Microwave heating of liquid foods in laminar flow through a circular tube has been modeled. In particular, skim milk as a Newtonian fluid and apple sauce and tomato sauce as non-Newtonian fluids have been considered. The temperature profiles have been obtained solving the motion and energy equations in transient regime and Maxwell’s equations in the frequency domain. Numerical resolution of Finite Element Method has been implemented in Comsol Multiphysics. The generation term due to the microwave heating has been evaluated according both to Lambert’s law and Poynting theorem. Finally, a comparison between the two methods has been made in order to check to what extent the results obtained with the simpler Lambert’s law approximation are comparable with those deriving from the exact solution of Maxwell equations. Dielectric properties are considered to be temperature dependent.
Microwave heating has been utilized since the 1940s [
Microwave heating of solid foods has been largely investigated in the last years, writing the energy equation as a conductive heat transfer with a generation term.
The latter has been modeled by many authors, using two different approaches to evaluate the effects of the microwave distribution: by solving the Maxwell’s equations [
As regards fluids, there are fewer studies and the most are about batch processes, conducted only by means of Maxwell equations solution [
As regards continuous microwave heating, the problem of temperature distribution analysis in liquid foods that flow in cylindrical ducts was faced by Romano and Apicella [
Sheryl Barringer and others [
In the present paper, an analysis on liquid foods and a cylindrical geometry is made. The results of the two different models hypothesis are compared in terms of absorbed power and temperature profile.
The physical system to be modeled is a cylindrical horizontal tube with a length
L = 0.3 m and a radius R = 0.02 m (
The aim is to check the temperature profile and to compare results obtained with Lambert’s law and Maxwell equations resolution.
The mathematical model consists of the following three differential equations in vector form with their boundary conditions [
D D t ρ = − ρ ( ∇ _ ⋅ v _ ) (continuity equation) (1)
ρ D D t v _ = − ∇ _ p − [ ∇ _ ⋅ τ _ _ ] + ρ g _ (momentum equation) (2)
ρ c p D D t T = k ∇ 2 T + Q (energy equation) (3)
In cylindrical coordinate system, Equations (1) and (2) for the specific problem become:
∂ v z ∂ z = 0 (4)
ρ ∂ v z ∂ t = − ∂ ∂ r ( r τ r z ) + P i n − P o u t L (5)
with the following initial and boundary conditions.
As initial condition, the fluid is stationary.
@ t = 0 , v z = 0 , ∀ r , z (i)
As boundary conditions, we assume no slipping at the wall
@ r = R , v z = 0 , ∀ z , ∀ t > 0 (ii)
And we consider the symmetry on the axis.
@ r = 0 , τ r z = 0 , ∀ z , ∀ t > 0 (iii)
Similarly, Equation (3) turns out to be
ρ c p ( ∂ T ∂ t + v z ∂ T ∂ z ) = k [ 1 r ∂ ∂ r ( r ∂ T ∂ r ) + ∂ 2 T ∂ z 2 ] + Q (6)
With the following initial and boundary conditions.
As initial condition, the temperature of the fluid is assumed to be uniform
@ t = 0 , T = T 0 , ∀ r , z (iv)
As boundary conditions for the radial direction we consider the symmetry respect the z-axis and the heat exchange at the wall by external convection.
@ r = 0 , ∂ T ∂ r = 0 , ∀ z , t (v)
@ r = R , k ∂ T ∂ r = − h c ∗ ( T − T a i r ) , ∀ z , t (vi)
With initial and boundary conditions for the radial direction, so defined, the system results to be axial-symmetric.
For the z direction uniform temperature in the inlet and Danckwerts’s condition in the outlet section are imposed
@ z = 0 , T = T i n , ∀ r , t (vii)
@ z = L , ∂ T ∂ z = 0 , ∀ r , t (viii)
Initial temperature T 0 and input temperature T i n are both equal to environmental temperature T a i r .
Heat generation due to microwaves can be modeled as the QLR predicted by Lambert’s law or as the QMR provided by the Poynting theorem after the resolution of Maxwell equations.
Lambert’s law forecasts an exponential decay inside the sample, Romano et al. (2005)
Q L R = 2 α ⋅ ( R r ) ⋅ Q 1 ⋅ [ e − 2 α ( R − r ) + e − 2 α ( R + r ) ] (7)
where
Q 1 = Q 0 2 π R ( R + L )
α = 2 π λ 0 ε ′ { [ 1 + ( ε ″ ε ′ ) 2 ] 1 2 − 1 } 2
ε ″ ε ′ is the loss tangent, known as tanδ.
The attenuation factor α for each fluid has been considered as a temperature function, calculated by interpolation starting from graphic relationships for dielectric constant and loss tangent versus temperature in a range 10˚C - 90˚C [
Maxwell equations are a system of four vectory equations [
∇ × E = − ∂ B ∂ t Faraday’s law (8)
∇ × H = ∂ D ∂ t + J + σ E Ampère-Maxwell law (9)
∇ ⋅ D = ρ Gauss’s law for electric field (10)
∇ ⋅ B = 0 Gauss’s law for magnetic field (11)
Keeping in mind also the following constitutive equations:
D = ε E (12)
B = μ H (13)
Making the rotor of first and second member of (8)
∇ × ( ∇ × E ) = ∇ × ( − μ ∂ H ∂ t )
Hence
∇ ( ∇ ⋅ E ) − ∇ 2 E = − μ ∂ ∂ t ( ∇ × H )
Using Equations (9) and (10)
∇ ρ ε − ∇ 2 E = − μ ε ∂ 2 E ∂ t 2 − μ ∂ J ∂ t − μ σ ∂ E ∂ t
∇ 2 E = 1 ε ∇ ρ + μ σ ∂ E ∂ t + μ ε ∂ 2 E ∂ t 2
For source-free region ( ρ = 0 ) and media with negligible electrical conductivity ( σ ≈ 0 ) it becomes:
∇ 2 E − 1 c 2 ∂ 2 E ∂ t 2 = 0 (14)
that is the wave equation in which c = 1 μ ε is the speed of light (299,792,458 m/s).
It can be noticed that solution of the latter equation, that is second order respect to time, is equivalent to solution of the original system made by two first order equations. Passing from time domain to frequency domain:
∇ 2 E + k 2 E = 0 (15)
where
k = ω μ ε = ω c = 2 π f c = 2π λ
The equation assumes the following form:
∇ 2 E + ω 2 μ ε E = 0 (16)
Anyway, once solved for electric field, magnetic field H can be easily found using Equation (9).
After that electric and magnetic fields have been calculated, one can also determine the volumetric unit power generated by the correspondent wave thanks to the so called Poynting theorem.
Poynting vector S [ = ] W m 2 is defined as the vector product E × H and represents the directional energy flux of an electromagnetic field.
Poynting’s theorem is a statement of conservation of energy for the electromagnetic field. In differential form, defining energy density as e:
− ∂ e ∂ t = ∇ ⋅ S + J ⋅ E [ = ] W m 3 (17)
− ∂ e ∂ t = ∇ ⋅ ( E × H ) + J ⋅ E
− ∂ e ∂ t = H ⋅ ∇ × E − E ⋅ ∇ × H + J ⋅ E
Using (8), (9), (10) and (11)
∂ e ∂ t = μ H ∂ H ∂ t + ε E ∂ E ∂ t
∂ e ∂ t = 1 2 ∂ ∂ t ( μ | H | 2 + ε | E | 2 ) (18)
In dielectric materials, the magnetic permeability is usually small and the first term of the equation second side can be neglected; so, shifting to the frequency domain, considering that is ε = ε ′ − j ε ″ and taking only the real component:
∂ e ∂ t = P = 1 2 ω ε 0 ε ″ | E | 2 (19)
To solve the previous partial differential equations, a Finite Elements Method (FEM) [
shown.
Three fluid foods have been considered: skim milk, with a Newtonian behavior and constant viscosity, apple sauce and tomato sauce as non Newtonian fluids, modeled with a power law having different fluid consistency coefficient and flow behavior index [
As ε' and ε" are temperature functions, average values have been obtained by integrating in the entire domain and in the time (range 0 - 100 s). They are fundamental for microwave heating, because they determine respectively the energy absorbed and the fraction converted in heat power.
In the next pages, velocity and temperature profiles and generation term modeled with both Lambert’s law and Poynting theorem are reported.
In all cases, the average velocity is the same (4 mm/s). To practically realize that, as the three fluid present different rheological properties, different pressure drops, reported in
To ensure a comparability between the two heat generation terms, a proper value of electric field has been chosen as boundary condition in wave equation so that the volume integral over all the domain is the same:
∫ V Q L R d V = ∫ V Q M R d V
In this way, the total heat supplied in the different cases is the same. Values of total power P for all cases and correspondent values of electric field in case of Poynting generationare reported in
Skim milk | Apple sauce | Tomato sauce | |
---|---|---|---|
Density, ρ [kg m−3] | 1047.7 | 1104.9 | 1036.9 |
Specific heat, cp [J kg−1 K−1] | 3943.7 | 3703.3 | 4000.0 |
Thermal conductivity, k [W m−1 K−1] | 0.5678 | 0.5350 | 0.5774 |
Viscosity, μ [Kg m−1 s−1] | 0.0059 | ||
Fluid consistency coefficient | 32.734 | 3.9124 | |
Flow behavior index | 0.197 | 0.097 | |
Dielectric constant, ε' | 66.31 | 68.97 | 74.27 |
Dielectric loss, ε'' | 13.26 | 46.42 | 5.30 |
Skim milk | Apple sauce | Tomato sauce | |
---|---|---|---|
Pressure drop [Pa] | 0.1425 | 1080 | 129 |
All the results showed above are obtained for the same fixed time of 100 seconds. It can be seen that skim milk, that is a Newtonian fluid, shows a parabolic shape (
The three fluids show a very similar temperature distribution even though apple sauce values are smaller (
Skim milk | Apple sauce | Tomato sauce | |
---|---|---|---|
P [W] | 640 | 210 | 750 |
E [V/m] | 1060 | 1950 | 850 |
underestimation of heating in case of Lambert’s law use is confirmed also in this case with a different geometry.
Actually, in order to better understand the phenomenon, it has been considered that the fluid elements of each of the three liquids have different velocity profiles along the radial direction (Figures 3-5) and so different residence times depending to the radius. For the fluid elements having higher velocities, near the tube axis, the residence times are lower; for those near the wall, the velocities are lower and so the residence times are longer. In skim milk this difference is more important, whereas in the other two cases the profile is quite flat, especially in the case of tomato sauce. Anyway, outlet section (z = 30 cm) has been fixed and temperature profile evaluated for each position based on the real residence time are overlapped in
It is evident that in the three cases the trend is always the same: in the first part near the axis in which Lambert’s laws overestimates Maxwell results in term of temperature. This is due to Lambert’s generation term that goes to infinite for r = 0 that in turn is due to the pre exponential factor R/r. In the larger middle part, Lambert’s law curve is instead always under Maxwell predictions, save to slightly raise in the last part near the wall.
In both cases temperature increases near the wall because of the longer residence times, but in Lambert’s law case the effect is more evident. After all, apart few points, a good agreement between the two models has been found.
In conclusion, it can be stated that Lambert’s law allows to estimate power generated during a microwave heating process without solving for electric field and so in a relatively simply way, but it presents some problems like the singularity of infinite power for r = 0 and an appreciable underestimation of the real power and so of the temperature distribution. Moreover as it suppose an exponential decay, it can’t forecast the oscillations of the electric field and so of Poynting generation.
On the other hand, solution of Maxwell equations allows to exactly know the real electric field and then, thanks to Poynting theorem, the actual generated power, although it is obviously more expensive in terms of time and resource of calculations.
The authors declare no conflicts of interest regarding the publication of this paper.
Apicella, R. and Romano, V. (2019) Comparison between Lambert and Maxwell Approaches in the Modelling of Microwave Heating of Liquid Foods. Engineering, 11, 1-13. https://doi.org/10.4236/eng.2019.111001