metal100030 (2)

Revue de Mtallurgie 107, 195204 (2010)c EDP SciencesDOI: 10.1051/metal/2010022www.revue-metallurgie.org

Revue deMtallurgie

Text received 02 March 2010accepted 24 March 2010

A simplified approach to the simulationof direct reduction of iron ore

M. Vannucci1, V. Colla1, G. Corbo2 and S. Fera2

1 PERCRO-CEIICP, Scuola Superiore S. Anna, 56127 Pisa, Italy2 ILVA S.p.A., Genova Works, 16512 Genova, Italy

Abstract This paper describes a model for the simulation of the process of direct re-duction of iron ore for steel production. The model is implemented through stand-alonesoftware and the simulation results have been compared with real experimental data. Thevery good agreement between the actual and simulated data proves that the model, de-spite its relative simplicity, takes into account all the fundamental phenomena of iron orereduction.

Rsum Une approche simplifie la simulation de la rduction directe de mi-nerai de fer. Cet article dcrit un modle pour la simulation du processus de rductiondirecte de minerai de fer pour la production dacier. Le modle a t mis en application parlintermdiaire dun logiciel indpendant et les rsultats de la simulation ont t compars de vritables donnes exprimentales. Lexcellente concordance entre les donnes relleset simules montre que le modle, en dpit de sa relative simplicit, prend en considrationtoutes les variables fondamentales de la rduction du minerai de fer.

I n recent years, the eorts of the steelmaking industry toward a reductionof CO2 emissions have become more and more intensive. Continuoustechnical improvements of existing processes have contributed to low-ering such emissions, but drastic reductions can be achieved only throughthe development of breakthrough technologies. For this reason, an impor-tant project has been developed since 2004 by a large consortium includingmost of the European steel producers and many research institutions, withinthe 6th Framework Program, which is entitled Ultra-LowCO2 Steelmaking(ULCOS) [1, 2] and aims at investigating technologies capable of cutting theCO2 emission of the steelmaking industry by an amount in the order of 50%.Among the technologies that have been studied within ULCOS and are nowin an advanced experimental phase, Direct Reduction (DR) is one of the mostpromising ones [3]. The emissions of DR-EAFplants are in fact already close tothe 50% target reduction expected by the ULCOS project and DR is designedto be suitable for integration with other technologies, such as CCS, aiming forthe reduction of CO2 emissions, which could further improve the appeal ofthis technology. DR is going to be a valid alternative to traditional routes froman economic point of view as well, as in the future the price of NG is expectedto become highly competitive.

DR is an alternative route for the production of steel which was developedin the late 70s and it is already applied, in countries where there is abundanceof natural gas, in a route including a pre-reduction furnace and an EAF formelting [4, 5]. The DR process allows the production of Direct Reduced Iron(DRI) by means of a mixture of reducing gases mainly composed of hydrogenand carbon monoxide, which play the role of reducing agents.

The main part of the DR plant is the reduction shaft where the reductionreactions take place. During the production the shaft is charged from the topwith iron ore and the reducing gases are blown from the bottom in order toallow the reduction, while the produced DRI is collected from the bottom ofthe shaft.

Literature results can be found related to the modeling of chemicaland physical transformations involved in the DR processes and, in particu-lar, on the reduction kinetics [6,7]. One of themost ecient DR processes is the

Article published by EDP Sciences

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

MIDREX r process, which exploits bothlumps and pellets as raw material andrecycles the used gas, thus showing lowconsumption and reduced environmentalimpact. The shaft furnace reactor of theMIDREX r process has been simulated in [8].

In this paper a model is proposed, whichhas been developed within the ULCOSproject in order to simulate in a fast but reli-able way the reduction process which takesplace in the reduction shaft of a DRI produc-tion plant.

The driver for developing this modelhas been the need for a simple-to-use de-sign tool that could manage the new reduc-ing gas composition envisaged for withinULCOS and that could also quickly com-pare dierent burden materials. In order toachieve this goal, non-conventional reduc-tion testing procedures were developed anda software package was implemented en-compassing both test result interpretationand DR reactor modeling.

In particular, given the geometry of theshaft, and the composition, temperature andpressure of the inflated gas, the model esti-mates the hourly DRI production that canbe obtained by using a specified burdenma-terial. The reduction kinetics of the burdenmaterial is fundamental information for themodel and is characterized by means of an-other model named Ilva REduction Simula-tion (IRES), which is based on a series of lab-oratory tests carried out on the material.

The DR shaft model has been imple-mented as a one-dimensional Finite ElementModel (FEM), where the shaft is representedby a cylinder formed by 50 overlapping el-ements, where the conditions aecting thematerial reduction vary only along the ver-tical dimension.

The developedmodel of theDR shaft hasalso been implemented in a stand-alone ap-plication named SAILORS (Sant Anna ILvaOre Reduction Simulator), which is realizedin Visual Basic and combines a user-friendlyinterface and high computation capabilities.

The DR shaft model was tested in or-der to evaluate the accuracy of the producedsimulations. In order to validate the devel-oped model, some real data from the mostcommon DRI production plants were com-pared with the results of the correspondingsimulations performed through the devel-oped model. In particular, the tests refer to aset of dierent configurations both for thedimension of the shaft and for other pa-rameters such as inlet gas composition and

temperature,while the burdenmaterial usedis for all tests the same kind of commercialpellets. The variability in the parameters al-lowedus to test the simulator in awide rangeof conditions. The results of the comparisonshow the very good agreement between realand calculated DRI production and demon-strate the goodness of the developed model.

The paper is organized as follows: Sec-tion 2 describes the IRES model, Section 3 isdevoted to the description of the DR shaftmodel, Section 4 depicts the software thathas been developed in order to make themodels easy to use in an experimental con-text and, finally, Section 5 presents some nu-merical results, by comparing themodel pre-dictions with some experimental data fromreal industrial plants.

Characterization of the reductionbehavior of the burden material

The DR shaft model has been implementedas a one-dimensional FEM, which is basedon a subdivision of the shaft in 50 overlap-ping elements of the same diameter of theshaft. The basic assumption is made, that ineach element the conditions aecting thema-terial reduction canvary along the axes of thecylinder and not along its radius. This hy-pothesis, although quite schematic, allowsconsiderable simplification of the processrepresentation and, as a consequence, of therelated computations.

The reduction of burden material whichtakes place within each layer is simulated byusing the IRES model. IRES is a model forsimulating the reduction behavior of an ar-bitrary material depending on temperatureand gas composition, provided that a set ofkinetic characteristics related to the materialhave been determined through laboratorytests.

The general kinetic laws obtained fromtests performed at constant conditions oftemperature and gas composition were sub-sequently validated by means of othertests performed at variable temperature andgas composition conditions. For the char-acterization of a material various isother-mal reducibility tests were run at dier-ent temperatures and dierent reducing gascompositions. During the tests, the weightloss was continuously registered and subse-quently converted into a reduction percent-age (indicated by R in the following). Thevariability ranges of conditions aecting re-duction aswell as someparameters of the lab

196

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

900C 700C

Fig. 1. Sample results of the reduction tests. These tests were carried out at constant temperature and gas composition.Fig. 1. Rsultats dchantillon des tests de rduction. Ces essais ont t eectus la composition en temprature et en gaz constante.

tests are listed in Table 1, while some sampletest results are shown in Figure 1.

As depicted in Table 1, the tests refer totypical DRI conditions [9]. The pellets usedfor the tests were provided by one of the in-dustrial partners of the ULCOS project andthey belong to the common types of pelletsused in DRI plants [10]. On the basis of theresults of the experiments, the kinetic lawsof the reduction reaction were assessed. Thefollowing general kinetic law was consid-ered [6]:

R (t) = 100(1 eKt

)(1)

where t is the time and K is the kinetic factordepending on the material type and size, thereducing environment and the temperature.The process of correlating the experimentalresults with the main test conditions (suchas temperature and gas composition) under-went various stages. As a conclusion it wasfound that:

the reduction behavior of all materialstested can be adequately described byequation (1);

Table 1. Variability in the test conditions for the main parameters.Tableau 1. Variabilit des conditions dessai pour les paramtres principaux.

Parameter Min Max

Temperature 700C 900CH2 45% 85%CO 7% 48%CO2 2.5% 4.5%Gas speed (STP) 0.057 m/sGas flow (STP) 9.21 L/minReduction section diameter 59 mmInitial sample weight 307 gPellet size 12.5 mm 15 mm

the gas composition can be representedby introducing the concept of equivalent% CO content: 1 mole of H2 = 2 moles ofCO (so that, for instance, 50% CO + 10%H2 behaves as 70% CO);

The parameter K can be linearly corre-lated to the gas composition and temper-ature, i.e. K = aX+ b, where: a and b are

197

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

Fig. 2. Comparison between a laboratory test and its IRES predicted behavior.Fig. 2. Comparaison entre un essai en laboratoire et son comportement prvu par IRES.

constants to be determined by fitting theexperimental results and X is a parame-ter defined as:

X = 100( T1000

)2.5 [CO]eq1000

(2)

where T is the absolute temperature and[CO]eq is the percentage of equivalent CO.

Usually the root mean square error be-tween actual test data and those obtainedfrom the above correlation lies within a fewpoints in percentage.

Figure 2 depicts the comparison betweena simulation of material reduction and itsmeasured behavior, showing the good per-formance of the model for isothermal andconstant gas composition tests. The simu-lation shown refers to an extreme situationwithin given test ranges and highlights thegoodness of the proposedfitting also for bor-derline conditions.

The kinetic parameters determinedthrough the above-described tests are storedin a database which describes the reduc-tion behavior of several materials in dier-ent conditions. Such parameters are used tosimulate the reduction of the tested materi-als as well as mixtures of them for varyingtemperature and gas composition in orderto be in line with the conditions of the DRIproduction shaft.

Varying conditions are simulated byconsidering any temperature-gas profile asa sequence of steps whose duration is

suciently short to be assumed as isother-mal and with constant gas composition.Moving from a step to the following onerequires the concept of virtual time, i.e. thetime required to reach under a hypotheticalhistory with the temperature and gas com-position of the next step, the same reduc-tion index already reached during the realhistory. The extra reduction gained in thesubsequent step is equivalent to that whichwould be obtained in a step of equal dura-tion beginning at virtual time.

The concept is described in Figure 3,which refers to a 2-step history: the first oneconsisting of a 100-minute reduction timewith 20% CO followed by 100 minutes with50% CO.

The DR shaft model

IRES is used within SAILORS as a reduc-tion modulewithin a more general model de-signed to simulate the shaft of a DR furnace.In particular, IRES is used to simulate the re-duction of burden material in each elementof the 1-dimensional FEM.

The following main assumptions weremade for the development of the model:

In order to exploit FEM simulation theshaft is divided into 50 cylindrical over-lapping elements. In each element condi-tions are assumed to vary only along thevertical axis of the cylinder and not in the

198

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

Fig. 3. Illustration of the virtual time concept.Fig. 3. Illustration du concept de temps virtuel .

radial direction. Elements are numbered1 to 50 starting from the bottom.

Reducing gas and burdenmaterial movein opposite directions: gas enters fromthe bottom of the shaft while material isput in from the top.

Reducing gas is formed by: H2, H2O, COand CO2.

A scheme of the described shaft includinginput and output of gases and materials isshown in Figure 4.

The main inputs needed by the model,which correspond to the main input vari-ables of the shaft, are the following:

Dimension of the reduction shaft (heightand diameter).

Flow rate, composition, temperature andpressure of inlet gas.

Inlet burden material temperature. Target reduction of outlet material.

On the other hand, the model provides thefollowing outputs:

Produced DRI outlet flow. Temperature profile for both gas andma-terial in the shaft.

Reduction profile of material in the shaft.

Within the FEM, representing a single layerof the shaft, several physical processes andrelations among variables are considered. Inthis framework, all reactions are heavily af-fected by the composition of the reducinggas. From the thermal point of view, there isheat exchangebetweengas andmaterial (gascools as it reaches the upper part of the shaft,

whilematerial getswarmer as it descends to-ward the bottom of the shaft itself). From thechemical point of view, gas reduces the min-eral by subtracting oxygen which combineswith H2 and CO and forms CO2 and H2O.Such reactions lead to an energy exchangewhich must be taken into consideration (forinstance, the reaction with H2 absorbs en-ergy while the reaction with CO releases en-ergy).

The functioning of the 1-dimensionalFEMmanaging these thermal and chemical-physical interactions can be described ac-cording to the following steps:

1. A first simulation of reduction is carriedout, by ignoring all thermal exchangesbut assigning an arbitrary initial temper-ature value to the burden material. Thesimulation, through an iterative processwhich is based on the exploitation of theIRES model, calculates an inlet materialflow which is compatible with the targetreduction.

2. Given the inlet material flow calculatedin the previous step, the thermal balanceis calculated in order to obtain the tem-perature of gas andmaterial in each layerof the shaft. To this aim, both the thermalbalance within each layer and boundaryconditions must be taken into account.In particular, two boundary conditionshave to be respected: the first one for theupper border concerning inlet materialwhose temperature is known; the otherone on the lower border concerning inletgas temperature,which is known aswell.

199

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

Burdenmaterial

Outletgases

Reducedmaterial

Reducinggas

Element #1Element #2

Element #50

Fig. 4. The DR shaft FEM representation exploited by the SAILORS model.Fig. 4. La reprsentation du modle dlment fini de laxe exploite par le modle.

3. Given the thermal profiles calculated instep 2, a new inlet material flow is cal-culated as in step 1 but, in this case,new thermal conditions are exploited.The computation terminates when thecalculated inletmaterial flowvalue is sta-bilized according to a set of convergencecriteria established by the user.

Figure 5 depicts the flow chart of the above-described calculation.

Within the simulation IRES is used forthe inlet material flow calculation. Materialflow calculation starts from element No.1and goes on element by element from thebottom to the top of the shaft. In this sit-uation and at each step of the computation,thematerial temperature is known (from thecurrent thermal profile) as well as the gascomposition in the shaft element, the reduc-tion degree on the bottom of the elementand the residence time of mineral in the el-ement. IRES is used each time to calculatethe reduction degree on the top border of theconsidered element,which represents the re-duction degree of the mineral in a previousmoment as the mineral goes from the top

to the bottom of the shaft. To this purpose,IRES calculates the reduction profile of thematerial until the reduction of the mineralat the exit of the element is reached underthe conditions of temperature and gas com-position of the considered element, then itlooks backward to a time t which representsthe residence time ofmaterial in the element.The obtained result is the reduction degreeof thematerial once it enters the element. Thereduction degree of the 50th element is usedin order to verify the mass balance for thecurrent situation: if there is convergence thecalculation can stop, otherwise parametersare modified until convergence is reached.

The developed software

On the basis of the above-described 1DFEM-based description of the DR shaft, softwarefor the simulation was implemented. The re-alization of a stand-alone application wasjudged necessary for two main reasons: thefirst one is related to the complexity ofmathematics involved in the calculus andto the heavy computational burden. Such

200

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

Fig. 5. Flow chart describing the computational structure of SAILORS.Fig. 5. Organigramme dcrivant la structure informatique du modle SAILORS.

computational eort needs to be aorded bydedicated software expressly designed andcompiled in order tominimize the computa-tional time. The second reason is related tothe need for having user-friendly softwarewhich makes the selection and modificationof the numerous parameters involved in thesimulation easy and provides a clear visual-ization of results.

For the implementation of the softwareMicrosoft c Visual Basic was chosen becauseit gives at the same time the possibility of:

1. writing code to be compiled by thestandard VB compiler. The code com-pilation gives the advantage of creatingportable stand-alone software compati-ble with MS Windows c-based systems.Moreover, the compilation drastically in-creases the speed of the whole computa-tion.

2. Designing in a very natural way a user-friendly interface similar to most soft-ware running under Microsoft operativesystems.

201

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

Fig. 6. The main window of the software interface.Fig. 6. La fentre principale de linterface du logiciel.

3. Easy interfacing with other programsused for the loading and storage of dataused by the model.

The software was designed according toa modular structure, in order to facili-tate subsequent modifications or code addi-tions: there are, for instance, procedures forthe mass balance, for the thermal balanceand some for the management of the FEMengine.

The main interface window of the de-veloped software is shown in Figure 6 andincludes an area for the input of plant andmaterial parameters inwhich themainprop-erties of the shaft and the burden materialare chosen and an output area for the moni-toring of main information concerning eachof the 50 elements into which the shaft isdivided and the main results of the DRIsimulation. Moreover, a picture showing theprogress of reduction and temperatureof thematerial and reducing gas inside the shaft isincluded.

The complete list of the inputs to themodel inserted through the interface and

taken into account by the model is the fol-lowing: material bulk density; guess value on burdenmaterial flow rate(needed for the FEM algorithm);

gas flow rate; inlet gas pressure; outlet gas pressure; burden material inlet temperature; reduction chamber height; reduction chamber diameter; target discharge metallization; average burden material size; inlet gas composition; burden material type.In addition to these inputs, when a burdenmaterial is chosen for the simulation, thesoftware exploits some information previ-ously stored in a separate (independent andeditable) Excel file. This file contains infor-mation used by the IRES model embeddedin SAILORS and describes the kinetics of thereduction of the specific material. The list ofburden materials can be extended by sim-ply adding information to the Excel file and

202

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

in an analogous manner material propertiescan be modified.

SAILORS allows the user to save andsubsequently recall an input set in order toform a library of principal input situations.

The software, after the calculation is com-plete, returns the following information onthe main window:

material inlet flow; material outlet flow; outlet gas composition; for each element of the shaft: gas tem-perature, burden temperature, reductionrate, residence time (in a table).

When the calculation is completed, the pic-ture area is updated and it is possible to showboth a graphic describing the progress of thereduction rate in function of time or (by se-lecting the specific radio-button) the burdenandgas temperature in functionof the heightof the shaft. It is also possible to export thesefigures in the JPG format.Moredetailed dataare available by accessing a further window(see Fig. 7) which provides the following in-formation:

total production rate of reducedmaterial; metallization degree; total iron; total metallic iron; residence time; oxides; total oxygen transferred.

All the above-described output data can beexported in a plain ASCII file which is eas-ily readable by most common text editors.The exported file also includes informationconcerning the input parameters of the sim-ulation.

The computation time is in line with theexpectations. A single run of the simula-tion depends on the whole set of previouslymentioned parameters. Normally, simula-tions on a 2.4 GHz processor computer with2GB ram take from 5 to 20 seconds.

Numerical results

In order to validate the developed model,some real data from the most common DRIproduction plants were compared with thecorresponding simulations performed bySAILORS. Such information is confidentialand comes from the work carried out withinthe ULCOS project. In particular, tests referto a set of dierent configurations both for

Fig. 7. The window providing detailed information onthe result of the calculation.Fig. 7. La fentre qui fournit des informations dtailles sur lersultat du calcul.

Fig. 8. Comparison between calculated and actual hourlyproduction rate of DRI for the performed tests.Fig. 8. Comparaison entre la cadence de fabrication horaire cal-cule et relle du DRI pour les essais raliss.

the dimension of the shaft and for other pa-rameters such as inlet gas composition andtemperature,while the burdenmaterial usedis for all tests the same kind of commer-cial pellets. The variability in such param-eters, whose ranges are shown in Table 2, is

203

M. Vannucci et al.: Revue de Mtallurgie 107, 195204 (2010)

Table 2. Variability ranges for the main param-eters for the reduction tests exploited for themodel validation.Tableau 2. Gammes de la variabilit pour des para-mtres principaux pour les essais de rduction exploi-ts pour la validation modle.

Parameter Min MaxShaft diameter 5 m 7 mShaft height 9 m 11.6 mGas temp. (C) 900C 1078CInlet H2 38% 70%Inlet CO 15% 36%Inlet CH4 2% 10%

important as it has allowed testing the sim-ulator in a wide range of conditions.

The real plant hourly DRI productionrates are compared with the ones calculatedby the SAILORsimulation. The results of thiscomparison, which are depicted in Figure 8,show the very good agreement between realand calculated DRI production and demon-strate the goodness of the developed model.

Conclusions

A model of a DR shaft has been developedwithin the ULCOS project, in order to sup-port the development of the new DR plant,which is one of themost promising routes forreducing theCO2 emissions involved in steelproduction. The mono-dimensional FEM ofthe shaft was exploited coupled with a re-duction module in order to simulate the re-duction of burden material in each elementof the FEM. Stand-alone software was alsoimplemented in order to make the modeleasy to use and to speed up the computa-tion.

The model was validated by exploitingreal data from the most common DRI pro-duction plants, which were compared withthe results of the model. The tests refer toa set of dierent configurations which arerepresentative of a wide range of possiblecases. The good agreement between the re-sults provided by the model and the experi-mental data encourage the use of the devel-oped software in the design phase of newconcept DR plants.

Acknowledgements

The work described in the present paper wasdevelopedwithin the project entitled Ultra-Low

CO2 Steelmaking (ULCOS), which has receivedfunding from the European Community withinthe 6th Framework Program. The sole responsi-bility for the issues treated in the present paperlies with the authors; the Commission is not re-sponsible for any use that may be made of theinformation contained therein.

The authors also wish to thank Dr. E Knopfor having provided the experimental data thathave been used to validate the model and Dr. E.Burstrm for the fruitful discussions which con-tributed to the development of the described re-search work.

References

[1] K. Meijer, M. Denys, J. Lasar, J.P. Birat, G.Still, B. Overmaat, ULCOS: ultra-low CO2steelmaking, Ironmaking & Steelmaking 36(2009) 249-251

[2] J.P. Birat, F. Hanrot, ULCOS: the EuropeanSteel Industrys Eort to Find BreakthroughTechnologies to cut CO2 EmissionsSignificantly, EU/Asia Workshop onClean Production and Nano Technologies,Seoul, Korea, Oct. 2006

[3] K. Knop,M.Hallin, E. Burstrm,ULCOREDSP 12 Concept for minimized CO2 emission,Rev. Mtall. 10 (2009) 419-421

[4] M. Small, Direct Reduction of Iron Ore, J.Met. 3 (1981) 67-75

[5] J. Feinman, D.R. Mac Rae, Direct ReducedIron, The AIST Foundation, Pittsburgh, PA,1999

[6] K. Mondal et al., Reduction of iron ox-ide in carbon monoxide atmosphere re-action controlled kinetics, Fuel ProcessingTechnology 86 (2004) 33-47

[7] K. Piotrowsky et al., Eect of gas compo-sition on the kinetics of iron oxide reduc-tion in a hydrogen production process, Int.J. Hydrogen Energy, 2004

[8] D.R. Parisi, M.A. Laborde, Modeling ofcounter current moving bed gas-solid reac-tor used indirect reductionof ironore,Chem.Eng. J. 104 (2004) 35-43

[9] F.N. Griscom, G.E. Metius, J.T. Kopfle,Ironmaking technology for the new mil-lennium, SEAISI Conference 2000, Perth,Australia, May 2000

[10] N. Eklund, A. Dahlstedt, The choice of pel-lets in amixedblast furnace burdenandhowit aects process conditions, 14th conferenceon Hungarian pig iron and steel making,Sept. 2002

204

A simplified approach to the simulation of direct reduction of iron oreCharacterization of the reduction behavior of the burden materialThe DR shaft modelThe developed softwareNumerical resultsConclusionsAcknowledgementsReferences