Dynamic Simulation and Control of Conveyorbelt Dryers

This article was downloaded by: [Virginia Tech Libraries]
On: 25 February 2015, At: 08:39

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Drying Technology: An International Journal

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/ldrt20
DYNAMIC SIMULATION AND CONTROL OF CONVEYOR-

BELT DRYERS
a a a
C. T. Kiranoudis , Z. B. Maroulis & D. Marinos-Kouris
a
Department of Chemical Engineering , National Technical University , Athens, GR-15780,
Greece
Published online: 31 May 2007.
To cite this article: C. T. Kiranoudis , Z. B. Maroulis & D. Marinos-Kouris (1994) DYNAMIC SIMULATION AND CONTROL OF
CONVEYOR-BELT DRYERS, Drying Technology: An International Journal, 12:7, 1575-1603, DOI: 10.1080/07373939408962188
To link to this article: http://dx.doi.org/10.1080/07373939408962188
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
DRYING TECHNOLOGY, 12(7), 1575-1603 (1994)
DYNAMIC SIMULATION AND CONTROL

O F CONVEYOR-BELT DRYERS
C.T.Kiianoudis, Z.B.Maroulis, D.Marinos-Kouris

Downloaded by [Virginia Tech Libraries] at 08:39 25 February 2015
Department of Chemical Engineering, National Technical University,

GR-15780.Athens, Greece
I
ABSTRACT
The dynamic behavior of conveyor-belt dryers involving externally controlled
heat and mass transfer phenomena has been studied via digital simulation. The
investigation concerned an indusmal dryer used for the moisture removal from wet
raisins. The dryer consisted of three drying chambers and a cooling section, all
involving the same conveyor belt. For each chamber, perfect temperature control
was assumed for the drying air temperature, while its humidity was left
unconw1led. The effect of material temperature and moisture content at the entrance
of the dryer and the drying air temperature on material tempenture and moisture
content at the exit of the dryer and the corresponding drylng air humidity, have been
explored by step forcing the disturbance and manipulated variables in the non-linear
dryer model simulator. Results showed that material moisture content at the exit of
the dryer is greatly affected by material moisture content at the enannce as well as
by the drying air temperature. Reliable aansfer functions for each process module
were obtained by fitting sevenl transfer function models on the simulated data
using a least-squares approach. It was found that when input material moisture
content could be instantly measured, the system responded slowly enough so that
excellent conml could be achieved for material moisture content at the exit of each
chamber. In this case a PI-feedback cascade tempenture conmller was used. When
a 15 sec delay measuring sensor was introduced, poor performance was observed.
A simplified lead-lag feedforward controller, added to the system, in conjunction
with the primary PI-feedback cascade controller, resulted in good control
performance of the delay sensor system.
INTRODUCTION
During the last decade, an increase in energy costs as well as the adoption of
more strict safety and environmental regulations have sparked a growing interest in
designing energy-saving systems throughout the conventional food industry.
1575
Copyright Q 1994 by Marcel Drkker. Inc.

I576 KIRANOUDIS, MAROULIS, AND MARINOS-KOURIS
Furthermore, in the case'of dehydration plants, the wide variety of final products,
which are today available to the consumer (snacks, dry mixes and soups, dried
fruits and vegetables, etc.) increases the concem to meet high quality specifications.
Therefore, the need of optimal management of energy during drying, together with
the demand for the production of dehydrated foods of excellent quality, leads to the
development of control strategies for the drying plants studied. In most cases
encountered in practice, material moisture content is the crucial process variable
affecting both energy conservation efficiency and final product quality. As a result,
the decisions made by the controllers are greatly influenced by the way this
important variable varies during drying.
The control of dryers is probably one of the less studied areas of process
control. This may be atmbuted to various reasons. To begin with, the dynamic
modeling of drying processes, tends to be elaborate due to the complexity of '1ood
composition and structure. The heat and mass aansfer models describing the
dynamic behavior of drying equipment, comprise systems of highly non-linear
partial differential equations, which involve m s f e r coefficients and thermophysical
properties strongly related to the material temperature and moisture content. As a
result, they have to be solved using complex numerical techniques. Moreover,
direct on-line measurement of material moisture content for a wide range of
dehydrated products is difficult from the practical point of view, due to the lack of
accurate and reliable sensors. In this case, we must always take into consideration
the possible sensor system inadequacies (Hambert, 1974, Fuentes and Luyben,
1983, Goldstein et al., 1991, Nickolls and de Los Santos, 1991, Kiranoudis et al.,
1993).
Shinskey (1979) proposed an inferential control algorithm in order to
compensate for the ill-measured material moisture content in commercial fluid bed
dryers. Cazaubon et al. (1980) presented a control scheme for a discontinuous
dryer with aucks and trays for fruits. An optimization problem that had to satisfy
'
some process constraints was solved on-line, and the microprocessor-system used
changed the set points of the controller according to the results obtained by the
optimization system. The entire adaptive algorithm was further explained and
tested. Yliniemi et al. (1982) proposed a feedforward control scheme for drying, in
the case of rotary dryers. The experiments were carried out i n a pilot-plant dryer
and showed that even 2 0 4 of total energy savings could be achieved. Forbes et al.
(1984) outlined the development of control stntegies for commercial corn dryers. It
was shown that model based controllers performed better than conventional
feedback and feedforward control systems, and proposed a method for improving
these model based controllers. Mittal et al. (1984) reviewed the management and
control systems employed and investigated, for low and ambient temperature corn
drying. Characteristics and performance of a microcomputer based conuol system
CONVEYOR-BELT DRYERS 1577
were also described. Ouhab and Le Pourhiet (1985) used a dynamic model and an
optimization algorithm in order to determine on-line optimum operating conditions
and total processing time for a multistage gnin dryer. Eltigani and Bakker-Arkema
(1987) reviewed all previous conaol efforts in the case of crossflow grain dryers.
They also developed an unsteady-state mathematical model of the process, and
fitted the results obtained in an empirical parameter adaprive equation. They used
both feedback and feedforward control configurations as well as a combination of
them, in order to propose a reliable material moisture content controller which
avoids the long deadtimes and the frequent and large load upsets. Panda and
Ramachandra Rao (1991) reviewed the prevailing modeling, identification
techniques, and control strategies on fluidized bed h n g .
The purpose of this paper is to explore the dynamics and controllability of
conveyor-beJt dryers involving externally conuolled drying processes. The dynamic
behavior was investigated by linear analysis and by digital simulation of a nonlinear
mathematical model. The dynamic responses of the open-loop system for changes
in various manipulated and disturbance variables were fust studied in order to gain
some qualitative and quantitative insight into the dynamic difficulties associated
with the conaol of these dryers. Several types of closed-loop control systems were
then invesagated.
CONVEYOR-BELT DRYER SYSTEM
Drying is externally controlled if heat and mass uansfer takes place in the air
phase only. In this case, heat and mass transfer depend mainly on the
corresponding transfer coefficients, which, in contrast to internally controlled
drying, must be accurately known. Take for instance the case of dryers in which air
passes through the wet material which consists of very small panicles so that
resistance to heat and mass transfer within them can be ignored. In other words, the
diffusion of heat and mass in the particles, is very large, and the process is
conaulled by the heat and mass mansfer in the air boundary layer at the solid-air
interface. In this case, both material moisture content and temperature are uniform
in the solid, while the corresponding quantities for air water activity and
temperature are not.
The dryer system examined in this paper, concerns moisture removal from
wet raisins undergoing a sulfidation procedure in order to be sterilized. The dryer
consists of four equal sized drying chambers. The perforated conveyor belt moves
through all chambers in which air is circulating. Drying of wet raisins takes place in
the f m t three chambers, in which air is heated by hot oil heat exchangers. The last
chamber cools the material down to ambient temperature by circulating air cooled by
cold water heat exchangers. Recirculating air is heated by passing over oil heat
1578 KIRANOUDIS. MAROULIS, AND MANNOS-KOURTS
exchangers and is fed under the conveyor belt The dryer configuration dong with a
typical chamber interior is shown in Figure 1.
Each chamber is equipped with an elementary temperature control system
which is also shown in Figure 1. A typical feedback control configuration is used in
order to adjust the drying air temperature to set point levels. We assume here small
time lags for temperature measurement and response of hot oil valve in the
exchanger, so that perfect temperature control is achieved. This is practically m e in
real world problems since air dynamics are essentially negligible with respect to
those of dried material, due to the fact that the extremely high air circulation levels,
which guarantees uniform axial drying of product. impose very small retention
times for air throughout the dryer. Therefore, no capacitance is assumed for air
streams participating in the process.
Since conveyor-belt dryers are made up of several drying chambers, the
overall mathematical model wiIl be formed by repetition of the individual
component ones. The mathematical model of a drying chamber involves heat and
mass balances of air and product streams, as well as heat and mass transfer
phenomena that take place during drying.
On the basis of the forementioned assumptions, the mass transfer equation for
the dried product is expressed as follows:
The water activity difference between the bulk air mass, aw, and the interface,
a w ~ was
, chosen as the mass transfer driving force in the air boundary layer
(Kiranoudis et al., 1990, Maroulis et al., 1991). The corresponding heat transfer
equation for the dried product is given below:
The mass balance in the d i n g comp&ent is given by the equation:
The average spatial material moisture content derivative implies integration of

material moisture content over the entire belt length.
The overall heat balance in the drying compartment is given by the equation:
The average spatial material specific enthalpy derivative implies integration of

material specific enthalpy content over the entire belt length.
CONVEYOR-BELT DRYERS
Air Outlet
I I
I Controller
TAc(manipulotive variable)
~TP, Ts
(controlled rmable)
FIGURE 1. Conveyor-bell dryer for moisture removal from wet nisins

1580 KIRANOUDIS, MAROULIS, AND MARINOS-KOURIS
Humidiry of mixed recirculation and fresh air streams can be calculated by the
following mass balance:
Temperature of mixed recirculation and fresh air streams can be calculated by

the following heat balance, through its specific enthalpy expression:
(FA, - F,,)hA + FAhAo = FAC~AM (6)
Heat balances at the heat exchangers of the chamber, are given by the
following equations:
The drying air water activity, awe, involved in Equation (1) is calculated as
follows:
XACP
awc =
(1,+ X A c ) p O
The specific enthalpy of an air stream, h ~as, a function of its temperature,
TAPand absolute humidity, XA, is given by the following equation:
h* = cpATA + XA (mo+ C p v T A )
(11)
The specific enthalpy of a product stream, h A , as a function of its

temperature, Ts, and moisture content, Xs,can be calculated as follows:
hs = cmTS+ XscWTs
(12)
Equations (1)-(Z), include the volumetric transfer coefficients for mass, kM

and heat, kH, which determine the drying rate (Maroulis et al., 1991).
The interfacial quantity awE (the water activity of air in equilibrium with the
wet material) is in general a function of material moisture content and temperature.
In the case of raisins, the fairly well-known GAB (Guggenheim-Anderson-deBoer)
equation can be used (Maroulis et al., 1988, Tsarni et al., 1990):
2 + (X, / X, - l)C - ([2 + (X, I x, - 1)c12- 4(1- c))1'2

am = 2K(1- C)
C = C,exp(AH, / RT, ) (14)
K = K,exp(AH, / RT,) (15)
Water vapor pressure can be computed by means of an Antoine equation of

the following form:
The latent heat of vaporization is given by the

Clausius-Clapeyron equation which makes use of the previous vapor pressure
d l n P ~ equation:
AH, =-R-
d(l I T * )
(17)
Equations (1)-(17) constitute the mathematical model of the m s i e n t behavior

of the dryer. It contains both algebraic and partial differential equations (PDEs). In
the simulation runs throughout the paper, the PDEs were discretized using the
control-volume method described by Patankar (1980). The remaining odrinary
differential equations (ODEs) were integrated with respect to time, using the GEAR
procedure for systems of stiff ODEs, as implemented in the computer code
DGEAFUIMSL. Within each gndient function evaluation requested by DGEAR, the
system of the remaining algebraic equations was solved by means o l a Newton-like
method using a Broyden-type approximation for the calculation of the Jacobian
mamx for the system. The average spatial material moisture content and specific
enthalpy derivatives throughout each drying chamber were caIculated for each step
by numerical integration using the method of Simpson (Finlayson. 1980).
The dryer examined, involved equal sized drying chambers which were 2 m
wide and 4 m long. Each chamber involved a series heat exchanger units totaling 50
m'. The conveyor-belt through them moved with a speed of 0.028 4 s . The raisins
were brought in each chamber with a rate of 0.167 kg/s db, while the drying air
circulation through the product was 1.67 kg/s db. The hot oil, entered the heat
exchangers of each chamber at a temperature of 200C,while the corresponding
cold water temperature was 10C. The total heat ~ a n s f e coefficient
r for hot oil
operation was 0.45 kW/mzK, while for cold water operation was 0.4 kW/m2K.
The operaiional process variables at the nominal steady-state operation point are
sumrnarised on Table 1.
TABLE 1. Nominal steady-state operational variables for the dryer exknined.

Variable Chambers
1 2 3 4
Xso (kglkg db) '
Xs (kglkg db)
Tso ("C)
Ts (OC)
XAC(kg/kg db)
TAC( O C )
XA (kglkg db)
TA(OC)
TAM("C)
FA (kglrnin db)
Q (kllmin)
To2 ("C)
Tw2 ("C)
Fo(kg/min)
Fw (kgimin)
OPEN-LOOP DYNAMIC BEHAVIOR
The open-loop dynamic behavior of dryingchambers was indentified by step

forcing the load and manipulative variables in the non-linear model simulation.
Tests were performed for load process variables (material moisture content and
temperature) and the manipulative process variable (drying air tempenture), while
output process v~ables~(conuolled output material moisture content, uncontrolled
output material tempenture and drylng air humidity) were simultaneously recorded.
The relative deviation from nominal steady-state of each recorded variable (bXs,
6Ts, bXAC)was subsequently calculated. The relative deviation from nominal
steady-state of variable XAC(axAC),is given by the following equation:
where xACNSS is the nominal steady-snte value of variable XAC.In the same way
all other relative nominal deviations can be calculated. Both positive and negative
steps were performed, at forcing levels of 10%. 5% and 1%, for all manipulative
and load variables and for every drying chamber involved. In all examples, the
recorded relative nominal deviation for the output variables, was exactly the same.
regardless of the input forcing level. The responses were also symmemc as far as
the sign of forcing is concerned. Therefore we conclude that as long as the input
variable flucmations remain relatively small, linear performance of the system can
be assumed for the specific operating point examined.
Both positive and negative forced relative nominal deviations tha! were
recorded for the fmt drying chamber are presented in Figure 2. The variable which
greatly influences all others is the load material moisture content The one that plays
a minor role on the dynamic deviation of the system is the load material
temperature. The influence of the manipulative variable of the system. that is to say
of the drying air temperature (assumed perfectly controlled), lies somewhere in the
middle, Drying air humidity is greatly and irregularly affected by input variables.
fluctuations. Output material moisture content and temperature seem to depend on
input variables in a more straightforward way. More specifically, the steady-state
gain for output material moisture content exceeds 120% for load material moisture
content fluctuations, and is throttled down to 30% for drying air temperature
changes. We note that dead time is observed in many output responses. This can be
attributed to the fact that change in material moisture content or temperature is not
generally expected to be sensed immediatelly at a certain point of the conveyor-belt
because of the large retention time of solids in the dryer. However, certain entrance
fluctuations can be sensed by measuring exit quantities. This can be explained by
the fact that drying air humidity, that is lumped for every point in the chamber,
serves as a medium of information transfer between disnetized beat and mass
balance equations. The fluctuations of this quantity are indeed complicated. They
tend to strongly amplify the input signal at very high frequencies (corresponding to
initial immediate responses). This may turn out to be a problem in the controller
design procedure, since this variable is uncontrolled and may lead to unexpected
crash features of the control system. An overall performance of the conwller
should reduce and throttle this dangerous system characteristic.
In Figure 3, the positive step responses for all chambers are presented
together, so that they can be compared. Negative step responses are not included
since they are symmetrical to the latter. We notice that the output material moisture
content is niainly affected by the load material moisture content (which is the load
variable of the SISO system) and drying air temperature (which is the manipulative
variable). The responses of these variables are not affected by the working
operating point They are essentially the same for all chambers examined.
A linear aansfer function model was then determined in order to design the
closed-loop c o n m l system. First-order transfer functions with dead-time, and
second-order transfer functions with and without dead-time were fitted to each
response using a least-squares approach. The gains in the transfer functions were
not fitted but were determined from the difference between the initial and find
values. In determining the best transfer function model to be used. it was desired to
KIRANOUDIS, MAROULIS, AND MARINOS-KOURIS
0 2 4 6 8 10
t (min)
4
3
0 2 4 6 8 10
t (rnin)
FIGURE 2.Open-loop system positive (solid line) and negative (dashed line)
responses for process vviabIes of the 1st chamber.
CONVEY OR-BELT DRYERS
-0.02
0
2
4
t
FIGURE 2. (cont.)
(rnin)
6
8
10
KIRANOUDIS,MAROULIS,AND MARINOS-KOWS
0.4 .
- _ _ _--- - - - - - _ _ -
0.2 . /
I
,/
/
6XI
bT*c 0 .
-0.2 -
-0.4
0 2 4 6 8 10
t (min)
2 .
6TAc \
\
-.
-1 -
\
---- - - _ - -_ _ _ _ _ _ _ _ .
-2
0 2 4 6 8 10
t (min)
2 -
6XAC hTAc o -
-1 .: /
___------------
--'
-2
0 2 4 6 8 10
I (rnin)
1.5
6x2 6X,
I (rnin)
t (min)
t (rnin)
FIGURE 3.Open-loop system posinve responses for process variables of the 1st
(solid line), 2nd (dashed line) and 3rd (dotted line) chamber.
FIGURE 3. (cont.)
KIRANOUDIS, MAROWLIS, AND MARINOS-KOURIS
-0.41
0
2
4
t
FIGURE 3. (cont.)
6
(min)
8
1590 KIRANOUDIS,MAROULIS,AND MARINOS-KOURIS
have a model with the smallest number of fitted parameters which would adequately
represent the drying chamber dynamics. The F-distribution method discussed by
Weischedel and McAvoy (1980) was used to discriminate between various transfer
function models which were fitted. The level of significance chosen for the F-test
was 99.5%.
The proposed procedure resulted in a second-order transfer function with
dead-time for load material moisture content and in a first-order transfer function for
drying air temperature. Gains and time constants are given in the following
formulation, which can be safely used for all drying chambers involved: ,
where all time constants are given in minutes.

The open-loop responses of Figure 2, give us some insight into the control
problems to be expected within drying chamber units. The fast dynamic response
of these units, requires a fast control system in order to avoid large changes in
product material moisture content. The unexpected behavior of uncontrolled drying
air humidity must be accounted for, in order to avoid a control system crash.
Unfortunately, in most cases material moisture content measurements depend on
devices which introduce a significant deadtime into the control loop (ie. 15 s). Thus
a basic problem is one of being unable to detect the disturbance fast enough for the
controller to act
CLOSED-LOOP CONTROL
Feedback PI controllers were designed for each chamber involved, so that
material moisture content at the exit of each chamber is controlled by the
corresponding drying air temperature. The closed-loop system used for each
chamber is shown in Figure 4. The process, load, feedback controller, and sensor
transfer functions are also included in Figure 4. In this section no dead-time is
assumed for the material moisture content sensor device. A more realistic approach
for the process sensor system will be examined in the following secrion.
Feedback PI controller tuning was performed using the maximum closed loop
modulus technique described by Luyben (1990). It is a frequency response
technique which examines the behavior of the overall servo closed-loop system
transfer function. If it was possible to achieve perfect conwl, the ideal closed-loop
servo fuction would be unity, and the closed-loop uansfer fuction would be zero.
In other words, we want the output to hack the setpoint perfectly for all
frequencies. and we want the output to be unaffected by the load disturbance for all
frequencies. However, this is practically impossible for all control systems
examined. In this case, most empirical controller design techniques
FIGURE 4. Feedback closed-loop control system used for each chamber (all time
constants in uansfer functions are in minutes).
(Ziegler-Nickols tuning, Cohen-Coon settings) try to approximate the closed-loop

servo m s f e r function as a second order underdamped system. This suggests that
the log modulus plot of the closed-loop servo transfer function will exhibit a
maximum value at a high resonant frequency and will decay rapidly as frequency
increases. In Ziegler-Nickols tuning, criterion to this critical frequency is the
open-loop crossover frequency of the system examined. In most systems, the
closed-loop servo log modulus curves move out to higher frequencies as the gain of
the feedback controller increases. This is desirable, since it means a faster
closed-loop system. However, the height of the resonant peak increases as the
controller gain is increased. This means that the closed-loop system becomes more
underdamped. A commonly used maximum closed-loop log modulus specification
is +2 dB. The controller parameters are adjusted so as to give a maximum peak in
the closed-loop servo log modulus curve of +2 dB. This corresponds to a
magnitude ratio of 1.3 and is approximately equivalent to an underdamped system
with a damping coefficient of 0.4.
Since the process transfer function is of the first order, we can select various
controller settings that would satisfy the maximum closed-loop log modulus
criterion presented in the previous paragraph. For frequencies greater than 0.83
minmlthe corresponding best controller gain and integral reset times are presented in
Figure 5. While conmller gain increases monotonically while resonant frequency
increases, the corresponding reset time exhibits a maximum of 0.37 min at a
resonant frequency of approximately 1.7 min-I and subsequently decays to zero.
KIRANOUDIS, MAROULIS, AND MARINOS-KOUIUS
FIGURE 5. Optimum conmolier gain (solid line) and integral reset time (dashed
.line) as a function of loop resonant frequency for the 1st chamber
control loop.
20 1
Log Modulus (dB)
-40.
-60.
-80 -
0.1 1 10
\
100
o (min-')
FIGURE 6. Closed-loop servo (solid line) and load (dashed line) transfer functions
for a loop resonant frequency of 1.7 mine'.
0 4 8 12 16 20
t (min)
.FIGURE 7. Closed-loop system behavior for a -5% step impose of materid

moisture content at the entrance of the system using no dead-time
sensor.
KIRANOUDIS. MAROULIS, AND MAFUNOS-KOUFUS
0 4 8 12 16 20
t (min)
0 4 8 12 16 20
t (min)
FIGURE 7. (cont.)
-0.2
0 50 100 150
t (min)
FIGURE 8. C losed-loop system behavior for a -5% step impose of material

moisture content at the enmnce of the system using 15s delay sensor.
The controller parameters used, are very loose.
-8
0 50 100 I50
t (min)
FIGURE 8. (cont.)
FIGURE 9. Feedback closed-loop and feedforward control system used for each
chamber (all time constants in m s f e r functions are in minutes).
Resonant frequencies below 0.83 min-' suggest negative controler gains and are,
therefore, physically unrealistic. The log modulus curves for both servo and load
closed-loop system transfer functions for this critical resonant frequency are
presented in Figure 6. These plots are typical of systems encountered in practice.
The load transfer function also exhibits a maximum at the selected resonant
frequency, the gain is lower than that of the corresponding servo transfer function,
and decays smoothly at lower and higher frequencies.
On the basis of the above, very aght controller settings could be used when
no dead-time for the sensor system is taken into consideration. We have chosen a
gain of 94.3 and an integral reset time of 0.0603 min, which correspond to a
resonant frequency of 20 rnin-l. The system response to a -5% step input of
material moisture content for the first drylng chamber is presented in Figure 7. Four
system variables are considered, namely, material moisture content and temperature
at the exit of the chamber, and drying air tempenture and humidity in.the chamber.
Material moisture content is perfectly controlled while all variables exhibit an
underdamped behavior which decays smoothly very quickly. We note that both
drying air temperature and humidity are greatly affected by the conml system and
the final control performance was influenced by both parameters in a simuItaneous
way. Furthennore, the drying air humidity does not exhibits great fluctuations as in
the case of open-loop system response, so that all h m f u l frequencies are filtered
by the control system.
KIRANOUDIS,MAROULIS,AND MARINOS-KOURIS
.O 4 8 12 16 20
t (min)
FIGURE 10. Closed-loop system behavior for a -5% step impose of material
moisture content at the e n m c e of the system using 15s delay sensor
(dashed line) and non dead-time sensor (solid line).
FIGURE 10.(cont.)
REALISTIC SENSOR SYSTEM
The introduction of 15 s dead-time in the material moisture content measuring

system makes the problem more realistic from the practical point of view.
However, the controller used in the previous section exhibits unfavorable
performance when tested in the new process configuration. In particular, the
temperature controller tracked tightly the material moisture content fluctuations at
the exit of the chamber and created enormous deviations in drying air temperature
and humidity in the chamber. The system was led in a highly non-linear region of
performance which definitely was not desired. A remedy, might be to skillfully
loose the feedback controller settings, but this action imposed undesirable
oscillatory behavior as shown in Figure 8. In this example, a very loose resonant
frequency of 1.67 min-' was chosen, resulting in a controller gain of 0.806 and an
integral reset time of 0.074 rnin.
We then introduced a feedfonvard controller which affected the drylng air
temperature by measuring the material moisture content at the entrance of the
chamber. The total control scheme is presented in Figure 9. The feedfonvard
transfer function used was selected by fitting a lead-lag controller module in the
perfect feedforward controller function. The resulting transfer function is also
included in Figure 8.
The system was then tested by imposing a -5% step in material moisture
content at the entrance of h e chamber. The recorded responses of system variables
are presented in Figure 10, together with the responses of no dead-time tight control
system for comparison. The previously mentioned loose conh-oller settings for the
feedback PI controller were used. The performance was greatly impoved and is
comparable to the one recorded for the no dead-time system. The underdamped
frequency response is smaller, but this is definitely atmbu~edto the loose controller
settings for the feedback system
CONCLUSION
This study indicates that feedback PI controllen can be used for the material
moisture content control in the case of conveyor-belt dryers performing
externally-controlled drying. Individual con~ollerscan be introduced for each
drying chamber studied. When no dead-time is considered for the material moisture
content measuring system, the controller setting c m be very tight and this results is
excellent material moisture content control performance and minor fluctuations for
drylng air temperature and humidity within the drying chamber. Introduction of
dead-time sensor systems necessitates the use of a feedfonvard controller in
addition to the feedback one.
NOMENCLATURE
Area of heat exchanger in each chamber (m2)

Water activity of air srrearn entering the prcduct
Equilibrium water activity of air s m leaving the product
Feedback conuoller transfer function
Specific heat of air (kl/kgK)
Specific heat hot oil (WikgK)
Specific heat of vapor (kJ/kgK)
Specific heat of water (IdkgK)

Flowrate of fresh air stream (kgfs db)
Flowrate of product saearn (kgls db)
Flowrate of drying air stream (kds db)
Flowrate of steam (kgls)
Load aansfer funcaon
Feedforward controller mansfer function
Sensor transfer function
Process aansfer function
Specific enthalpy of wet air stream (kJkg)
Specific enthalpy of drying air stream (kJ/kg)
Specific enthalpy of mixed recirculation and fresh air streams
Wkg)
Specific enthalpy of wet product (kT/kg)
Heat transfer coefficient at the b o u n d 6 layer (kWJkgK)
Mass transfer coefficient at the bound& 1iyer (11s) '
Length of drying chamber conveyor-belt (m)

Total pressure (!@a)
Water vapor pressure (kPa)
Exchanged heat rate at each chamber (kW)
Ideal gas constant (kJ1molK)
Residence time (s)
Temperature of mixed recirculation and fresh air streams (T)
Temperature of fresh air stream ( O C )
Temperature of rejected air stream (OC)
Temperature of drying air stream ( O C ) ,
Temperature of hot oil at the entrance of heat exchanger (OC)

Temperature of hot oil at the exit of heat exchanger ('0
Temperature of product stream on entering the chamber (OC)
Temperature of product stream on leaving the chamber (OC)
Overall heat transfer coefficient at heat exchangers of the chamber
OrW/m2K)
Conveyor-belt valocity (rn/s)
Spatial coordinate (m)
Absolute hurnidty of fresh air stream (kg~kgdb)
XA Absolute humidity of rejected &r scream (kgkg db)

xAC Absolute humidity of drying air stream (kg& db)
XSO Material moisture content of product stream on entering the
chamber (kg/kg db)
XS Material moisture content of product stream on leaving the
chamber (kg/kgdb)
Greek Letters
AHs Latent heat of vaporization ( W g )

AH0 Latent heat of vaporization at reference temperature ( W g )

XB Mass ratio of water and air molecules
Equation parameters
XM,CO,KO. Parameters of the GAB equation (13)-(15)

AHc,AHK
AlsA2,A3 Paramaters of the Antoine equation (16)
REFERENCES
Cazaubon, C., Benin, R. and Blazquez, M., 1980, %ward automatic control
by microprocessor for convective dryer systems, pp. 107-1 11, in A.S.Mujumdar
(ed) Drying '80, Hemisphere Publishing Corporation, NY.
Eltigani, A.Y. and Bakeer-Arkerna, F.W., 1987, Automatic control of

commercid crossflow grain dryers. Drying Technology, 3 4 ) 561-575.
Finlayson, B., 1980, Non-Linear analysis in chemical engineering,
McGraw-Hill Publishing Company, NY, 48 pp.
Forbes. J.F.,Jacobson, B.A., Rhodes, E. and Sullivan, G.R., 1984, Model
based control strategies for commercial grain drying systems. Canadian Journal of
Chemical Engineering, 62(5) 773-779.
Fuentes, C. and Luyben, W.L., 1983. Control of high-purity distillation
columns, Indusmal Engineering Chemistry Process Design and Development. 22
(3) 361-366.
Goldstein, N., Bien, F. and Gersh, M.E., 1991, Multipoint fiber optic
humidity monitor, Drying Technology. 9(4). 833-847.
Hambert, F.C., 1974, Automatic control of drying processes. Moisture
measurement and control by the teinperature difference method, Chemical
Engineering Science, 29(5), 888-890. '
Kiranoudis, C.T., Maroulis, Z.B. and Marinos-Kouris, D., 1990, Mass

transfer modeling for Virginia tobacco curing. Drying Technology, 8(2), 351-366.
CONVEYOR-BELT DRYERS I603
Kiranoudis, C.T., Dimitratos, J., Maroulis, Z.B. and Marinos-Kouris, D.,

1993, State estimation in the batch drying of foods, Drying Technology, 11(5),
1053-1069.
Luyben. W.L., 1990, Process modeling, simulation and conuol for chemical
engineers, McGraw-Hill Publishing Company, NY, 474 pp.
Maroulis, Z.B., Tsami, E., Marinos-Kouris. D. and Saravacos. G.D., 1988,
Application of the GAB model to the moisture sorption isotherms for dried fruits.
Joumal of Food Engineering, 7(1) 63-78.
Maroulis, Z.B., Kiranoudis. C.T. and Marinos-Kouris, D., 1991,

Simultaneous estimation of heat and mass transfer coefficients in externally
controlled drying, Journal of Food Engineering, 14(3) 241-255.
Mittal, G.S., Otten, L. and Brown, R.B., 1984, Management and control
systems for low temperature corn drying, pp. 210-217, in A.S.Mujumdar (ed)
Drying '84, Hemisphere Publishing Corporation, NY.
Nickolls. C.I. and de Los Santos. A., 1991, Hydrogen transient nuclear
magnetic resonance for industrial moisture sensing, Drying Technology, 9(4),
849- 873.
Ouhab, R. and Le Pourhiet, 1985, Optimal drying conditions of grains, pp.
428-433, in A.S.Mujumdar (ed) Drying '85, Hemisphere Publishing Corporation,
NY.
Panda, R.C. and Ramachandra Rao, V.S., 1991, Fluidized bed dryers:
Dynamic modelling and control, Chemical Engineering Technology, 14(2)
307-3 10.
Patankar, S.V.. 1980, Numerical heat transfer and fluid flow, Hemisphere
Publishing Corporation. NY.
Shinskey, F.G.,1979. Process control systems, McGraw-Hill Publishing
Company. NY. 322 pp.
Tsarni, E., Marinos-Kouris. D. and Maroulis, Z.B., 1990, Water sorption
isotherms of raisins, currants, figs, prunes and apricots, Journal of Food Science,
55(6) 1594-1597.
Weischedel, K. and McAvoy, T.J., 1980, Feasibility of decoupling in
conventionally controlled distillation columns, Indusmal Engineering Chemistry
Fundamentals, 19(4) 379-384.
Yliniemi. L.. Uronen. P. and Leiviska, K., 1982, Control and economical
examination of a rotary dryer, pp. 230-237, in A.S.Mujumdar (ed) Drying '82,
Hemisphere Publishing Corporation, NY.

Dynamic Simulation and Control of Conveyorbelt Dryers

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dynamic Simulation and Control of Conveyorbelt Dryers

Uploaded by

Copyright:

Available Formats

This article was downloaded by: [Virginia Tech Libraries]

On: 25 February 2015, At: 08:39

Drying Technology: An International Journal

DYNAMIC SIMULATION AND CONTROL OF CONVEYOR-

To link to this article: http://dx.doi.org/10.1080/07373939408962188

PLEASE SCROLL DOWN FOR ARTICLE

DYNAMIC SIMULATION AND CONTROL

C.T.Kiianoudis, Z.B.Maroulis, D.Marinos-Kouris

Department of Chemical Engineering, National Technical University,

Copyright Q 1994 by Marcel Drkker. Inc.

CONVEYOR-BELT DRYER SYSTEM

The mass balance in the d i n g comp&ent is given by the equation:

The average spatial material moisture content derivative implies integration of

The average spatial material specific enthalpy derivative implies integration of

FIGURE 1. Conveyor-bell dryer for moisture removal from wet nisins

Temperature of mixed recirculation and fresh air streams can be calculated by

The specific enthalpy of a product stream, h A , as a function of its

Equations (1)-(Z), include the volumetric transfer coefficients for mass, kM

2 + (X, / X, - l)C - ([2 + (X, I x, - 1)c12- 4(1- c))1'2

C = C,exp(AH, / RT, ) (14)

K = K,exp(AH, / RT,) (15)

Water vapor pressure can be computed by means of an Antoine equation of

The latent heat of vaporization is given by the

Equations (1)-(17) constitute the mathematical model of the m s i e n t behavior

TABLE 1. Nominal steady-state operational variables for the dryer exknined.

Xso (kglkg db) '

OPEN-LOOP DYNAMIC BEHAVIOR

The open-loop dynamic behavior of dryingchambers was indentified by step

CONVEY OR-BELT DRYERS

where all time constants are given in minutes.

(Ziegler-Nickols tuning, Cohen-Coon settings) try to approximate the closed-loop

Log Modulus (dB)

.FIGURE 7. Closed-loop system behavior for a -5% step impose of materid

FIGURE 8. C losed-loop system behavior for a -5% step impose of material

REALISTIC SENSOR SYSTEM

The introduction of 15 s dead-time in the material moisture content measuring

Area of heat exchanger in each chamber (m2)

Specific heat of water (IdkgK)

Length of drying chamber conveyor-belt (m)

Temperature of hot oil at the entrance of heat exchanger (OC)

XA Absolute humidity of rejected &r scream (kgkg db)

AHs Latent heat of vaporization ( W g )

AH0 Latent heat of vaporization at reference temperature ( W g )

XM,CO,KO. Parameters of the GAB equation (13)-(15)

Eltigani, A.Y. and Bakeer-Arkerna, F.W., 1987, Automatic control of

Kiranoudis, C.T., Maroulis, Z.B. and Marinos-Kouris, D., 1990, Mass

Kiranoudis, C.T., Dimitratos, J., Maroulis, Z.B. and Marinos-Kouris, D.,

Maroulis, Z.B., Kiranoudis. C.T. and Marinos-Kouris, D., 1991,

You might also like