Journal

of Food

Engineering

28 (1996) 27 I -2X2

Copyright D 1996 Elsevier Science Limited Printed in Great Britain. All rights reserved 0260-X774/96/$ IS.00 ELSEVIER 0260-8774(95)00053-4

Design and Control of Drum Dryers for the Food Industry. Part 1. Set-Up of a Moisture Sensor and an Inductive Heater
G. Rodriguez, J. Vasseur & F. Courtois

ENSIA-INRA Food Engineering Department. 1, Av. des Olympiades, F-91305 Massy Cedex, France (Received 5 March 1995; accepted 23 August 199.5)

ABSTRACT The aim of this work is to analyse different ways for controlling the final moisture content of a product dried on a drum dryer; and to reduce its unevenness in order to obtain a high-quality product and to increase dryer productivity A new way to determine product moisture content by means of product temperature measurement is presented, which is convenient for a moving and low thickness film. The product temperature is directly related to moisture content, because drying on a drum dryer takes place by boiling: this relation comes from the experimental desorption isobal; or boiling curve of the product. This measurement allows the detection of wet zones, and the correction of the moisture profile across the width by means of an additional inductive heater Copyright 0 1996 Elsevier Science Limited.

NOTATION Water activity Concentration of noncondensible gases (kg air/kg water) Dry matter load (kg DMlm) Film thickness (mm) Outlet mass flow rate (kg/h) Inlet mass flow rate (kg/h) Inlet steam flow rate (kg/h) Nominal inductive power (kW) Pressure of saturated steam at Tpe (bar) Total pressure (bar) *Author to whom correspondence should be addressed.
271

272

G. Rodriguez, J. Vasseul; E Courtois

Steam pressure in the cylinder (bar) Temperature measured by an infrared thermometer Boiling temperature (K) Product temperature (C) Local final value of product temperature, just above the knife (left, middle, right) (C) Drum speed (rpm) Dipping satellite drum speed (rpm) Final value of moisture content (kg water/kg DM) Local final moisture content vs width, just above the knife (left, middle, right) (kg water/kg DM) Emissivity Feed product viscosity (Pa s) Gap between drum and dipping satellite (mm)

DRUM DRYING

SURVEY

Drum drying is currently used in the food industry to dry heavy pastes and thick liquids, such as cooked starch, baby food, mashed potatoes, concentrated liquids, caseinates, maltodextrins, yeast creams, fruit pulps, etc. The obtained dried product is porous and easy to rehydrate, ready to use. Nevertheless, problems are sometimes encountered with the output product quality because of some perturbations in the drying processes, such as fluctuations in the initial moisture and thickness of product (Trystram, 1988; according to the degree of starch gelatinisation, temperature, for example), accumulation of noncondensible gases in the drum with local bad heat transfer, e.g. on both sides (Abchir, 1988). These perturbations produce wet zones on the drying film, and unevenness in the moisture content of the final dried product. The same problems are observed in the paper industry (Udino, 1983; Vanot & Develey, 1988). The usual way to overcome these problems is to overdry the product so that even the wetter zones are nevertheless dry, resulting in a lower productivity due to the reduction in drum rotation speed, a lower product quality, and a higher cost than is necessary. Phenomenological studies of the process (Abchir et al., 1988; Kim & Piyarat, 1986; Trystram et al., 1988; Vasseur & Loncin, 1983) show the complex interaction between all operating parameters. These studies allowed us to select and to classify both the input and the output variables (Fig. 1). Input variables. Drum speed Vrc and steam pressure pv are the most important variables of the process with regards to the final product moisture content Xf and outlet mass flow rate m,. For an increase in Vrc, the influence on Xf is due to the reduction in the drying time available and change in the quantity of product smeared over the drum Cs (see Fig. 2(a)). Final moisture content is also conditioned by steam pressure pv: Xf obviously decreases with increase in the steam pressure (pv) due to the higher drum surface temperature profile. Figure 2(b) shows the nonlinear relation between WC and TV for a constant Xf value (TV is the saturated steam temperature at pv). Dipping satellite speed Vrs increases slightly the quantity of product smeared on the drum Cs, but the effect on Xf is not considerable because of the auto regulation effect of the cylinder (increase of steam flow). All these

Design and control of drum dryers for the food industry

273

Unmeasured Disturbances non-condensible gases changes in Xo, pv, p, .. .

Input Variables

Output variables

Prc Sysren

(Dryer -_

Fig. 1. Schematic representation

of the influence of variables variables.

(input) on objectives

(output)

variables induce nonlinear responses with response times and delay depending on the set points. Variations in initial moisture content (X,), product viscosity (p), thickness of product layer (e) and the gap between dipping satellite and drum drier (rj) are considered as disturbances. The variations of these factors produce an unevenness in film thickness, meaning that dry matter load (Cs) is not the same everywhere. Vasseur & Loncin (1983) have proved that the evaporation rate increases at low thickness of product smeared on the drum (Fig. 3). Output variables. The most important are moisture content X, which will be measured by means of product temperature (T C) and mass flow rate (mr) linearly depending on Cs and Kc. Figures 4 and 5 show the influence of WC and pv on final moisture content and mass flow rate. Several models of the process have been proposed (Abchir et al., 1988; Trystram et al., 1988; Vasseur & Loncin, 1983). Simulation results have shown that WC and PLJ are the most important input variables for controlling product moisture content vs time. However, this global action is not sufficient to correct unevenness in moisture content across the drum width. The wetter zones are commonly on the drum edges. To reach the desired moisture content in these zones, it is necessary to overdry the central zone, resulting in a lower productivity and a higher cost because of the drum speed decrease. A method to overcome this kind of problem is proposed.
State of the art

of moisture measurement

Detection of a small change in the product moisture content is not easy to realise due to the small thickness of the film and the motion of the product. The traditional method used to measure the final moisture content is to take product samples and dry them in an oven at 105C for 24 h or in an infrared heated balance, but the response time is too long for control. In industry, on-line measurement of -Yf is sometimes made by capacitive sensors, conductivity sensors, microwave sensors or

274

G. Rodriguez, .I. Vasseur;1;: Courtois

infrared sensors (Goldstein et al., 1991; Richard et al., 1988; Slight, 1976). Most of these methods are difficult to calibrate and work on ground product in thick layers after having left the drum. The results may be affected by the particle size, the product temperature and thickness, the colour (Jones, 1986; Swift, 1971), and besides, these methods are quite expensive. This is why real time moisture measurement techniques are rarely encountered for industrial drum dryer control.

m
l

10

15

20

25

Drum Drier VIC (m/r&)

10

15

20

25

30

Dmm speed VR: (rpm) Fig. 2. (a) Influence of drum speed over Xf (Trystram, 1988). (b) Relation between steam temperature (TV) and drum speed (WC) when keeping Xfconstant (Xf-0.02 kg/kg). Influence of Vie on flake thickness (Trystram, 1988).

Design and control of drum dryersfor the food industry

275

tnife
wet product

91

\\ lzl

ilf

Fig. 3.

Schematic diagram of the pilot drum dryer.

We have looked for cheaper sensors, able to measure on-line the moisture content in order to control the process with precision and rapidity. Previous studies on the drying mechanisms (Abchir, 1988; Vasseur & Loncin, 1983) have shown that drying on a drum dryer takes place by boiling, since air velocity around the dryer is under 6 m/s. That means that the product attains its

m Oswins Model 0 X Experimental by Bassals Method

100

110

120

130

140

150

Product TempeI;lture(C)
Fig. 4.

Desorption isobar (1 bar) obtained in the laboratory of product (maltodextrin D.E. 17(20%), native wheat starch (3.3%) and water (76.7%)) fitted with Oswins Model.

216

G. Rodriguez, J. Vasseq l? Courtois

100

110

120

130

140

150

Product temperature (C) Fig. 5. Validation of moisture estimation from direct (oven) and indirect measurement (temperature measurement at two emissivities and isobar equation).

boiling temperature Tpe, which is a function of X, defined by the fact its steam pressure at Tpe (i.e. PTpe) is equal to the total pressure P, over the product (generally atmospheric pressure). Therefore, if product temperature is measured, the corresponding moisture content X is known through knowledge of the desorption isobar. This property facilitates the use of cheaper sensors, able to measure on-line the moisture content with precision and rapidity.

MATERIALS

AND METHODS

The product to dry is smeared in a very thin film on the hot metallic surface of a rotating drum (kc), internally heated with steam @v). In all our tests, a dipping satellite was used for distribution on the drum (see Fig. 3) determining the specific load Cs, leading to a stationary state defined by its input m,, m, and output mf mass flow rates. The product dries during contact with the hot wall, for about 314 of a rotation, before being scraped by a knife. The product to be dried in our tests was a model product of maltodextrin D.E. 17(20%), native wheat starch (3.3%) and water (76.7%). The DUPRAT pilot dryer is made of cast iron, 600 mm long, 414 mm diameter, and 22 mm thick. The speed range is O-10 rpm, the useful drying surface is 0.44 m2, and the evaporating flow rate is 30 kg/h maximum. The energy consumption varies
*Constructor of drum dryers. SociCtC DUPRAT, 20, AV. daubibre, 63800 Cow-non, France.

Design and control of drum dtyers for the food industry

277

classically from 1.15 to 1.2 kg steam/kg water removed. the pilot drum drier is shown in Fig. 3.

The schematic

diagram of

Principle of moisture estimation from temperature measurement In boiling type drying, the product temperature reaches 100C at atmospheric pressure as long as activity of water in the product is a,=l, which is usually the case at the beginning of the drying process; the heat flux is then very high. At the end of drying, for a, < 1, the boiling temperature of the product goes over 100C and the temperature difference between product and metallic wall decreases, leading to a low heat flux at the end of drying. The desorption isobar (1 bar) was obtained using Bassals method (Bassal & Vasseur, 1992). This method consists of drying a product in a pure steam atmosphere until it reaches equilibrium. This curve determines the relation between moisture content X of a product and its boiling temperature Tpe. To describe the desorption isobar, we used the transformed (Bassal & Vasseur, 1992) Oswin model to take into account the effect of temperature (eqn (1)).

X=(kl +kzTpe)

___ [ 1 -a,

k,, kZ, k3 and k4 are the parameters to be identified P, is the total pressure over the product (bar), P.l.pe is pressure of saturated steam, at Tpe (bar) and Tpe is the boiling temperature of product (C). Remote temperature sensor In order to measure on-line the product temperature without contact, we used an infrared thermometer. Different factors were taken into consideration: 0 the film is thin, and in motion with the metallic wall; 0 the product emissivity has to be as high as possible (E >OG30) for the chosen sensor waveband (opaque product), and quite independent of moisture content variations (it is a temperature measurement). The selected sensor was a near infrared thermometer: range 0-2OOC, with a 7-14 pm spectral range, emissivity can be set from 1;=0.4 to 1. Its target is 20 mm distant and the response time is about one second. Due to the emissivity difference between the wall and the product, this sensor also detects the absence of product over the drum, when some problems occur upstream.

a, 1

(k,~tXJ7pt2)
aw=<

p,

(1)

RESULTS This section describes moisture measurements heater. successively the results obtained for temperature and for the correction of wet zones with an additional and local

278

G. Rodriguez, J. Vasseq l! Courtois

TABLE 1 Optimized Parameters used to Fit the Isobar Desorption with the Oswin Model for a Model Product of Maltodextrin D.E. 17(20%), Native Wheat Starch (3.3%) and Water (76.7%) Modified Oswin Model Parameters

k,
k2

6.284369 -9*687127x10-4 6~616002~10~~ 5.537222~10-~

Validation of moisture estimation method Figure 4 shows the desorption isobar (1 bar) of our product obtained from equilibrium laboratory experiments using Bassals method and described with Oswins model (1992). A nonlinear optimisation method (Simplex) was used to estimate the parameters of the model from the experimental results (Table 1). We observe in Fig. 4 that the newly adjusted model is close to the experimental points especially between 118 and 145C, which is the temperature working range at the output of the drum dryer. We can thus deduce the moisture content Xf from the measurement of Tpe with a good precision. In order to validate the use of the desorption isobar with the dried product coming out of the drum, samples at different temperatures were taken just before the knife and their moisture contents were determinated in the oven. Two different emissivities were tested for the IR thermometer: &=0.85 and 0.95. Figure 5 shows a good agreement between the experimental moisture content obtained from drum drying experiments through classical oven methods and through the isobar equation and temperature measurement for ~=0*85. We can also check that the emissivity variation has little influence on the moisture content deduced from the temperature. Conclusion, the final product moisture content can be achieved by an infrared thermometer T=TIR=f(Xj), giving through the isobar a local value of Xf, which is precise (error 7% for T=118-148(Z), without contact and quite cheap compared to other methods of moisture measurement. Correction of unevenness in moisture content across the drum width 6 is a recording of product temperature at three points across the width of the drum: left, middle and right (Ti, T,, T,). We can see both variations vs time and vs width: the left temperature for example is lower than the middle one, which means that the product is wetter in this zone. The same situation is observed in the paper industry (Udino, 1983; Vanot & Develey, 1988), where nonuniform moisture profiles are attributed to basic weight variations, felt nonuniformities, uneven dryer roll conditions, etc. In order to correct these uneven moisture content distributions, process modifications have been made by adding complementary energy sources. Several additional sources of energy have been proposed:
Figure

Design and control of drum dryers for the food industry

279

3200 Fig. 6.

3400

3600
time s

3800

4ooo

4200

Evolution of final product temperature at three locations on the width of the drum: T,, T,, T,. (Vrc=1.23 rpm, Vrs=30 rpm,pv=S bar.)

0 high-frequency

dielectric heating has been used directly over the paper sheet (Jones, 1986); 0 a modular inductive moisture profile corrector has been situated outside a heating cylinder (Udino, 1983); l a modular infrared heating system has been located inside the cylinder (Manfredi & Raymond, 1989) or most often, directly over the paper sheet at the end of the process. In the last two cases, the modules are individually commanded to control the local moisture content. Nevertheless, this kind of solution has not yet been used in the food industry, using a single-step drum dryer. The following results show that such a process modification can also be implemented in the food industry, in order to correct the moisture profile and to increase the product quality and productivity. In this work, two additional energy sources have been tested: a short infrared heater and an inductive electric heater. The results obtained with the infrared heater (4 kW over a 40 cm width) were not satisfactory: the energy is not well absorbed by the product which is transparent to this wavelength IR radiation. Also the flux density transferred to the drum wall and later released to the product was not sufficient. The results with the inductive electric heater are given below. Inductive electric heater The inductive electric heater was used to correct the unevenness in moisture content on the left side. It is a low-frequency (50 Hz) inductor, with a ma netic circuit, producing up to 3.4 kW over a 20 cm width (heated surface 0.26 m #), for a wide range of gaps (3-10 mm), without impedance adaptation.

280

G. Rodriguez, J. Vasseur;I? Courtois

Fig. 7.

Zone C Induction heating of dryer.

Figure 7 shows the inductor installed near the outside surface of the drum dryer (3 mm gap). When the inductor is energized, an alternating electromagnetic field intersects the portion of the drier shell adjacent to it. Since the wall dryer is electrically conductive, an eddy current is induced in the wall, and heat is generated from the resistance to this eddy current. Three zones are possible for installing the inductor (Fig. 7). However, this technique brings heat to the wall, which must be heated before increasing the local heat flux (left side), thus requiring the time of several rotations. Thus, the inductor location is rather a question of practicability, washability, etc. The environment in zone A is very humid because of the steam produced by the drying of product. That is why the actuator was placed under the drum in zone C, between the feed zone and the knife. Figure 8 shows the process response when the local actuator is switched on. In the first part of the test, the system is used without the inductor. The initial unevenness between left side and middle is obvious, the left side being 15% wetter. When the inductor is started, the left moisture content decreases rapidly, changing from Xf=4-2% to Xf=2*8%. The response time is about 100 s, with an applied power of 3 kW. The inductor efficiency is about 75%, i.e. 2-25 kW is transferred to the wall. The additional evaporating flux does not consume 2.25 kW (on 20 cm width). In fact, most of electrical power input also results in a reduction of the energy coming from the heating steam condensation; only a small part (10%) is net energy to increase the local mass flux. We must choose an inductive power adapted to the unevenness to control and to the size of the drum.

CONCLUSION We can conclude that drying on a drum dryer takes place as a boiling phenomenon mechanism; consequently, an infrared temperature sensor can be used as a local moisture sensor for the dried product (Xf. This measurement works with a thin

Design and control of drum dryers for the food industry pv = 4

281

bar, Vrc = 3 trlmin

2 z! 00 6 fg

Pind = 3000 W
Inductive Heating

2 (. .,, 0

,,, 100

., ,,, 200

,,, 300

,,,

,,. 400

,.!

,., 500

*, ,,, 600

,.( 700

,.,

,,. 800

.,,

,,, 900

,. ,,. 1000

,,, 1100

.,.

, 1200

time (s)

Fig. 8.

Step response of product moisture content (Xi=left and X,,=middle) when the local
actuator is switched on.

film, without contact, is precise and quite cheap compared to other methods. It can be used to detect moisture unevenness and location of wet zones in real time, and eventually to detect anomalies such as absence of product on the dryer. The feasibility of unevenness correction of Xf has been demonstrated, using a complementary heating source, i.e. an inductive heater. It increases the drum dryer productivity, by avoiding over-drying which is the conventional way to overcome unevenness in industry. Of course for adequate control, the power of the additional local heating should be related to the moisture unevenness and to the drum size. It should be taken into account that the controlled electrical power is partially used for increasing the moisture evaporation rate. The next step is to include the above sensor and actuator in a control loop in order to adjust the local additional power to the objective in moisture content. In fact, the industrial control can combine several actuators - steam pressure @v), rotation speed (Vrc) and local induction heating (Pind) - with specific dynamics and nonlinear response, which is a more complex strategy. This will be the subject of part 2 of this paper.

ACKNOWLEDGEMENTS The authors would like to thank DUPRAT Company and EDF, which provided the material and invaluable aid in the preparation of this work. The authors would also like to thank Dr. G. Trystram and Dr. A. Bassal whose comments led to significant improvements of the text.

282

G. Rodriguez, J. Vasseul; F Courtois

REFERENCES
F. (1988). Modelisation du sechage sur cylindre. Thesis, ENSIA-Massy. R., Vasseur, J. & Trystram, G. (1988). Modelisation and Simulation of Drum Drying. International Drying Symposium, IDS88, Versailles, pp. 435-439. A. & Vasseur, J. (1992). Measurement of water activity at high temperatures. In Drying 92, ed. A. S. Mujumdar. Elsevier Science Publishers B.V., Amsterdam, The Netherlands, pp. 313-21. Bimbenet, J. J. & Trystram, G. (1992). Process control in the food industry. Trans. ZChem E, 70C (September), 115-25. Domanico, C. (1982). On-line moisture control system features *0.05% accuracy. Food Processing, 43 (12) 116-7. Goldstein, N., Bien, R., Gersh, M. E., Lee, J. & Zakin, M. R. (1991). Multipoint fiberoptic humidity monitor. Drying Technol., 9 (4) 833-47. Jorrr$_t4L. (1986). High frequency dielectric heating in paper making. Drying Technol., 4 (2) Abchir, Abchir, Sixth Bassal, Kim, K. & Piyarat, W. (1986). Drying characteristics of minced fish on drum dryers. In Drying 86, ed. A. S. Mujumdar. Hemisphere Publishing Corporation, Massachusetts, pp. 408-12. Manfredi, M. G. & Raymond, D. (1989). Cylindre chauffe interieurement par rayonnement infrarouge. Journal Fraqais de IElectrothermie, 39 (May), 27-31. Richard, A., Tomezak, M., Aubrun, M. & Ronat, E. (1988). An Approach to Automation of Rotary Drum Dryers. Comparative Study of Industrial Applications. Sixth International Drying Symposium, IDS88, Versailles, pp. 225-32. Slight, H. A. (1976). Process Control. ht. Food Sci. and Technol., 9 (4) pp. 147-51. Swift, J. R. (1971). Measures moisture with infrared. Food Eng., 43 (12), 83-4. Trystram, G. (1988). Contribution a lautomatisation dun pro&de industrial de sechage sur cylindre. In Automatic Control and Optimisation of Food Processes, eds. M. Renard & J. J. Bimbenet. Elsevier Applied Science, pp. 265-83. Trystram, G., Meot, J. M., Vasseur, L., Abchir, F. & Couvrat-Desvergnes, B. (1988). Dynamic Modelling of a Drum Dryer for Food Products. Sixth International Drying Symposium, IDS88, Versailles, pp. 13-18. Udino, S. (1983). Magnetic moisture profile corrector. Tappi Journal, 66 (6) pp. 49-51. Vanot, S. & Develey, G., (1988). Chauffage par induction dun cylindre de sechage.
J. fkarqais de lelectrothermie, 33, 47-53.

Vasseur, J. & Loncin, M. (1983). High heat transfers coefficient in thin film drying: application to drum drying. In Engineering and Food, ed. B.M. M. Elsevier Applied Science, Barking, Essex, pp. 217-25.