increased reactor performance versus reactor safety aspects in acrylate copolymerization

Increased Reactor Performance versus Reactor Safety Aspectsin Acrylate Copolymerization

By Stefan Erwin, Kathrin Schulz, Hans-Ulrich Moritz, Christian Schwede, and Hermann Kerber*

Dedicated to Professor Dr. Karl-Heinz Reichert on the occasion of his 65th birthday

The aim of reducing cycle times of semibatch-polymerization processes requires systematic investigations of the kinetics, carefuladjustment of the desired polymer properties, proper thermal reactor design and reliable reactor safety assessment [1]. As aconcrete example, a semibatch-copolymerization was carefully examined with respect to four different aspects. Thermo-kineticsof the reaction were investigated with isoperibolic reaction calorimetry and GC. In order to obtain reliable values for the overallheat transfer coefficient of the production scale reactor, cooling experiments were carried out with solvent and final copolymersolution as reactor content. For consistent reactor safety assessment additional investigations are necessary including casestudies of breakdown incidences. These simulations were performed with a mathematical model based on the GC data andexperimental vapor pressure curves. As a result of these calculations, a reduction of reaction time from 10 to 6 hours was possible.To convert into practice, it must be ensured that even in this shortened time a product of the same quality is produced.

1 Introduction

During exothermic reactions carried out in semibatch-operation safeguarding considerations may frequently be inconflict with a large reactor performance. The systematicprocedure for the reduction of reaction time consideringsafety aspects is demonstrated at the example of a semibatch-copolymerization at industrial scale.

Therefore, the overall reaction enthalpy and the timedependent heat flow rate are important quantities which canbe properly obtained by isoperibolic calorimetry. Thechemical heat flow rates are controlled by gas chromatogra-phy and time dependent monomer concentrations becomeavailable thereby. For setup of the reactor heat balance, thecooling capacity of the production vessel must be known. Thiswas achieved by cooling experiments in the production plant.In addition, simulations of a variety of scenarios were carriedout for assessing the thermal safety of the process even in thecase of a cooling failure and monomer accumulation.

The investigated reaction system is a free radical copolymer-ization of styrene, 2-hydroxyethyl methacrylate (HEMA),n-butyl acrylate and acrylic acid in solution. A mixture ofhydrocarbon solvents, including xylene and n-butyl acetate isusedassolvent.Acombinationoftwocommonperoxo-initiatorswith different half-life times are used as the initiator system.

2 Experimental Part

All thermo-kinetic investigations of the copolymerizationwere carried out in an automated isoperibolic reactioncalorimeter [2,3]. The main part of this calorimeter is shownin Fig. 1.

Figure 1. Schematic draft of the main part of the isoperibolic calorimeterCalWin.

The thermostated jacket as well as the ballast vessel arefilled with Marlotherm oil, since the operating temperature is140 �C. The ballast vessel is used to limit the increase inreaction temperature to a typical level of about 1 K. There-fore, reaction conditions are regarded as quasi-isothermal.The main advantage of this calorimeter design is the setup oftwo heat balance equations, one for the reactor1)

(1)

and one for the ballast vessel:

(2)

Significant changes in overall heat transfer coefficient Udue to increasing dynamic viscosity of the reaction mass [4]

Chem. Eng. Technol. 24 (2001) 3, Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0930-7516/01/0303-0305 $ 17.50+.50/0 305

±

[*] Dipl.-Chem. S. Erwin; Dipl.-Chem. K. Schulz; Prof. Dr. H.-U. Moritz,Universität Hamburg, Institut für Technische und MakromolekulareChemie, Bundesstraûe 45, D-20146 Hamburg, Dr. C. Schwede; Dipl.-Ing.H. Kerber, DuPont Performance Coatings, Christbusch 25, D-42285Wuppertal, Germany.

0930-7516/01/0303-0305 $ 17.50+.50/0

±

1) List of symbols at the end of the paper.

Full Paper

and occasionally reactor fouling are well known inpolymerization processes. In addition, the effective heatexchange area A changes in semibatch-reactors as the resultof reactant feed, volume contraction due to changingdensities from monomers to polymers and changes in theshape of the stirrer vortex, which varies with reactor fillingheight and viscosity. However, combination of these twoheat balance equations eliminates the heat transfer term ofthe reactor UA� �

Rand subsequently, the calculation of the

heat flow rate generated by the chemical reaction, isindependent from the UA term of the reactor:

(3)

The accumulation terms in Eq. (3) may be summed up as areasonable approximation, if quasi-isothermal conditions areconsidered:

(4)

Consequently, the chemical heat generation rate can bemeasured without the need of measuring the reactortemperature. When both, reactor and ballast vessel temper-ature are known, the reactor heat transfer term UA� �

Rcan be

calculated by rearranging Eq. (2)

(5)

as a function of time, which offers valuable information forthermal reactor design. Data acquisition and dosing arecomputer based. The data acquisition and control program isbased on the software TestPoint. At the beginning of eachexperiment, 240.1 g of the solvent mixture were filled into the1.1 liter autoclave. While the system was heated up, the reactorwas purged from oxygen. After determination of thestationary temperature and a first calibration, the polymer-ization was started by feeding of the comonomers and initiatorsolution at the desired mass flow rates. The total mass of bothwas kept constant for all measurements at 373.8 g ofmonomers and 22.4 g of initiator solution. All chemicals usedwere of technical quality.

Total monomer conversion and comonomer concentrationswere determined by gas chromatography. The final conversionwas determined for each polymerization separately. Inaddition, samples were taken during the entire polymerizationprocess as the basis for calculation of comonomer profiles.Since the reaction temperature is above the boiling point ofsingle components, overpressure is built up inside the reactor.Therefore, sample collection had to be performed with specialtechniques. Samples were cooled immediately by ice andprepared for GC-measurements by dissolving approximately500 mg of the sample in about 5 ml THF and adding 100 mg ofn-butyl methacrylate as internal standard. A WGA-SB-11-FFAP column was used for these investigations. Quantitativeresidual monomer determination was obtained by propercalibration for each of the comonomers.

Large differences exist between the calorimeter autoclaveand the 15 m3-production reactor with regard to thepossibilities of heat removal. For evaluation of theseparameters, especially the cooling capacity, cooling experi-ments were carried out in the production reactor. Therefore,temperature sensors were installed at the in- and outlet of thecooling water jacket in addition to one sensor measuring thereactor temperature, and a mass flow controller was built intothe jacket loop as well. Since the original process controlsystem is not suitable for export and evaluation of data, datawere acquired with the aid of a personal computer andmeasurement electronics by HiTech Zang [5]. Heating of thereactor was conducted by medium pressure steam in aseparated jacket loop. Both, cooling and heating jacket loopsconsist of external welded half-pipes. In order to get morereliable values, two different methods were used for thisdetermination:

Method 1: Cooling with constant temperature of the coolingwater at the jacket inlet.

If only cooling water is responsible for changes of thereactor temperature, UA can be determined by fittingaccording to the course of measured reactor temperatureTR(t). Eq. (6), which is derived from Newton's cooling law, isgiven in the VDI-Wärmeatlas [6]:

TR�t� � TJ;in ÿ �TJ;in ÿ TR;0��

exp ÿ _mJ cP;J

MR cP;R

1ÿ exp ÿ UA_mJ cP;J

!" #t

( )(6)

Eq. (6) is limited to the following constraints:± Heat transfer coefficient may be described by a suitable

average value.± Heat losses to the surroundings as well as power input by the

stirrer or any chemical reaction can be neglected.± Heat capacity of the reactor is small compared to its

content.This method was performed at the end of a polymerization

run with the final copolymer solution. Hence, viscosity andany reactor fouling had production like conditions.

Method 2: Heat balance of cooling water at stationary state.Stationary state was achieved by cooling and heating of the

reactor at the same time. This ends in a thermal balance. Whilestationary conditions are realized, the conductive heat flowrate through the reactor wall

(7)

should be equal to the convective heat flow rate of thecooling water through the jacket,

(8)

if losses from lateral heat conduction inside the reactor walland other secondary heat flow terms of minor importance areneglected. The reactor temperature is kept constant by

306 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0930-7516/01/0303-0306 $ 17.50+.50/0 Chem. Eng. Technol. 24 (2001) 3

Full Paper

heating, which equals the conductive heat flow. The overallheat transfer coefficient U is estimated by combination ofEqs. (7) and (8)

U � _mJ cP;J�TJ;outÿTJ;in�A �TLM

(9)

with DTLM describing the logarithmic mean temperaturedifference between reactor and cooling water jacket [7]. Thesemeasurements were carried out in the production scalereactor filled with water and solvent, respectively. In general,the reactor temperature during these tests should be as closeas possible to the reaction temperature of about 140 �C.However, boiling has to be avoided, so that these measure-ments were limited to about 90 �C for water and about 120 �Cfor the solvent mixture.

3 Results and Discussion

3.1 Thermo-Kinetics of the Semibatch-Copolymerization

In order to obtain information about the chemical heatflow rate of the reaction, the measured heat flow rate has tobe corrected by the convective heat flow rate, which isneeded to heat up the cold added reactants to reactiontemperature. Fig. 2 shows the resulting chemical heat flowsrates and overall heats of reaction for different dosingperiods.

Figure 2. Heat flow rates for different dosing periods and overall heat ofreaction.

All experiments were reproduced with similar results.Furthermore, the measured values for the heats of reaction,which were obtained by integration of the heat flow ratecurves correspond for dosing periods of 4, 5 and 6 hourswith data from literature, if the overall heat of copolymer-ization is assumed to be additive with regard to theindividual heats of homopolymerization of the comono-mers. For known values of final monomer conversion, thefractional thermal conversion was determined by calorim-etry according to Eq. (10) [8]:

X�t� �

Rt0

_Qchem dt

R10

_Qchem dt(10)

For control of thermal conversion and to determine themaximum accumulation of monomers, which is a crucialmoment for reactor safety, samples were taken during theentire polymerization process and analyzed by gas chroma-tography. Fig. 3 shows good agreement of both, calorimetricand chromatographic data. In the beginning, calorimetryseems to be more reliable, since the gas chromatographicresult depends largely on the exact sample time. On the otherhand, the error of conversion determined by integration of theheat flow rate increases with time because small errors, e.g. adrift in baseline, are accumulated. Monomer conversionobtained by calorimetry presupposes unchangeable copoly-mer composition.

Figure 3. Comparison of overall monomer conversion obtained by calorimetryand gas chromatography.

Furthermore, monomer accumulation and MTSR ± values(Maximum Temperature of Synthesis Reaction) are impor-tant quantities for thermal safety assessment, as shown byStoessel [9]. The MTSR is defined as follows

MTSR = TR,0 + DTad with DTad = YaccQRCR

(11)

and can be calculated from gas chromatographic measure-ments for determination of accumulated monomer part Yacc

and known feed rates. Fig. 4 shows conversion, accumulationand calculated MTSR ± values for dosing periods of 6, 4 and2 hours.

Monomer accumulation increases with reduced dosingperiods as expected and reaches its maximum at about17 % in case of a 2 hours dosing period. However, not onlythe comonomer accumulation is increasing but overallmonomer conversion and the total reaction mass as well,when comonomer feed rates are increased. Therefore,numerator and denominator in Eq. (11) increase at thesame time and the resulting maximum MTSR value for a 2


Full Paper

hours dosing period is only 10 K higher than for a dosingperiod twice as long. Nevertheless, it has to be kept in mind,that a reasonable small difference between MTSR and thedesired level of reaction temperature is insufficient in orderto ensure that isothermal reaction conditions may berealized.

3.2 Cooling Capacity of the Industrial Reactor

In order to achieve isothermal reaction conditions, thecooling capacity of the 15 m3-production reactor has to belarger than the heat generation rate of the reaction during theentire process. The cooling capacity consists of the convectiveterm (direct cooling by dosing of reactants) and theconductive heat transfer to the jacket. While the convectivecooling is almost constant and can be calculated easily, the UAterm of the conductive cooling changes significantly duringsemibatch-polymerization and has to be determined experi-mentally. In Fig. 5 measured reactor temperatures arecompared with fitted data estimated from Eq. (6). A goodagreement is obvious and the correlation coefficient iscalculated to be 0.9968.

The experimentally determined overall heat transfercoefficients are± Method 1 122 W / m2 K± Method 2 97.7 W / m2 K

Since Method 2 was carried out only with pure water andsolvent, the result had to be recalculated considering thedifference in viscosity by using the heattransfer characteristic (Nusselt equation).With regard to numerous assumption andapproximations made for the calculationsand the significant different measurementconditions, the difference between bothestimates of about 20 % seems to reason-able small.

3.3 Heat Balance

Since the measured heat flow rate scales with the reactorvolume, measured data are normalized to 1 kg of reactorcontent for better comparison and summarized in Tab. 1. Inaddition, the corresponding cooling capacities are listed. Asimple comparison of maximum and removable heat flowrates for each dosing period shows that for dosing periods of 6and 4 hours isothermal reaction conditions can be realized.

Subsequently, isothermal conditions can not be realizedduring the entire process for a dosing period of only 2 hours.

3.4 Simulation of Cooling Failures

For an extended reactor safety analysis of the polymeriza-tion process the statement of isothermal reaction conditions atnormal operating conditions is insufficient. In addition, theeffects of possible breakdowns, especially cooling failures, onthe large scale reactor have to be calculated in order to preventthermal runaway situations. Consequently, a variety ofbreakdown incidences and resulting scenarios have beensimulated on the basis of deterministic mathematical models,which help to identify critical effects on the thermal safety ofthe polymerization process. In fact, some different failures,e.g. a broken stirrer shaft, can be simulated with the samemodel as a cooling breakdown. The following conservativeassumptions were made:


Figure 4. Monomer conversion, instantaneous monomer accumulation andMTSR values versus reaction time for different dosing periods.

Figure 5. Measured and fitted temperatures during cooling of the productionreactor.

Table 1. Produced and removable heat flow rates for different dosing periods.

dosingperiod [h]

Max. heatflow rate[W/kg]

conductivecooling capacity[kW]

convectivecooling capacity[kW]

overallcooling capacity[kW]

removable heatflow rate[W/kg]

6 19 238.2 70 308.2 25.4

4 25.2 238.2 105 343.2 28.3

2 55 238.2 209.6 447.8 36.9

Full Paper

± produced heat is completely used to rise the temperature ofthe reactor content(adiabatic reactor behavior)

± pressure over the reaction mass can be estimated from theintegrated form of the Clausius-Clapeyron equation.The Clausius-Clapeyron equation is used in its integrated

form for the calculation of the resulting pressure as a functionof temperature (Fig. 6). The overall heat of vaporization andthe integration constant are obtained from temperature-pressure measurements and consecutive fitting of theseparameters to the experimental data of the pure solvent.The resulting equation shows excellent conformity withexperimental results. The overall heat of vaporization(50 kJ/mol) needed to achieve this conformity seems to be alittle high in comparison with data from literature for o-xylene(41,8 kJ/mol [10])

Figure 6. Temperature ± pressure measurement for different polymer contents.

As expected from Flory-Huggins theory, the resultingpressure decreases with increasing volume fraction ofpolymer contained in the solvent. Again the worst case,pure solvent, is used for further simulation. Severe break-downs incidences are cooling failures which are investi-gated predominantly. Two different scenarios are presentedhere:± Scenario 1 is based on a stopped reactants dosing if the

cooling breaks down.± Scenario 2 simulates an unstopped dosing even when a

cooling breakdown occurs.

3.4.1 Scenario 1

Fig. 7 shows the adiabatic runaway behavior of theproduction scale reactor in case of a an immediate stop ofmonomer and initiator feeding pumps when the breakdown inreactor cooling occurs. The time chosen for the coolingbreakdown is the moment of maximum comonomer accumu-lation, which represents the worst case study with respect toreactor safety. Since the reactor stays within its characteristicdesign limits of 3 bar even if the cooling failure appears at this

most critical point in time, the process will be safe as far asthermal aspects are considered, if the dosing period is notshorter than 4 hours.

Figure 7. Simulated temperature and pressure curves for adiabatic reactorbehavior for different dosing periods when dosing is stopped after coolingfailure.

3.4.2 Scenario 2

Unstopped dosing of reactants without any jacket coolingleads to temperature and pressure curves as shown in Fig. 8.The remaining dosing period is set to 4 hours for both feedrates, which means that in case of 4 hours dosing period, thepolymerization runs at ªadiabaticº conditions right from thebeginning. Independently of the dosing period, the plantexceeds its design limits in both cases. In case of the shorterdosing period, a slight decrease of temperature and pressurecan be observed during the first hour. This effect results fromthe direct cooling by cold reactants fed into the reactor. Whenthe reaction starts, most of the accumulated monomer will beconverted into copolymers in a very short period of time ofapproximately 30 minutes. Subsequently, temperature andpressure climb up steeply, which makes it even more difficult


Figure 8. Simulated temperature and pressure in case of ªadiabaticº reactorbehavior for different dosing periods with ongoing dosing after cooling failure.

Full Paper

to stop the reaction and cool down the reactor before its designlimits are exceeded and reactor venting will occur.

3.5 Largest Controllable Batch-Reactions

The reaction is carried out in semibatch-operation mode.Therefore, the mass of the reactor content increases con-stantly with reactant feedings and in particular, the amount ofaccumulated monomer is enlarged accordingly. The increasein molar number of monomers result in an increase in latentheat, which may exceed the reactor design limits. For theseinvestigations again two scenarios can be distinguished. Thelargest controllable batch-reaction with a cooling failure at thesame time is defined by the resulting adiabatic temperaturerise. It must be restricted to 40 K, since a MTSR over 180 �Cwould lead to pressures above 3 bar. Fig. 9 shows themaximum amount of accumulated monomer, which can beaccepted in comparison with the measured amount.

Figure 9. Comparison of acceptable and measured accumulation.

For a dosing period of 6 hours, there is always a largedifference between the maximum acceptable accumulationand the measured one. This difference becomes smaller withreduced dosing periods. The accumulation at 4 hours dosingperiod even reaches the maximum tolerable accumulation.Any further reduction of dosing periods leads, e.g. in the caseof 2 hours, to a small period of time, where the monomeraccumulation exceeds the acceptable amount. Because of thefast increase in mass, this period is surprisingly short, but is notacceptable with regard to reactor safety.

The determination of the maximum acceptable latent heat,which is given by the maximum tolerable monomer accumula-tion times the heat of polymerization, is more difficult in caseof a working jacket cooling. Calculations for these scenariosrely on the experimentally determined cooling capacity andsubsequently, they have the same error margins. Thesescenarios simulate the effect of uncontrolled fast dosing ofmonomers or ineffective initiator while the cooling is stilloperating. Since isothermal conditions can not be ensured for

a dosing period of 2 hours anyway, these scenarios concen-trate exclusively on dosing periods of 4 hours. Fig. 10 showsthe thermal behavior of the production reactor for differentlevels of monomer accumulation. The temperature is calcu-lated from the start of the reaction while the dosing ofreactants continues. Isothermal conditions at 140 �C wereassumed until the reaction actually starts.

Figure 10. Temperature curves for different monomer accumulation times.

For an accumulation level of 12.5 % of total monomerwhich corresponds with half an hour of dosing without anyreaction, the resulting temperature curve reaches its max-imum after about 1 hour at approx. 172 �C and stays below180 �C, which represents the reactor design limit. In contrast,the reaction of monomer accumulated in 0.75 h alreadyexceeds this reactor safety limit and has to be avoided. Inaddition the temperature climb up is significantly faster andreaches its maximum of 190 �C already after about 30 min-utes. As expected, the temperature rise is even more rapidly, ifmonomer accumulation during 1 hour is assumed.

3.6 Product Quality

If safety aspects do not prevent a reduction in reaction time,then a product of at least the same quality must be produced.In general, shorter dosing periods result in higher molarmasses due to an increase in monomer concentrations. Thiseffect was observed in the present investigation as well. Twopossible measures may compensate this. Changes in the dosingprofiles or shifts in the dosing times will establish the formerproduct quality. A change in the recipe, e.g. the use of adifferent initiator system, can be a suitable alternative thatworked well in the presented investigations.

4 Conclusions

For dosing periods from 6 to 4 hours the reactor will remainwithin its design limits, if reactant feedings are blocked whenthe jacket cooling breaks down. Since pressure may rise up to


Full Paper

3 bar in the case of adiabatic behavior, any further reductionof the dosing period with increasing latent heat should beavoided. 4 hours seems to be the lower limit of the dosingperiod under the constraints respected.

In combination with a reduction of 2 hours of consecutivebatch-reaction time, which is needed to increase the conver-sion after monomer feed is completed, the total reaction timecan be reduced from 10 to 6 hours. This reduction of 40 % ofthe original reaction time is achieved without additionalinvestments. An equivalent product quality can be obtainedby optimization of the recipe within the reduced reaction time.Additional time may be saved by an initial charge ofcomonomers which shortens time to reach quasi-stationarystate during the dosing period. Furthermore, adiabaticreaction conditions after the end of monomer dosing mayresult in additional increase of reactor space time yield. Sinceproduction costs may vary not only with reduced reaction time(e.g. if different chemicals must be used) overall economicviews have to compare the cost advantage of technicaloptimization with the new, possibly changed raw materialcosts in order to evaluate the minimum production costs.

Received: March 17, 2000 [CET 1215]

Symbols used

A [m2] heat exchange areaC [J/K] effective heat capacitycP [kJ/kg K] specific heat capacityMTSR [K] Maximum Temperature of Synthesis

Reactionm [kg] mass_m [kg/h] mass flow rate

Pstirr [J] power input by the stirrerQ [kJ] overall heat of reaction_Q [J/s] heat flow rate

U [W/(m2K)] overall heat transfer coefficientT [K] temperature_T [K/s] time derivative of temperaturet [s] time

DTad [K] adiabatic temperature riseDTLM [K] logarithmic mean temperature

differenceX [±] conversionYacc [±] accumulated monomer part

Subscripts

0 initial valueacc accumulationB ballast vesselchem chemical reactioncond conductiveconv convectivein inletJ jacketout outletR reactor, reaction

References

[1] Steinbach, J., Safety Assessment of Chemical Processes, John Wiley &Sons, New York 1998.

[2] Stockhausen, T.; Prüû, J.; Moritz, H.-U., An Isoperibolic Calorimeter ± ASimple Apparatus for Monitoring Polymerization Reactions, 4th Interna-tional Workshop on Polymer Reaction Engineering (Reichert, K.-H.;Moritz, H.-U. Eds.) DECHEMA-Monographie 127 (1992) pp. 341±349.

[3] Stockhausen, T., Entwicklung isothermer und isoperiboler Reaktionska-lorimeter und ihre Erprobung am Beispiel der Emulsionspolymerisationvon Styrol, VDI-Fortschritt-Berichte, Reihe 3, No. 429, VDI-Verlag,Düsseldorf 1996.

[4] Moritz, H.-U., Increase in Viscosity and its Influence on PolymerizationProcesses, Chem. Eng. Technol. 12 (1989) pp. 71±87.

[5] HiTech Zang, email: [email protected][6] VDI-Wärmeatlas, 8. Auflage, Springer Verlag, Berlin 1997.[7] Incropera, F. P.; Dewitt, D. P., Introduction to Heat Transfer, 3rd edition,

John Wiley & Sons, New York 1996.[8] Lopez, F., Experimentelle Ermittlung formalkinetischer Parameter,

DECHEMA-Seminar ªSicherheit von chemischen Reaktionenº, Berlin,1997.

[9] Stoessel, F., What is Your Thermal Risk? Chem. Eng. Progress 89 (1993)No. 10, pp. 68±75.

[10] CRC Handbook of Chemistry and Physics, 65th edition, CRC Press Inc.,Boca Raton, Florida.


_______________________

Full Paper

increased reactor performance versus reactor safety aspects in acrylate copolymerization

Documents