Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: A model describing the heat and mass transfer involved in food drying is presented. The aim is to deter-
Received 6 August 2007 mine the effect of air temperature on the performance of the drying process applied to fresh-cut vegeta-
Received in revised form 28 February 2008 ble slices, but other effects can be easily incorporated in the model. The model allows to disregard one of
Accepted 4 May 2008
the most limiting parameters in such modeling, i.e. the average heat and mass transfer coefficients at the
Available online 15 May 2008
food/drying substrate interface, which are generally taken from the literature. Such assumptions are lim-
iting in the sense that they are referred to average transfer conditions and general geometries. The pre-
Keywords:
sented model relies upon a finite-element solution of time-dependent differential equations for
Forced convection
Food dehydration
simultaneous and conjugate heat and moisture transfer in a two-dimensional domain, without any infer-
Conjugate model ence in such empiricism.
Temperature and moisture evolution A special formulation for drying kinetic of the substrate is also exploited, and a treatment of the
Local heat and mass transfer dependence of the properties upon the residual moisture content is included. After proper validation with
the available experimental measurements, the numerical solution is discussed by presenting each
involved field variables, emphasizing on the conjugate nature of the drying process. Due to its flexibility
and generality, the model can be used in common industrial driers’ optimization, even in the assumption
of a laminar flow field.
Ó 2008 Elsevier Ltd. All rights reserved.
0260-8774/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jfoodeng.2008.05.008
M.V. De Bonis, G. Ruocco / Journal of Food Engineering 89 (2008) 232–240 233
Nomenclature
and properties variation on moisture and temperature), an innova- sion) and (2) by capillary forces, while the motion of water vapor
tive, further approach is to solve a model in which the mass and is (3) by diffusion in air spaces within the substrate caused by va-
energy interface fluxes vary seamlessly in space and time as the por pressure gradients (Fellows, 2000).
solution of the field variables. With this objective Oliveira and Hag-
highi (1997) obtained the temperature and the moisture contours 2.1. Assumptions
for the drying of wood, but their work was affected by the limita-
tion given by considering a laminar boundary layer flow over the The following assumptions are considered in this work:
substrate. This approach was later complemented by Murugesan
et al. (2001) for a timber block using a full Navier-Stokes formula- (1) the flow is laminar; the dryer is two-dimensional, and a
tion for the flow field, allowing for the buoyancy term. While over- small portion in the vicinity of the product is studied only,
coming the limitations of the boundary layer assumptions, their for sake of simplicity;
work was focussed on Nusselt and Sherwood numbers on the ex- (2) due to the adopted flow regime, no body force is accounted
posed substrate surface. For the first time a full conjugate model for;
of a drying food was presented by De Bonis and Ruocco (2007), (3) the thermophysical food properties are moisture-dependent,
yet for a specific exchange configuration (a thin baking product as reported by Ruiz-López et al. (2004), and reported in Table
to be dried by an impinging turbulent jet draft), where the focus 1, while the air and water properties are temperature-
was on residual local water activity. dependent and are taken from Perry et al. (1997): for sake
A similar approach is carried out in the present work for a drying of simplicity their dependency are not reported in the
vegetable substrate, by employing a finite-element approach. formulation;
Residual water and temperature fields are computed locally within (4) the effect of capillary forces is included in liquid water
the substrate, when this interacts with a forced, laminar air flow. diffusivity;
The later assumption allows to focus on the basic aspects of flow (5) the diffusivity of vapor in the substrate is the same than the
transport, focussing upon the vapor and liquid water production/ diffusivity of liquid water, as implied for example by Braud
depletion and transport, which is dealt with by an ad-hoc first-or- et al. (2001).
der irreversible kinetics. Such kinetics is included to solve for tran-
sient, two-dimensional flow, temperature and moisture fields. The following simplifying assumptions are adopted:
Realistic transfer exchanges are inherently considered that vary
with process time and surface location, eliminating the need for (1) the viscous heat dissipation in the drying medium and the
empirical heat and mass transfer (averaged) coefficients evaluation. heat generation within the moist substrate are neglected;
2. Problem formulation
Table 1 various parameters that describe both the inherent water phase
Food properties functions, dependent on the moisture content X (Ruiz-López et al., conversion and interface conditions.
2004)
In this paper a modified exponential model of evaporation has
Property Function been adopted, based on an Arrhenius first-order irreversible kinet-
qs (kg/m3) 440:001 þ 90X ics formulation. Several works have been presenting such an ap-
X
cps (J/kgK) 1750 þ 2345ð1þX Þ proach, as Panagiotou et al. (1999), Azzouz et al. (2002) and
ks (W/mK) 0:49 þ 0:443 expð0:206XÞ Roberts and Tong (2003). It is seen here that the inherent (volu-
10
Dls , Dvs (m2/s) 2:8527 10 expð0:2283369XÞ
metric) evaporation physics (Roberts and Tong, 2003) must be
joined to interface conditions (Panagiotou et al., 1999; Azzouz et
al., 2002), such that the thermal, fluid dynamic and concentration
(2) due to the nature of the interacting species, no diffusion
regimes could be all be represented in the mass source term.
fluxes are accounted for in the energy equation;
The present work is focussed upon the additional dependence
(3) neither shrinkage nor deformation of drying substrate are
on process temperature variation, so that the basic Arrhenius-type
accounted for.
relationship can be modified as follows:
Process inlet ðx ¼ 0; 0 < y < Ha Þ for concentration, velocity, temperature and pressure gradients
resolution in the boundary layer and within the substrate’s ex-
cv ¼ cv0 ; u ¼ ua ; v ¼ 0; T ¼ Ta ð11Þ
posed surface, induced by the heating and evaporation. Execution
Bottom plate, air interface ð0 < x < L0s and L0s þ L00s < x < Lp ; time for t = 18000 s elapsed time has been approximately 20 min
y ¼ 0Þ on a Pentium Xeon PC (WindowsXP Pro OS, 3.0 GHz, 2 GB RAM).
ocv;l oT Specific underrelaxation factors have been employed to solve the
¼ 0; u ¼ v ¼ 0; ¼0 ð12Þ Navier-Stokes equations in the start-up phase of drying.
oy oy
Bottom plate, substrate interface ðL0s < x < L0s þ L00s ; y ¼ 0Þ
3. Results and discussion
ocl oT
¼ 0; ¼0 ð13Þ
oy oy 3.1. Model validation
Upper open surface ð0 < x < La ; y ¼ Hp Þ
The available literature data are rather limited in order to vali-
ocv ou ov oT date the model and its numerical treatment, as geometry and flow
¼ 0; ¼ ¼ 0; ¼0 ð14Þ
oy oy oy oy regimes were always left unspecified and transfer coefficients were
Process outlet ðx ¼ La ; 0 < y < Ha Þ assumed from empirical correlations, except in Murugesan et al.
(2001) (who dealt with a non-food substrate). However, the exper-
ocv ou oT imental average residual moisture reported by Ruiz-López et al.
¼ 0; ¼ 0; v ¼ 0; ¼0 ð15Þ
ox ox ox (2004) has been first compared with the present numerical solu-
Finally, continuity is ensured by enforcing the following positions: tion and reported in Fig. 3. A 4 h baking process of a thin carrot
slice with Ls ¼ 0:06 m and Hs ¼ 0:0050 or 0.0075 m (for Data Set
Across the horizontal sub-domains interface ðL0s < x < L0s þ L00s ; 1 and 2, respectively) was configured with the following driving
y ¼ Hs Þ parameters: T a ¼ 343 or 323 K (for Data Set 1 and 2, respectively),
T0 = 298 K, U 0 ¼ 0:87, and inlet air relative humidity of 45%. Care
ocl
cva ¼ cvs ; ¼ 0; u ¼ v ¼ 0; Ta ¼ Ts ð16Þ was exercised to adapt the present model so that the inlet air
oy velocity ua was 2.0 m/s, but still in the laminar regime, with
Across the upwind ðx ¼ L0s ; 0 < y < Hs Þ and downwind ðx ¼ La ¼ 0:20 m and Ha ¼ 0:10 m. The reference constant K 0 for the gi-
L00s ; 0 < y < Hs Þ (vertical) sub-domains interfaces ven configuration was 7 103 , while the temperature factor a was
found to be 0 and -10 for Data Set 1 and 2, respectively.
ocl
cva ¼ cvs ; ¼ 0; u ¼ v ¼ 0; Ta ¼ Ts ð17Þ For Data Set 1 there is a good agreement at the beginning of
ox treatment, while a maximum difference of approximately 15% is
detected after 2 h. At the end of drying the measured and com-
puted moisture are again very similar. In Data Set 2 the drying con-
2.6. Numerical method and additional considerations dition are milder therefore the kinetic parameters (in absence of a
thickness adjustment) underestimate the measurements, the max-
A finite-element commercial solver has been employed to inte- imum difference being less than 10% after 3 h.
grate the partial differential equations system (COMSOL Multi- A second such benchmark has been found in the numerical data
physics User’s Guide, 2007). A preliminary grid independency from Aversa et al. (2007) and reported in Fig. 4. A similar process
test has been carried out with 3 different grids of approximately (baking of a carrot substrate) was configured, with Ls ¼ 0:06 m
2000, 4000 and 6000 triangular elements, respectively, and the and Hs ¼ 0:015, and with the following driving parameters:
second grid was selected as the local heat flux across the interface T a ¼ 353, 343 or 333 K (for Data Set 3 to 5, respectively),
vary less than 2% in all locations with respect to the one computed T0 = 303 K, U 0 ¼ 0:64, and inlet air relative humidity of 75%. Care
with the third grid. The mesh was distorted locally (Fig. 2) to allow was exercised, as well, to adapt the present model so that the inlet
Fig. 3. Average U evolution during process: comparison with Ruiz-López et al. (2004) measurements for 2 different Data Sets.
Fig. 4. Average U evolution during process: comparison with Aversa et al. (2007) computations for 3 different Data Sets.
air velocity ua was 0.3 m/s, in the laminar regime, with La ¼ 0:20 m cases a good agreement is detected between the two different
and Ha ¼ 0:10 m. The reference constant K 0 for the given configu- models. Small discrepancies (less than 5%) are found after 1 hr of
ration was 90, while the temperature factor a was found to be 0, treatment only, due to the condensation phenomenon reported
17 and 10 for Data Set 3 to 5, respectively. in the benchmark work, which remains unjustified for empirical
The same limitations with the earlier benchmark were found, as transfer coefficients such as the ones reportedly employed in Aver-
no information was available on employed configuration. In all sa et al. (2007).
M.V. De Bonis, G. Ruocco / Journal of Food Engineering 89 (2008) 232–240 237
3.2. Flow and temperature field In addition to the available Data Set 3, the present model has
been exercised by varying the nominal value of velocity. Fig. 7
The simulation results for Data Sets 3 configuration (Aversa shows the new flow field generated with ua = 0.3 m/s, 10 times
et al., 2007) are then briefly presented in the form of velocity, tem- higher. The velocity distribution is very similar to the previous
perature or moisture distributions. Fig. 5 shows first the vector and one, but the velocity local values are much higher indeed. These
scalar distributions of velocity in the drying air. Due to the flow in turn reflect on the higher thermal regime, reported in Fig. 8,
field contraction and speed-up, the action of the drying air is stron- where the product center temperature increase by 4 K with respect
gest on top of the substrate, while the front and back faces are sub- to Data Set 3 comparison. A more dynamic flow situation dictates
ject to stagnation and recirculation flow regions, respectively. This an overall more even side-to-side treatment, therefore the slowest
justifies the adoption of a fully conjugate model for a detailed heating point is almost perfectly centered this time.
description, as transfer properties vary considerably with exposed
surface location. 3.3. Moisture and vapor removal
Depending on the flow field, the temperature distribution in Fig.
6 presents a related non-homogeneous behavior, due to the non- Based on the above flow field and temperature maps, it is ex-
uniform heat transfer, which will then reflect upon the residual pected that (1) the evaporation occurs non-homogeneously within
moisture distribution. On the three exposed substrate sides, due the substrate, and (2) the vapor mass transfer across the fluid-sub-
to the conjugate nature of the model, the isotherms are obviously strate interface also occurs non-uniformly. Consequently, the
inclined. The substrate is found to be more than 3 K warmer on the moisture will be non-homogeneously removed within the
leading edge, with respect to the trailing edge, and its left side is substrate.
being heated more effectively (as expected) than the right one. Residual moisture distribution after the treatment is reported in
The lowest temperature of about 340 K is detected on substrate Fig. 9 for Data Set 3. The evaporation and depletion of water is
bottom, by the adiabatic floor, with the slowest heating point being more effective where the temperature is the highest (Fig. 6) at
located slightly in the flow direction. the leading edge (a triangular chunk, one-fifth of the entire prod-
Fig. 5. Close-up of flow field (vector field and streamlines) in the vicinity of substrate for Data Sets 3 to 5 fluid dynamic configuration, after a 5 h drying. juj values range from
0 to 0.21 m/s.
Fig. 6. Close-up of temperature field (isotherms) in the substrate and its vicinity for Data Sets 3 configuration, after a 5 h drying. T values range from 341 to 352 K.
238 M.V. De Bonis, G. Ruocco / Journal of Food Engineering 89 (2008) 232–240
Fig. 7. Close-up of flow field (vector field and streamlines) in the vicinity of substrate for a higher air velocity, after a 5 h drying. juj values range from 0 to 2.7 m/s.
Fig. 8. Close-up of temperature field (isotherms) in the substrate and its vicinity for a higher air velocity, after a 5 h drying. T values range from 345 to 352 K.
Fig. 9. Close-up of residual moisture concentration field (isolines) in the substrate for Data Set 3 configuration, after a 5 h drying. cl values range approximately from 3.19 to
3.33 104 mol/m3.
uct), but the trailing edge is dried more than the average too, due Fig. 10 describes the effect of ten-fold velocity increment on
to the favorable momentum transport in its vicinity. residual moisture. The drying process is stronger, so the humidity
M.V. De Bonis, G. Ruocco / Journal of Food Engineering 89 (2008) 232–240 239
Fig. 10. Close-up of residual moisture concentration field (isolines) in the substrate for a higher air velocity, after a 5 h drying. cl values range approximately from 3.10 to
3.23 104 mol/m3.
Fig. 11. Close-up of vapor excess field (contours) in the vicinity of substrate for Data Set 3 configuration, after a 5 h drying. Representation in 10 levels of gray, for a cv range
from approximately 6.05 to 6.20 mol/m 3.
Migliori, M., Gabriele, D., de Cindio, B., Pollini, C.M., 2005. Modelling of high quality Perry, R.H., Green, D.W., Maloney, J.O., 1997. Perry’s Chemical Engineers’ Handbook.
pasta drying: mathematical model and validation. Journal of Food Engineering McGraw-Hill, New York.
69, 387–397. Roberts, J.S., Tong, C.H., 2003. Drying kinetics of hygroscopic porous materials under
Murugesan, K., Suresh, H.N., Seetharamu, K.N., Aswatha Narayana, P.A., isothermal conditions and use of a first-order reaction kinetic model for
Sundararajan, T., 2001. A theoretical model of brick drying as a conjugate predicting drying. International Journal of Food Properties 6, 355–367.
problem. International Journal of Heat and Mass Transfer 44, 4075–4086. Ruiz-López, I.I., Córdova, A.V., Rodríguez-Jimenes, G.C., García-Alvarado, M.A., 2004.
Oliveira, L.S., Haghighi, K., 1997. Finite element modeling of grain drying. In: Turner, Moisture and temperature evolution during food drying: effect of variable
I., Mujumdar, A.S. (Eds.), Mathematical Modeling and Numerical Techniques. properties. Journal of Food Engineering 63, 117–124.
Marcel Dekker, New York, pp. 309–338. Wang, Z.H., Chen, G., 1999. Heat and mass transfer during low intensity convection
Panagiotou, N.M., Stubos, A.K., Bamopoulos, G., Maroulis, Z.B., 1999. Drying kinetics drying. Chemical Engineering Science 54, 3899–3908.
of a multicomponent mixture of organic solvents. Drying Technology 17 (10),
2107–2122.