Professional Documents
Culture Documents
www.elsevier.com/locate/jfoodeng
Abstract
The focus of this research is on the concentration of orange juice in agitated thin ®lm evaporators (ATFE). The AspenPlusä
simulation program was used to develop a model of the ATFE. A rigorous heat exchanger model, Heatx followed by the rigorous 2-
phase ¯ash model, Flash2, was used to simulate the dominant eects of the ATFE. The thermo-physical properties of orange juice
are not available in the AspenPlusä databank. They were, therefore, determined experimentally and modelled as functions of
temperature and solid content. Heat transfer coecients were predicted using correlations and measured from process measure-
ments. Experimental and simulation results are presented. The AspenPlusä simulation model using experimentally determined
thermo-physical properties of orange juice compares well with the experimental data from the ATFE pilot plant. Process mea-
surements were reconciled using the optimisation features of AspenPlusTM . Ó 2000 Elsevier Science Ltd. All rights reserved.
much higher ¯ow rates. The speci®cations of the evap- by the rotating wipers and the liquid ¯ows downwards
orator are the same in each case and are as follows: under gravity. The liquid is concentrated by steam that
enters the outer shell of evaporator. The concentrated
Material stainless steel (0.0034 m) liquid ¯ows into tank R1 and leaves the system via valve
Height 4.5 m
Diameter 0.108 m
V5. The vapour formed during the process is condensed
Heat transfer 0.251 m2 in the Condenser and ¯ows to the tank R2 and leaves
area the system via valve V7. A vacuum pump is used to
Agitator blades 4 Luwa ®xed clearance (smooth or meshed) create a vacuum in the ATFE system, valve V6 is used to
control the vacuum pressure in the ATFE system.
The dual system was used to test the model on two
dierent-sized systems as a test for generality. The
ATFE systems were operated under vacuum conditions 3. AspenplusTM simulation of the ATFE process
to reduce the boiling point of the orange juice. The dilute
orange juice feed solution was pumped from the Feed An AspenPlusTM model of the ATFE process was
Tank; the ¯ow rate was controlled by adjusting valve V1. developed using a rigorous heat exchanger model, Heatx
The solution ¯ows through a ¯owmeter and enters the followed by a rigorous 2-phase ¯ash model, Flash2 as
ATFE and is distributed over the inner circumference of shown in Fig. 2, Chuaprasert et al. (1999). The feed to
the heated ATFE cylinder. A thin liquid ®lm is formed the Heatx block represents the dilute orange juice±water
N. Chawankul et al. / Journal of Food Engineering 47 (2001) 247±253 249
feed to the process. Steam enters the Heatx and leaves as These thermo-physical property equations were used
condensate. The product from the Heatx block is a 2- to determine the value of overall heat transfer coe-
phase mixture of water vapour and concentrated orange cient, U, using the models developed by Sae Tae (1999)
juice±water liquid; this stream does not exist in the real as follows:
process. The Flash2 model is used to separate the con- 0:0112 0:02
NuDe 0:00538 Re0:285
De
1 FrDe Pr0:539 K 1:643 ;
6
centrated orange juice from the water vapour. The
products from the Flash2 represent the two products where
from the real process. The pressure in the Flash2 unit " #0:25
was assumed to be the same as in the Heatx unit and m_ avg gavg
De 5:492 ;
7
adiabatic operation was assumed. Adiabatic operation is q2avg B
a reasonable assumption since the feed to Flash2 is a 2-
phase mixture of water vapour and concentrated juice 8 m_ avg
and the purpose of the Flash2 is merely to separate the ReDe ;
8
De gavg g B
two phases.
N 2 D2
FrDe ;
9
De g
4. Physical property equations of orange juice
Cp;avg gavg g
The thermo-physical properties of tangerine orange Pr ;
10
kavg
juice were developed as a function of temperature and
total dissolved solids over the temperature range of 32± L
K ;
11
80°C and dissolved solids range 5±40% w/v by Boons- Cp;avg Tw ÿ Tf;avg
riudomsuk (1999). The correlations are as follows:
% total dissolved solids; S 1:294B 3:2167;
1 Q_ evap
Tw Tsteam ÿ ;
12
A U0
2
density; q 1:001 0:348S ÿ 0:002S ÿ 2:947
1
10ÿ7 T 2 ;
2 U0 ;
13
1
hs
rkww
viscosity; g 3:269 ÿ 2:4592S 17:0113S 2
hp D e
ÿ 0:0673T 3:8 10ÿ4 T 2 ;
3 NuDe ;
14
kavg
specific heat capacity; Cp 1 1 1 rw
ÿ5 2 :
15
3:8325 ÿ 5:423 10 T 0:01027 U hs hp k w
ÿ 3:0486S 2 ÿ 0:68486S;
4 Eqs. (6)±(15) were developed based on the ATFE pilot
plant data operating over the following range of oper-
thermal conductivity; k 0:54689 ÿ 6:886 ating conditions: feed ¯ow rate 60±120 kg/h, steam
10ÿ6 T 2 0:00206T pressure 1±2 bar and agitator speeds of 0±1000 rpm
and using a sugar-syrup solution, Sae Tae (1999). In
ÿ 0:15732S 2 ÿ 0:2776S:
5 our experiments we used tangerine orange juice and
250 N. Chawankul et al. / Journal of Food Engineering 47 (2001) 247±253
operated both the pilot plant and lab scale ATFE under however, in batches 9±13 (pilot plant evaporator), the
the following conditions: simulated product ¯ow rates are somewhat higher than
· rotational speed of the agitator 200±800 rpm; the experimental results. If one compares the dierence
· concentration of feed 3±10 Brix; between experimental and simulated results on a per-
· feed ¯ow rate: centage basis then the high ¯ow rate results are also in
lab scale (batches 1±8) 13±21 kg/h, very good agreement.
pilot scale (batches 9±13) 60±117 kg/h; Fig. 4 shows the vapour ¯ow rate results. Again there
· absolute pressure of steam 1±2 bar. is good agreement in batches 1±8 (lab scale) and some
larger discrepancies in batches 9±13 where we ®nd that
the simulated vapour ¯ow rates are lower than the ex-
5. Results and discussion perimental vapour ¯ow rates. It is worthwhile to note
that all the experimental results were measured inde-
5.1. Predicted heat transfer coecients pendently and therefore may exhibit some mass balance
errors when compared with the feed ¯ow rate and the
Eqs. (1)±(5) were used to determine the physical liquid ¯ow rate.
properties of the orange juice and then the overall heat Fig. 5 presents the experimental and simulated
transfer coecient, U, was predicted from Eqs. (6)±(15). product concentrations using heat transfer coecients
The predicted U-value was then used in the Aspen- predicted using Eqs. (6)±(15). The results are in rea-
PlusTM Heatx model. Although we expect the general sonable agreement for batches 1, 8 and 11±13, however,
form of the heat transfer coecient model, Eq. (6), to in batches 5, 6 and 7 we ®nd large discrepancies between
remain the same for orange juice it is expected that the the simulation and experimental results where the pre-
coecients in Eq. (6) will be dierent for orange juice. A dicted product concentrations are very high. These three
new model is currently under development. experiments were run at the lowest feed ¯ow rates (13±
Figs. 3±5 show the comparison between the predicted 16 kg/h); in this range we had diculty in maintaining
and measured values of product ¯ow rate, vapour ¯ow the ¯ow rates and small absolute errors in the ¯ow rates
rate and product concentration, respectively. From Fig. will lead to large errors in concentration. In addition
3, we can see that the simulation and experimental because the ¯ow rate was so low the orange juice resi-
product ¯ow rate results compare well. In batches 1±8 dence time in the ATFE was high and scalding of the
(lab scale evaporator), the agreement is very good, orange juice on the heat transfer surface of the ATFE
occurred. As a result the measured concentration of
orange juice in the liquid product was abnormally low; it
should have been much higher due to the high heat
transfer rates and large residence time.
of the orange juice concentration is small compared to ®t was obtained from the data reconciliation. One can
the standard deviation of other variables. Therefore we think of the reconciled values (d) as improved experi-
set the weighting factor for orange juice feed concen- mental values that are both close to the actual experi-
tration to be a high value, 50, relative to the other mental values (m) and also satisfy all the mass and
weighting factors to force the optimiser to adjust other energy balance constraints. Once the data have been
process variables rather than the orange juice feed reconciled to satisfy the heat and material balances, we
concentrations. There was not enough information to can use these adjusted data for further studies. For ex-
calculate the standard deviations therefore it was de- ample, the operation of the evaporator may now be
cided to ignore standard deviations and use only studied to determine the eect of alternate operating
weighting factors. The objective function was therefore procedures to improve production or reduce energy
written as consumption.
2
OF 1
feed flow rate ÿ Ff
50
feed concentration of orange juice ÿ Xf 2 8. Conclusions
2
1
steam flow rate ÿ Fs
1
product flow rate ÿ Fp 2 1. A steady-state simulation model of an ATFE for con-
centrating tangerine orange juice was developed using
1
product concentration of orange juice ÿ Xp 2 AspenPlusTM .
2
1
vapour flow rate ÿ Fv :
18 2. The simulation results show good agreement with the
experimental results when using overall heat transfer
One cannot directly manipulate all decision variables in coecients predicted from an independent correla-
AspenPlusTM . For example, in our case one cannot di- tion and better agreement when using eective heat
rectly manipulate the product ¯ow rate and product transfer coecients calculated from experimental
concentration. Therefore manipulated variables that will process measurements.
indirectly adjust or manipulate these variables must be 3. Low and dicult to control feed ¯ow rates lead to
selected. Four manipulated variables were chosen to large errors in the prediction of the product concen-
minimise the objective function, Eq. (18): trations due to scalding of the orange juice on the
· ¯ow rate of water in feed, heat transfer surface of the ATFE.
· steam ¯ow rate, 4. The steady-state AspenPlusTM simulation model
· heat transfer coecient, combined with a built-in optimisation routine was
· ¯ow rate of orange juice component in feed. used to reconcile the experimental data gathered from
The adjustment of ¯ow rate of the orange juice the evaporator. After data reconciliation a signi®cant
component in the feed and the ¯ow rate of the water in improvement in the ®t of the measured data was ob-
the feed has the combined aect of adjusting the total served. On average the ®t between measured values
feed ¯ow rate and composition of the feed simulta- and the process model was increased by about 60%.
neously. The steam ¯ow rate was adjusted to try and
match the evaporator temperature and product con-
centration. The overall heat transfer coecient was ad-
justed to try and match the concentration of orange Acknowledgements
juice in the product stream. Therefore, solution to the
problem using AspenPlusTM can be written as The authors are very appreciative of research funding
from The National Science and Technology Develop-
choose ¯ow rate of water in the feed ment Agency of Thailand (NSTDA) which enabled this
¯ow rate of orange juice component in the research to be undertaken.
feed
steam ¯ow rate
heat transfer coecient
to minimise Eq. (18) such that h(x) 0 References
Piccolo, M., & Douglas, P. L. (1996). Data reconciliation using ment of Chemical Engineering, King Mongkut's University of
AspenPlusTM . Developments in Chemical Engineering & Mineral Technology Thonburi, Bangkok, Thailand, (in Thai).
Processing, 4(3/4), 157±182. Tjoa, I. B., & Biegler, L. T. (1991). Simultaneous strategies for data
Sae Tae, A. (1999). Heat transfer coecients in an agitated thin ®lm reconciliation and gross error detection of nonlinear systems.
evaporator for concentrating sugar syrup. Masters Thesis, Depart- Computers and Chemical Engineering, 15(10), 679±690.