You are on page 1of 6

Pergamon

Chemical Engineering Science, Vol. 51, No. I 1, pp. 3163--3168, 1996


Copyright O 1996 Elsevier Science Ltd Printed in Oreat Britain. All rights rtmerved 0009-2509/96 $15.00 + 0.00

S0009-2509(96)00214-X

A NEW HEATING-COOLING
CONTROLLABILITY

SYSTEM TO IMPROVE OF BATCH REACTORS

Z. L O U L E H , M. CABASSUD, M.V. LE LANN, A. CHAMAYOU and G. C A S A M A T T A Laboratoire de 06nie Chimique - URA CNRS 192 Ecole Nationale Sup6rieure d'Ing6nieurs de G6nie Chimique 18, Chemin de la loge - 31078 TOULOUSE Cedex - FRANCE Tel. : (33) 62 25 23 68 - Fax : (33) 62 25 23 18 - email : Michel.Cabassud@ensigct.fr

Abstract - The mathematical modelling and the development of a numerical simulator of a pilot-plant batch reactor are presented. In this paper, dynamic simulation and advanced control strategy are given. This work presents the development and characterisation of a multipurpose batch reactor with a variable heating/cooling structure for temperature control. This strategy is based on the use of the thermal flux as the manipulated variable. The controller (the Generalised Predictive Controller) computes at each sampling time the thermal flux to be exchanged with the thermal fluid flowing in the jacket to achieve the desired reactor temperature profile. This information is then used to select the thermal fluid on the basis of the maximal and minimal thermal flux capacities computed on-line for each utility. This methodology allows to overcome the discontinuities due to utility fluid switching frequently encountered during the batch reactor operation.
INTRODUCTION A large number of industrial processes, such as the production of polymers, speciality and fine chemicals, pharmaceuticals, bioproducts, as well as other products for which continuous production is not feasible, are operated batchwise. In many cases this mode of operation is used to manufacture a variety of products that need significantly different characteristics such as conversion time, heat of reaction ...etc. Good control in such reactors is quite often difficult to achieve (Juba and Hamer, 1986) due to their flexible and multipurpose character (different operating configurations and the use of these reactors for different products). To guarantee batch-to-batch reproducibility and improve yields and selectivities, automation of batch reactors must be widely increased. Due to the complexity of the reaction mixture and the difficulty to perform on-line composition measurements, control of batch reactors is essentially a problem of temperature control (Friedric and Peme, 1995). Moreover, the control performances are mainly dependent on the heating-cooling system associated with the reactor (Berber, 1995) Two main types of heating-cooling systems are commonly used in industry : the alternate system or multifluid system and the monofluid system. The well-known alternate system makes utility fluids flow alternatively in the jacket. These thermal fluids are available at given temperature (plant utilities). It represents the most widely used system in industry (more than 90%) due to the relative ease of design and low cost (direct use of plant utilities). Nevertheless, besides the work undertaken in our laboratory CLe Lann et al., 1995) only a few papers have been published on the subject (Marroquin and Luyben, 1972 ; Chylla and Randall, 1993). Therefore, the control task of this type of process is rather difficult and can be divided into two parts : firstly the choice of the right fluid and secondly the appropriate action on the flowrate of this utility fluid in order to track satisfactorily the desired temperature profile. Thus, to go from heating to cooling, a changeover of fluid is required which results in a discontinuity in the operation. An alternative configuration is the monofluid system. This system uses only one fluid whose temperature can be modified to achieve the desired reactor temperature by an intermediate thermal loop which may include heat exchangers, power heaters, etc ... Nevertheless, the dynamics of this external thermal loops can be penalising, especially in the case of urgent need of a rapid cooling or heating. In this paper, we describe a new heating-cooling system which takes advantage of the large heating (cooling) capabilities of steam (glycol/water) when rapid heating (cooling) is needed. On the other hand, for normal operating conditions, the use a single fluid circulating at a sufficiently rapid flow-rate to ensure good heat transfer coefficients is preferred. A new methodology for control and supervision has been developed. It is based on the on-line computation of the maximal thermal flux capacities of each configuration (i.e.: steam, intermediate fluid, glycol/water) and the computation by the controller of the thermal flux (manipulated variable) needed to track the reactor temperature set-point. DYNAMIC S I M U L A T I O N AND P R E D I C T I V E C O N T R O L L E R The accuracy and complexity of the mathematical process model needed in developing a control system depend upon the processing objectives, the control objectives, and the degree of automation required. A dynamic simulation 3163

3164

Z. LOULEH et al.

programme of batch reactors fitted out with a multifluid heating-cooling system has been developed and presented elsewhere (Cabassud et al., 1994). It integrates both the modelling of the dynamic behaviour of a jacketed batch reactor and the supervisory control strategy for the operation of this heating/cooling system and the temperature control by an advanced controller. In this paper, the simulated jacketed reactor corresponds to a real pilot plant fitted out with the new heating-cooling system under consideration. A description of the overall process is given in the following section.
The dynamic simulation model The set of differential equations representing the dynamic behaviour has been established from mass and energy balances on the reactive mixture with the description of the heat transfers between the circulating fluid in the jacket, the jacket wall and the reactive mixture. Data and models supplied to the dynamic simulator include the reaction kinetics and the necessary correlations to calculate the heat transfer coefficients and the physical properties. It should be emphasised that this simulation software allows scaling-up since it integrates the geometrical structure of the simulated reactor. The model has been established assuming that the reactor is perfectly mixed with uniform temperature and concentration of reactants. The composition of the reaction mixture, its temperature and the jacket wail temperature are computed by a numerical solution of the system of differential and algebraic equations. The adaptive control algorithm An adaptive and predictive controller (the Generalised Predictive Controller with Model Reference : GPCMR) is used to compute the necessary flux (manipulated variable) to be exchanged between the reaction mixture and the jacket in order to track the desired reactor temperature profile. This adaptive controller is based on the linear input-output representation of the process. The on-line estimation of the model parameters allows to follow the changes in the dynamics occurring during the different steps : heating, reaction, cooling. It is also a predictive controller which computes the manipulated variable minimising the square errors between the future set-points (reference profile) and the output predictions (temperature in the reactor) on a receding horizon. Details of this algorithm and its use for temperature control of batch reactor can be found in Le Lann et al., 1995. PROCESS DESCRIPTION The characteristics of the simulated reactor correspond to a real pilot plant consisting of a batch reactor, its jacket and the heating-cooling hybrid system described above. The reactor is a 16 liters-DeDietrich type glass-lined reactor with a maximum operating volume of 12 liters. Three utility fluids are available at given temperatures : steam (6 bars), cold water (at about 15 C), mixture of monopropylene glycol/water (50/50 weight% at a temperature of -10 C). The intermediate fluid is obtained by direct mixing of cold water and steam, with a maximum reachable temperature of 70 C. The reactor has the following physical specifications : internal diameter of 0.3 m, external diameter of 0.4 m, a height of 0.3 m, a jacket heat transfer area of 0.2 m 2, a jacket volume of 0.012 m 3, a wall thickness of 0.01 m with a thermal conductivity of 0.005 kcal s-1 m ' l K -1 and a weight of 270 kg. A sketch of the pilot plant is given in Fig. 1.

uid
I I

~~st

~ "'-~ntermediate ] Glycol/Water ~ f l u i d I

Fig 1. Scheme of the pilot plant


A NEW METHODOLOGY FOR SUPERVISORY AND CONTROL OF BATCH REACTORS A new strategy integrating supervisory and control together is proposed. The controller (GPCMR) computes the thermal flux necessary to achieve the required reactor temperature profile. The maximal and minimal thermal capacities of each thermal fluid (steam, intermediate fluid, glycol/water) are then determined and used to choose the "right" fluid according to these thermal limit capacities with a priority to the fluid present in the jacket. The limit capacities of heating and cooling of each thermal fluid are calculated on-line by a procedure involving the reaction

Controllability of batch reactors

3165

mixture temperature, the jacket inlet and outlet temperatures, the physical properties of the fluids and the manipulated variable value delivered by the controller. When this desired flux overshoots the limit capacities of the present fluid, a changeover of thermal fluid is performed. An air purge of the jacket is carried out before feeding the jacket with the "right" fluid. In comparison with most of the previous works, the changeover of fluids is thus not performed according to a predefined temperature alarm system but according to the physical capacity of the heat exchanges. Moreover, to overcome the strong discontinuities due to the changeover from hot to cold utilities as done traditionally with multifluid systems, we propose to use an intermediate system obtained by mixing cold water and steam. This new heating-cooling system is an hybrid configuration (Cabassud et aL, 1995) which integrates the advantages of both the monofluid and multifluid systems : - during normal conditions, a unique fluid (intermediate fluid) circulates at a fixed flowrate (high enough to ensure good heat transfer coefficients) and its temperature is modified according to the required thermal flux computed by the controller by acting on the quantity of injected steam. - for extreme temperature conditions steam or glycol/water are directly used (as in the multifluid system) ; the utility flowrate is changed according to the required thermal flux. It may be noted that the advantages of this strategy, based on supervisory control according to thermal flux limits, not only makes the heat exchange capacity vary continuously but enables an adequate change of system configuration, from multifluid to monofluid configuration. At each sampling period the following two steps procedure is performed : in a first step, the thermal flux limit capacities in each configuration are computed (steam, intermediate fuid, glycol/water). This leads to the choice of the "right" fluid. Then, in a second step, the real control action value (valve opening degree for steam and glycol/water) is computed depending on the utility involved.
Model based supervisory control The thermal flux delivered by the thermal fluid to the reactor is expressed by : Q1 = f Cp(TJin-Tjout) The thermal flux exchanged by the thermal fluid and the reactor is given by : Q2 = Ua A0m A0m is the mean temperature difference between the jacket and the reactor :

(1) (2) (3)

a0m = {(Tjin - Tr) + (Tjout - Tr)} 2 The whole methodology relies on the assumption that both thermal flux are equal : Q1 = Q2
Determination of the thermal flux limit capacities for each utility fluid In this case, the thermal fluid inlet temperature (Tjin) and its flow-rate (f) are known.

Its maximum value corresponds to a zero flowrate so : Qmaxgw = 0 (4) The minimum value corresponds to the maximum flowrate, fmaxg w, of the glycol/water stream available at a temperature Tugw (-10C in practice).

The flowrate (fir) of the intermediate fluid is obtained by mixing the cold water stream (fixed flowrate fcw) with file steam flowrate : fif=fcw+fst (5) It is assumed that the intermediate fluid has the same properties (Cp) as the cold water (main water). The steam flowrate fst is computed as following : fst =1~.c v s . ~ s t - P c w (6) where Pst is the steam pressure (6 bars in practice), Pcw is the pressure of the cold water (main water), 13the valve opening degree, cvs the valve characteristic coefficient. The intermediate fluid temperature (after mixing) is given by:

(7) Cpcwfif where Tucw is the inlet temperature of cold water (main water, in practice about 15C). Tust is the temperature of the inlet steam (boiling temperature at Pst). Tim Tucw-~
From Eq. (1) ,(2) and (3) we obtain Tjout : Tjout = {Tif [2fif.Cpcw-U.a]+ 2Tr.U.a} (2. fir .Cpc w + U.a) (8)

fst (Cpcw[Tust- Tucw ]+ Lv)

Qmaxif is determined for fifmax corresponding to fslmax, i.e. 13=1.In practice, the steam control valve has been chosen in order to get a maximum temperature of 70 C. The minimum thermal flux corresponds to the case where only cold water is used as intermediate fluid, so : Qminif = fcwCpcw (Tucw-Tjout) (9)

steam The thermal flux exchanged by steam in the jacket and the reactor mixture is proportional to the latent heat of condensation :

3166 We have also the heat exchange expression : The steam flowrate is given by :

Z. LOULEHet aL Qst=fst Lv Qst=U.a (Tjst-Tr) fst =[~.cvs. NVP-~- Pj (10) (11) (12)

The maximum value is obtained for [3=l.The pressure Pj can be derived from the pressure of air in the jacket and the partial pressure of the steam in the jacket according to: pj = Pair' Tjst + PJst (Tjst) (13) 293 The steam temperature in the jacket, Tjst, is determined as the boiling temperature corresponding to the partial pressure PJst in the jacket. So, to determine the steam temperature corresponding to the maximum flowrate of steam, we have to solve a non-linear function by an iterative procedure. The minimal thermal flux is obtain when the valve is closed : Qminst=0 (14)
Computation o f the t h e r m a l fluid f l o w r a t e

Different manipulated variables have to be controlled. This problem is solved by cascading the control loop. The master controller computes the heating/cooling flux to be delivered/removed to the reaction mixture. This value is then used, in a slave-like loop, to evaluate the opening value of the control valve. This computation is made in two steps : in the first one, it is determined whether the thermal flux can be feasibly realised with the fluid circulating in the jacket or whether a changeover to another fluid is needed ; in a second step, the computation of the control valve opening degree is carded out. The flux to be exchanged by the chosen thermal fluid (Eq. 1 for liquid fluid, Eq. 11 for steam) is computed by the controller (Qcon0. Thus, contrary to the previous procedure, the objective of the computation is to determine the flowrate of the fluid by using the equations given above. Let us recall that, for the intermediate fluid, this flowrate corresponds to the amount of injected steam in the cold water. The actual manipulated variable is the valve opening degree computed according to the calibration characteristics of the valve.
R e c y c l i n g o f h o t w a t e r at the outlet o f the j a c k e t

Energy can be saved by recycling the intermediate fluid and therefore by using exact quantities of steam or cold water necessary to get the desired temperature at the jacket inlet. In this case, a comparison between the calculated inlet temperature, Tjin, and the measured outlet jacket temperature Tjout allows to choose the fluid to be injected and to determine the corresponding valve opening degree. Moreover, in practice the flowrate of the intermediate fluid is kept constant (fifc) by a trap. - If Tjout _>Tif injection of cold water is needed. The flow of fresh cold water (fw) to be introduced is given by : fw = fife. Tjin - Tjut Tucw - Tjin - If Tjout < Tif, the quantity of steam to be introduced is given by : Tjin - Tjout fst fife. Cpcw' [L v_ CPcw (Tjin - Tust)] (15)

(16)

SIMULATION RESULTS To demonstrate the good performance of this new methodology, simulation studies have been performed. The simulated reactor corresponds to the pilot plant previously described. The reactor is assumed to be fed with 12 1 of water at 22 C. Reactor temperature control, according to a 4 steps temperature set-point profile, has been studied for various conditions : - 1st phase : heating from 22 C to the desired temperature in 1500 s - 2rid phase : constant temperature set-point during 4500 s - 3rd phase : cooling to 30 C in 2800 s - 4th phase : constant temperature set point at 30 C during 1200 s. The first simulation run deals with a desired constant temperature of 45C (Tcons). Figure 2 gives the time evolution of the reactor temperature, the set-point and the actual manipulated variable (the valve opening degree). Let us recall that there is not unique valve but four : one for the steam when it is used alone, one for the steam injection in cold water, one for cold water when the intermediate fluid is recycled and finally one for glycol/water. The plotted valve opening degree (~) corresponds to the simulation run performed with recycled intermediate fluid. Figure 3 gives the time evolution of the inlet and outlet jacket temperatures. In Fig. 4, time evolutions of the limit capacities for each fluid and of the thermal flux required (Qcont computed by the controller) are plotted. To get a more readable figure, the maximal steam thermal flux has not been plotted (~- 5 keal/s). These two figures slaow that the intermediate fluid is used during almost the entire profile. Just at the end of the preheating step (around 1750 s), a changeover of utility has been performed from intermediate fluid to steam, as the required thermal flux computed by the controller overshoots the maximum capacity of the intermediate fluid (Fig. 4). This changeover is accompanied by an air purge which corresponds to an inlet jacket temperature of 20C (Fig. 3). As soon as the constant temperature step is reached, the computed thermal flux becomes negative (Fig.4). Thus, there is another changeover from steam to intermediate fluid with an air purge (Fig. 3). When the intermediate fluid is used, one can notice a smooth evolution of the inlet jacket temperature (Fig. 3). This example shows the double role of the

Controllability of batch reactors

3167

intermediate fluid which can be used both for heating and cooling. In Fig. 5, the flowrate of steam injected in water to produce the intermediate fluid has been plotted for both conditions (with and without recycle). It can be noticed that the recycle allows high energy savings. In addition, this arrangement enables a good adjustment of the hot water temperature at the jacket inlet by feeding small amounts of steam or cold water as a function of the required thermal flux.
Temperature (*C) valve opening 15 50 F--~--~--~-~...*.~.~.~..~r~`.~.-~-~----r-~-.~.~.~.:~.~:~~ 1 Temperature (C) 80 ".................. ' ................. ................. r................. : ..................
70 60 / 3035!/ . ----Tcons . . . . . [~ X '!~~
/, " ~ J~

time(s)
'
'

I ; '

0.6 0.4

50 40 30 20

i/ ii
f"~--4
0

i:Vt.,
2000

:.

.............. ] !.
4000 6000 S000

hi

......i0
10000

10

'2--000'

'

' 4000 ~ . . . 6000 . . . . . . 8000 . .

10000

Figure 2 : Temperature and manipulated variable

Figure 3 : Inlet and outlet jacket temperatures


Steam flowrate injected in the intermediate fluid (kg/s) x 10"

Thermal flux (kcal.s -t) 2; ' : 1,5 k \ 1 .~ %

1.6 1.4

..... Qm~x I . . . . . Krmm. f

~ ...... ..""

1.2

--without

recycle I

~--

in'

~] 1.(3 !
0.8 0.6

. . . . recycle

~
0.4 0.2

time(s)
-10 2000 4000 6000 8000

............~ ~ . : n . . , _ ~ . g ~

10000

Figure 4 : Limits and control value of the thermal flux

Figure 5 : Flowrate of steam with and without recycle

To study the flexibility of this strategy, another simulation run has been performed for a different constant temperature set-point : 60 C. Let us notice that a change of this temperature implies greater heating and cooling rates. Fig. 6 shows that the reactor temperature follows perfectly the set-point profile. As previously mentioned, the opening valve degree (6) corresponds to the four different valves and to the case of recycled intermediate fluid. Fig.7 gives the evolution of thermal flux limits for each utility fluid and the thermal flux calculated by the controller. As expected, due to the higher temperature level, the required capacities of heating and cooling are larger. Compared to the previous conditions (Fig. 5), all the different utility fluids have had to be used. In particular, during the heating phase, steam is required instead of intermediate fluid. Similarly, at the end of the cooling phase (around 8000 s), the higher cooling rate implies a changeover from intermediate fluid to glycol/water. These changeovers of fluids are decided automatically by the supervisory control strategy based on the thermal flux computation.
Temperature (C) 70 t . . . . . . . . . t ~Tr
t . . . . I

60 L

. . . . . Tcons |

.~

.~ i l, .~ \t

valve opening 13 " ' ' i 1 ~ [ tl I


l~

Thermal flux limits (kcal.s t )


2 "--~"-, ......... r " ' ~ ' ' ~ ........ r " ' ' ~ ' ' ~
........ r ' " ~ ' " ~ ' " ~ " ' r ......... " ' " ~ " "

-~o8 , I

1,5 ~

I --Qcont [ . . . . Qmaxit

't
1 "[~f~

I ..... Qmax I
..... Qmm 4 - -Qmin .-" ....

,o

ii

of/,
3o
~/

I,

fl
s _. . . . . . . . . . . . . 4000

~,,

i!

4 ,6

IN
0,5 "

",1

'="

!ti

"k. ; i

.io,,
I
^ 10000 u

..,...~"

,,,"

gv zv~^ r
0

![i J i , "~
2000

i!, kd
6000

. . . . . 8000

i!Li j ':.iv' L.:'

-0,5 " " " " " ,


. . . . . . . . . . . . .

"

4~

-1

2000

4000

6000

time { s ) 8000

10000

Figure 6 : Temperature and manipulated variable

Figure 7 : Limits and control value of the thermal flux

3168

Z. LOULEHet al.

CONCLUSION This study demonstrates the feasibility of the developed supervisory control strategy based on the thermal flux computation. The results exemplify the multipurpose character of the proposed methodology which is a necessary condition for a successful automation of batch chemical reactors. The strategy presents the advantages of a multi-fluid thermal system (possibility to use utility fluids directly available on a plant) and those of monofluid thermal system (characterised by high dynamic performance and continuity of thermal control). The model based supervisory control allows an adequate changeover of thermal utility fluid as a function of the feasible capacity of the thermal configuration. NOTATION
-a - Cp -

cvs -f - Lv - Pcw

-Pj
- Pst

-Q
- Qcont - Qmax, Qmin -T

-xj
-Tr -Tu -U

heat transfer area specific heat valve specification mass flowrate latent heat of steam main water pressure steam pressure in the jacket feed steam pressure thermal flux exchanged between jacket fluid and reaction mixture thermal flux computed by the controller maximal and minimal values of the thermal flux temperature jacket temperature reaction mixture temperature source temperature of utility fluid heat transfer coefficient valve opening degree mean temperature difference cold water glycol/water intermediate fluid inlet outlet steam

(m2) (kcal Kg -1 K -1) (kg s-1 bar -1/2) (kg s -1) (kcal kg -1) bar bar bar (kcal s-I) (kcal s -I) (kcal s-1) (K) (K) (K) (K) (kcal m -2 s-1 K -1)

-8
-

Greek

characters

AOm Subscripts cw gw if in out st


REFERENCES

(%) (K)

Berber R., 1995, Control of batch reactors : a review, Methods of Model Based Process Control, R. Berber (ed.), Kluwer Academic Publishers, Dordrecht, 459-493. Cabassud M., Le Lann M. V., Ettedgui B. and Casamatta G., 1994, A general simulation model of batch chemical reactors for thermal control investigations, Chem, Eng. Technol. 17, 225-268. Cabassud M., Chamayou A., Pollini L., Louleh Z., Le Lann M.V. and Casamatta G., 1995, Proc6de de contr61e thermique d'un r6acteur discontinu polyvalent ~ partir d'une pluralit6 de sources de fluides thermiques, et dispositif de raise en oeuvre de ce proc6A6, Patent N95.03753. Chylla R. W. and D. Randall Hasse, 1993, Temperature control of semi-batch polymerization reactors, Comput. Chen~ Engng 17 N 3, 257-264. Juba M.R. and J.W. Hamer, 1986, Progress and challenges in batch process control, Proc. CPC III. CACHE, Elsevier, Amsterdam. Le Lann M.V., M. Cabassud and Casamatta G., 1995, Adaptive model predictive control, Methods of Model Based Process Control, R. Berber (ed.), Kluwer Academic Publishers, Dordrecht, 426-458 Liptak B. G., Controlling and optimizing chemical reactors, Chem. Engr., May 26, 1986, 69-81 Marroquin, G. and Luyben, W.L., 1972, Experimental evaluation of nonlinear cascade controllers for batch reactors, Ind. Eng. Chem. Fundam. 11 N 4, 552-556. Friedric M. and R. Perne, 1995, Design and control of batch reactors : an industrial viewpoint, Comp. Chem. Engng 19, $357-$368.

You might also like