Effects of Clogging, Argon Injection and Casting Conditions on ...ccc.mines.edu/s/Publications/00_ISS 99.mymac.ccc.pdfaspiration is less. The optimal argon flow rate depends on the

83rd Steelmaking Conference Proceedings, (Pittsburgh, PA, March 26-29, 2000), Vol. 83, Iron and Steel Society, Warrendale, PA, 2000. pp. 183-197.

2

Effects of Clogging, Argon Injection and Casting Conditions on Flow Rate and Air Aspiration in Submerged Entry Nozzles

Hua Bai and Brian G. Thomas

Depart. of Mechanical and Industrial Engineering University of Illinois at Urbana-Champaign,

1206 West Green Street, Urbana, IL USA, 61801 Ph: 217-333-6919; Fax: 217-244-6534;

Key Words: clogging, nozzle, multiphase flow, air aspiration, interpolation, numerical models

ABSTRACT

The inter-related effects of nozzle clogging, argon injection, tundish bath depth, slide gate opening position and nozzle bore diameter on the flow rate and pressure in tundish nozzles are quantified using three-dimensional multiphase turbulent numerical fluid flow models. The results are validated with measurements on operating steel continuous slab-casting machines, and presented for practical conditions with the aid of an inverse model. Initial clogging is predicted to enhance the flow due to a potential streamlining effect before it becomes great enough to restrict the flow channel. The clogging condition can be detected by comparing the measured steel flow rate to the theoretical value predicted under the same conditions by the inverse model presented here. Increasing argon injection may help to reduce air aspiration by increasing the minimum pressure below the slide gate. More argon is needed to do this at intermediate casting speeds and in deeper tundishes. Argon flow should be reduced during shallow tundish and low casting speed conditions

(such as encountered during a transition) in order to avoid detrimental effects on flow pattern. It should also be reduced at high casting speed, when the slide gate is open wider and the potential for air aspiration is less. The optimal argon flow rate depends on the casting speed, tundish level, and nozzle bore diameter and is quantified in this work for a typical nozzle and range of conditions.

INTRODUCTION

Nozzle clogging is one of the most disruptive phenomena in modern continuous casting operation. Argon injection into the nozzle is widely employed to reduce nozzle clogging. Both clogging and argon injection greatly affect flow through the nozzle, altering both the flow rate and flow symmetry. In many slab casting operations, the slide-gate opening is adjusted to compensate for these changes and maintain a constant flow rate. However, clogging, argon injection, and slide position all greatly affect the flow pattern in the mold and related quality issues, even for a fixed flow rate. Thus, there is incentive to understand quantitatively how these parameters are all related.

Air aspiration through cracks and joints into

the nozzle leads to reoxidation, which is an important cause of inclusions and clogging [1, 2]. Air aspiration is more likely if the pressure inside the nozzle drops below atmospheric pressure, creating a partial vacuum. While regulating the liquid steel flow, the slide-gate creates a local flow restriction which generates a large pressure drop. This creates a low-pressure region right below the throttling plate, which often falls below 1 atm (0 gauge pressure). The minimum pressure is affected by argon injection, tundish bath depth, casting speed, gate opening and clogging. Predicting when a partial vacuum condition exists and choosing conditions to avoid it is one way to prevent this potential source of reoxidation products and the associated clogging and quality problems.

3

In this paper, a mathematical model isdeveloped to relate argon injection, tundish bathdepth, casting speed, and gate opening for practicalslab casting conditions. The influence of nozzleclogging and nozzle bore size are also investigated.This model is derived from interpolation of thenumerical simulation results of a three-dimensionalmodel of liquid steel-argon bubble two-phaseturbulent flow in tundish nozzles. Modelpredictions are compared with plant measurements.The model is then extended to predict the minimumpressure in the nozzle as a function of the samecasting conditions. Finally, the model is applied toinvestigate operating conditions to avoid partialvacuum pressures, such as the optimal flow rate ofargon gas.

MODEL DEVELOPMENT

A model to investigate the interrelatedeffects of casting variables on the minimumpressure in the nozzle is developed in five stages.First, a 3-D finite-difference model developed inprevious work is used to perform a parametricstudy. Then, the output pressure drops areconverted to tundish bath depths and the results arecurve fit with simple equations. Next, theseequations are inverted to make the tundish bathdepth an independent variable and to allowpresentation of the results for arbitrary practicalconditions. Finally, the predicted minimumpressure results are combined with the inversemodel, so that they also can be presented forpractical casting conditions.

3-D Finite Difference Model

A three-dimensional finite-difference modelhas been developed to study steady two-phase flow(liquid steel with argon bubbles) in slide-gatetundish nozzles using the K-ε turbulence model. Inthis model, the chosen slide gate opening isincorporated into the computational domain duringmesh generation. The casting speed and argoninjection flow rate are fixed as inlet boundaryconditions at the top of the nozzle and the gasinjection region of the UTN respectively. This was

done because fixing the pressure drop causesnumerical stability problems. For each 3-Dsimulation, the numerical model calculates the gasand liquid velocities, the gas fraction, and thepressure everywhere in the nozzle. The modelequations are solved with the CFX4.2 codedeveloped by AEA Technology [3]. Further detailsof the model are described elsewhere [4].

The accuracy of flow predictions near theport outlet has been verified both qualitatively bycomparison with experimental observations andquantitatively by comparison with velocitymeasurements using Particle Image Velocimetry [5].In this work, the model is employed to simulate thetypical slide-gate nozzle shown in Figure 1, and toperform an extensive parametric study of variousvariables, including casting speed, gate opening,argon injection flow rate, and bore diameter.

Liquid Inlet from tundishnormal liquid velocity = constantK=constant�ε =constant�Liquid volume fraction =1��

Gas Injectionnormal gas velocity = constant�Argon volume fraction =1�

Outlets (both ports)pressure = constant�zero normal gradients �for velocities, K and ε��

��

Tundish Well (Nozzle Top)

UTN(Upper Tundish Nozzle)

Slide-Gate Opening

SEN(Submerged Entry Nozzle)

Nozzle Ports

Shrould Holder

Figure 1. Outline of the standard nozzle andboundary conditions for parametric studies

Parametric Study with 3-D model

Over 90 simulations were performed in theparametric study with the 3-D finite differencemodel. All of the runs were based on the standardnozzle in Figure 1 with the standard geometry andoperating conditions given in Table I. This nozzle istypical of a conventional slab casting operation. It

Table I Nozzle dimension and conditions

4

Dimension &Condition

StandardNozzle

ValidationNozzle A

ValidationNozzle B

Casting speed(m/min, 8”x52”slab)

1.21

Tundish depth (mm) 1125 927Argon injectionflow rate (SLPM)

7~10 14

Argon bubble diam.(mm)

1. 1. 1.

UTN top diameter(mm)

114 115 100

UTN length (mm) 241.5 260 310Gate thickness(mm) 63 45 45Gate diameter(mm) 78 75 70Gate orientation 90° 90° 90°Gate opening(FL) 52%Shroud holderthickness (mm)

100 100 66

SEN length (mm) 748 703 776SEN bore diameter(mm)

78 91~96 80

SEN submergeddepth (mm)

200 120~220

165

Port width Xheight(mmXmm)

78X78 75X75 78X78

Port thickness(mm) 30 30 28.5Port angle (down) 15° 35° 15°Recessed bottomwell depth (mm)

12 12 12

Table II Simulation Conditions for Standard Nozzle

Variables Value NotesCasting Speed

VC (m/min)0.2, 0.5, 1,1.5, 2.0, 2.3

For 8”x52”slab

Gate OpeningFL (%)

40, 50, 60, 70,100

Linearopening

Argon Flow Rate QG

(SPLM)0, 5, 10 “cold” argon

Nozzle BoreDiameter DB (mm)

60, 70, 78, 90 Also simulatesclogging

has a 90° orientation slide-gate, in which the slide-gate moves in a direction perpendicular to the wideface of the mold. Thus, the right and left sides ofthe mold are nominally symmetrical. Thisorientation has the least bias flow between the twoports, so is widely adopted in practice. The effect of

different orientations of the slide-gate has beenstudied elsewhere [6].

The simulation conditions for the parametricstudy are listed in Table II. Casting speed VC refersto a typical size of the continuous-cast steel slab(8”x52”) and can be easily converted into liquidsteel flow rate through the nozzle or to castingspeed for a different sized slab.

Slide gate opening fraction FL is a linearfraction of the opening distance, defined as the ratioof the displacement of the throttling plate (relativeto the just-fully closed position) to the borediameter of the SEN. This measure can beconverted to many other definitions of gate opening,such as the displacement relative to a differentreference position. The most relevant way todenote gate opening is via the area fraction, FA,found by:

F F F FA L L L= −( )− −( ) − −( )−2 1 2 1 1 11 2

π πcos (1)

Argon is injected into the upper tundish nozzle(UTN) at the “cold” flow rate QG measured atstandard conditions (STP of 25˚C and 1 atmospherepressure). Calculations show that argon gas injectedthrough the “hot” nozzle wall heats up to 99% ofthe molten steel temperature even before it hits theliquid steel [6]. Thus, the argon flow rate used in thenumerical model is the corresponding “hot” argonflow rate. This is simply the product of the QG andthe coefficient of gas volume expansion due to thetemperature and pressure change [6], which is about5 [7]. The most relevant measure of gas flow rate isthe hot percentage. This measure is defined as theratio of the hot argon to steel volumetric flow ratesand changes with casting speed and strand width.

Nozzle bore diameter D B refers to thediameter of the circular opening in the slide-gate,which is assumed to be the same as the innerdiameter of the SEN and bottom of the UTN.Decreasing DB also approximates the effect ofsevere clogging when alumina builds up uniformlyin the radial direction. Four different nozzlediameters are simulated in this work, shown inTable II. In order to isolate the effect of DB and

5

better approximate the uniform clogging buildup,all nozzles keep the same axial dimensions as thestandard nozzle. The ports are proportionally scaled,however, to keep the same square shape for all boresizes.

0 20 40 60 80 100

0

200

400

600

800

1000

FL=40%, HT=1.545m

FL=50%, HT=0.904m

FL=60%, HT=0.653m

FL=70%, HT=0.583m

FL=100%, HT=0.513m

(b) Pressure (KPa)

Dis

tanc

e fr

om th

e po

int O

at t

he c

ente

r lin

e (m

m)

Argon injection QG=10 SLPMCasting speed VC=1m/minNozzle bore diameter DB=78mm

Gate opening FL Tundish level HT

FL=40%

50%60%FL=100%

FL=70%

z

x

o

HighPressure

LowPressure

(a)

z

xo

Figure 2. Pressure distribution in the standardnozzle, predicted by the 3-D FD model (a) Shadedcontour plot at the center-plane (b) Pressure profilealong the centerline (from top to outlet port)

Figure 2(a) shows a typical shaded contourplot of the pressure distribution in the standardnozzle from the 3-D finite-difference modelsimulation. Figure 2(b) shows the pressure profilealong the nozzle, for a few cases with different gateopenings. The path follows the nozzle centerlinefrom the nozzle top to point O at the center of theport section and then along the line from point O tothe port outlet. It can be seen that the biggestpressure drop occurs across the slide gate, due to thethrottling effect. The lowest pressure is found wherethe slide gate joins the SEN, so joint sealing is veryimportant there to avoid air aspiration if a vacuum

occurs. Increasing gate opening results in smallerflow resistance and thus less pressure drop. Thepressure drop is also affected by other factors suchas flow rate, rounding of the refractory surfaces andclogging. Pressure at the outlet ports is independentof these variables and depends mainly on SENsubmerged depth.

Multivariable Curve Fitting

In order to interpolate the results of theparametric study over a continuous range ofoperating conditions, equations were sought tocurve-fit the data points from the parametric studieswith the 3-D model described in the previoussection. Flow through the nozzle is driven bygravity so the pressure drop calculated across thenozzle corresponds to the pressure head given bythe tundish bath depth, HT. The relationshipderived from Bernoulli’s equation, is

Hp gH U U

gTl SEN l B C

l

=+ + −∆ ρ ρ

ρ

12

2 2( ) (2)

where ∆p is the overall pressure-drop across thenozzle which can be directly output from CFXsimulation, HSEN is the SEN submerged depth, UB

is the average velocity at the top inlet of the nozzleandUC is the average jet velocity at the nozzle port,which is a weighted average of the liquid flowexiting the port [4].

The calculated tundish bath depths (HT) areplotted as a function of the other process variables,in Figures 3(a-d). Each point in these plotsrepresents one simulation case. Equations to relatetundish bath depths (HT) with those variables wereobtained by fitting the points in Figures 3(a-d) usinga multiple-variable curve fitting procedure, which isnow briefly described.

First, the form of the equation is chosen foreach variable. Figure 3(a), shows that the HT vs. VC

data fits well with a quadratic polynomial function.The HT vs. QG data shown in Figure 3(b) fits well

6

0

0.5

1

1.5

2

0 0.5 1 1.5 2 2.5

Tu

nd

ish

bat

h d

epth

HT (

m)

Casting speed VC

(m/min, for 8"x52" slab)

Argon injection QG=10 SLPM

FL=10%

20%

30%

40%50%

60% 70% 80%

90%

100%

Gate opening FL

Nozzle bore diameter DB=78mm

0

0.5

1

1.5

2

2.5

3

0 5 10 15

Tu

nd

ish

bat

h d

epth

HT (

m)

Argon injection flow rate: QG(SLPM)

Gate opening FL=50%

casting speed Vc (m/min, for 8"x52" slab)

Vc=2.3

Vc=2.0

Vc=1.5

Vc=1.0

Vc=0.5

Vc=0.2 Nozzle bore diameter DB=78mm

(a) HT vs. VC at different FL and QG =10SLPM (b) HT vs. QG at different VC and FL=50%

0

0.5

1

1.5

2

2.5

3

50 55 60 65 70 75 80 85 90

Tu

nd

ish

bat

h d

epth

HT (

m)

Nozzle bore diameter DB (mm)

Argon injection QG=10SLPM

Gate opening FL=50%

Casting speed VC=1m/min for 8"x52" slab

Casting speed Vc (m/min, for 8"x52" slab)

Vc=2.0

Vc=1.5

Vc=1.0

Vc=0.5

0

0.5

1

1.5

2

0 20 40 60 80 100

Tu

nd

ish

bat

h d

epth

HT (

m)

Gate opening FL(%)


Vc=2.5

Vc=2.3

Vc=2.0

Vc=1.5

Vc=0.5

Vc=1.0

Vc=0.2



(c) HT vs. DB at fixed VC, QG and FL (d) HT vs. FL at different VC and QG =10SLPM

Figure 3. CFX data (points from Equation 1) and fitting curve (lines of Equation 2) showing effects of castingspeed VC, gate opening FL, argon injection QG and nozzle bore size DB on tundish bath depth HT

with a simple linear function and the HT vs. DB datain Figure 3(c) fits well with a cubic function. Asingle simple function could not be found to fit theHT vs. FL data in Figure 3(d) over the whole FL

range. Thus, these data were split into two regions,with a quadratic function for FL ≤ 60% and a linearfunction for FL≥60%. Putting these relationstogether yields the overall relation:

H a V a V a a F a F aT C C L L= + +( ) + +( )12

2 3 42

5 6

a Q a a D a D a D aG B B B7 8 93

102

11 12+( ) + + +( )for FL ≤ 60% (3a)

H a V a V a a F aT C C L= + +( ) +( )132

14 15 16 17

a Q a a D a D a D aG B B B18 19 203

212

22 23+( ) + + +( )for FL ≥ 60% (3b)

where the ai are 23 constant coefficients. Becausethere are more data points than coefficients, a least

7

squares curve fitting technique was used to find theai values that minimize the distance of each datapoint from the curve. More details about the fittingprocedure are reported elsewhere [4, 8]. The closematch in Figures 3(a-d) between the lines fromEquation 3 and appropriate points from the 3-Dmodel indicates the accuracy of this fit.

Inverse Models

For a given nozzle geometry and cloggingstatus, the four basic casting process variables ofcasting speed, argon injection flow rate, gateopening and tundish bath depth are related.Choosing values for any three of these variablesintrinsically determines the fourth. During a stablecasting process, tundish bath depth and argoninjection are usually kept constant. Gate opening isregulated to compensate for any unwanted effects,such as nozzle clogging and changes in tundish bathdepth, in order to maintain a constant casting speed.

Plots in Figure 2 are inconvenient to apply inpractice because tundish bath depth is generally nota dependent variable. In order to determine andpresent the results in arbitrary practical ways,Equation 3 is inverted into three other forms witheither VC, QG, or FL as the dependent variable(instead of HT) [4, 9]. The derivation of theseinverse models is very easy due to the simple formof Equation 3. Figure 4 shows typical plots withtwo of the inverse models.

The following observations can be made fromexamination of Figures 3 and 4:

• For a given nozzle geometry and gas flow rate,higher casting speed results from a deepertundish bath depth (constant gate opening) or alarger gate opening (constant bath depth).

• Casting speed is more sensitive to a change inbath depth at low casting speed than at highcasting speed.

• Casting speed is more sensitive to a change inbath depth at large gate opening than at smallgate opening.

0

1

2

3

4

5

0 20 40 60 80 100

HT=1.6m

HT=1.4m

HT=1.2m

HT=1.0m

HT=0.8m

HT=0.6m

HT=0.4m

Cas

tin

g s

pee

d V

C(m

/min

, 8"

x52"

sla

b)

Gate opening FL

(%)

Argon injection: QG=5 SLPM

Tundish bath depth: HT


(a)

0

0.5

1

1.5

2

0 2 4 6 8 10

cast

ing

sp

eed

VC

(m/m

in,

for

8"x5

2" s

lab

)


Tundish bath depth: HTH

T=1.6m

HT=1.4m

HT=1.2m

HT=1.0m

HT=0.8m

HT=0.6m

HT=0.4m

HT=0.2m Gate opening F

L=50%


(b)

Figure 4. Inverse model plots showing effect of (a)gate opening, (b) gas injection and tundish bathdepth on casting speed

• Casting speed is more sensitive to gate openingwhen maintaining a high casting speed.

• For a fixed tundish bath depth, increasing argoninjection will slightly slow down the castingspeed (shown in Fig. 2(b)) unless the gateopening increases to compensate.

8

• For a fixed gas flow rate, the percent gasincreases greatly at low casting speeds, resultingin large buoyancy forces which reduce theeffectiveness of the gate opening and make itdifficult to drain the tundish.

• The extent of clogging condition can be inferredby comparing the measured steel flow rate withthe value predicted by the inverse model for thegiven geometry, tundish bath depth, gas flowrate and percent gate opening.

Combined Model

The same multivariable curve-fitting methodused to find Equation 3 can be employed to developequations to predict trends for other importantnozzle flow characteristics under practical operatingconditions. Such characteristics include the lowestpressure in the nozzle (air aspiration), bias flow dueto the slide-gate throttling, and the properties of thejets exiting the nozzle ports.

As an example, a model is now developed topredict the lowest pressure in the nozzle. When thelowest pressure in nozzle is below atmosphericpressure, air aspiration may occur if the joints arenot properly sealed. In the 3-D numericalsimulations, the reference ambient pressure is set tozero. Therefore, a negative pressure predicted in thesimulation implies the existence of a partial vacuum(less than one atmosphere) which suggests atendency for air aspiration.

For each 3-D simulation case in Table II, thelowest pressure in the nozzle is recorded. Theresults are then curve-fit to produce an equation forthe lowest pressure, PL, as a function of the fourindependent variables, VC, FL, QG, and DB.

As shown in Figure 5, the PL vs. VC data fitswell with a quadratic function, the PL vs. QG datafits well with a linear function, the PL vs. DB datafits well with a quadratic function, and the PL vs. FL

data must be split into two different linear regionsfor FL≤70% and FL≥70%. The overall relationshipcan be written as

P b V b V b b F bL C C L= + +( ) +( )12

2 3 4 5

b Q b b D b D bG B B6 7 82

9 10+( ) + +( ) for FL ≤ 70% (4a)

P b V b V b b F bL C C L= + +( ) +( )112

12 13 14 15

b Q b b D b D bG B B16 17 182

19 20+( ) + +( ) for FL ≥ 70% (4b)

The fitting constants bi (i=1-20) are obtainedusing the same least square curve fitting procedureas for Equation 3. The close match in Figures 5(a)-(d) between the lines from Equation 4 and some ofthe points from the computational model indicatesthe accuracy of this fit. Using two different linearfunctions to fit the PL vs. FL data produces sharptransitions, observed at FL = 70% in Figure 5(d). Asmoother transition could be obtained if more databetween FL = 70% and FL = 100% were generatedand a higher-order fitting model were employed forPL vs. FL.

It should be cautioned that all of the curves inFigures 5(a)-(d) correspond to varying tundish bathdepths. This makes this presentation of the resultsdifficult to interpret. In practice, the tundish bathdepth is usually kept at a relatively constant level.It is the gate opening that is continuously adjustedto compensate for changes in the other variables,such as clogging, tundish level and gas flow rate inorder to maintain a constant casting speed. Tobetter present the minimum pressure results in Eq. 4under these practical conditions, it is combined withone of the inverse models derived in last section.Specifically, the inverse model which rearrangesEq. 3 to express FL as a function of VC, HT, QG, andDB is simply inserted to replace FL in Eq. 4. Thisyields the combined model expressing PL as afunction of these same four practical independentvariables. The results are presented in the latersection on air aspiration.

9

-100

-80

-60

-40

-20

0

20

0 0.5 1 1.5 2 2.5

Lo

wes

t p

ress

ure

in n

ozz

le P

L (

KP

a)

Casting speed VC


40%

50%

60%

70%

FL=100%


Gate opening FL


(a) PL vs. VC at different FL and QG =10SLPM

-80

-60

-40

-20

0

20

50 55 60 65 70 75 80 85 90

Nozzle bore diameter DB(mm)

Lo

wes

t P

ress

ure

in

No

zzle

PL(K

Pa

) Vc=0.5Vc=1.0

Vc=1.5

Vc=2.0


Gate opening FL=50%



(c) PL vs. DB at fixed VC, QG and FL

-100

-80

-60

-40

-20

0

20

0 2 4 6 8 10

Lo

wes

t p

ress

ure

in n

ozz

le P

L (

KP

a)


Vc=0.2

Vc=0.5

Vc=1.0

Vc=1.5

Vc=2.0

Vc=2.3

casting speed Vc (m/min, for 8"x52" slab)

Gate opening FL=50%


(b) PL vs. QG at different VC and FL=50%

-100

-80

-60

-40

-20

0

20

30 40 50 60 70 80 90 100

Lo

wes

t p

ress

ure

in n

ozz

le P

L (

KP

a)

Gate 0pening FL(%)

Vc=0.2

Vc=0.5

Vc=1.0

Vc=1.5

Vc=2.0

Vc=2.3

Vc=2.5


Argon injection QG

=10 SLPM


(d) PL vs. FL at different VC and QG =10SLPM

Figure 5. CFX data (points) and fitting curve (lines) showing effects of casting speed VC, gate opening FL, argoninjection QG and nozzle bore size DB on the lowest pressure PL in nozzle (under varying tundish bath depth)

Comparing with Plant Measurements

To verify the model, the predictions of theinverse model are compared with measurements onan operating steel slab casting machine. Usingvalidation nozzle A in Table I, gate openingpositions were recorded for different steelthroughputs over several months [10]. Figure 6shows the several thousand data points thus

obtained. Only first heats in a sequence wererecorded in order to minimize the effect ofclogging. The tundish bath depth was held constant(HT=1.125m) for these data, and the argon injectionranged from 7 to 10 SLPM. Since the measurementswere recorded with different units from the Table IIfor the inverse model, the model predictions requireconversion of FL to the plant definition of gate

10

opening FP and casting speed to steel throughputQFe by

FP = (1-24%)FL + 24% (5)

and

QFe(tonne/min)=1.8788 VC(m/min) (6)

The geometry of the Validation Nozzle A isnot exactly the same as the Standard Nozzle whichthe inverse model predictions are based on, but it isreasonably close. In addition to the inverse modelprediction, additional CFX simulations wereperformed for the actual conditions of theValidation Nozzle A in Table I. These results alsoare shown in Figure 6 as 3 big dots.

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.50.0

10

20

30

40

50

60

70

80

0

Steel throughput (tonne/min)

Slid

e-g

ate

op

enin

g F

p (

%, p

lan

t d

efin

itio

n)

Measured data legend: A = 1 obs, B = 2 obs, etc. All data for the first heat

AAAAAAAAAAAAAAAAAAAAAAAA

AAAAAAAAAAAAAAAAAAAA

AAAAAAAAAA

Inverse Model (standard nozzle)

CFX simulation (Validation nozzle A)

Measured data fitting curve

Measured dataA ~ Z

AA

Slid

e-g

ate

op

enin

g F

L (

%)

0

10

20

30

40

50

60

70

0.0 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25

Casting speed VC (m/min, for 8”x52” slab)

Figure 6. Comparison of measurements forValidation nozzle A and model predictions

Figure 6 shows that the CFX results are veryclose to the inverse model predictions, despite theslight difference in nozzle geometry. In addition tovalidating both models, this suggests that theinverse model derived from the Standard Nozzle isapplicable to other practical conditions, as long asthe nozzle geometry is reasonably close. Bothpredictions from the inverse model and CFXsimulation match the larger extreme of the range ofmeasured gate opening percentage for a given steelthroughput. The decreased gate opening often

experienced in the plant is likely due to thefollowing reasons:

• Less argon flow in the plant (7~10 SLPM vs. 10SLPM), needs smaller opening to accommodatethe same liquid flow.

• Rounded edge geometry likely found in theplant nozzles may cause smaller pressure dropthan the sharp edge in new or simulated nozzles,so need less opening to achieve the same flow.

• The initial clogging experienced during the firstheat may reduce the gate opening required for agiven steel throughput. This is because, before itstarts to restrict the flow channel, thestreamlining effect of the initial clogging mayreduce the overall pressure loss across nozzle.The last two factors will be further discussed inthe next section.

EFFECT OF CLOGGING

Initial Clogging

In both numerical simulations andexperiments, three recirculation zones are observedin the vicinity of the slide-gate [6, 8, 11]. One iscreated in the cavity of the slide-gate itself and theother two are located just above and below thethrottling plate. In these recirculation zones, theflow is turbulent and the gas concentration is high.These recirculation zones and the sharp edges of theslide gate surfaces both may create an extraresistance to flow. Slight erosion by the flowingsteel may round off the ceramic corners. Inaddition, it is known that clogging tends to buildupinitially in the recirculation regions [12]. Because ofthis, the initial might not impede the flow andinstead may decrease the flow resistance bystreamlining the flow path. This may decrease thetotal pressure drop across the nozzle.

To investigate these phenomena, foursimulations were performed using the 3-D K-εnumerical flow model for the cases illustrated inFigure 7. The geometry and casting conditions,given in Table I for Validation Nozzle B, were

11

AAAA

AAAAAAAA AA

AAAA

AAAAAAAAAAAA

(a) Sharp edges (b) Rounded edges (c) Sharp edges with (d)Sharp edges with initial clogging more inital clogging

Figure 7. Schematic of initial clogging and roundededges in the vicinity of the slide-gate (ValidationNozzle B)

chosen to match conditions where measurementswere available for comparison [13]. All four casesare the same, except for the geometry near the slidegate. The first case, Figure 7(a) has sharp edgessimilar to the standard nozzle simulated in theprevious parametric study. The next case, shown inFigure 7(b), has the four slide gate edges roundedwith a 3mm radius. The final two cases have therecirculation regions partially filled in to representtwo different amounts of initial clogging withalumina reinforced by solidified steel. One case,Figure 7(c), has solid clog material in the gatecavity and around the throttling gate and smoothsurfaces in the upper SEN. The final case, Figure7(d), has extra clogging above and below the gate.and the other with clogging at the same places butwith more buildup around the gate.

From the numerical simulation results, thecorresponding tundish bath depth for each case wascalculated using Equation 2. These values arecompared in Figure 8 with the measured tundishbath depth. The standard sharp-edge case with noclogging has the largest pressure drop, so requiresthe greatest bath depth. Rounding the edges of thethrottling plates reduces the pressure drop across thegate plates and lowers the required tundish head by18%. The initial clogging is even more effective atstreamlining the liquid steel flow around the slide-

gate, decreases the recirculation loops and lowersthe pressure loss. The initial clogging of Figure 8(c)reduces the required tundish bath depth by 24%,relative to the standard sharp, non-clogged case.Further increasing the initial clogging, case Figure8(d), decreases the required tundish bath depth by36%, which is even lower than the measured valueof 0.927m. The clogging condition for themeasurement is unknown. The measurement wastaken at the first heat, so it is likely to have someinitial clogging buildup around the slide-gaterecirculation regions.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 . 9 2 7 m

1 . 3 4 0 m

1 . 1 0 2 m

1 . 0 1 5 m

0 . 8 5 6 m

Tu

nd

ish

bat

h d

epth

(m

)

Predicted with Equation 1

Sharp edges

Roundedges

Sharp edgeswith initial clogging

Sharp edgeswith moreinitial clogging

Measured( a ) ( b ) ( c ) ( d )

Figure 8. Effects of initial clogging and roundededges on tundish bath depth (Validation Nozzle B)

The clogging condition and edge roundnessaffects not only the pressure drop across the nozzlebut also the flow pattern exiting the ports into themold. Figure 9 shows the simulated flow pattern atthe center plane parallel to the narrow face.Difference such as edge roundness and cloggingaround the slide gate greatly change the flow patternin the SEN as well as the jets out of the ports. Thejets is seen to to vary from two small symmetricswirls to a single large swirl which can switchrotational directions. Thus, a slight change inclogging can suddenly change the jet characteristicsexiting the port. This will produce a transient

12

fluctuation in flow in the mold cavity which couldbe very detrimental to steel quality. This resultprovides further evidence of problems caused evenby initial nozzle clogging.

Figure 9. Effects of initial clogging and roundededges on nozzle flow pattern (center plane parallelto the narrow face) for Validation Nozzle B.

Severe Clogging

With increasing alumina buildup, theclogging, instead of streamlining the flow, begins torestrict the flow channel and to create more flowresistance. The gate opening then must increase tomaintain constant liquid steel flow rate through thenozzle. The effect of clogging on the flow depends

0

20

40

60

80

100

50 60 70 80 90 100

Gat

e o

pen

ing

FL (

%)


HT=0.6m

HT=0.8m

HT=1.0m

HT=1.2m

HT=1.4m

HT=1.6m




(a) Gate opening changes to accommodateclogging (decreased nozzle bore size) for fixed gasflow rate and casting speed

0

0.5

1

1.5

2

2.5

50 60 70 80 90 100

Cas

tin

g s

pee

d V

C (m

/min

, 8"

x52"

slab

)


HT=0.6mH

T=0.8m

HT=1.0m

HT=1.2m

HT=1.4m

HT=1.6m


Gate opening FL=50%


(b) Casting speed changes caused by clogging (ornozzle bore size change) for fixed gate opening

Figure 10. Effects of clogging or nozzle bore size

both on the clogging status (how much aluminadeposits) and on the clogging shape (where and howthe alumina deposits). It has been observed thatclogging often builds up relatively uniformly in theradial direction and acts to reduce the diameter ofthe nozzle bore [12]. Based on this fact, a way toinvestigate the effect of this type of clogging is

13

simply to reduce the bore diameter. Figure 3(c)shows that decreasing the bore size, ie. increasingclogging, requires the tundish liquid level toincrease in order to maintain the same flow rate atconstant gate opening. Using the inverse model forgate opening FL, the effect of the clogging /decreasing bore size is quantified for practicalconditions in Figure 10.

Figure 10(a) shows how gate opening mustincrease to accommodate clogging (or decreasingbore size) in order to maintain a constant flow ratefor a fixed tundish level. Figure 10(b) shows howthe steel flow rate decreases if the gate openingpercentage does not change. It can be seen that thegate opening much less sensitive to clogging whenthe bore diameter is large. Thus, clogging may bedifficult to detect from gate changes until it is verysevere and the gate opening increases above 60%.

EFFECTS ON AIR ASPIRATION

One of the suggested mechanisms for thebeneficial effect of argon injection in reducingnozzle clogging is that the argon generates positivepressure in the nozzle [12]. Avoiding a partialvacuum in the nozzle should make it less likely forair to be drawn in through any cracks, joints, orsealing problems in the nozzle, with the benefit ofavoiding reoxidation. Numerical simulation in thiswork, Figure 5(b), and water modeling [2] bothshow that the minimum pressure in the nozzle candrop below one atmosphere in some circumstances,and that argon gas injection can raise that pressureabove zero. The lowest pressure in the nozzle is alsoaffected by the casting speed, gate opening, tundishbath depth, nozzle bore size (or extent of clogging),as shown in Figure 5. The combined fitting model(Equation 4) is now applied to study the effects ofthese variables on minimum pressure for thepractical conditions of fixed tundish bath depth.

The lowest pressure in the nozzle is presentedas a function of casting speed in Figure 11 and as afunction of argon flow rate in Figure 12. Both ofthese figures fix the tundish bath depth and allowgate opening to vary, which reflects realistic

operation conditions. The corresponding gateopenings, along with both “cold” and “hot” argoninjection volume fractions, are also marked onFigures 11-12 for easy reference.

Low

est

pres

sure

in n

ozzl

e P

L(K

Pa)

Casting speed (Vc) and argon injection volume fraction (fAr)

Vc(m/min) 8"x52"slab

1020

10203040

30

fAr(%,hot)

fAr(%,cold)

10090807060504030

20 90807060504030

20 90807060504030

20 807060504030

20 706050403010

20 6050403010

-60

-50

-40

-30

-20

-10

0

10

20

0 0.5 1 1.5 2 2.5

HT=0.6m

HT=0.8m

HT=1.0m

HT=1.2m

HT=1.4m

HT=1.6m

Argon injection flow rate QG=10SLPM



Gate opening FL(%)

FL@HT=0.6m

FL@HT=0.8m

FL@HT=1.0m

FL@HT=1.2m

FL@HT=1.4m

FL@HT=1.6m

Figure 11. Effect of casting speed on minimumpressure in the nozzle for constant tundish bathdepth and argon injection flow rate.

The results in Figures 5, 11 and 12 quantifyhow increasing argon injection and decreasingtundish bath depth both always tend to decrease thepressure drop across the slide gate, thereby raisingthe minimum pressure in the nozzle and making airaspiration less likely.

The effect of casting speed is complicatedbecause of several competing effects. Higher flowrate tends to increase the pressure drop and vacuum

14

Argon injection flow rate (QG) and volume fraction (fAr)

QG(SLPM)

1 2

102 4 6

58 706052 54 56

5048

44 4846

444240

424038

3634

52

38

8 12 14

3

-15

-10

-5

0

5

10

15

0 2 4 6 8 10

Casting speed Vc =1m/min

Gate opening FL(%)

Low

est

pres

sure

in n

ozzl

e P

L (

KP

a)

fAr(%,hot)

fAr(%,cold)


FL@HT=0.6m

FL@HT=0.8m

FL@HT=1.0m

FL@HT=1.2m

FL@HT=1.4m

FL@HT=1.6m

HT=0.6m

HT=0.8m

HT=1.0m

HT=1.2mHT=1.4m

HT=1.6m


Figure 12. Effect of argon injection flow rate onminimum pressure in the nozzle for constanttundish bath depth and casting speed.

problems. At the same time, increasing the flowrate allows the gate to open wider, which tends toalleviate vacuum problems. The worst vacuumproblems occur with the gate at about 60% open bylength or 50% open by area fraction, regardless ofcasting speed. Above 70% linear gate opening, theeffect of decreasing the throttling effect withincreased gate opening dominates, so that vacuumproblems are reduced with increasing casting speed.Below 50% gate opening, the effect of loweringcasting speed dominates, so that vacuum problemsare reduced with decreasing speed. A further effectthat helps to reduce vacuum problems at lowercasting speed is that the gas percentage increases(for a fixed gas flow rate).

The common practice of employing oversizednozzle bores to accommodate some clogging forcesthe slide gate opening to close. Although thismakes the opening fraction smaller, the openingarea actually may increase slightly. Thus, thetendency for air aspiration due to vacuum problemswill also decrease, so long as the linear openingfraction stays below 50%. However, this practicedoes generate increased turbulence and swirl at thenozzle port exits, so should be used with caution.

When the pressure drop across the gate issmall and there is no vacuum problem, theminimum pressure in the nozzle moves to thenozzle ports. The port pressure depends mainly onSEN submerged depth.

Optimal Argon Flow

The minimum argon flow rate required toavoid any vacuum in the nozzle can be obtained byletting PL=0 in Equation 4 and solving for QG. Theresults are plotted in Figure 13 as a function ofcasting speed at fixed tundish bath depth and nozzlebore size. The top of this figure shows thecorresponding slide gate opening. The resultssuggest ways to optimize argon flow to avoid airaspiration conditions in the nozzle.

Injecting argon gas sometimes enables thetransition from an air aspiration condition topositive pressure in the nozzle. The minimumargon flow rate required to avoid a vacuumcondition can be read from Figure 13. It increasesgreatly with tundish bath depth. For a given tundishbath depth, the minimum argon flow rate firstrapidly increases with increasing casting speed, andthen decreases with increasing casting speed. Themost argon is needed for linear gate openingsbetween 50-70% for the reasons discussed earlier.

At low casting speed, (below 0.5m/min), or atlow tundish levels (below 0.6m), no vacuum ispredicted in the nozzle. Thus, argon injection is notneeded under these conditions. During ladletransitions and at other times when either castingspeed or tundish level is low, argon flow should beturned off or at least severely reduced. Besides

15

saving argon, this avoids flow problems in the moldand possible gas bubble entrapment.

0

20

40

60

80

100

0 0.5 1 1.5 2 2.5

HT=0.6m

HT=0.8m

HT=1.0m

HT=1.2m

HT=1.4m

HT=1.6m

Co

rres

po

nd

ing

gat

e o

pen

ing

FL (

%)



0

5

10

15

20

25

30

35

40

0 0.5 1 1.5 2 2.5

Min

imu

m a

rgo

n f

low

rat

e (S

LP

M)

req

uir

ed

for

po

siti

ve lo

wes

t p

ress

ure

in n

ozz

le

Casting speed VC


HT=0.6m

HT=0.8m

HT=1.0m

HT=1.2m

HT=1.4m

HT=1.6mNozzle bore diameter D

B=78mm


Figure 13. Effect of casting speed and tundish depthon minimum argon flow rate required for positivepressure in nozzle (bottom) and the correspondinggate opening (top)

Figure 13 shows that very large argon flowrates (over 20 SLPM) are needed to avoid a vacuumcondition for high tundish level (deeper than 1.2m)and high casting speed (above 1.5m/min).Specifically, a 0.2m increase in tundish bath depthtypically requires an additional 5 SLPM of argon tocompensate the vacuum effect at high castingspeeds. In practice, the argon injection flow rate islimited to a maximum of about 15 SLPM. This isbecause argon injection greatly changes the flowpattern in the mold [7]. Excessive argon injectionmay cause a transition from “bubbly flow” to“annular" flow in the nozzle [14], create boilingaction at the meniscus [1 ] and cause qualityproblems. Therefore, it is not feasible for argoninjection to eliminate the vacuum in the nozzle

when the tundish bath is deep and the casting speedis high. Other steps should be taken to avoid airaspiration. Besides improving the sealing at thejoints (especially the joints between the slide-gate,the lower plate, and the SEN holder), other methodssuggested by the model (Equation 4) include:

• Choose bore diameters according to the steelflow rate in order to avoid linear gate openingsnear 60%. To increase gate openings above60%, a smaller nozzle bore diameter could beused, but this allows too little accommodationfor clogging. To decrease gate openings tobelow 60%, a larger bore diameter is needed.

• Decrease tundish bath depth. A shallowertundish level has less pressure drop, sogenerates less vacuum tendency.

CONCLUSIONS

The turbulent flow of liquid steel and argonbubbles in a slide-gate nozzle have been simulatedwith a verified three-dimensional finite differencemodel. The results are further processed usingadvanced multivariable curve fitting methods torelate casting speed, argon injection rate, slide-gateopening position, nozzle bore diameter and tundishbath depth to clogging and air aspiration potential.

Both rounding the nozzle edges due to erosionand initial clogging buildup are found to enhancethe steel flow rate due to a streamlining effect.Only after severe clogging builds up is the floweventually restricted so that the gate opening mustincrease to maintain the casting speed. The extentof clogging can be predicted by comparing themeasured steel flow rate to the prediction of theinverse model presented in this work.

The pressure drop generated across thepartially-closed slide gate creates a partial vacuumjust below the slide gate which tends to entrain air,leading to reoxidation problems. The worst vacuumappears to occur for 50-70% linear gate opening(about 50% area fraction). Increasing argoninjection helps to raise the lowest pressure andsometimes may avoid this vacuum. For shallowtundish bath depths or low casting speeds, the

16

pressure is always positive, so argon should not beused. Less argon is needed if the nozzle bore size ischosen to avoid intermediate casting speeds so thatthe gate is either nearly fully open or is less than50%. For high casting speeds, a 0.2m increase intundish bath depth typically will require anadditional 5 SLPM of argon to compensate thevacuum effect. In practice, argon injection islimited by its effect on the flow pattern, and maynot be able to fully compensate the vacuum effect.

ACKNOWLEDGMENTS

The authors wish to thank the NationalScience Foundation (Grant #DMI-98-00274) andthe Continuous Casting Consortium at UIUC,including Allegheny Ludlum, (Brackenridge, PA),Armco Inc. (Middletown, OH), Columbus Stainless(South Africa), Inland Steel Corp. (East Chicago,IN), LTV Steel (Cleveland, OH), and Stollberg,Inc., (Niagara Falls, NY) for their continued supportof our research, AEA technology for use of theCFX4.2 package and the National Center forSupercomputing Applications (NCSA) at the UIUCfor computing time. Additional thanks are extendedto technicians and researchers at LTV Steel andInland Steel for the measured plant data.

REFERENCES

1. S.M. Dawson, "Tundish Nozzle BlockageDuring the Continuous Casting ofAluminum-killed Steel" (Paper presented at73rd Steelmaking Conference, Detroit, MI,1990, Iron and Steel Society, Inc.), 73, 15-31.

2. H.T. Tsai, "Water Modeling on PressureProfile in the Tundish Shroud at Flo-Con",private communication, Inland Steel, 1986.

3. AEA Technology, "CFX4.2 Users Manual,"(1997).

4. H. Bai and B.G. Thomas, "Two Phase Flowin Tundish Nozzles During ContinuousCasting of Steel" (Paper presented atMaterials Processing in the Computer AgeIII, TMS Annual Meeting, Nashville, TN,2000).

5. S. Sivaramakrishnan, H. Bai, B. Thomas, P.Vanka, P. Dauby, M. Assar, "Transient FlowStructures in Continuous Casting of Steels"(Paper presented at 83rd SteelmakingConference, Pittsburgh, PA, 2000, Iron andSteel Society, Warrendale, PA), 83.

6. B.G. Thomas, "Mathematical Models ofContinuous Casting of Steel Slabs" (Report,Continuous Casting Consortium, Universityof Illinois at Urbana-Champaign, 1998).

7. B.G. Thomas, X. Huang and R.C. Sussman,"Simulation of Argon Gas Flow Effects in aContinuous Slab Caster," MetallurgicalTransactions B, 25B (4) (1994), 527-547.

8. B.G. Thomas, "Mathematical Models ofContinuous Casting of Steel Slabs" (Report,Continuous Casting Consortium, Universityof Illinois at Urbana-Champaign, 1999).

9. H. Bai and B.G. Thomas, "Inverse Modelfor Casting Process Variable Relationship"(Report, Continuous Casting Consortium,1999).

10. R. Gass, private communication, InlandSteel, 1998.

11. Y.H. Wang, "3-D mathematical modelsimulation on the tundish gate and its effectin the continuous casting mold" (Paperpresented at 10th Process TechnologyConference, Toronto, Ontario, Canada,1992, Iron and Steel Society, Inc.), 75, 271-278.

12. K.G. Rackers and B.G. Thomas, "Cloggingin continuous casting nozzles" (Paperpresented at 78th Steelmaking Conference,1995), 723-734.

13. M.B. Assar, private communication, LTVSteel, 1998.

14. M. Burty, M. Larrecq, C. Pusse, Y.Zbaczyniak, "Experimental and TheoreticalAnalysis of Gas and Metal Flows inSubmerged Entry Nozzles in ContinuousCasting" (Paper presented at 13th PTDConference, 1995), 287-292.

For information on this paper, contact B.G. Thomasat 1206 West Green St., Urbana, IL 61801; Ph 217-333-6919; email: [email protected]

Documents

Effects of Clogging, Argon Injection and Casting Conditions on ...ccc.mines.edu/s/Publications/00_ISS 99.mymac.ccc.pdfaspiration is less. The optimal argon flow rate depends on the