Seminar on

Reduction Kinetics of Iron ore- Coal Composite pellets

Presented on 3rd December, 20 3 !" Rac#na $ripat#" % 2&$' R2()

*nder t#e +uidance of Prof, -,-,Ro" . Prof, P,K,Sen

Department of Metallurgical and Materials Engineering Indian Institute of Technology, Kharagpur 7213 2 India

M!DE""I#$

The current work aims to model the reduction kinetics of Iron ore-Coal Composite pellets reduced in a RHF. This chapter deals with the details of two different types of mathematical models, which are used to determine the reaction rate constants and temporal evolution of different phases during the reduction of pellets at any given temperature. The first model is the I !TH"R#$% #!&"%, taking into consideration the kinetics of reduction at a constant temperature. The '!'-I !TH"R#$% #!&"% incorporates the time and kinetics starting with room temperature ()*+ ,- upto the reduction temperature. This model utili.es the difference in the theoretical value weight fraction of phases and those o/tained e0perimentally to estimate the values of the rate parameters such as the activation energy values and the fre1uency factors associated for each step of reduction, /y employing a glo/al optimi.ation tool, 2enetic algorithm in con3unction with mass /alance ordinary differential e1uations.

1.1

Assumptions in the model

everal assumptions are made to e0pressed simplified the /ehavior of pellets in the proposed mathematical model. 4. The input ore is assumed to /e entirely hematite. ). The composite pellet is considered to /e a perfect sphere and pellet shape and diameter remain constant throughout the reduction. 5. The reactions considered are assumed to proceed in the forward direction only. It means that the thermodynamic conditions that will favor forward reaction prevailing during the reduction period. 6. $ll the wustite is converted to iron while fayalite formation is not taken into account. 7. $s heat is entering into the system only form top and heat flu0 e0ist only in 8-direction.

1.2 Theoretical Basis of Kinetic Modeling of Iron ore-coal composite pellets in RHF
The reduction of iron ore in an ore- car/on composite system proceeds /y the intermediate o0ides formation, which include conversion from hematite to magnetite, magnetite to wustite and finally wustite to iron (4.4 to 4.5- and reduction takes place /y gaseous intermediates mainly /y the regeneration of C! /y 9ouduard reaction as shown /y e1uation (4.6-.

5 Fe)!5 : C! ; ) Fe5!6 : C!) Fe5!6 : C! ; 5 Fe< : C!) Fe! : C! ; Fe : C!) C!) : C ; ) C! (4.5-

(4.4(4.)-

(4.6-

$ssuming that the reactions are of first order the rate of evolution of intermediate o0ides can /e descri/ed /y following e1uations 4.7, 4.=, 4.>, 4.+ and 4.*.

The a/ove e1uations represent four first order ordinary differential e1uations with = unknowns kinetic parameters ($H, $#, $?, "H, "#, "?- and four primary unknowns like concentrations of Hematite (CH-, #agnetite(C#-, ?ustite(C?- and Iron(CF-. $H, $#, $? (pre e0ponential factorsand "H, "#, "? ($ctivation energy- represent the rate parameters for Hematite, #agnetite and ?ustite transformations respectively. R is the universal gas constant while T is the reduction temperature in ,elvin. ince there are si0 e0cess varia/les than the num/er of e1uations, to get uni1ue solution from four e1uations, these varia/les, the rate parameters, are re1uired to /e estimated first. In the present work those parameters are estimated /y applying an optimi.ation procedure, the 2enetic $lgorithm. Thereafter, the evolution of phases is predicted /ased on these optimi.ed rate parameters. In the present study e0periments were carried out with three layers of composite pellets in RHF. The two diameters of spherical pellets considered were 4< mm and 4= mm. ,inetic models were developed for all the layers separately for /oth the I !TH"R#$% as well as '!'- I !TH"R#$% case. I !TH"R#$% model has /een developed /y solving "1ns. 4.6 @ 4.* /y R,-6 method. However for modeling the '!' @I !TH"R#$% system, an additional e1uation of heat transfer is solved along with the a/ove 7 e1uations ("1ns. 4.6- 4.*- , thus the output also includes the rate of change of Average Temperature of the pellet /ed with time . The e1uation is as followsA

?here B ;density of a phaseC Cp; pecific heatC D ; tefanEs contant ; 7.=><6 0 4< -+ C F ; porosity of /edC G; volume of the pellet /edC HH; heat of reaction for each of the reactions 4.44.6C Ta ; am/ient temperature (46>5 , or 47>5 , in this study-C T; Temperature of the pellet /ed.

The theoretical degree of reduction (I- can /e calculated from rate of o0ygen loss from o0ides.

?here, I.C. is the total remova/le o0ygen and !.% is the total o0ygen removed per cu/ic meter

of the pellet up to time t,

represents the o0ygen removal rate per cu/ic meter of the

pellet through hematite to magnetite transformation

imilarly,

and

represents the rate of o0ygen removal rate per cu/ic meter of

pellet through magnetite to wustite and through wustite to iron respectively. Rate of o0ygen removal is directly related to the rate of mass loss of hematite, magnetite and wustite according to e1uations given /elowA

?here JH, J#, J? are stoichiometric factors defined /y the e1uations (4.4=-, (4.4>- and (4.4+-.

Here

and

represent the molecular weight of hematite, magnetite,

wustite and iron respectively. The consumption of car/on can /e related to rate of o0ygen removal as followsA

1%3 Modeling &arameters

Heats of reaction This model re1uires estimates the heats of main of reactions occurring in reduction. These temperature dependent formulae are /ased on data from &onskoi et al.. $ polynomial temperature dependent function is used

H ; $ : (9 0 4<-5 T- : (C 0 4<-= T)- : (& 0 4<-* T5Coefficients $,9,C,& are given in Ta/le 4.).

(4.)<-

pecific heat &onskoi and #c"lwain have developed a methodology, also /ased on a polynomial in a/solute temperature, to estimate the specific heats of hematite, magnetite, wustite, iron, and car/on. The polynomial function is as followsA

Cp ; $ : (9 0 4<= T)- : (C 0 4<-5 T- : (& 0 4<-= T)- : (" 0 4<-* T5Thevalue of co-efficients $,9,C,&," are given in Ta/le 4.5

(4.)4-

!ensit" &ensity of solid constituents are taken constant /y assuming that the effect of temperature and pressure negligi/le. The values of density of solids are given in Ta/le 4.4.

Ta#le.1.1 !ensit" of different components

$omponent Fe)!5 Fe5!6 Fe! Fe C

!ensit"% &g'm() 7)<< 74=< 77<< >+<< ))=>

Ta#le 1.2 Heats of reaction for reactions %*.1)+%*.(), &- mol for reaction %*./) these 0alues are
.1

per 1 mole of $1 and for reaction %*./) the" are per 1 mole of car#on

Ta#le1.( $oefficients used to estimate temperature dependent specific heats of Fe 21(, Fe(1/, Fe, and $ using formula %22), - K mol
.1 .1

4.5 Methodolog" for estimation of rate parameters

$ genetic algorithm uses the general principle of evolution of Ksurvival of the fittestL. 2enetic $lgorithm (2$- randomly generates an initial population of individuals each of which represents a set of unknown parameters to /e estimated. Then 2$ sends each individual to kinetic model which evaluates the 1uality of individual /y comparing the predicted phases with the e0perimentally determined phases through MR& analysis of reduced pellets. The kinetic model solves the set of five ordinary differential e1uation /y RN'2$ ,NTT$ method using the rate parameters provided /y the 2$ individual.
The 1uality of the individual is 3udged /y fitness parameter which is defined as

?here CH, C#, C?, CF represent the concentration of the hematite, magnetite, wustite and iron respectively. The su/script p and e represent the predicted and e0perimental concentration of different phases respectively. Here higher the fitness parameter, minimum is the difference /etween the predicted and e0perimental value of the degree of reduction, and /etter is the individual. The kinetic model send /acks the information regarding the 1uality of the individual through fitness parameter to the 2$ routine. $fter evaluating all the individuals of the population /y kinetic model, 2$ generates a /etter population for the ne0t generation. $ set of genes from the population that provided the /est fitness is chosen upon /y some stochastic methods and operators. 2enerally these operators include reproduction, crossover and mutation. 2radually the population is upgraded /y the /est genes o/tained /y the operation of earlier genes. $ ma3or advantage of this algorithm is that it utili.es the information availa/le in current generation to form /etter genes for ne0t generation.

Indi/idual De+ree of Reduction

01perimental

$#eoretical de+ree of reduction

0rror bet2een t#e e1perimental and predicted Concentration of p#ases

3ptimi4ation of error b" -5 and calculation of rate parameters

Iterate till +eneration 6 &a1imum +eneration

3btain optimi4ed /ariable as output

fif% 1%1 'olution methodology adopted to sol(e for the e(olution of different phases )ith time during reduction or iron o*ide%

7or t#e current problem a real coded +enetic al+orit#m 2ritten b" Deb #as been utili4ed, $#e parameters used in t#e -5 for findin+ t#e optimum /alues of t#e rate parameters are pro/ided in $able ,8
1%+ &arameters used for the optimi,ation pro-lem and result of a run 3aria#le Total population si.e of a 2$ run Total 'um/er of 2enerations to /e run Cross !ver Oro/a/ility #utation Oro/a/ility 'um/er of real coded varia/les $ctivation "nergy Range Fre1uency Factor Range 3alue

4)<<
4<< <.+ <.<<7 = 4<< - 4<<< kPQmol 4<45 @ 4<4* s-4

$ real coded genetic algorithm written in C language has /een used for the current pro/lem. The optimi.ed parameter o/tained after studying the response to the current pro/lem is presented in Ta/le RRR. &e/ has emphasi.ed on the proper selection of range of the varia/les. The o/3ective function that is to /e minimi.ed is the R.#. (Root #ean 1uare- value of the error /etween the predicted and e0perimental degree of reduction at various temperatures of interest. ince the e1uations are a set of first order differential e1uations they are solved using any numerical method to solve a ordinary differential e1uation @ Initial Galue Oro/lem (!&" @ IGO-. The 2$ code used to o/tain the current result is clu//ed with a su/routine written in C language for solving the e1uations 4.7- 4.* using the Runge ,utta 6 (R,6- algorithm to find the concentration of various phases as reduction proceeds, i.e the variation of H, #, ?, F with t. olution of the simultaneous e1uations 4.7 @ 4.+ yields the amount of phases (hematite, magnetite, ?ustite and iron- at various time steps. The time step is taken as 4 second. The parameters are selected after carefully studying the variation of fitness parameter with the value

of varia/les such that the optimum parameters could /e reached. The fitness value decreases as the num/er of generation improves.

.ig% 1%2 /ariation of a(erage fitness )ith num-er of generation for I'!T0E1M2" model

.ig% 1%3 /ariation of a(erage fitness )ith num-er of generation for #!#3 I'!T0E1M2" model

$ plot of fitness value against the num/er of generations for sphere shape pellet respectively is provided in Fig. 4.) and 4.5. It is o/served that the average fitness value sta/ili.es after 7< generations. The converging plot gives a confirmation that the varia/les used during the simulation of model through 2$ has /een chosen correctly, and if these varia/les are made to change, the converging nature of the plot will not /e o/served.

$ plot showing the e0perimental as well as the predicted data points for hematite, magnetite, wustite and Iron is shown in Figure. 4.6 for the top layer simulation of '!'-I !TH"R#$% model at 4)<< oC. It can /e easily concluded from the following plots that during the initial generations in 2$, there is a wide margin of error in e0perimental and predicted values due to random selection of individuals from a larger set of population. ?ith each generation, the error is decreasing i.e. the fitness parameter is decreasing till the 2$ gives a converging set of parameters, which is visi/le after appro0imately 7< generations for this case.

.ig% 1%3 'chematic presentation sho)ing -etter fitness )ith $enerations% Kinetic Model /alidation
The kinetic model estimates the degree of reduction /ased on the estimated rate parameters which was derived at with the help of the e0perimentally measured degree of reduction. &ifferent results can /e o/tained if the input in the genetic algorithm is not selected properly. Hence it is necessary to validate the values o/tained from the model /y tallying it with the e0perimental data In the present study, the rate parameters of the relevant reactions have /een estimated /ased on e0perimental phases as determined /y the MR& analysis of the reduced pellets. It is found that predicted phase evolution matches with e0perimental value with S4.<T error. The predicted and e0perimental value of weight percent of different o0ide phases and iron for 4<mm si.e pellets are shown in Ta/le 7.5. Ta-le 4%3 5omparison -et)een predicted and e*perimental )eight 6 of different phases of the reduced pellet at 12 o5

4a"ers

5hase %6t 7)

5redicted #" model

Measured from 8R!

Fe)!5
T!& "27E1

44.*> =<.6* 46.*6 47.=) )7.+> 76.57 >.=5 4).4= 57.>7 65.+) >.=*+ 44.7)

4<.>=+ 7*.< 4=.)57 4=.< )5 7> *.4 47 55.< 7<.< >.<7++ 45

Fe5!6 Fe! Fe Fe)!5 Fe5!6 Fe! Fe Fe)!5 Fe5!6 Fe! Fe

MIDD"E "27E1

8!TT!M "27E1

1E'9"T' 2#D DI'59''I!#'

2%1 Estimated 1ate &arameters
$s discussed earlier, a kinetic model has /een developed to estimate the evolution of various iron o0ide phases, as well as metallic iron during the course of reduction . The reduction of iron o0ide (hematite- in an iron ore car/on composite system takes place in three consecutive steps. The amount of various iron o0ides and metallic iron present in the pellet changes with time. Calculating the evolution of phases is important to calculate instantaneous thermal properties, heat effects of transformations re1uired for thermal model for temperature calculation. 9esides, knowledge of instantaneous phases may focus light on the incipient fusion, sticking of pellets etc. The evolution of various phases can /e calculated only when the kinetics of each step is known. $ssuming first order kinetics, the estimation of rate parameters, namely, the activation energy values and pre e0ponential factors, for each step of transformation are re1uired to evaluate kinetics of all steps separately. The optimi.ed rate parameters o/tained for the three layers of pellets is provided in Ta/le ).4 Ta-le 2%1 Estimated rate parameters o-tained from the output of the $2 code

M!DE"

"ayer

20 : 1 ;s31< 9:,( :0,; ';, 9:,2 ';,'

17

2M : 1 ;s31< 9',( '0 8;,8 9',: '3,3

1=

2> : 1 ;s31< 93,33 9(,8 8;,: 98,( ':,3

11

E0 ;?@Amol< ' ' ' 9 '2 ' ( ' (

EM ;?@Amol< '02 (9; (9 '00 (9;

E> ;?@Amol< 88; 880 832 88' 83(

T" I'! M" 8" #!#3 I'! T" M"

8"

;0,

'0,3

(8,;

' 9

(9'

83,

It is obser/ed t#at t#e estimated acti/ation ener+" is least for top la"er for all transformation steps, $#e relati/el" lo2er of acti/ation ener+" for top la"er is because of less resistance for t#e #eat transfer, $#ere is a /er" sli+#t /ariation in terms of t#e rate parameter for bot# t#e models, $#is ma" be due to t#e fact t#at t#e time ta<en for pellet bed temperature to reac# t#e ambient temperature, in case of t#e =on Isot#ermal &odel is /er" less, around 3-( minutes, 5fter t#e temperature is reac#ed, it be#a/es in t#e same 2a" as t#e Isot#ermal model, >ence t#e parameters are nearl" same, $#e /ariation of temperature 2it# time durin+ t#e consideration of =3= IS3$>0R&5? &3D0? is s#o2n in 7i+ure, 2, ,

.ig% 2%1 Temperature /ariation of pellet -ed ;T"< in case of the non isothermal model The plot shows a rapid rise in temperature of the /ed from room temperature tp the furnace temperature in a very short time of 4<< sec (U)min-, thereafter /ecoming constant after which the actual reduction process starts. The optimi.ed rate parameters thus o/tained are then utili.ed to solve the four differential e1uations representing the three stages of reduction in order to predict the phase evolution with

time. The amount of different phases o/tained on solving the four simultaneous e1uations at each time step can /e the utili.ed to predict the theoretical degree of reduction at different time.

2%2 &redicted &hase E(olution

It was not possi/le to o/tain the evolution of different phases with time during the reduction. Thus a model is employed to mathematically achieve the phase evolution with time for reduction of the composite pellets. However, some researchers such as Vang et al. have measured the phase evolution with time for the reduction of composite pellets via in-situ MR& analysis and the results display the evolution similar in nature to that o/tained /y the mathematical model employed in this study.

.ig% 2%2 &redicted concentrations of (arious o*ide phases, as )ell pure iron, as a function of time during reduction at 12 B5of 1 mm pellet% .or ;2< 8ottom layer, ;8< Middle "ayer, ;5< Top "ayer respecti(ely for I'!T0E1M2" M!DE"

.ig% 2%3 &redicted concentrations of (arious o*ide phases, as )ell pure iron, as a function of time during reduction at 12 B5of 1 mm pellet% .or ;2< 8ottom layer, ;8< Middle "ayer, ;5< Top "ayer respecti(ely for #!#3I'!T0E1M2" M!DE"

.ig% 2%3 &redicted concentrations of (arious o*ide phases, as )ell pure iron, as a function of time during reduction at 13 B5of 1 mm pellet% .or ;2< 8ottom layer, ;8< Middle "ayer, ;5< Top "ayer respecti(ely for I'!T0E1M2" M!DE"

.ig% 2%3 &redicted concentrations of (arious o*ide phases, as )ell pure iron, as a function of time during reduction at 13 B5of 1 mm pellet% .or ;2< 8ottom layer, ;8< Middle "ayer, ;5< Top "ayer respecti(ely for #!#3I'!T0E1M2" M!DE"

$#e plots of p#ase e/olution at

200 oC considered for a time of 30 min, s#o2s a

decrease in #ematite concentration %from :,( <+ mol@m3 to 2,: <+ mol@m3 for top la"er), >o2e/er at 200 oC, complete consumption of #ematite does not occur, $#is is also e/ident from t#e ARD anal"sis 2#ic# s#o2s #ematite pea<s in t#e reduced sample, $#e ma+netite concentration increases, reac#es its pea< a little before 30 min, and b" t#e end, a ne+ati/e trend in slope is obser/ed, $#e rate of #ematite depletion is /er" #i+# in case of top la"er in comparison to t#at of middle and bottom la"er, $#ere is a stead" increase in 2ustite and iron 2it# time, $#is indicates t#at t#e rate of formation of 2ustite from ma+netite is more t#an compensated for subseBuent reduction of 2ustite to iron, Reduction of 2ustite reBuires lar+e reduction potential of +as %more t#an ;0C C3 in t#e C3, C32 +as mi1ture), 2#ic# starts at later sta+e 2#en si+nificant C3 is +enerated b" !oudouard reaction, However at 45<< oC, a complete reduction of pellets is o/served. The hematite concentration has reached .ero, /oth the magnetite and wustite first increased, reached its peak, and then decreased to .ero, there/y allowing complete formation of Fe. It is also seen that complete reduction of Fe is faster in the Top layer and the slowest for the 9ottom layer. The plot /elow shows the e0trapolated phase evolution showing the complete reduction of pellet and the overall time taken for 4<<T Iron to form.

2.( !egree of reduction 5rediction The Oredicted degree of reduction o/tained from the 2$ is plotted graphically in Figure ).= for Isothermal #odel and Figure ).> for 'on-Isothermal model.

.ig% 2%= &redicted degree of reduction at 12 I'!T0E1M2" model

B5 C 13

5 for different layers in the

.ig% 2%7 &redicted degree of reduction at 12 #!#3 I'!T0E1M2" model

B5 C 13

5 for different layers in the

From the a/ove plots, it can easily /e generali.ed that for the /ottom layer, degree of reduction the least, and it is ma0imum for the top layer. This is due to the fact that heat flu0 enters from the top, whereas the /ottom layer is insulated. ince the top layer is su/3ected to ma0imum heat flow, hence the time taken for complete reduction was also less than the other layers. $lso, car/on from the /ottom reacts with o0ygen, generating C!, which is itself a reductant, hence as it moves up, the percentage of reduction increases successively. The only anomaly was o/served during the 'on Isothermal model for 45<< oC, where very little difference was o/served /etween the percent reductions of consecutive layers. The reason may /e attri/uted to the fact that for a 4< mm diameter pellet, the height of the /ed is relatively less. $t 45<< oC, the rate of reduction is very high and 4<< T reduction can /e achieved within 47 mins or less, for all the three layers, as can /e 3ustified /y the comparative study given later in Ta/le ).). Hence, all the three curves in fig. ).> (/- are nearly coinciding.

.ig% 2%D 5omparison of the predicted degree of reduction -et)een the t)o models for top layer

$lso, another comparison can /e seen /etween the two #odels which have /een used in this study. In case of the non-Isothermal model, the rate of increase of T reduction is slower than in the Isothermal case. $s a result, 4<<T reduction can /e achieved in a relatively shorter time, if

the pellets are reduced in the RHF isothermally. From the graph, it can /e predicted that time for 4<<T reduction in case I is around 7 min, and for case II is 47 mins.

Ta#le 2.2 Time ta&en for (2, 92 and 122 7 reduction of different la"ers of the pellet #ed At 1222o$:
30C %min) Isothermal T" M" 8" T" M" 8" 2:,( 38,:: 3(,;3 2;,9; 3(,(( 3:,33 ;0C %#rs) ,03 ,3 ,32 ,09 ,33 ,82 00C %#rs) ,88 ,:( ,:; ,( ,:; 2,0

#on3 isothermal

At 1(22o$:
Isothermal T" M" 8" T" M" 8"

30C %min) ,8( 2 2,(3 3,:3 8,0; 8,

;0C %min) 3,(( 8,(; (,8 ;,0( ;,83 ;,83

00C %min) (,0: ',33 ;, ; (, ; (,;( ',2(

#on3 isothermal

The Ta/le ).) shows the time taken for the reduction to complete its various stages. It can /e easily deduced that time taken for reduction is higher, when the conditions of operation of RHF are assumed to /e 'on-Isothermal. The assumption of a prevailing average temperature that was taken into consideration while modeling may /e the reason for the anomaly, and it may /e prefera/le to consider a temperature gradient along the /ed from top to /ottom, for the accurate understanding of the kinetics of reduction.