 s when 200 double frames were captured.
fully developed [37]. Fig. 7 contains a description of all arrangements The recordings were performed at least three times for each case being

Fig. 6. The velocity profiles compared for the grid sensitivity study along the jet axis (left) and in the radial direction from the sphere surface (right).

8
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

Fig. 7. Comparison of the velocity profiles in m s−1 along the axis for different orifice arrangements with their geometrical description and dimensions indicated in mm.

considered. The collected data were analysed using Dantec Dynamic- Table 3
Operational conditions for the cases used for validation.
Studio software. Several procedures were executed in order to improve
the quality of the recorded snapshots and the steady-state velocity Test Case 𝑑 (mm) 𝐻 (mm) 𝑇 (◦ C) 𝑚̇ (kg s−1 )

analyses. Firstly, the Sobel filter was used to increase the contrast, 1 3 50 17.05 0.04086
2 3 60 17.05 0.04086
which improved the seeding particles detection. Then, the average
3 3 70 17.05 0.04086
correlation was applied to obtain a single scalar map of the refrigerating 4 3 50 17.05 0.05162
medium velocity field. The averaging is beneficial for this analysis, 5 3 60 17.05 0.05162
because some random defects can be eliminated, e.g. air bubbles flow 6 3 70 17.05 0.05162
within the field of observation, laser light reflections, occurrence of the 7 5 50 16.73 0.1144
8 5 60 16.74 0.1132
seeding particle with the size significantly different from the median 9 5 70 16.74 0.1132
value. For that reason, average correlation method increases the fidelity
of the steady-state analysis. In addition, to investigate the fluctuations
of the velocity, the recordings were analysed using adaptive correlation
method. Values of velocity fluctuations were significant, especially in refrigerating medium local velocity within regions close to the liquid
the area close to the sphere surface. Random defects caused by the air free surface in the tank were affected by the free surface waving. It
bubbles, seeding particles or the light reflections affected the velocity was caused by the liquid intensive flow from the orifice. Therefore,
fluctuations, because they could not be eliminated using that method. measured values at Points 1–3 are most reliable.
Nevertheless, the general trends of the velocity fluctuations were used The comparison of results from both turbulence models with the PIV
to analyse the results in terms of this quantity. results for Point 4 again shows the tendency that values obtained using
The model validation is based on the measurements performed the 𝑘-𝜔 SST model were in better agreement with measured velocity
at the different operating conditions presented in Table 3. All the values than the RSM model. The mean relative difference for all test
numerical results used for the validation of the model were obtained cases was equal to 11.6% and 26.7% for those turbulence models,
for the same operational conditions as described measurements. The respectively. It is worth mentioning that this point could be mostly
validation covers a comparison of the refrigerating medium velocities affected by the sphere and the laser beam position, since its distance
at 6 points, for which the locations are presented in Fig. 8. The same from the sphere surface was the smallest among all considered points
figure presents the measured contours of the fluid velocity distribution and the velocity gradients are very high near the wall.
that will be described in detail in further paragraphs. Points 1–3 are Two points located above the area of fluid detachment from the
located along the jet axis, 3 mm, 4 mm and 5 mm below the stagnation sphere surface are Points 5 and 6. It is seen that the values obtained
point, respectively. Point 4 is located by the left side of the sphere, at using the 𝑘-𝜔 SST model are in good agreement with the PIV results
the same height as its centre but 1 mm away from its surface. Points 5 for Point 5, except for Test Cases 1 and 9. The mean value of the
and 6 are located 3 mm and 6 mm exactly above the previous point, relative difference excluding them is equal to 4.1%. The RSM results do
respectively. Table 4 presents a set of velocity values obtained from all not show such a tendency; the relative differences for that model are
9 measurement configurations as well as relative differences comparing equal to 23.0% and 22.9%, including and excluding the mentioned test
the CFD results with the PIV measurements, as defined in (8). cases, respectively. However, the comparison for Point 6 shows that the
results for both turbulence models differ from the PIV measurements.
| 𝑣𝑃 𝐼𝑉 − 𝑣𝐶𝐹 𝐷 |
𝛿% = || |
| (8) In addition, the fluid velocity contours from both turbulence models
| 0.5 ⋅ (𝑣𝑃 𝐼𝑉 + 𝑣 𝐶𝐹 𝐷 |
)
were compared with two test cases, namely, 5 and 9. That comparison
As seen in Table 4, relative differences between measurements and is presented in Fig. 8. The field of observation for these velocity
CFD results are generally clearly higher for RSM model than for 𝑘-𝜔 contours is approximately 62 mm high and 42 mm wide. It includes
SST. For Points 1–3, the mean value of relative difference for all the the closest area of the sphere, partially with the fluid jet created by
measurements is equal to 16.3% for RSM model and 11.5% for 𝑘-𝜔 the orifice (nearly 30 mm of the area below the stagnation point is
SST. It can be noticed that relative difference for Test Cases 3 and 6 at presented in the contours). It is seen that the values of velocity and
Point 2 exceeds 60%. The mean value of relative difference in that point the thickness of the jet core are closer to the measurements for the 𝑘-
excluding mentioned measurements is equal to 4.6% for the 𝑘-𝜔 SST 𝜔 SST model. In addition, the refrigerating medium flow within the
model. In addition, the Test Case 6 presents high divergence for Point region above the sphere is simulated much better using that model,
3 comparing with the results obtained from the 𝑘-𝜔 SST model. Points which confirms that the 𝑘-𝜔 SST model performs well for flows with the
1–3 were located in the area below the sphere. The measurements of the stream detachment from the wall [25]. The RSM model predicts a more

9
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

Table 4
Comparison of the velocity magnitude values in m s−1 with the values of the relative difference for both turbulence models at 6 points indicated in Fig. 8 for the operating
conditions described in Table 3.
Point Test Case 1 Test Case 2 Test Case 3
𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀 𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀 𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀
1 1.57 ± 0.26 12.4% 28.2% 1.56 ± 0.06 15.0% 0.4% 1.08 ± 0.05 1.2% 2.6%
2 1.91 ± 0.05 3.0% 18.8% 1.64 ± 0.06 8.0% 6.7% 1.16 ± 0.03 60.6% 8.5%
3 1.98 ± 0.07 6.2% 21.8% 1.72 ± 0.04 6.1% 8.7% 1.21 ± 0.04 7.5% 11.9%
4 0.59 ± 0.06 12.6% 35.5% 0.70 ± 0.07 1.6% 35.3% 0.57 ± 0.04 0.0% 2.1%
5 0.73 ± 0.13 12.6% 35.1% 0.81 ± 0.10 1.6% 27.1% 0.75 ± 0.10 0.0% 8.9%
6 0.80 ± 0.15 18.0% 12.4% 0.85 ± 0.14 6.9% 5.8% 0.92 ± 0.11 5.1% 6.1%
Point Test Case 4 Test Case 5 Test Case 6
𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀 𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀 𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀
1 2.07 ± 0.31 7.8% 21.8% 1.89 ± 0.13 11.2% 2.6% 1.20 ± 0.07 10.8% 14.2%
2 2.43 ± 0.17 2.2% 16.3% 2.08 ± 0.09 9.6% 0.7% 1.25 ± 0.04 74.1% 22.6%
3 2.46 ± 0.13 7.5% 21.4% 2.16 ± 0.08 6.4% 2.8% 1.31 ± 0.10 21.6% 25.5%
4 1.08 ± 0.10 0.7% 64.8% 0.81 ± 0.06 5.9% 24.1% 0.72 ± 0.10 4.2% 1.9%
5 1.04 ± 0.49 0.7% 41.4% 1.05 ± 0.13 5.9% 26.8% 0.91 ± 0.14 4.2% 2.9%
6 1.00 ± 0.65 18.2% 7.2% 1.15 ± 0.18 0.6% 9.9% 1.13 ± 0.18 3.4% 3.0%
Point Test Case 7 Test Case 8 Test Case 9
𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀 𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀 𝑣𝑃 𝐼𝑉 (m s−1 ) 𝛿%𝑘−𝜔 𝛿%𝑅𝑆𝑀
1 4.33 ± 0.20 11.9% 18.8% 2.58 ± 0.20 5.9% 28.4% 1.96 ± 0.20 7.7% 30.1%
2 4.48 ± 0.23 6.2% 21.8% 2.96 ± 0.18 2.6% 24.0% 2.36 ± 0.04 0.5% 22.1%
3 4.52 ± 0.27 1.0% 25.0% 3.32 ± 0.14 2.3% 18.3% 2.55 ± 0.17 0.2% 21.1%
4 1.39 ± 0.42 7.8% 50.9% 1.06 ± 0.12 8.3% 18.2% 1.04 ± 0.18 24.0% 7.5%
5 1.43 ± 0.66 7.8% 34.4% 1.34 ± 0.21 8.3% 19.7% 1.09 ± 0.15 24.0% 9.7%
6 1.47 ± 0.56 19.0% 8.8% 1.56 ± 0.33 7.2% 8.2% 1.23 ± 0.21 26.1% 21.7%

Fig. 8. The velocity contours of the jet and of the area around the sphere in m s−1 obtained using the 𝑘-𝜔 SST or the RSM turbulence models compared with the PIV results for
Test Case 5 (left pair of contours) and Test Case 9 (right pair of contours). The location of the PIV observation region is presented in Fig. 1.

dispersed flow in that area. In Fig. 9, the velocity values were compared exactly 5 mm above Profile B. In that case, the PIV and the CFD
along the three characteristic Profiles A, B and C for the previously results are different, especially close to the sphere surface. Within
discussed Test Cases 5 and 9. Those values along Profile A in the jet that area, the measurements were most sensitive for all the factors
axis confirm that the 𝑘-𝜔 model predicts the refrigerating medium flow mentioned before. However, the shape of the velocity profile starts to
within that region with high accuracy. Depending on the case, the be comparable 2 mm from the sphere surface.
simulated profiles start to diverge from the measured values around To summarise, the numerical model predicts the refrigerating
12–20 mm below the stagnation point. However, those differences are medium velocity with satisfactory accuracy for both turbulence models
not significant from the point of view of the HF freezing method. within the points and profiles being examined. However, the velocity
The velocity profiles obtained using the RSM model show that this values obtained using the 𝑘-𝜔 SST model are more accurate than the
model overpredicts the velocity within the area of the jet core. Similar RSM in comparison with the PIV measurements, especially within
situation was described by Shademan et al. [29]. They compared the jet the jet region, which affects the liquid flow over a spherical object.
centreline velocity values simulated using 𝑘-𝜔 SST and RSM turbulence Moreover, the shape of the jet obtained using the 𝑘-𝜔 SST and presented
models with the measurements performed by Rajaratnam et al. [39]. in Fig. 8 suggests that this model performs better for the flow within
In that comparison, 𝑘-𝜔 SST predicted the velocity profile more accu- such a system. For this reason, this turbulence model has been selected
rately. Also in their study, the measured velocity values were lower for the presentation of results in further paragraphs.
than the numerical results in some distance below the stagnation point.
The authors indicated, that the nozzle geometry could be a reason of 5. Results
this discrepancy.
Velocity comparisons along Profile B, located in the radial direction 5.1. Investigated configurations
starting from the stagnation point, show that both turbulence models
accurately predict the tendency of the velocity to decline along this The range of conditions being studied using numerical models em-
profile. Fig. 9 also presents velocity values along Profile C located ploying the 𝑘-𝜔 SST turbulence model covers different configurations of

10
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

Table 5
Parameters of the considered cases.
Case number 𝑑 (mm) 𝐷 (mm) 𝐻 (mm) 𝑚̇ (kg s−1 ) 𝑣𝑜 (m s−1 ) Re
Case 0 5 20 50 0.08333 4.28 34,365
Case 1 5 20 50 0.05000 2.57 20,619
Case 2 5 20 50 0.11673 5.99 48,112
Case 3 5 20 50 0.1500 7.71 61,858
Case 4 5 5 50 0.083330 7.71 15,464
Case 5 5 10 50 0.08333 4.28 17,183
Case 6 5 15 50 0.08333 4.28 25,774
Case 7 5 25 50 0.08333 4.28 42,957
Case 8 5 30 50 0.08333 4.28 51,548
Case 9 5 20 20 0.08333 4.28 34,365
Case 10 5 20 40 0.08333 4.28 34,365
Case 11 5 20 60 0.08333 4.28 34,365
Case 12 5 20 80 0.08333 4.28 34,365
Case 13 2 20 50 0.01333 4.28 34,365
Case 14 3 20 50 0.03000 4.28 34,365
Case 15 4 20 50 0.05333 4.28 34,365
Case 16 3 20 30 0.03000 4.28 34,365
Case 17 3 20 45 0.03000 4.28 34,365
Case 18 3 20 60 0.03000 4.28 34,365
Case 19 3 20 40 0.05000 7.14 57,276
Case 20 4 20 60 0.05000 4.01 32,218
Case 21 4 20 80 0.05000 4.01 32,218
Case 22 5 20 75 0.08333 4.28 34,365

Due to the vast amount of data, the results for only the representative
variants given in Table 5 are presented in this study. Case 0 can be
considered as a reference case, as it is compared with the other cases
commonly.

5.2. Boundary layer separation

The enhancement of convective heat transfer in the IJF method is a

result of the boundary layer breaking up [40]. Heat transfer coefficient
profiles along a spherical surface are dependent on the location of
the boundary layer separation point. Values of this coefficient sud-
denly drop at the point of separation; however, behind that spot, they
increase due to the occurrence of fluid wakes [11].
Boundary layer separation points have been found using coordinates
where the velocity vector changes its direction dramatically within
cells near the sphere surface. Coordinates are given using the polar
Fig. 9. Validation of the velocity in m s−1 along profiles A (top), B (centre), and C
(bottom). angle of a spherical coordinate system, where 0◦ corresponds to the
stagnation point position. The mean value with its standard deviation of
the separation angle from all 5304 cases is 146.2 ± 1.1◦ . Fig. 10 presents
the effects of sphere position, sphere diameter, orifice diameter and
geometrical parameters like the orifice diameter, the sphere diameter
and the sphere position above the inflow plate given as the distance refrigerating medium mass flow rate on the separation angle. It can be
between the orifice and stagnation point. The refrigerating medium seen that these parameters have a minor impact on its value; however,
mass flow rate was also investigated in this study. The orifice diameter when the sphere has a diameter less than 15 mm, a decrease in the
varied from 2 mm to 5 mm with steps of 1 mm. The diameter of the separation angle is clear.
sphere was 5 mm to 30 mm with steps of 5 mm. The last investigated For the range of Reynolds number being investigated in this study,
geometrical parameter was the distance between the orifice and the the separation angle value is close to values reported in the literature.
sphere stagnation point, which varied from 20 mm to 80 mm with steps According to Cengel, the boundary layer separation angle of 140◦ for
of 5 mm. It has to be mentioned that the mass flow rate range was de- the flow over a sphere was noticed for the Reynolds number range that
pendent on the orifice diameter in order to keep the same level of mean corresponds with this study [35]. In the paper of a similar system by
jet velocity. Therefore, for the orifice diameter of 2 mm, the range was Peralta et al. the value of the angle was approximately 138◦ [19]. Those
from 0.002667 kg s−1 to 0.024 kg s−1 , while for the diameter of 3 mm, authors pointed out that the delay of the boundary layer separation for
it varied from 0.006 kg s−1 to 0.054 kg s−1 . For an orifice diameter of such a flow can be a result of Coanda effect [41].
4 mm, the range varied from 0.01067 kg s−1 to 0.096 kg s−1 . Finally, for
a diameter of 5 mm, it ranged from 0.01667 kg s−1 to 0.15 kg s−1 . For 5.3. Pressure coefficient
each orifice diameter, 17 values of the mass flow rate were investigated,
and the values of the step change were 0.00133 kg s−1 , 0.003 kg s−1 , In many papers describing the flow over some objects, the distri-
0.00533 kg s−1 and 0.00833 kg s−1 for the mentioned orifices diameters bution of the pressure coefficient on the surface is presented. This
in ascending order. The refrigerating medium temperature was set as quantity, defined according to Eq. (9), is dimensionless. Therefore it
0 ◦ C for all cases. Reynolds number, defined taking into consideration is convenient for comparison, even for cases with different geometry.
the mean fluid velocity in the orifice and the sphere diameter, covered a In the numerical study of similar system, the pressure coefficient was
range of 1700–93,000. There were 5304 single cases simulated in total. compared for different values of the H/d ratio, Reynolds number and

11
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

Fig. 10. Values of the boundary layer separation angle for different values of the parameters being investigated: 𝐻, 𝐷, 𝑑 and 𝑚.
̇

the liquid temperature [19]. It was proven, that the distance between 5.4. Velocity profile along the jet axis
the orifice and the stagnation point has a significant effect on the
pressure coefficient if the orifice diameter is fixed. The authors also The velocity distribution within the jet has characteristic features.
confirmed that similar profiles are observed for cases with the jets Within the region of the jet potential core, the refrigerating medium
impinging curved or flat surfaces. velocity remains constant along the axis after flowing through the
𝑝 − 𝑝∞ orifice. The velocity starts to decrease until reaching the stagnation
𝐶𝑃 = (9) point, where the jet impinges the sphere surface.
1
2
𝜌𝑣0
The parameters being investigated in this case, namely, the orifice
where 𝐶𝑃 is the pressure coefficient, 𝑝 is the static pressure at the point diameter, the sphere diameter, the sphere position above the orifice and
of the pressure coefficient evaluation, 𝑝∞ is the static pressure in the the refrigerating medium mass flow rate, may have an impact on the
free-stream and 𝑣0 is the average velocity in the orifice. jet profile, which finally affects the fluid flow over a spherical artificial
Fig. 11 presents the distribution of the pressure coefficient on the food product.
sphere surface in the function of the angular position. In the region To compare the effects of all the mentioned parameters on the
of stagnation point, the pressure coefficient reaches the values in the velocity profile, the velocity values as well as the profile length were
range of 0.42 and 0.52. At the position of 30–50◦ , it reaches the shown using dimensionless quantities. The dimensionless velocity was
value being slightly below 0 and then remains stable. The cases being defined as the ratio of the local refrigerant velocity to the mean velocity
compared with the literature [19] are characterised by the same H/d at the orifice. The dimensionless orifice-stagnation point profile length
ratio of 10 and similar Reynolds numbers. As seen, Case 16 is in a good was defined as the ratio of the local position measured from the orifice
agreement with the literature, while the pressure coefficient for Case to the total length of this profile.
0 is significantly higher. Both cases have the same Reynolds number As seen in Fig. 12, the refrigerating medium mass flow rate has a
and the H/d ratio, however, the orifice diameter for both cases is negligible impact on the jet profile shape. Within a Reynolds number
different. In the papers describing the jet impinging curved [38] or flat range from 21,000 to 62,000 the flow pattern is the same. The velocity
surfaces [29], Reynolds number is defined using the orifice diameter as profile tends to decline closer to the stagnation point for cases with a
higher mass flow rate. However, the increasing mass flow rate magni-
a characteristic length. However, in the HF method, the frozen product
fies the fluid velocity; therefore, this tendency is insignificant. Fig. 13
dimension, e.g. the sphere diameter, is considered as the representative
presents the same comparison of profiles for cases with various sphere
length [19]. Hence, when different orifice diameters are compared, this
diameters. It can be seen that this parameter does not affect the flow
matter has to be taken into account.
within the jet potential core. Reynolds number change from 15,000 to

Fig. 11. Pressure coefficient on the sphere surface for two cases with different orifice
diameter compared with the case from the literature. Fig. 12. Velocity profiles along the jet axis for different mass flow rates.

12
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

Fig. 13. Velocity profiles along the jet axis for different sphere diameters. Fig. 15. Velocity profiles along the jet axis for different orifice diameters.

52,000 has not affected the flow along the axis. The only difference can of the orifice-stagnation point distance to the orifice diameter seems
be observed directly below the stagnation point. In addition, the case to be essential for this analysis. Fig. 16 compares configurations with
for which of the velocity profile differs from the others is Case 4, with different orifice diameters and sphere positions using a dimensionless
the smallest sphere diameter of 5 mm and the lowest Reynolds number velocity and distance between the orifice and the stagnation point.
of approximately 15,000. The combination of these variants brings three different values of the
Different parameters that have an influence on the jet potential core orifice-stagnation point distance to the orifice diameter ratio. It is
diameter and length are the orifice diameter and the sphere position.
noticed that keeping the same value of this ratio results in having
As seen in Fig. 14, a change in the orifice-stagnation point distance
comparable velocity profiles along the jet axis, although changes in
significantly affects the velocity profile. A comparison of the presented
individual parameters affect the profile shape, as was shown previously.
profiles suggests that the velocity along the jet axis is not affected by
It is worthy to refer to the study of Shademan et al. where the authors
the sphere until it approaches the region of stagnation. Refrigerating
extensively described the velocity profiles along the axis for cases with
medium velocity decreased approximately 5–20 mm below the stag-
different H/d ratios [29]. According to this study, the H/d value of 18.5
nation point, depending on the specific case. For variants where the
is sufficient for the full development of all the jet sub-regions.
sphere is placed higher, the jet stagnation tends to be smoother. In
Case 9 and Case 10 this region has been reached before the velocity
started to decrease due to the jet potential core decay. Similar situation 5.5. Velocity distribution around the sphere surface
was observed in the numerical and experimental study of similar case,
where the maximum H/d ratio was 4 [42]. As was mentioned, the Refrigerating medium flow was also investigated along an orbital
orifice diameter also plays a major role. For variants with a smaller profile located 0.5 mm away from the sphere surface. Both the axial
orifice, the jet potential core is significantly shortened, as shown in and radial components of velocity were compared. All given parameters
Fig. 15. were tested in terms of characterising the flow. However it was ob-
The paper describing similar configurations of jets impinging spheri- served that the only parameter that did not affect the fluid flow profile
cal surfaces [38] and the HF system analysis [17] suggest that the ratio along the profile was the orifice diameter. For that reason, profiles of
the velocity components were compared for cases with different orifice

Fig. 14. Velocity profiles along the jet axis for different orifice-stagnation point
distances. Fig. 16. Velocity profiles along the jet axis for 𝐻∕𝑑 ratio values of 10, 15, and 20.

13
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

diameters. The comparison was performed again using a dimensionless Table 6

Operational conditions for the cases used for cooling analyse.
velocity, defined as the ratio of the local radial or axial velocity to its
Case number 𝑑 (mm) 𝐻 (mm) 𝑇 (◦ C) 𝑚̇ (kg s−1 ) 𝐷 (mm) 𝑇0 (◦ C)
value reached at the profile’s maximum value.
Case HT.1 5 50 −0.4 0.05000 30 14.9
It can be seen in Fig. 17 that the orifice diameter affects the axial Case HT.2 5 50 −1.1 0.08333 30 13.8
velocity near the stagnation point region. This is in good agreement
with the comparison of the velocity profiles along the jet axis for cases
with different orifice diameters, where a similar tendency was noticed.
5.6. Preliminary cooling analysis of food sample
The highest axial velocity value is reached at the same point located
60◦ from the stagnation point for all the considered cases. All the
In order to assess the heat transfer in the presented system, a
profiles again diverge above that point. This suggests that both the
preliminary cooling analysis has been performed. Due to high thermal
orifice diameter and the refrigerating medium jet affect the fluid flow
conductivity, a copper sphere used for a fluid flow analysis would
above the sphere. Negative values of the axial velocity are reached at be cooled down in seconds and, as a consequence, results would be
angles of 130–140◦ , depending on the case. Wakes start to occur there, not accurate. Therefore, at this stage, an artificial object has been
and the refrigerating medium starts to recirculate, which results in the replaced by a real food product. The product selected for this analysis
convective heat transfer being enhanced. was a potato because it is a common product with a uniform internal
Fig. 18 presents the radial velocity component along the same structure. To achieve repeatable results, spherical samples having a
profile. For all cases, the velocity rises from zero near the stagnation diameter of 30 mm were cut out from a whole potato piece.
point and reaches a maximum value at approximately 30◦ . The profiles In this study, only the cooling of these samples is presented, there-
start to diverge after reaching an angle of 100◦ . Again, the wakes are fore the temperatures were held above the freezing point of potatoes.
noticeable above the sphere where the values of the radial velocity are The exact process of the phase change within the foods is more complex
and was not implemented in the model presented in this paper. The
positive.
characteristic feature for a freezing of foods is that it does not occur
at a fixed temperature or a specific range of temperatures, but it is a
process progressing gradually below an initial freezing temperature of
a food product. It takes place in such a manner due to the concentration
change of the solids being dissolved in a liquid water within the food
tissues during the freezing [43].
As mentioned before, the samples were cooled in temperature above
the initial freezing point of samples. The cases being compared cover
experimental results with the numerical results obtained with the 𝑘-𝜔
SST and RSM turbulence approaches. The operating conditions used for
this comparison are described in Table 6.
The results of numerical simulations and experimental measure-
ments of the spherical potato sample cooling in the HF unit are pre-
sented in Fig. 19. For both cases, 500 s of the cooling process was
enough to cool the samples down to the temperature being close
to the refrigerating medium. The potato samples centre temperatures
after this time were 1.8 ◦ C and 1.6 ◦ C higher than the refrigerating
medium temperatures for cases HT.1 and HT.2, respectively. It should
be mentioned that the solution temperature was relatively high for both
cases, there was only one orifice installed in the system and the sample
Fig. 17. Axial velocity profiles along the profile around the sphere for variants with
had relatively high diameter in comparison with the food products
different sphere diameters. being more appropriate for the HF method. At the industrial scale, food
products cooling would be much quicker.
The differences between the results obtained numerically using dif-
ferent turbulence approaches are not significant. In the process, where
the heat transfer realised by the forced convection is very intensive, the
heat transfer by conduction within the food product causes the major
thermal resistance. This relation may be different and depend on the
refrigerating medium flow conditions and on the food product itself,
i.e. its properties, the shape and the size.
For the case HT.1, differences between the simulated and measured
temperatures reach up to 2 K, which is higher value than the thermo-
couple measurement error. The difference depends on the turbulence
model as well as on the value of the specific heat of the food product.
The time required to cool down the potato sample centre point to
the temperature of 2 K is about 70 s higher for the measurements
than for the numerical results with the 𝑘-𝜔 SST turbulence approach
employed with the specific heat value of 3670 J kg−1 K−1 . The highest
difference in time required to reach that temperature with respect to
the experimental results is for the case with 𝑘-𝜔 SST model and the
specific heat of 3450 J kg−1 K−1 , while the lowest is for the case with
Fig. 18. Radial velocity profiles along the profile around the sphere for variants with RSM model and the specific heat assumed as 3670 J kg−1 K−1 . These
different sphere diameters. differences are 90 s and 40 s, respectively.

14
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

Fig. 19. The temperatures of the spherical potato samples HT.1 and HT.2 during the HF cooling.

For the Case HT.2, differences in the temperatures are slightly affected by the parameters that were investigated and was equal to
higher and reach up to 3 K. Comparing the time required to cool down approximately 147◦ for all cases.
the centre point of the potato to the temperature of 2 ◦ C, the higher Moreover, the preliminary study of a real food product cooling
difference concerning the measurements is about 80 s and the lowest is has been presented. The process has been conducted above the initial
40 s for the same simulation settings as for the first case. As seen, for freezing temperature of a potato. The results give an overview of the
the cases HT.1 and HT.2, the assumed value of the specific heat is at time required to start freezing in the presented simplified HF system.
least as much important as the turbulence model used for the numerical The measurements of the potato centre point temperatures have been
simulation. compared with the numerical results obtained with different values of
Taking into consideration the fact of handling with real food prod- specific heat and with both approaches of the turbulence modelling
uct in such a flow, the presented results are satisfactory. Apart from the used in the proposed HF system model. The accuracy of the CFD
physical properties evaluation, the measurements were very sensitive to results is satisfactory in comparison with the measurements, taking into
the sample shape, its positioning and the positioning of the thermocou- consideration the difficulties in performing such an experiment and in
ple measuring the temperature at the centre point of the sample. The the evaluation of the real food product physical properties.
reason for selecting quite high diameter of the sample was to minimise The refrigerating medium flow in the HF unit being studied was
the error based on the thermocouple positioning in the centre point of properly modelled with appropriate assumptions and a turbulent
the spherical sample. model. This has been confirmed by comparing the numerical results
with the performed PIV measurements for 9 cases varied by the refrig-
6. Conclusions erating medium mass flow rate, the orifice diameter and the sphere
placement.
In this study, a CFD model of refrigerating medium flow within a
laboratory-scale HF system has been proposed. A numerical domain Declaration of competing interest
has been defined based on the HF unit designed for this work. Real
food products have been replaced by a copper sphere in order to study The authors declare that they have no known competing finan-
the fluid flow around an artificial object having a regular shape. One cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
arrangement of the single orifice and single sphere was used to study
the refrigerating medium flow within a wide range of the geometrical
Acknowledgements
variations in such a system. For this type of HF system, two turbulence
models were examined: 𝑘-𝜔 SST and RSM. The results of the validation
Financial assistance was provided by grant no. DEC-2016/22/
performed using the PIV technique suggested that the 𝑘-𝜔 SST model
E/ST8/00517 funded by the National Science Centre, Poland, and
predicts the fluid flow with higher accuracy, especially within the area
is here acknowledged. The work of MS was supported by Silesian
of the highly turbulent refrigerating medium jet.
University of Technology through project No. BKM-551/RIE6/2019.
The performance of numerous simulations with various geometrical
configurations and operating conditions was conducted by employing
References
the CoolFoodPL application designed for automatically preparing and
solving multiple cases. [1] K.A. Fikiin, Emerging and novel freezing processes, in: Frozen Food Science and
The main purpose of this study was to develop, verify and validate Technology, Blackwell Publishing Ltd., Oxford, UK, 2008, pp. 101–123, chapter
a numerical model for simulating the turbulent flow of refrigerating 5.
medium within an HF system. The obtained results proved that the [2] C. James, G. Purnell, S.J. James, A review of novel and innovative food freezing
technologies, Food Bioprocess Technol. 8 (8) (2015) 1616–1634.
velocity profiles along the axis of the fluid jet core are mostly affected [3] T. Marazani, D.M. Madyira, E.T. Akinlabi, Investigation of the parameters
by the orifice-stagnation point distance and the orifice diameter. In ad- governing the performance of jet impingement quick food freezing and cooling
dition, the profiles are similar if the ratio of those geometric parameters systems - A review, Procedia Manuf. 8 (2017) 754–760.
is maintained. The shapes of the axial and radial velocity profiles also [4] A.E. Delgado, A.C. Rubiolo, Microstructural changes in strawberry after freezing
and thawing processes, LWT - Food Sci. Technol. 38 (2) (2005) 135–142.
show similar trends along the profiles close to the sphere surface, with
[5] D. Góral, F. Kluza, Physical changes of vegetables during freezing by
exceptions in some characteristic areas. The angle of separation point conventional and impingement methods, Acta Agrophys. 7 (1) (2006) 59–71.
of the boundary layer near the sphere surface was not significantly [6] R.H. Mascheroni, Operations in Food Refrigeration, CRC Press, 2012.

15
M. Stebel et al. International Journal of Thermal Sciences 151 (2020) 106284

[7] P. Dempsey, P. Bansal, The art of air blast freezing: Design and efficiency [24] M. Palacz, W. Adamczyk, E. Piechnik, M. Stebel, J. Smolka, Experimental
considerations, Appl. Therm. Eng. 41 (2012) 71–83. investigation of the fluid flow inside a hydrofluidisation freezing chamber, Int.
[8] S.E. Zorrilla, A.C. Rubiolo, Mathematical modeling for immersion chilling and J. Refrig. 107 (2019) 52–62.
freezing of foods: Part II: Model solution, J. Food Eng. 66 (3) (2005) 339–351. [25] H.K. Versteeg, W. Malalasekera, An Introduction to Computational Fluid
[9] D. Góral, F. Kluza, Heat transfer coefficient in impingement fluidization freezing Dynamics: The Finite Volume Method, Pearson Education, 2007.
of vegetables and its prediction, Int. J. Refrig. 35 (4) (2012) 871–879. [26] E.E.M. Olsson, L.M. Ahrne, A.C. Trägårdh, Flow and heat transfer from multiple
[10] F. Erdogdu, M. Ferrua, S.K. Singh, R.P. Singh, Air-impingement cooling of boiled slot air jets impinging on circular cylinders, J. Food Eng. 67 (3) (2005) 273–280.
eggs: analysis of flow visualization and heat transfer, J. Food Eng. 79 (3) (2007) [27] T.J. Chung, Computational Fluid Dynamics, Cambridge University Press, 2010.
920–928. [28] S.B. Pope, Turbulent Flows, Cambridge University Press, 2000.
[11] C. Dirita, M.V.D. Bonis, G. Ruocco, Analysis of food cooling by jet impingement, [29] M. Shademan, R. Balachandar, R. Barron, CFD analysis of the effect of nozzle
including inherent conduction, J. Food Eng. 81 (1) (2007) 12–20. stand-off distance on turbulent impinging jets, Can. J. Civil Eng. 40 (7) (2013)
[12] A. Dewan, R. Dutta, B.V. Srinivasan, Recent trends in computation of turbulent 603–612.
jet impingement heat transfer, Heat Transfer Eng. 33 (4–5) (2012) 447–460. [30] E.E.M. Olsson, H. Janestad, L.M. Ahrné, A.C. Trägårdh, R.P. Singh, Determination
[13] A.G. Fikiin, Method and System for Hydrocooling of Fish, Invention Certificate of local heat-transfer coefficients around a circular cylinder under an impinging
No. 40164, Bulgarian Patent Agency INRA, 1985. air jet, Int. J. Food Prop. 11 (3) (2008) 600–612.
[14] K.A. Fikiin, W. Ming-Jian, M. Kauffeld, T.M. Hansen, Direct contact chilling and [31] F. Bode, K. Sodjavi, A. Meslem, I. Nastase, Comparison of turbulence models in
freezing of foods in ice slurries, 2005, pp. 251–271. simulating a cruciform impinging jet on a flat wall, in: Proceedings of the 15th
[15] P. Verboven, N. Scheerlinck, B.M. Nicolai, Surface heat transfer coefficients International Heat Transfer Conference, 2014.
to stationary spherical particles in an experimental unit for hydrofluidisation [32] V. Gesellschaft, VDI Heat Atlas, Springer Berlin Heidelberg, 2010.
freezing of individual foods, Int. J. Refrig. 26 (3) (2003) 328–336. [33] M.S. Rahman, Food Properties Handbook, CRC press, 2009.
[16] K.A. Fikiin, O.B. Tsvetkov, Y.A. Laptev, A.G. Fikiin, V.S. Kolodyaznaya, Thermo- [34] ASHRAE, ASHRAE handbook: Refrigeration : SI Edition, American Society of
physical and engineering issues of the immersion freezing of fruits in ice slurries Heating, Refrigerating and Air-Conditioning Engineers, 2006.
based on sugar-ethanol aqueous solutions, 2001, pp. 10–15. [35] Y.A. Cengel, Fluid Mechanics, McGraw-Hill Education, 2010.
[17] J.M. Peralta, A.C. Rubiolo, S.E. Zorrilla, Design and construction of a hydroflu- [36] F. Russo, N.T. Basse, Scaling of turbulence intensity for low-speed flow in smooth
idization system. Study of the heat transfer on a stationary sphere, J. Food Eng. pipes, Flow Meas. Instrum. 52 (2016) 101–114.
90 (3) (2009) 358–364. [37] G. Xu, R. Antonia, Effect of different initial conditions on a turbulent round free
[18] A.G. Fikiin, New method and fluidized water system for intensive chilling and jet, Exp. Fluids 33 (5) (2002) 677–683.
freezing of fish, Food Control 3 (3) (1992) 153–160. [38] D.H. Lee, Y.S. Chung, D.S. Kim, Turbulent flow and heat transfer measurements
[19] J.M. Peralta, A.C. Rubiolo, S.E. Zorrilla, Mathematical modeling of the heat on a curved surface with a fully developed round impinging jet, Int. J. Heat
transfer and flow field of liquid refrigerants in a hydrofluidization system with Fluid Flow 18 (1) (1997) 160–169.
a stationary sphere, J. Food Eng. 99 (3) (2010) 303–313. [39] N. Rajaratnam, D.Z. Zhu, S.P. Rai, Turbulence measurements in the impinging
[20] J.M. Peralta, A.C. Rubiolo, S.E. Zorrilla, Mathematical modeling of the heat and region of a circular jet, Can. J. Civil Eng. 37 (5) (2010) 782–786.
mass transfer in a stationary potato sphere impinged by a single round liquid [40] L.D. Kaale, T.M. Eikevik, T. Rustad, K. Kolsaker, Superchilling of food: A review,
jet in a hydrofluidization system, J. Food Eng. 109 (3) (2012) 501–512. J. Food Eng. 107 (2) (2011) 141–146.
[21] E.E. Belis, S.E. Zorrilla, J.M. Peralta, Effect of the number of orifices and [41] T.E. Faber, Fluid Dynamics for Physicists, Cambridge University Press, 1995.
operative variables on the heat and mass transfer in a hydrofluidization system [42] E. Baydar, Y. Ozmen, An experimental and numerical investigation on a confined
with static spheres, J. Food Eng. 153 (2015) 96–107. impinging air jet at high Reynolds numbers, Appl. Therm. Eng. 25 (2–3) (2005)
[22] J.D. Oroná, S.E. Zorrilla, J.M. Peralta, Computational fluid dynamics combined 409–421.
with discrete element method and discrete phase model for studying a food [43] R.P. Singh, D.R. Heldman, Introduction to Food Engineering, Chapter 7 - Food
hydrofluidization system, Food Bioprod. Process. 102 (2017) 278–288. Freezing, Academic Press, San Diego, 2014, pp. 521–563.
[23] J.D. Oroná, S.E. Zorrilla, J.M. Peralta, Sensitivity analysis using a model based on
computational fluid dynamics, discrete element method and discrete phase model
to study a food hydrofluidization system, J. Food Eng. 237 (2018) 183–193.

16