Current problems in the modelling of chemical reactors

CURRENT

PROBLEMS

IN

THE

MODELLING

D. Dept

of

Chemical

OF

CHEMICAL

REACTORS

Thoenes

Engineering, Emdhoven,

Unlverslty Netherlands

of

Technology.

ABSTRACT

Modelllng of chemical reactors 1s reviewed with an emphasis on process development and scale-up A dlstlnctlon 1s made between modelllng of chemical klnetlcs, of rate processes In volume elements and of whole reactors Examples are mainly taken from papers presented at the Sixth International Symposium on Chemical Reaction Englneerlng Special attention 1s given to the modelllng of single phase reactors, fixed beds, trickle beds, fluld beds, and gas bubble reactors

KEYWORDS

Reactor modelllng, fluid bed, bubble

1

THE

reactlon and dLffuslon, column, stirred gas-llquld

CONCEPT

OF

mlxlng in contactor

reactors,

flxed

bed,

trickle

bed,

MODELLING

The concept of modelllng 1s used both In experlmental research and In englneerlnq We use models for systematically descrlbzng experimentally observed phenomena, In order to understand them better, or to explain them In terms of accepted theories FJe also use models for predlctlng the effects of phenomena that are not yet there, e g In deslgnrng equipment or In optlmlslng an exlstlng productlon process In any case a model 1s a descrlptlon of the characterlstlcs of a number of phenomena that take place In a process It IS convenient to present the descrlptlon In the form of one or several mathematical equations A model may also have the form of qraphs or of series of numerlcal data Sometimes a model has the form of an experlmental set-up, where phenomena take place that can be described by the same mathematical equations as the orIgIna process One could also say, that then both processes are described by the same "model", meaning the mathematical equations and the physlcal assumptions that lay behlnd them I shall use the word model In this sense Experimental phenomena, If observed accurately enough, are always complex and of a nonlinear nature For practical reasons we want models to be relatively simple, they are therefore mostly approxlmatlons or dellberate slmpllflcatlons of the physlcal reality However, this ImplIes that the range of condltlons for which the model holds with a certain accuracy 1s llmlted. Apparently, the making of a model always Involves a compromise between slmpllclty, accuracy and appllcablllty Which compromise one settles for, depends on the purpose of the modelllnq excerclse

2

PURPOSES

OF

In the last Internatzonal excellent review on this functions of models, and tlve models"

MODELLING Symposium on Chemzcal Reactzon Englneerlng In Houston, an sub-Ject was given by Shlnnar (1978) He elaborated on the prlmarlly dlstlngulshed between "learnrng models" and "predlc-

I want to dlstlngulsh further between simple and complex IearnIng models, that may Simple learning models may be practical for formulating geneserve different purposes ral theories, such as one may use In teaching One may want to Isolate a certain phenomenon from a more complex process, and model it seperately for more general use

Current problems

m the modehng of chemical reactors

An example IS the combined effect of dlffuslon and rapid The simplest model 1s based on the surround gas bubbles lrreverslble reaction and dlffuslon In a stagnant phase and it can be checked by experiments under teal effects,

19341

reaction in llquld films that assumptions of first order This model may explain essen"model" condltlons

Complex learning models ~111 have to be used for studying the combined effect of a large number of phenomena, especially If some are highly non-linear Several examales ~111 be discussed later Here I want to point at a very general dilemma one 1s often faced with In making ccmplex learning models. The studled process may Indeed be so complicated, that an accurate descrlptlon of all observed phenomena may require an Such a enormous amount of experiments and may result In a model of vast complexity model ~111 often contazn a large number of parameters, that are only valid for a rather lImIted range of condltlons As a consequence, for applying the model In other condltlons, one first has to determlne values for all parameters This 1s not always practical or indeed posszble In maklng complex learning models, one may end up, as the saying goes, almost everything about almost nothlng" by "knowing It 1s therefore Important to dlstlngulsh, as Shlnnar (1978) does, between "learning It might appear that any model that describes a models" and "predlctlve models" what practical use has the model process has a predlctlve value However, the point 1s 3 The practical use will mostly Involve extraIn predlctlng the effects of a process at the same time a certain accuracy and rellabllzty are required The purpose polatlon, but to let It happen of using such a model 1s not only to predict what ~111 happen, And the consequences are often to be expressed In terms of proflts and In eventually Predlctlve models may be drvlded Into "design models", "process control terms of safety models" and "optlmlsatlon models" These are In fact qnlte different, even If they because the avaIlable knowledge 1s different at the would describe the same process, time each model 1s made The use of design models often calls for large extrapolations, where rellablllty may be more important than accuracy Also, a design model has to be valid for wide ranges of condrtlons, because In the practrce of deslgnlng a whole plant one may have to adapt especially when there are large the condltlons for each piece of equipment frequently, recycle streams In the plant After the plant has been built and put into ooeratlon, One of the first things the dynamic behavlour may be quite different from expectations Parameters from this model have one has to do 1s find a sultable process control model to be determined from measurements In the plant It_ 1s often worth whzle to make a much After more operating experience has been galned, At this time, one may have more accurate and more complzcated model of the reactor more detalled knowledge on the various rate processes taking place In the reactor, from laboratory and pilot plant studies By conductrng experrmental measurements on the plant reactor during a certain period of time, for different operating condltlons, one may eventually determIne the necessary parameters for a sophlstlcated optlmlsatlon model Such a model may have an accuracy that 1s much greater than a design model ever can be The makzng of a good optlmlsatlon model may require some Investments for addltlonal measurement facllltles In the plant, and for a process computer The expected savings may be several percent of the variable cost for a well deslgned plant For a plant of a kind" the savings may be much more that 1s "first

3

MODELLING

RATE

PROCESSES

ON

DIFFERENT

SCALES

In reactor modelllng, one may or one may not seperate physical from chemical rate processes One may also dlstlngulsh between various scales on which one observes gradients connected with these rate processes Chemical reactions occur on the molecular scale, either In a homogeneous phase or at the Interface between two phases. In either case the klnetlcs of the process 1s descrlbed In terms of concentrations and rate constants, that take into account the relevant linear dlmenslons Kinetic models for homogeneous reactzons may be used as such for descrlblng chemical reactions taking place In any volume, zrrespectlve of scale, as long as there are no concentration or temperature gradients wlthln these volumes The same may apply for heterogeneous reactlons, as long as the specific surface area 1s a known constant In most reactors at least one reactant 1s Introduced the catalyst The reactants have to reach each other physical transport processes before they can react

separately from the others or from or the catalyst by a number of The nature of these transport

D THOENES

1842

processes may be such, that it that enables us to look at the

1s practical to various transport

dlvlde the processes

reactor into separately

smaller

parts,

Micro mlxlng, lntrapartlcle dlffuslon and particle to fluid mass transfer may be consldered on the scale of volume elements, that In each dIrectIon are small compared to any of the reactor dlmenslons. Also on this scale can be observed lntrapartlcle heat conduction and particle to fluid heat transfer Axial dlsperslon and countercurrent flow of two phases may be consldered on the scale of a reactor segment that m two dlrectlons 1s small compared to the reactor Radial dispersion and heat transfer to tube walls may have to be consldered on the scale of a reactor segment, that in only one dlmenslon 1s small compared to the reactor. Macro mlxrng, bubble coalescence and slnular phenomena have to be considered on the scale of the reactor Itself. Many chemical reaction englneerlng studies are concerned with rate processes on the scale of volume elements and reactor segments Others are concerned with physical processes on the scale of the reactor A number of dzfferent modelllng sltuatlons 1s shown m Table 1. For the next dzscussion I wrll dlvlde the varzous sorts of models into four types - Models for chemical klnetlcs - Volume element models - Partial reactor models - Overall reactor models In the next chapters I shall discuss these various types of models, with examples mainly taken from contrlbutlons to this symposium Partial and overall reactor models will be treated together It 1s important to note that there 1s a certain hierarchy m these types of models In the development of a process, one needs a more restrlcted model first to make a more comprehenslve model later It 1s therefore useful to consider In each case what the following modelllng efforts have to be, In order to arrive at an overall reactor model, which essentially 1s the ultzmate obJectlve of reactor modelllng

TABLE

TYPES OF REACTOR MODELS

1,

MOLECULES

VOLUME

REACTOR

ENTIRE

ELEMENTS

SEGMENTS

REACTORS

PHENOMENA FOR

MODELS

CHEMICAL

CHEMICAL

KINETICS

PHYSICAL PARTIAL REACTOR

CHEMICAL AND SOME PHYSICAL

ALL

MODELS

RELEVANT REACTOR

4. 4

1.

Separation

MODELS of

FOR Chemical

CHEMICAL and

KINETICS PhysIcal

Rate

Processes

For determlnlng the mechanisms and klnetlcs of chemical reactions, the effects of chemical and physrcal rate processes have to be separated completely This 1s done either by using gradientless laboratory reactors, or by arranging klnetlc experiments In such a way, that mass transfer and reactlon may be consldered as processes In series,

Current problems

mthe

modelmg

of chemical reactors

1843

and by measuring mass transfer rates seperately several examples are presented for reactlons at solld catalysts In thus symposium, who used a Examples of gradzentless reactors are described by Bub et al (H-441, reactor) for model studies of a fIxed bed Fischer small laboratory reactor (a "Berty" and by Paspek and Varma (A-5) who used a splnnlng basket reactor for Tropsch process, the oxldatlon of ethylene to carbon dloxlde Examples of studylnq mass transfer and reactlon In series are given In the following three presentatrons Chandrasekharan and Calderbank (H-43) demonstrated the use of a "tube wall catalytic reactor" The catalyst has the form of a thin powder layer adhering to the wall, with Gas flow and mass transfer to the tube a negllglble resistance to internal dlffuslon Mass transfer parameters wall were modelled carefully, using a two dlmenslonal model and subsequently used In a simpler one-dlmenszonal were estimated with this model, model, with greater practical appllcablllty some gas-llquld reactlons may be consldered as By carefully controlling the klnetlcs, lnvolvlng mass transfer and reaction rn series Klnetlc modelllng studies that use this method are presented by Komlyama and Inoue (D-20) for the absorption of nitrogen oxldes and by Satterfleld and Huff (~-25) for a Frscher Tropsch slurry process Their In water, flndlngs ~111 be discussed In the next sectlon Another lnterestlng method for determrnlng chemical was proposed by Abdul-Kareem and coworkers (F-35) By assuming effects of forced composltlon cycling. occurs 1s a measure for the relaxation time of the that In a speclflc case a desorptlon step was rate

4.2

Krnetlc

klnetlcs for gas-solad reactions These authors lnvestlgated the that the period at which resonance rate controlling step, they showed controlling

Models

In chemical reaction englneerlng, klnetlc modellrng 1s frequently used for processes where several reactlons take place at the same time For Instance, nztrogen, oxygen and water can combine In many different ways, depending Two papers deal with klnetlc modelllng of this reactlon on the reactlon condltlons Komlyama and Inoue (D-20) present a study of the mechanism and klnetlcs of the system. It was shown how the chemical absorption of nitrogen oxides In weak alkaline solutions klnetlcs change when the gas phase concentrations become lower, which 1s Important for tall gas purlflcatlon processes_ Lefers and coworkers (D-19) InvestLgated the reactions between nitrogen oxldes and nLtrlc acid, wrth the same obIectlve Their results ~111 be discussed In section 5 4 Another type of sltuatron that requires klnetlc modellzng occurs when there are so many reactions taking place at the same time, that they have to be lumped for practical purposes Three sltuatlons occur frequently &e starting material 1s converted-into a complex mixture of products, because of an Inherently low selectlvlty of the process Examples are mostly reactlons wath a radical mechanism, e g the thermal cracking of ethane, the partial oxldatlon of hydrocarbons In the llquzd phase, and the Fischer Tropsch synthesis of hydrocarbons or alcohols from synthesis gas In all of these processes, the products are separated Into fractions with narrow bolllng ranges Polymerlsatlon processes are similar In that a complex mixture results from a simple starting material, however, the product mixture 1s seldom separated Into more than two fractions, The desrred end product 1s essentially a complex maxture A complex mixture may be converted Into another complex mixture Many petroleum treating processes fall Into thus category, e g lsomerlsatlons Another example conversions of fatty 0~1s. An example of the first kind 1s presented by Bub et al (H-44) for a fixed bed Frscher Tropsch process Thus 1s a chain type reactlon, where each lntermedzate complex may either grow by adding one CO-molecule, or else decompose rnto a product with the same number of C-atoms The product may be a paraffin, an olefln or an alcohol The assumption of this mechanism results In a relatively simple model and a speclflc product dlstrlbutlon One has to detennlne as many klnetlc constants as one wishes to dlstlnguzsh products The other study on a Fischer Tropsch process by Satterfleld and Huff (E-25) concerns a slurry process. Thus study 1s prlmarlly aimed at modelllng the rates of absorption of CO and H2, so the klnetlc model 1s much simpler Polymerlsatlon reaction englneerlng was the sublect at the Heidelberg symposium by Gerrens (1976)

of

a

comprehensive

review

presented

D

THOENFS

in thrs symposium only one paper deals wzth a polymerlsatlon reaction However, Van Dlerendonck and coworkers (p-60) describe a copolymerlsatlon process with a very simple klnetlc model, that 1s used to predict the copolymer composltlon under condltrons of less perfect mlxlng that takes place m larger reactors (compare section 6.1 2) The catalytic reformrng of petroleum fractions 1s one of the most complicated processes m Industry and at the same time one of the most Important It 1s an example of the third type. Many research groups =n several large petroleum companies have spent years In the bulldlng of models for this process. The Importance of krnetlc modelllng 1s Increased by the circumstance that petroleum refznerles have to be able to adapt their processes to a varying supply of starting materials and to varying customer demands Usually the models used In practice are consldered proprietary knowledge and are not publlshed We can be thankful that at least one large petroleum company has allowed their models to be presented at our Reactlon Englneerlng Symposia Already in Evsnston Weekman (1975) presented a review on this sublect, which was followed In the Houston symposium by a paper by Grazlanr and Ramage (19781, and, In this symposium, by a paper of Ramage, Grazzanl and Krambeck These authors present a slmpllfred two-dlmenslonal reactlon scheme, lumping (n-6) the more than 200 ldentlfled components Into 13 species, and recognlslng some 30 reactions The model 1s aimed at descrlblng selectlvltles as functions of yields and temperatures Apparently the predlctlve value of thus model 1s so good, that It 1s used as an optlmlslng model throughout the company's plants It 1s Interesting to A-6) concern processes

5 51

SolId

VOLUME

note, that both papers without mass transfer

ELEMENT

on modelllng llmltatlons

of

complex

klnetlcs

(H-44

and

MODELS

Catalysts

The modelllng of drffuslonal transport wIthIn porous catalyst pellets 1s quite compllcatto take into account the various transport mechanisms and also the ed, if one wants complexity of the geometrical structure (compare Luss, 1976) If only bulk dlffuslon has to be consIdered, the problem of geometric modelllng can be avolded by conslderlng the porous catalyst as a quasi homogeneous phase, and rntroduclng an effectrve dlffuslvrty constant and a volumetric reactron rate constant, that include all structural effects Both constants can be determined experimentally However, It was found by McGreavy and Slddlqul (A-l) that a transient method for determlnrng the effective dlffuslvlty gives consistently lower values than a steady state method It 1s not self evident which of the two values 1s the most suitable one for use rn models descrlblng reactlon and dlffuslon In the steady state The differences were attributed to drfferent relative effects of small and large pores, and of open and dead end pores Carbonell and Whltaker (A-2) present a theory for calculating the effective Ryan, dlffuszvaty on the basis of a certain pore model The model recognlses the effect of anlsotroplc structures, but does not Include the effect of a large variety In pore sizes. It appears that the pore size drstrlbutlon 1s Important enough to be included In practlcal models However, In the most simple model that would take Into account two pore sizes, the number of parameters denoting the pore structure 1s increased from 2 to 5, which already requires a considerable extra experimental effort Another lnterestlng aspect of reaction In solid catalysts was described by Jensen and Ray (F-311, These authors descend to the level of micro-crystallltes that are found These crystallltes have dlmenslons on the pore walls In supported metal catalysts. between lo-' and 10m8m The authors present a steady state model for reactlon, heatand mass transfer at these crystallltes, and they find that stable osclllatlons may occur under certarn condltlons These osclllatlons may add up to osclllatrons In that result In measurable varlatlons In local certain zones In the catalyst partzcle, convers=ons, Particle to fluld heat and mass transfer 1s often studled separately from chemrcal reactions, with the assumption that these processes generally do not Influence one as was shown by Paspek and another. This 1s not so for large temperature gradients, Varma (A-5) The authors interpreted their experimental results In terms of a reactor

Current

problems

m the modehng

of chemxcal reactors

They deduced values for heat and mass lncludlng axial temperature gradients model, transfer coefflclents that were 5 to 7 times higher than predlcted from measurements wlch was attributed to the severe temperature gradients at the without reactlon, In another paper about reaction and mass transfer with severe gradients, solld surface Juza and Cerm&k (I-54) demonstrate that one should take Into account the effects of (compare section 5 2). thermal dlffuslon under these condztlons. It 1s recommended that tbls aspect 1s consldered for other gas solid processes with large temperature gradients Compllcatlons arIsIng sectrons 6 2 and 6 3

5 2

Non-Catalytrc

from

Solid-Gas

flxed

and

fluld

bed

arrangements

will

be

dlscussed

In

ReactIons

Processes for converting solrd reactants are flndrng an lncreaslng chemical reactron engineers. A number of papers presented at this solld-gas reactlons, notably the combustion of coal lncludlng the of sulfur and nitrogen oxldes (D-22) (G-37) (G-38), the gaslflcatlon the gaslflcatlon of black shale (p-561, and var=ous other processes

Interest symposium chemical of coal (D-23)

from deal with conversion (G-37) (G-39), (E-27) (I-52)

The modelllng of solId-gas reactlons 1s often complicated, because there are several ways the reaction zone may proceed through the solld phase, depending on circumstances If the solld 1s partzally converted Into a volatile compound, a porous solId may be formed The reactants diffuse through the porous zone to the unreacted core and react at the surface of that core This sltuatlon 1s described by the "shrlnklng core" model (Yagl and Kunrl, 1955) Sometlrnes this model 1s also used If a second solid compound If this takes up less space than the startzng material, or If both solid 1s formed, compounds are present XI a porous form An lnterestlng example of this sltuatzon was recently described by Klmura, FItsgerald and Levensplel (19791, concerning a reactor and a regenerator with a solid reclrculatlon In the reactor, Iron oxzde reacts with hydrogen sulfide to iron sulfide, In the regenerator the sulfide 1s burned to form the oxide It has been demonstrated that under partzcular condltlons the porosity of the outer layer of the solid particle 1s reduced by chemical conversion of the solrd, which may eventually make complete conversion xmpossrble. A "pore plugging" model was discussed by Chrostowskl and Georgakls (1978) In the last symposium In this symposium, there 1s an example of the use of each of these Shlral (E-27) studled the hydration of calcium oxide pellets, which process Careful experlmentatlon revealed rather unusual temperature functions of time and radial posltlon. The authors prlmarlly report results A model 1s proposed based on the shrlnklng core prlnclple, account the compllcatlon of vapour fonnatlon In the pores, Lee, Hodges and Georgakrs (G-38) present particles with sulfur droxlde, based on structure of the mzneral determines the

models Dutta and 1s a very exothermal proflles as on experlmental but takzng znto

a model for the reaction of calcium the plugging pore Idea It 1s shown overall klnetlcs

carbonate that the

Processes for conversion of gaseous reactants into solid products are not common, but several technlcal examples exist, such as the oxldatlon of some volatile metal chlorides of Fe or TX) to solld oxldes and chlorine or hydrogen chloride Another process (e g in this category 1s the vapour deposltlon of sllrcon by reduction of slllcon tetrachlorzde with hydrogen, which 1s the subxect of a modellrng study by Juza and Cerm& (I-54) These authors demonstrate that a complete model, lncludlng launnar flow, heat conduction, molecular dlffuslon, thermal dlffuslon, and chemical reactlon, may be practical for design purposes, despite Its complexity.

53

Trickle

Flow

In trickle flow reactors, a number of different regimes may be reco9nlsed, depending on relative flow rates of the gas and the lrquld phase (compare Charpentler, 1978) Even at moderate flow rates, the flow pattern may be quite complex Only part of the particle surface may be wetted, which means that only part of the pores are filled with llquld The wetted parts of the surface may be dlvlded into stagnant and dynamrc zones. One needs then at least two mass transfer coefflclents for modelllng the liquid-soled mass

D

1846

transfer the liquid

So even the flow rate.

sxnplest

model

needs

THOENES

four

parameters,

that

all

are

dependent

on

The wettxng parameters may be determlned by purely physxcal methods, such as reported by Slcardl and coworkers (B-9). By measuring mass transfer rates w1t.h and wlthout chemical reaction at different lxquld flow rates, the other two parameters are estlmated The authors warn against the practxce of using the "liquid hourly space velocity" as a relevant crlterlon for scale-up Indeed, thus concept has no real physlcal meaning, when comparing reactors of different height. Because the wettxng phenomena are so complicated, reliable scale-up models can only be obtained by experlmental lnvestlgation of the process at the same speclflc llquld and gas flow rates as ~111 be used on the larger scale Consequently, with small laboratory columns one will have to measure very low relative conversions Van Kllnken and Van Dongen (B-8) advocate the use of laboratory trickle flow columns where the voids between the particles are fIlled with an Inert powder, In order to ensure wettxng at low llquxd flow rates Though this practice may be useful In studyrng klnetrcs under various chemical condltlons, I would not recommend it for scale-up studies Crzne, Marchot and L'Homme (B-7) present a phenomenologlcal model, that also contaxns four empxclcal parameters Since they all depend on lxquxd flow, they can be effectlvely replaced by one wetting parameter By flttlng the kznetlc constants, the model predicts the varlatlon of chemxcal conversion with flow rate for a number of condltlons, with a reasonable accuracy A general conclusion for trickle reactors that follows from these papers IS, that It would not be practxcal to try to make a model for general use Such a model would Instead, contaxn too many parameters for any given process, one may find practxcal as long as one sticks to rather narrow flow regimes models to be used for scale-up,

5.4

Gas

Absorption

In

Lxqulds

In gas absorptxon with rapid chemzcal reaction, the essential phenomena take place In a thin film close to the Interface Thus sltuatlon that occurs frequently In practice has several Important consequences. One consequence of practical Importance In process development ls, that this type of absorption can be studled In small laboratory reactors, even In ldeallsed reactors such as wetted wall columns, llquld 3et reactors and carefully stlrred cells, Another characterlswhich have well defLned surface areas tic of absorption w1t.h rapld chemxcal reactlon 1s. that the concentrations of some This may have reactants In the reactlon zone may vary from a maximum value to zero In consecutive reactions an Important effect on the selectlvlty of the reactions, e g In case of reversible reactlons with desorptxon of a reaction product, conversions When two may be obtalned that are higher than correspond to single phase equllxbrlum such reactions take place at the same time, dlffuslon resistance ~111 favour the selectxvlty with respect to the fastest reactxon In fact, the use of a two-phase reactlon system may have some advantages comparable to the use of a catalyst, especially In vxew of the possxbllltxes to increase reaction selectlvltles In process development, selectlvlty problems require as much attentxon the effect on product cost of 1 % as reactor capacxty In many large scale processes, varzatlon In selectlvlty may be slmllar to the effect on product cost of some 10 - 30 varlatLon 1n reactor volume Gas absorption modellrng 1s the sub3ect of several papers In this symposxum. HBrner, Abbenseth and Bergbauer (E-30) present a fundamental model for correlatrng mass Their transfer rates at lxquld surfaces wxth locally measured turbulence propertxes whxch xncluded the use of a laser-Doppler-anemometer, allowed experimental technique, the meaurement of local fluctuation velocxtles at distances from the surface of about This technxque could be lnterestlng for use In reacting systems 10e3 m Alper, Wlchtendal and Deckwer (E-28) lnvestlgated mass transfer from gas bubbles to Though llquxd suspensions of finely dxvlded catalysts, with two model reactions suspended powders In general do not Influence mass transfer rates much, the authors found that adsorbents may contrlbute slgnxflcantly to gas-llquxd mass transfer. A qualltatlve physical model predicts a greater enhancement of the mass transfer coefflclent the smaller the partxcles. It 1s recommended that thus effect 1s consldered for modelllng of mass transfer In catalytic slurry processes In general

%

Current problems

m the modehng

of chemical reactors

1847

An lnterestlng compllcatron arrses when absorption 1s accompanied by desorptlon of Joosten and coworkers (E-29) report reactlon products (compare Shah and Sharma, 1976). on a process where a reactlon product (HCl) acts as a catalyst and 1s desorbed It wrll accumulate to a certain extent In the reactlon zone, and raise the enhancement factor considerably, as compared to non-catalysed absorptzon The authors present one based on the film theory and the other on the penetration two quantltatzve models, The results of the numerlcal calculatzons show no slgnlflcant drfferences theory Thus 1s to be expected, srnce for bubble reactors there 1s no a prlorl knowledge Like the film available about the contact time to be used In the penetration model thickness in the film model, It has to be estimated from mass transfer measurements since It requires less computing Of course, the film model has to be preferred then, time In most gas absorptron models used In practice, only mass transfer of the chemically active components 1s taken into account However, Inert gases present In the bubbles or dissolved In the llquld, may have a retarding effect on the absorption of the active compound. This 1s demonstrated for szngle bubbles by Wild and Schliinder both by modelllng and by experiments. During the absorption process Inert gas (P-641, 1s stripped from the llquld and accumulates in the bubble If the actrve compound IS better soluble, It ~111 be removed from the bubble with a certain speed, leaving behind a bubble that 1s now relatively rich In the Inert compound In the final stage of the both components are absorbed Inert gas that was orglnally present In the process, bubbles has a slmllar effect In thus context I want to mention a paper about slmultaneous absorption and desorptlon by Cornellsse and coworkers (1980) The authors present a comprehensive model for the simultaneous absorption of two gaseous compounds, The numerical calculatfollowed by reversible chemical reactlons with different rates Ions can be used very generally for different sltuatzons, If klnetlc data are available. One example 1s especially interesting It concerns the s3multaneous absorption of hydrogen sulfide and carbon dloxlde In an smlne solution, a process that 1s used for flue gas purlflcatlon Both reactlons are reversible, but the hydrogen sulfide reacts much faster than the carbon dloxrde. In the transfer zone, the concentratlon of absorbed H2S may become so high, that the C02-amme-complex 1s decomposed, resulting In desorptlon of the CO under condltlons where the bulk composltlon of the llqurd would favour absorption T 2 1s 1s an example of the prlnclple that selectlvlty may be deternuned entirely by relative klnetlcs Another lnterestlng compllcatron arlses when the desorbed component reacts rn the gas phase with a reactant Lefers and coworkers (D-19) rnvestlgated the scrubbing of tall gases from nltrrc acid plants wrth concentrated nltrlc acid This process, that seems to run contrary to what our lntultron would predict, 1s based on the fact that nitrous oxide (NO) 1s very little soluble, but 1s easily oxldlsed by gaseous nltrlc acid to nitrogen dloxlde (NO21 and nitrogen tetroxlde (N204) The latter compound 1s readily soluble In nltrrc acid The oxldatlon takes place very rapidly In a thin layer close to the Interface. The model descrlblng this process 1s the well known model by Danckwerts (1970) for instantaneous reactzons.

6 6

1_

Single

PARTIAL Phase

AND

OVERALL

REACTOR

MODELS

Reactors

The znfluence of residence time drstrlbutlon on conversion 6.1.1. Langensrepen (P-62) compares a number of reactor types on the basis of Peclet numbers for axral dlsperslon Instead of us=ng a cascade of strrred tanks, a horrzontal tubular reactor with gas agktatlon may be practical, not only for gas-llquld reactrons, but also for single phase reactions Axial dlsperslon as function of gas and llquld flow rates have been studled extensively (compare Ottengraf, 1980). The narrowest reszdence time dlstrlbutlon 1s obtained In packed columns The risk of foulrng may be reduced by using pulsed packed columns, such as are employed for llquld-llqurd extractlons (compare Srmons, 1978). 6.1 2. Macro and micro mlxlng In stirred tank reactors Here we have one of the most complzcated problems In reactor modellrng Several regimes have to be dlstlngurshed, depending on the relatzve rates of various rmxlng processes and the rate of reactron We consider essentrally lamlnar flow condltlons In intermedrate vlscoslty ranges, the reszdence time dlstrlbutron may be close to the RTD of a perfectly mlxed reactor If the macro mlxlng rate 1s fast compared to the reaction rate, the macro mlxlng process 1s the llmltlng factor. Macro mlxlng 1s often described In terms of a "mlxlng time", whzch does not have a quantltatlve physlcal basis, but 1s easily measured and may be used as a crzterlon for scale up Le Cardinal

D THOENECS

1848

and coworkers (p-63) used the m~xlng time fox characterzsrng one finds wrth different impellers For each impeller type, the dlmenslonless mlxlng number (mrxlng tune trmes Impeller Reynolds number and the power number.

polymerlsatlon reactors a unrque relation between speed), the impeller

When the reactor contents are so VLSCOUS, that the RTD deviates consrderably from the one needs two parameters to characterrse the RTD perfectly mixed tank behavlour, A slmplzfLed model can be made If complete segregation 1s assumed, that means that there 1s no mlxlng of lrgurd elements with different composltlons Chemical reactlon 1s then only possible If the reactants are mixed before enterzng the reactor, and are sufflclently cooled to exclude reaction In the feed line Thus would require heat exchange within the reactor between volume elements of different ages, whrch might turn out to be rate detennlnlng Therefore I think that models for complete segregation are not very reallstlc for single phase reactlons A further refinement 1s made by recognlszng a mlxed zone around the Impeller, where macro mlxlng 1s very fast, and segregated zones In the remainder of the reactor, and assuming a certain lzguld clrculatzon through these zones. A model based on this Idea was presented by Valderrama and Gordon (1979). In their model the mrxed volume fraction The authors made calculatrons for conversions of reactions wrth IS a parameter. arbitrary reactzon orders, as functions of mlxzng condrtlons mlxed feed was However, assumed Van Dlerendonck and coworkers (p-60) present a slmrlar model, but w1t.h the Important Improvement that seperate feed streams are introduced Into different mIxed zones, whrch are assumed to be of equal volume Their model contains two parameters, the mrxed volume fraction and the llquld clrculatlon rate Plug flow 1s assumed for the segregated part of the reactor Both mlxlng parameters were estrmated from remalnlng, tracer response measurements The mlxlng model so obtained was then combined with a klnetlc model for copolymerlsatlon (compare sectlon 4 2) It was shown that marked dzfferences In copolymer composrtlon may be expected on scaling-up of the process A slmllar problem was rnvestlgated by Wehner (19781, concerning a copolymerlsatlon propylene and 1,4-hexadlene) In lamlnar flow In a tube reactor He (of ethylene, ethylene was locally depleted, so that rn thrs found that due to dlffuslon llmltations, region a polymer was formed with a higher propylene content (and a lower molecular weight) Thus appears to agree with the fIndIngs of Van Dlerendonck and coworkers (p-60) In reactors with hrghly viscous media and with seperate feed streams, the rate of macro mlxlng may be a rate determlnrng process An effectrve method for determrnlng the rate Llntz of micro mlxlng 1s based on the use of a chemical reactlon wrth known krnetlcs and Weber (E-26) present an lngenlous method, based on the use of an autocatalytlc The rate of this reactlon shows a sharp maximum at a certain degree of reactlon By almlng at maximum conversion at maximum mlxedness, any devlatlon from conversion maxlmum mrxedness will result In a lower conversion However, the authors found a Apparently, both macro complicated relation between conversion and impeller speed One would need a detailed model mlxlng and micro mlxlng were changed at the same time Such a model would need at least two parameters to explain the quantltatlve effects either In terms of residence time dlstrlbutlon to characterzse the macro mlxlng process, or In terms of a model such as was used by Van Dlerendonck and coworkers functrons, In addltron to that, one needs a micro mlxlng parameter, and one has to (P-60) recognlse two-way dlffusron and chemz.cal reactlon on the scale of the lamlnar eddles eddres as a parameter Geurden and Thoenes (1972) used the average thrckness of lamlnar Vrllermaux (1972-1979) ana coworkers (see Piasarr, David and Vlllerfor nmcro-nnxlng for t:-e earl;- phase 05 micro-mlxlng, and ,naux, 1978) used a shrlnklng aggregate model noth models have only one a mass transfer model for the later phase of micro-nrxlng (a characterlstlc trme). micro-mixing parameter Tubular reactors wrth radial heat flow Two papers In this symposium deal with 613 et al (H-44) consider a packed bed reactor for a this modelllng problem. Bub with plug flow and radial Fischer Tropsch process as a quasi homogeneous tube reactor, and heat transfer at the wall Data for thermal transport properties heat conduction, The obtalned klnetlc data were determined seperately were taken from the literature, (compare sectlon overall model was used succesfully for predlctrng product composltlons 4 2.1. Sundaram and Froment cracking of ethane local values of the

(R-46) simulated The authors used eddy dlffuslvlty

an Industrial tubular reactor for available correlations for velocity Klnetlc constants In turbulent flow

the thermal profIles and were

Current

problems

m the modelrng of chemical reactors

The overall model descrrbes radial temperature and concentration de termzned separately and axial profiles for cross-sectional average temperature and concentrations. proflles, The overall Some model parameters were estimated from radral temperature profIles and predrcts accurately reactor product model so obtarned 1s unusually sophrstlcated, that a one-dlmensronal model, which requires composrtlons The authors demonstrate, gives results of lesser accuracy For an accurate design much less computing effort, the use of the more complicated twoor optlmlsatron of an exlstlng plant reactor, dlmenslonal model appears to be well 3ustifled but used the penetration theory Ooms and coworkers (1978) studled the same problem, The use of this model for modelllng heat and mass transport at the reactor walls that for fast reactions gas pockets might introduced an artziflcral phenomenon, VIZ The authors rntroduced become depleted of reactant wLthrn the contact time at the wall whrch under certain condrtlons was of the an "enhancement factor" for heat transfer, order of 4 Two experlmental studies are 614 Transient behavrour of stlrred tank reactors that both use the hydrolyszis of an epoxlde as a model reactron. Creswell reported here, and Santos (F-36) studred the klnetrcs In a batch reactor and made a model to predict the transient response to changes In catalyst feed In a continuous pllot reactor Heemskerk, D-ers and Fortuln (P-55) demonstrated the phenomenon of limit cycles In a contrnuous reactor with accurate feed forward control of all rnput variables The Parameters were authors present a complex model contalnrng all pertinent varrables determined by experiments where temperatures were measured with extreme accuracy It This phenomenon may be was shown that llrnlt cycles may occur under certain condltlons especially during start up, because of practrcal Importance for lndustrlal plants, one ~~11 then pass through a wade range of condltlons Horde and JlrdEek (P-61) demonstrate ways to control continuous starred tank reactors In open-loop unstable states The authors studled the effect of the dynamrc propertres of all essential parts of the system Generally speaking, the safest control 1s obtained by manlpulatlng the fastest acting variable, such as the feed rate and the Inlet temperature of the feed Another aspect of transzent behavlour that 1s of practical rmportance 1s the varlatlon This problem was studred In yreld of a set of consecutrve and parallel reactions recently by Varma and DeVera (1979) Their calculations showed, that during start-up of a continuous reactor, under certarn condltlons the yield of a desrred product may A cycling operation might pass through a maxrmum, that exceeds the steady state value then result In Increased yrelds These phenomena need further experlmental verrflcatlon. 6

2

FIxed

Bed

Reactors

In the modelllng of flxed bed reactors one 1s faced wrth a geometrlcal problem, not unlike the problem of modelllng the structure of porous catalysts Both structures are of a stochastrc nature, that can not be adequately represented by a completely regular If one would consrder a packed bed as a regular arrangement structure For Instance, of spheres, a systematic error 1s Introduced, because the cross sectlonal area avaIlable to flurd flow In a regular packing varies consrderably w1t.h axial distance, whrch 1s not the case In random beds Thus results rn a systematic difference In flow pattern Another important aspect of real beds 1s a certain anlsotropy caused by the walls In a zone wIthIn several particle diameters from the wall the porosity 1s higher than In the core, which results In considerable differences In flow rates In a heat or mass transfer process, the fluId In the core ~111 reach equrlrbrrum wzth the particle phase much sooner than the fluid In the outer region The combined effect 1s conslderable, especially at low values of the Pdclet number (the product of Reynolds and Prandtl or Schmrdt numbers) In the previous symposium Schliinder (1978) presented an excellent review on packed beds, The effect 1s where this effect was dlscussed In detail reduced however and further complicated by radial drsperslon which 1s governed by the Reynolds number alone In order to take these effects into account, one needs a packed bed model with at least three porosity parameters, two equations for mass transfer, and another one for mass exchange between the two regrons All thus greatly Increases the complexity of a packed bed model However, by dzsregardlng these effects, one finds a conszderable scatter In mass transfer data (represented as Sh or 3D numbers as functions of Re), such as have been found In lrterature for many years Paspek taking axial

and Varma (A-5) present a good overall model Into account chemical reaction, partrcle to dlsperslon of heat ans mass. The authors did

for an adlabatrc packed bed reactor, flurd transfer of heat and mass, and not study the rnfluences of the tube

D

THOENES

dlsmeter and consequently drd not consider wall effects. The authors did succeed In accurately predrctlng the degree of conversion of ethylene oxldatzon on a solid they could derive values for catalyst, as a function of feed rate. At the same time, local heat and mass transfar coefflclents and axzal dlsperslon coefflclents. The transand the drsperslon coefflclents lower than predlcted fer coefflcAents were hzghers, from experiments without reaction (compare section 4 1). Several papers deal with the transient behavlour of adsorptzon and desorptlon In packed bed reactors HlavdEek and coworkers (F-33) studled the effect of temporary They found some polsonlng of exhaust purlflers by tetra ethyl lead and other compounds currous examples of hysteresis after polson=ng and de-polsonlngNo mathematical model 1s presented, but the graphs with experlmental data may serve well as "learning models" for lllustratlng this phenomenon Relnke and coworkers (D-23) measured for adsorbing SO2 and HCl from flue transient behavlour adequately The apparent klnetlcs.

breakthrough curves for fixed bed reactors used gases A one-dlmenslonal model describes the model 1s based on plug flow and sample first order

Cho, Carr and Arls (B-10) used the prlnclple of transient adsorptlon In a novel continuously operating gas chromatographlc reactor This appears to be a very lngenlous with a narrow spaczng concept, The bed 1s contarned between two coaxial cylinders, The carrier flurd 1s passed continuously through the bed, but the feed port by whrch Both reaction and separation take place at the reactants are Introduced rotates slowly. Thus reactor offers the same time Conversions greater than equlllbnum may be reached lnterestlng possrbllltles Practical appllcatlon IS of course llmlted by the low specsfzc production rate reactlon and axial flow are taken Into account In the reactor model, only adsorption, The condltrons are such, that the resrstance to particle to fluid mass transfer and lateral dlsperslon need not be consldered

63

FluId

Bed

Reactors

A bubbling fluld bed IS certainly one of the most complicated contacting devices used Despite a really huge volume of literature on this sub3ect, In process englneerlng. there s-t111 1s no model avaIlable for general appllcatlon We understand now that such The reason 1s that In a bubblrng fluId bed several a model would be lmposslble to make In the Houston complicated phenomena that are Interdependent, occur at the same trme that a physical model symposium, Van Swaal3 (1978) showed In a comprehensive review, that would take Into account the most Important effects that are experlmentally observed, Only a few of these could be determIned separatewould contain at least 13 parameters This about IO parameters would have to be derived from overall experiments ly, so that And even If one succeeded with t.h~~, such a complicated would of course not be feasible The matter 1s further complicated when model would not be sultable for practical use. both slow and fast reactlons take place In the bed, as we wzll see later Most lnvestlgators of fluld that contains not more than lImIted range of condltlons, reaction rates The problem

bed reactors attempt to make a slmpllfred physlcal model Such a model can only be applied for a 3 parameters speclflcally as concern scale of operation and relative to find where the llmltatlons lie IS, of course,

They used Yates and Gregolre (H-48) lnvestrgated the catalytic oxldatlon of o-xylene a slugging bed, with bubbles of a diameter equal to the diameter of the reactor (0 1 m). greatly reduced the The Inter-slug distance was estimated as a constant 0 2 m_ This The authors used an Improved version of a model proposed by other modelllng problem authors, descrlbmq mass transfer between the slugs, the wakes and the emulsron By rncludlng klnetlc data, the authors succeeded In correlating conversions wrth flow rates =n a satisfactory manner In a lecture at the the sltuatlon 1s far more complzcated For large bubbling beds, Krrshna (1980) presented a strategy for Reaction Englneerlng Colloquium In Ghent, He used a model with three parameters, based on deslqn and scale-up of large reactors bubble hold up, bubble rrse The parameters are theories of DavIdson and Darton Krishna proposed to velocity and mass transfer coeffrclent per unit bubble volume measure these by physical experiments In reactors up till 1 m In diameter and 4 m bed The overall reactor model takes into account back mlxlng In the emulsion phase, height which 1s Important for predxctlng conversions for relatively slow reactions

Current problems

m the modelmg

of chemical reactors

1851

Werther (H-47) presents a model that 1s based on different assumptions, that also He presents emplrlcal relations for estlmatlng contains three physlcal parameters these physical parameters, for two types of fluldlslng material The empIrIca correlatlons are based on the observatzon that bubbles coalesce and grow, but reach a stable diameter above a certain height of about 1 m, and that bubbles of equal size rise faster In beds with greater diameter The overall model 1s based on the assumption, that the emulsion phase 1s completely mlxed, and that mass transfer and reactlon can be described In a way slmllar to gas absorption, making use of the film theory This assumption appears quite unreallstlc, since there 1s of course no such thing as a stagnant layer close to the bubble surface, where mass transfer can be described by molecular dlffuslon only Even so, calculation of the Hatta numbers for the examples given by Werther, shows that they are all smaller than 1, which means that the mass transfer 1s rapld compared to chemical reactlon Nevertheless, Werther shows good agreement of this model wzth experiments, for the hydrogenation of ethylene, publIshed by other authors Werther's model correctly predicts In three other examples, the varlatlon of conversion with flow rate, for fltted values of klnetlc constants I do not think these flndlngs prove the valldlty of the proposed model However, I do agree with the author that this model may be used for predlctlve purposes, If sufflclent care 1s exercised I want to mention a recently publIshed model by Fan and Fan (19791, that recognlses The authors present a method for calculating three phases with interchange coefflclents in models of this complexity the conversions, even for transient condltlons However, that It 1s doubtful If these can be numbers of physical parameters 1s so great, estzmated with sufflclent accuracy. For design purposes, I prefer the models of Krzshna (1980) and Werther (H-47) In thus symposium, several papers are presented on appllcatlons of fluld bed reactors, (D-22)(G-37, 38, 39) (p-56). Lee, Hodges and Georgakls (G-38) lnvestlgated the reaction of sulfur dloxlde with lImestone In a fluId bed coal combustor They combined a model for the klnetlcs of the The authors gas-solid reactlon (see section 5 2) with a two-phase fluid bed model namely that sulfur 1s released uniformly make one assumption that may appear doubtful, throughout the emulsion phase as SO2 The agreement between the model and the expertments 1s not very accurate, but this may have a number of reasons If we consider the complexity of the process Park, Levensplel and Frtzgerald (G-37) present a very lnterestlng comparative study on various models that have been proposed for fluid bed combustion of coal The authors argue that devolatlllsatlon of coal particles takes place so rapidly, that the bed has to be dlvlded into two zones with a reducing and an oxldlslng environment, respectively Under certain condltlons, the "plume" conslstlng of a reducing envlronment, may extend to the top of the bed, which results In a considerable temperature Iump above the bed The gases, released by volatlllsatlon (Including H2S) ~~11 burn at the surface of the plume, a sltuatlon comparable to a dlffuslon flame The authors propose one dlmenslonless "plume group", which characterlses the shape of the plume This group does not contain any klnetlc constants, since It was assumed that the devolatlllsatlon 1s fast compared to radial dlsperslon Since the paper contains very few experImenta data, I find lt dlfflcult to Judge the general valldlty of the proposed model However, In a qualltatlve way, the model 1s convlnclng enough to serve as a challenge to other students of fluId bed coal combustion Furthermore, the model 1s of a strlklng slmpllclty

64

Bubble

Columns

and

StIrred

Gas-Llquld

Contactors

In many gas-llquld reactors the gas phase 1s dispersed as bubbles In the llquld phase. A sufflcrently fine dlsperslon 1s made either by forcing the gas stream through narrow holes or by us%ng an impeller with a high tip speed A sufflclently large rnterfaclal area 1s malntalned by either passing a large gas flow through the reactor, or by using an Impeller, mostly of the turbine type, that ~111 circulate the dlsperslon In the reactor The experimental study and the modelllng of the physical phenomena occurrlng In gas-llquld dzsperslons have been much more succesful than for the bubbling fluldlsed bed. The maln reason for this 1s. that gas bubbles XQ transparent llqulds can be observed better than bubbles In beds of solId partzcles Another reason XS, that under condltlons where little coalescence occurs such as one often meets z_n practice, gas bubbles In llqulds have a longer llfetlme and can be reproduoed In small scale equipment On the other hand, under condltlons where rapld coalescence occurs such as

1

1852

D

THOENES

in liquids of relative purity, the modellrng of the coalescence and break-up processes IS drfflcult (compare Koetsler and Thoenes, 1972). Other typrcal problems in gas-lrquld reactors are the Influence of concentration gradients on bubble behavlour, and the mrxlng in the llqurd phase between the zone where bubbles are formed and the bulk of the llquzd. Both phenomena are Important when mass fluxes across the interface are locally high A comprehensrve paper on modelllng of stlrred gas-llquld contactors 1s presented by Botton, Cosserat and Charpentzer (B-11). The authors present emplrlcal models for the hydrodynamics and the mass transfer rates under wade ranges of conditions Operatrng Comparisons wrth bubble columns zones sultable for practical appllcatlon are lndrcated are Included, based on the speclflc mass transfer capacity per unit power Input StIrred tanks are shown to be superror up till diameters of about 3 m The experimental measurement of mass transfer rates and surface areas has long been an area of some uncertainty, because most experlmental methods are rather rndlrect A common method 1s measurement of mass transfer with and wrthout a rapid chemzcal reactlon of well known klnetlcs. By assuming certain models for mass transfer and reactlon, both the mass transfer coefflclent and the surface area follow from these two measurements. There are at least several pitfalls along this route One 1s the possible Influence of the chemical reaction on the mass transfer coeffrclent The chemical reactron may be accompanied by stronger concentratron and temperature gradients, which Joosten and coworkers may cause InterfacIal turbulence and affect mass transfer rates (E-29) used an lnterestlng method for crreumventrng this problem They measured at the same time the absorption of the reactant and the desorptron of an inert gas. In this both the physlcal mass transfer rate and the chemically enhanced mass transfer way, rate are measured under the same condltlons_ Another possible pltfall was rndlcated by Lopes de Flguerredo and Calderbank (1979) In laboratory experiments one often uses a continuous gas flow, without replacement of the llquld For short periods of time a quasi steady state 1s often assumed, and models for mass transfer wzth chemical reactzon are based on thus assumption. The authors show that a transzent model should be used to account for changes In gas phase concentration It 1s believed that previously publlshed data on surface areas and mass transfer coeffrclents may be In error when this effect has not been taken Into account The authors used an optlcal method for measuring the surface area KaSt&nek and coworkers (P-57) propose a strategy for the scale-up of bubble reactors The authors recognrse the effect that chemical reactlons may have on the nature of the lnterfaclal area They assume that this effect ~~11 be vlslble m the gas hold-up It 1s further assumed that the mass transfer coeffrclent and the lnterfaclal area are functions of the gas hold-up The authors use a made1 that correlates these parameters for various condltrons and scales of operation It 1s then sufflclent to measure the gas hold-up under reacting and non-reacting conditions In a laboratory column, In order to find a design model for a large reactor Hatcher and Hart (B-12) demonstrate the succesful reactor, based on measurements In batch reactors on the assumption of llqurd srde mass transfer and the rate coefflclents were determined seperately the selectlvlty were predlcted accurately for the

7

SOME

CURRENT

PROBLEMS

IN

REACTOR

design of a continuous gas-llqurd The model that was used was based chemical reactlon rn series, and Both the degree of conversion and continuous operatron

MODELLING

When studying contrlbutrons to the art o f chemrcal reactlon englneerlng of the last I find some modelllng problems few years, lncludrng those presented at this symposium, that appear to require addrtronal effort It seems to me that the effort that 1s most needed 1s generally of an experImenta nature I want to recognlse the followrng categories of modellAng problems The effects of the simultaneous transfer of mass and heat across InterfacIal phenomena can only be modelled accurately rf the gas-lrquld interfaces under reactlon condltlons, combined effects are studled This applres also for reactrons at solld surfaces, when The Interthere are severe concentratron or temperature gradients near the surface actron of a gas phase and solrd surfaces may need addltronal attentron, especially for finely dzvlded solrds and particles with a high porosity. Solld-gas reactlons In the partial chemical converslon of solld reactants, several complicated phenomena may take place The way the reactzon zone moves through the solId

Current problems

m the modehng

of chemxal

1853

reactors

of the solid depends both on the klnetlcs and the structure There 1s a great need for models of a more general appllcabrllty Mass transport In reactors with complicated flow patterns In several reactors of the flow patterns are so complicated, practical Importance, that physlcal modelllng Examples are the gas fluldlsed bed reactor, the stlrred will always be a problem tank reactor with a lzquld of high vlscoslty, or with a coalescing gas dlsperslon, and the trickle bed These modelllng problems require experImenta studies of the entlre reactors lncludlng all relevant phenomena, preferably also on larger scales The need 1s even greater for processes with relative fast reactions, followed by slower reactIons Dynamics of chemical reactors Much research In this field has been in the area of "learning models" There IS a greater need for studies in support of process control models and optrmlsatlon models, with an emphasis on safety and on product yields As a general conclusion, I want to suggest that In many physical modelllng have been attalned There appears to modelllng studies concerning the simultaneous effect of 3_n real reactors

cases the llmlts be an increased several phenomena

of pure need for that occur

REFERENCES Letters followed by numbers International Symposium on - SC1 Chem 35_, 3-525. Eng

between Chemical

brackets ReactIon

refer to papers contributed Englneerlng (1980), publlshed

to

the In

Sixth

Charpentrer, J C (1978) Chem Reactlon Eng Reviews-Houston, ACS-Symposrum Serves 72, 223-261, (1978) Chrostowskl, J W and Georgakls, C Chem Reaction Eng -Houston, ACS-Symposium Series 65, 225-237 Cornellsse, R , A A C M Beenackers. F P H van Beckum, and W P.M van Swaal3 (1980) submrtted for publlcatlon to Chem. Eng Scl. Danckwerts, P V (1970) Gas-Llquld Reactions, McGraw-HI11 Fan, LT and L S Fan (1979) Chem SC1 34, 171-179 Eng (1976) Gerrens, H Chem ReactIon Eng -Heidelberg, Proc 4 ISCRE, Dechema, 585-614 Geurden, J M G and D Thoenes (1972) ReactIon Englneerlng-Amsterdam, Chem Proc 2 ISCRE, Elsevler, Amsterdam, B-l, 35-44 Grazlanl, K R and M P Ramage (1978) Reaction Eng.-Houston, Chem ACS-Symposium Series 65, 282-291. Klmura, S , T J Levensplel (1979) Fitzgerald and 0 Chem Eng Scl. 34, 1195-1201 Koetsler, W T and D_ Thoenes (1972) Chem ReactIon Eng -Amsterdam, 2 ISCRE, Proc. Elsevzer, Amsterdam, B-3, 15-24. Krishna, R (1980) Design and Scale-up of Gas Fluldlsed Bed Reactors, to be presented at NATO Adv Inst on MultIphase Reactors, Portugal, S13thoff-Noordhof. Lopes de Flguelredo, M.M and P H Calderbank (1979) Chem Eng. Ser. 34, 1333-1338. LUSS, D (1976) Reaction Eng -Herdelberg, Proc. 4 ISCRE, Dechema, 487-516 Chem Ooms, G . G Groen, D P de Grass and J F Balllntlln (1978) Chem Eng Scl 33, 357-363 Ottengraf, S (1980) Chem Eng. Scl 35, 687-707 Plasarl,

E , R David and J Vlllermaux (1978). Chem Reaction Eng.Houston, ACS-Symposium Series 65, 125-139 Reaction Eng. (1978). Chem Reviews-Houston, Schliinder. E U 1 lo-161 ACS-sy&osrum Series, 72, Shah, Y T. and M M Sharma (19761, Trans. Inst Engrs, 54, 1 Chem. (1978). Chem Shznnar, R Reaction Eng Reviews-Houston, ACS-Symposzum Series, 72, (1978) and Ind , 1978, 748-757. Slmons, J F Chem Reaction Eng van Swaal3, W P M (1973). Chem Reviews-Houston, ACS-Symposium Series, 72, 193-222. Valderrama, J 0 and A L Gordon (1979). Chem Eng. Sci. 34, 1097-1103. Varma, A and A-L DeVera (1979)_ Chem Eng_ Sc1. 34, 1377-1386 Vlllermaux J. and Devlllon J C Chem Reaction Eng.-Amsterdam, Proc 2 ISCRE, Elsevler, Amsterdam, B-1,13 Vlllermaux, J (1979) Chzmla 33, nr 10, 377-378 Weekman, V.W. (1975) Chem. Reaction Eng. Reviews-Evanston, Serres 148, 98-131 Adv Chem. Wehner, J.F (1978). Chem ReactIon Eng.-Houston, ACS-Symposium Series, 65, 140-152 Yagl, S and D 5th Int. Symposium on Combustion, Reinhold, New York, Kunl1 (1955)

l-36

231.