CCC Annual Report Transient Turbulent Multiphase Flow and Level Fluctuation Defects in Slab Casting

Transcription

1 CCC Annual Report 00 UIUC, August, 00 Transient Turbulent Multiphase Flow and Level Fluctuation Defects in Slab Casting R. Liu *, J. Sengupta **, D. Crosbie **, and B. Thomas * *Department of Mechanical Science & Engineering University of Illinois at Urbana-Champaign **Global R&D at Hamilton ArcelorMittal Dofasco Inc. OUTLINE Modeling flow rate using two models: stopperbased model and level-based model, to provide inlet boundary conditions for the transient simulations Multiphase Flow Computational Model Validation iva comparison with water model measured data Turbulent multiphase flow simulation of steel flow in and mold in No. CC caster of Dofasco, using RANS and LES models Future work University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu

2 -- st PART: model development for flow rate calculation Modeling Flow Rate Objectives to provide accurate inlet boundary conditions for the CFD transient simulations, since: flow rate cannot be accurately measured, especially for cases with sudden changes of the stopper rod position, thus needs modeling; flow rate change with time as inlet boundary condition is crucial to the accuracy of the transient CFD simulations to study clogging during casting process since clogging cannot be measured directly, but is critical to the flow patterns in the mold and the quality of the final products by comparing flow rates from zero-clogging model prediction with the plant trial measurements University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 3 Two Models to Predict Flow Rate Stopper-based Model using stopper-rod position to predict the flow rate in based on the analysis of Bernoulli s equation parameters including: measured stopper rod position stopper rod zero-flow position tundish fraction Level-based Model using measured mold level signal and casting speed to predict the flow rate in based on the mass balance at inlet, mold bottom and mold level parameters including: measured mold level measured casting speed University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 4

3 Sketch of the Stopper Rod System -- definition of Stopper Rod Opening p, v, z f*h T g Shell Nozzle Clogging p, v, z Shell V casting h SR h SRC hsro hsrc : : : h SRO Stopper Rod Position Stopper Rod Zero-Flow Position (closed position) Stopper Rod Opening hsro = hsr hsrc hsr reference location University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 5 Stopper-based Flow Rate Model -- Analysis of Bernoulli s Equation Bernoulli s Equation: p V p V + z + = + z + + hminor + hfriction + hclogging ρg g ρg g At location at tundish meniscus: p = atm; V 0 At location at port exit: NOTE: Minor loss at port exit is ignored in the model p = atm + ρ gh sen _ sub p p V + z z = + hminor + hfriction + hclogging ρg g V h _ + f h + L = + h + h + h g sen sub tundish tundish sen minor friction clogging Variables in the EQN h sen_sub f tundish h tundish L sen Physical Meaning submergence depth Tundish (weight) fraction Total height of the tundish = m/s; V = V Distance from tundish bottom to port center University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 6

4 Bernoulli s Equation gives: Stopper-based Flow Rate Model Three Sub-models for h friction, h minor andh clogging V h h h L h f h g = + ( _ ) minor friction clogging sub tundish tundish sub-model sub-model sub-model 3 Sub-model for friction head loss: h friction V = ξ g Sub-model for Stopper Rod Gap Head loss: h gap V = ξ g ξ ξ Sub-model 3 for Stopper Rod Gap Head loss: h V clogging = ξ 3 g ξ 3 : function of Re number, and surface roughness, diameter and length : function of stopper rod opening, inner corss-section area : can only be estimated by comparing predicted flow rate with estimated flow rate from plant trials University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 7 Original Bernoulli s equation: V Sub-Model Modeling Friction Head Loss h h h h f h L g = + + gap friction clogging sen_sub tundish tundish The friction head loss along the nozzle is modeled as: L V hfriction = C D g C is a function of Reif length and diameter are fixed Since:. Flow in the usually reaches the Re of 0 5 ;. The inner surface is not smooth (due to attachment of the alumina oxide inclusions) C = 0.07 ~ 0.08 Figure from S Beck and R Collins, University of Sheffield University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 8

5 Sub-Model Modeling Stopper Rod Gap Minor Loss V Contraction of Flow Area Expansion of Flow Area h = h + h gap 3 h h h h f h L g = + + Minor Loss in Contraction Case (reference []), from location to location : V A h = ξ 0 g A ξ = 0.5 Minor Loss in Expansion Case (reference []), gap friction clogging sen_sub tundish tundish from location to location 3: 3 3 h 3 = ξ A Reference []: 3 3 g ξ 3 = Z. Zhang, G. Cui. Fluid Mechanics. A Tsinghua University ISBN: /O0. pp330. University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 9 A, V A 3, V 3 V h Assume: h = h + h gap SRC 3 A h SRO = C h SRO h V A A = g ChSRO ChSRO SR V Final Form of Flow Rate Q from Stopper-based Model According to the friction head loss and loss models: Let V hgap hfriction hclogging hsen_sub ftundishhtundish L g = + + V A A L C h h f h L = + + g ChSRO ChSRO D h V clogging = C 3 g V A A L clogging sen_sub tundish tundish C + + C3 = hsen_sub + ftundishhtundish + L g ChSRO ChSRO D = ( sen_sub + tundish tundish + ) g h f h L A A L C + C C h C h D 3 SRO SRO The model consists of three parameters, C for friction loss along the nozzle C for minor loss at stopper rod gap C 3 for head loss due to clogging in the Q = A ChSRO ChSRO D 3 University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 0 ( sen_sub + tundish tundish + ) g h f h L A A L C + C

6 Model Parameters Calibration In the final model, the parameters include C, C, and C 3 :. Q = A ( sen_sub + tundish tundish + ) g h f h L A A L C + C C h C h D 3 SRO SRO As mentioned previously, according to Moody s chart, C = 0.07 ~ C can be calibrated using the measured Throughput vs. Stopper Rod Opening data, as shown in latter slides; 3. No Clogging Assumption C 3 is assumed to be zero for the data used to calibrate the parameters (provided by Dofasco); Variables Definition and Value Variables Definition and Value h sen_sub submergence depth (e.g m) f tundish Tundish (weight) fraction 0~ (e.g. 0.8) h tundish Total height of the tundish (e.g..45 m) L sen D A Length from tudish bottom to port upper edge: (e.g..59 m) inner diameter (e.g m) inner cross-section area (e.g m ) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu Model Parameters Calibration Since the tundish fraction influences the pressure head in the system, in order to calibrate the model using the measured data, an estimation of the tundish fraction is needed: Tundish Fraction Time (sec) During the process, it is observed that the average tundish fraction stays around 0.8, the calibration will take the following parameters: Variables Physical Meaning Value C Friction Loss Coefficient C 3 Minor Loss Coefficient due to Clogging 0 f tundish Tundish (weight) fraction 0.8 University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu

7 Stopper-based Model Prediction vs. Measured Data--Effect of C Stopper Rod Opening (mm) h SR C =0.075 C 3 =0 f tundish =0.8 Measured Data C =.7 C =.5 C =.3 C =.0 Linear Interpolation Throughput (ton/min) Increase of the C coefficient leads to an upper-shift of the Stopper Rod Opening vs. Throughput curve; (shifting curve) Measured data points above the blue curve indicate extra head loss from the measurement (maybe due to clogging) Choose the Stopper Rod gap minor loss coefficient C as.7in following slides University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 3 Influence of Tundish Fraction on Flow Rate By keeping C at.7 and C at as mentioned on last slide, varying tundish fraction from 0. to 0.8 at an interval of 0., the influence of the tundish fraction on the flow rate gives: 3 Increase of the tundish C 30 = fraction leads to an C =.7 lower-shift of the 8 Stopper Rod Opening vs. Throughput curve; Stopper Rod Opening (mm) h SR Measured Data C =.7 Tundish Fraction 0. Tundish Fraction 0.4 Tundish Fraction 0.6 Tundish Fraction 0.8 Tundish Fraction.0 For most processes, the tundish fraction keeps between 0.4~0.8, varying from the blue line to the pink line in the plot (shifting curve) Throughput (ton/min) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 4

8 Influence of C on Flow Rate By keeping C at.7 and tundish fraction at 0.8, varying friction factor C from 0.0 to 0.08 at an interval of 0., the influence of the friction factor on the flow rate gives: Stopper Rod Opening (mm) m) h SR Tundish Fraction: 0.8 C =.7 Measured Data C = 0.0 C = 0.04 C = 0.06 C = 0.08 Little influence is found if the throughput is less than 3.5 ton/min; For throughput larger than 3.5 ton/min, increase of the friction factor shifts the stopper rod opening vs. throughput curve upper. (changing curve slope) Throughput (ton/min) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 5 Influence of C 3 on Flow Rate By keeping C at.7 and tundish fraction at 0.8, friction factor C at 0.075, the influence of the minor loss coefficient due to clogging on the flow rate gives: Stopper Rod Opening (mm) h SR C = C =.7 Tundish Fraction: 0.8 Measured Data C 3 = 0 C 3 = C 3 = 4 C 3 = 6 C 3 = 8 For throughput larger than.5 ton/min, increase of the clogging minor loss coefficient shifts the stopper rod opening vs. throughput curve upper. (changing curve slope) Throughput (ton/min) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 6

9 Level-based Flow Rate Model Flow rate based on measured casting speed: Flow rate based on mass conservation from the moldlevel signal: ( ) m ( ) ( ) Q i = V i W T V cast is casting speed h m i + hm ( i ) π d, outer QE ( i ) = W T + Qm i t 4 cast ( ) h m is measured mold level Parameters Physical Meaning Parameters Physical Meaning h m mold level d,outer outer diameter of W mold width Q E flow rate prediction T mold thickness Q m Throughput from measured casting speed t time interval between data points i i th time step University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 7 Measured Data for Example Transient Case -- Heat 96078, segments ~4 90 measured stopper rod position measured mold level Stopper Rod Position (mm) Mold Level (mm) Measured Stopper Rod Position Data Cubic Spline Interpolation Time (sec) Measured Data Cubic Spline Interpolation Time (sec) Symbols represent the data recorded by the PI system, Curves are generated by Cubic Spline Interpolation (to smoothe the data points) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 8

10 Model Parameters Change during Clog Releasing C 3 change during release of clogged material due to bumping.3 C change Loss Coefficient C 3 Clogging Head Time (sec) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 9 Stopper Rod Gap Min nor Loss Coefficient C Time (sec) Sudden reduction of clogging results in a decrease of the clogging head loss coefficient in the model Reduction of clogging around stopper-rod tip will increase the stopper rod gap area, thus increase the gap coefficient Stopper-based vs. Level-based Flow Rate Model --C 3 decreasing, C increasing, h SRC = 46 mm clog released t =. sec Curve to put into the transient model as unsteady inlet B.C. Stopper rod starting position: 46 mm C = C = increasing from.8 to. C 3 : decreasing from 6 to 0 translation of estimated flow rate curve is translated from the estimated flow rate curve by. sec ahead. sec is the response time for the meniscus from observation University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 0

11 Conclusions (for part ) The model predictions (from stopper-based model) are compared with measurement from Dofasco. Reasonable match is achieved. Three parameters in the model are calibrated to match with the measured data: Friction loss coefficient varies from 0.07~0.08 Stopper rod gap minor loss coefficient is calibrated between.0~.7 The clogging minor loss coefficient is assumed 0in the calibration University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu Conclusions (for part ) Influences of the three coefficients on the flow rate is studied: Increase of either the stopper rod gap minor loss coefficient C or the tundish fraction will shift the stopper rod vs. throughput curve lower Increase of either the friction loss coefficient C or the clogging loss coefficient C 3 will increase the slope of the curve This calibrated and validated model can be used to predict clogging in the system by calibrating the clogging minor loss coefficient C 3 according to the measurement University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu

12 -- nd PART: Computational Model Validation Model Validation by :.3 Water Model Objective: to validate current computational models with the water model measurement water flow rate 4.3 m 3 /h Air gas injection rate Nozzle submerging depth Nozzle inner diameter Mold width Mold thickness Domain width Domain thickness Domain length Density (air) Density (water) Dynamic viscosity (air) Dynamic viscosity (water) Bubble Diameter (mm) CASE : 0 LPM CASE : 6.7 LPM (room temperature) 76 mm 33.6 mm 77 mm 00 mm mm 50 mm 700 mm.5 kg/m3 (at velocity inlet) 998. kg/m e-05 kg/m-s 0.00 kg/m-s.7 mm University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 3 Geometry and Mesh for :.3 Water Model quarter mold Total: 0.3 million structured hexahedral cells University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 4

13 Computational Details and B.C. Settings: Models and Schemes Turbulence Models Multiphase Model Model for Shell Growth Gas Escaping from Meniscus Advection Discretization Pressure Discretization Computational Details and Boundary Conditions Name k-epsilon with std. wall function Mixture Model No Mass Sink for Argon Phase st order upwinding for k-epsilon model Body Force Weighted Scheme Parameters for the transient run: Time marching scheme Time before collecting statistics Time for the stats st order implicit, 0.05 sec time step 60 sec 60 sec Domain Boundaries B.C. Domain Boundaries Meniscus Free-Slip Wall (no slag layer) Outlet Pressure Outlet University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 5 B.C. Single Phase Flow Water Velocity Distribution Mold Heigh ht m/s velocity magnitude (m/s) velocity magnitude (m/s) m/s meniscus velocity distribution 0. m/s Mold Width Double Roll Flow Pattern (at center plane between wide faces) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 6

14 Water Velocity Distribution at Port Exit Back Flow Region Half port centerline Y m/s X Z velocity magnitude (m/s) Port Width University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 7 Port Height Jet swirling at port corner 0.5 m/s velocity magnitude (m/s) Meniscus Horizontal Velocity V x (m/s) Turbulent Kinetic Energy along Meniscus CL (m /s ) x0-4.0x0-4 Horizontal Velocity along Meniscus Centerline (Prediction vs. Measurement) Distance from Mold Center 6.0x sec time-averaged velocity (RMS of the mean velocity is negligible, 0-4 m/s magnitude) Meniscus Horizontal Velocity V x (m/s) Simulation Measurement Distance from Mold Center Distance from Mold Center 3 University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu Estimate Fluctuations from calculated turbulent kinetic energy: k = u + v + w ( ) u k

15 Water-Air Two-Phase Flow Water Velocity Distribution -0. water velocity field at center plane between wide faces velocity magnitude (m/s) Mold Heigh ht m/s velocity magnitude (m/s) Mold Width 0 0. m/s Mold Width University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 9 Mold Thickness m/s water velocity field on meniscus Water Velocity Distribution at Port Exit Y X m/s Z velocity magnitude (m/s) Port Height Streamline showing jet swirling at port corner port center line 0.5 m/s velocity magnitude (m/s) Port Width University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 30

16 Water-Air Two Phase Flow Air Velocity Distribution f = 0.0 f = 0.03 f = 0. Mold Height m/s Air Volume Fraction 5.00E-0.00E E-0.00E E-03.00E E-04.00E E-05.00E E Mold Width air velocity field on center plane between wide faces 0 0. m/s same contour legend as above D view of air volume fraction Iso-surface air velocity field on meniscus University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 3 Air Velocity Distribution at Port Exit port center line Y X m/s Z Port Height air volume fraction Port Width University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 3

17 Turbulent Kinetic Energy along Meniscus CL (m /s ) Meniscus Horizontal Velocity V x (m/s) Simulation VS. Measurement Water Velocity at Meniscus Centerline Distance from Mold Center sec time-averaged velocity (RMS of the mean velocity is negligible, 0-4 m/s magnitude) Meniscus Horizontal Veloc city V x (m/s) k = u + v + w u k Distance from Mold Center 3 University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu Simulation Measurement SE EN Distance from Mold Center Estimate Fluctuations from calculated turbulent kinetic energy: ( ) Single Phase VS. Multiphase Flow Water Velocity Distribution 0. m/s 0. m/s Mold Height -0.4 Mold Height Mold Width Mold Width Exiting of gas at port exerts a drag force acting on the liquid pointing towards meniscus, which will change the double-roll flow pattern into complex or even singleroll flow pattern, depending on the liquid/gas flow rate, mold width and bubble size distribution. University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 34

18 Conclusions (for part ) By comparing simulation results with water model measurements, for single phase flow as well as the multiphase flow with relatively low gas volume fraction (8% gas), current model is able to match nicely with experiment data gas injection into the mold tends to make more single-roll of the flow pattern, depending on the liquid/gas flow rate simulation results give lower RMS of velocities than measurements, due to: use of URANS (model is diffusive) use of st order upwinding for advection terms (diffusive scheme) use of quarter mold as domain (suppressing the bias flow between left and right part of the mold) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu rd PART: Computational Model Validation Geometry and Mesh: Simulation of Dofasco No. Steel Caster --Heat 96078, Time 9955 sec Total Mesh: blocks for structured grid meshing 800,000 hexahedral cells Model Geometry Parameters Value Mold Width.47 Domain Length 3.0 Mold Thickness at Meniscus 0.5 Outer/Inner Diameter 0.30/0.075 Submergence Depth 0.66 Solidification Constant for Shell Thickness (in/sqrt(min)) 0.98 University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 36

19 Computational Schemes and Boundary Conditions Computational Details and B.C. Settings: Models and Schemes Name Turbulence Models. k-epsilon with std. wall functions. LES with Wale Model Multiphase Model Mixture Model Model for Shell Growth Gas Escaping from Meniscus Mass and Momentum Sinks for Liquid Steel Mass Sink for Argon Phase Advection Discretization. st order upwinding for k-epsilon model. Bounded Central Diff. for LES Time Marching Scheme. st order implicit scheme for k-e model, 0.05 sec time step. nd order implicit scheme for LES, 0.00 sec time step Pressure Discretization Body Force Weighted Scheme Time before collecting statistics Time for the averaging 0 sec 0 sec Domain Boundaries B.C. Domain Boundaries Meniscus No-Slip Wall (with slag layer) Outlet Pressure Outlet University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 37 B.C. Argon Gas Injection and Process Parameters argon injection at upper tundish nozzle 4.03 SLPM plate argon injection 8.0 SLPM argon injection.73 SLPM argon injection at stopper rod tip.73 SLPM Process Parameters Values Casting Speed (m/min). m/min Argon Injection Method Material Properties shown on the left Values Liquid Steel Density (kg/m3) 750 Argon Density (kg/m3). At 93 K, At 83 K, 0.9 Liquid Steel Dynamic Viscosity (Pa*s) Argon Dynamic Viscosity (Pa*s) Argon Mean Bubble Dia (mm) At 93 K,.86*0-5. At 83 K, 8.85*0-5.5 University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 38

20 Liquid Steel Velocity Distribution at Center Plane between Wide Faces by k-epsilon Model Length Mold Liquid Steel Velocity (m/s) Length Mold L 0.5 m/s Liquid Steel Velocity (m/s) Mold Width Contour for Liquid Steel Velocity Distribution Mold Width Vector Plot for Liquid Steel Velocity Distribution Partial double roll flow pattern is observed (complex flow pattern) Jet exiting the port is split into an upward portion heading to meniscus, and a remaining portion impinging the narrow face University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 39 Argon Gas Velocity Distribution in the Mold by k-epsilon with Mixture Model argon volume fraction distribution at meniscus (same legend as below) argon volume fraction distribution at center plane between wide faces 0.5 m/s Argon Volume Fraction Mold Length E-0.7E-0.3E-0.0E-0.7E-0.3E-0.0E-0 6.7E-0 3.3E-0.0E-05 Gas gathers near the outer surface, with a maximum gas velocity of.0~. m/s Mold Width University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 40

21 Steel/Argon Velocity Distribution at Port Exit by k-epsilon Model Time-averaged Liquid Steel Velocity 0.5 m/s Y Time-averaged Argon Velocity 0.5 m/s Y Mold Length Z X Liquid Steel Velocity (m/s) m/s Liquid Steel Velocity (m/s) Mold Length Z X Argon Volume Fraction Jet swirling at port lower corners Part of the jet bends towards meniscus due to the gas drag force Maximum jet velocity is found at the port bottom Gas gathers both near the top edge of the port and in the lower middle part of port Gas jet is split into two parts, one moving upward to meniscus and the other part moving with the liquid steel jet Mold Thickness Mold Thickness University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu m/s Argon Volume Fraction Liquid Steel Velocity Distribution at Center Plane between Wide Faces by LES with Wale Model Instantaneous Steel Flow Field at Center Plane Mold Leng gth Liquid Steel Velocity (m/s) m/s Mold Width Contour for Instantaneous Liquid Steel Velocity Distribution Vector Plot for Instantaneous Liquid Steel Velocity University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 4

22 Argon Gas Velocity Distribution in the Mold by LES with Wale Model Instantaneous Argon Flow Field at Center Plane Mold Length Argon Volume Fraction 6.0E-0.7E-0 5.0E-0.4E-0 4.E-03.E E-04.0E Mold Width Contour for Instantaneous Argon Gas Velocity Distribution University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 43 Mold Length ( m/s Mold Width Vector Plot for Instantaneous Argon Gas Velocity Argon gas is dragged further down the mold by liquid steel, and then floats to the mold top surface Steel/Argon Velocity Distribution at Port Exit by LES with Wale Model Instantaneous Liquid Steel Velocity m/s Z X Y Liquid Steel Velocity (m/s) m/s Liquid Steel Velocity (m/s) Instantaneous Argon Velocity m/s Z X Y m/s Argon Volume Fraction Maximum liquid steel velocity is found at the port corner Velocity field is not symmetric for the left and right part of the port, unlike RANS Gas gathers both near the top edge of the port and in the lower part of the port University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 44 m/s 0.5 m/s Argon Volume Fraction

23 Mean Argon Gas Volume Fraction Distribution --RANS VS. LES Argon Volume Fraction 6.0E-0 5.0E-0 4.0E-0 3.0E-0.0E-0.0E-0.0E-04 Features in Common for RANS and LES mean:.gas sheet attaching at inner surface is formed.higher gas concentration at the upper region of port exit Argon Volume Fraction 6.0E-0 5.0E-0 4.0E-0 3.0E-0.5E-0.0E-0.5E-0.0E-0 5.0E-0 3.0E-0.0E-0 6.0E-03.0E-03 Difference:.LES has a gas region spreading to the lower portion of mold, while RANS does not.rans has higher gas volume fraction at port exit, and the gas sheet in Time-averaged field over 0 sec University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 45 Mean Liquid Steel Velocity Distribution --LES VS. RANS Time-averaged Steel Flow Field at Port Exit m/s m/s Port Height velocity magnitude (m/s) Port Height velocity magnitude (m/s) sec is not long enough for the LES port mean velocities to achieve symmetry Port Width mean argon volume fraction 4.7E-0 4.0E-0 3.3E-0.7E-0.0E-0.3E-0 6.7E-0.0E Port Width mean argon volume fraction 4.7E-0 4.0E-0 3.3E-0.7E-0.0E-0.3E-0 6.7E-0.0E-04 major difference observed in gas volume fraction distribution at port exit --by LES Wale Model, averaged over 0 sec --by k-epsilon Model University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 46

24 Mean Liquid Steel Velocity Distribution --LES VS. RANS Instantaneous Argon Flow Field at Center Plane between Broad Faces m/s 0.5 m/s Velocity Distribution by LES Wale Model -- time averaged field over 0 sec Velocity Distribution by k-epsilon Model University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 47 Conclusions (for part 3) By comparing simulation results from LES and URANS, the following features are observed: LES shows less gas gathering at upper port exit, thus less drag force for liquid steel; LES shows a larger gas region, down the liquid pool; Reasonable match between LES mean velocity field (average over 0 sec) and the RANS results is obtained Due to the difference of the two models in predicting argon gas volume fraction distribution, the flow patterns from LES and URANS are slightly different: LES tends to generate double-roll flow patterns, while URANS tends to generate complex flow patterns (especially at high argon injection rate) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 48

25 Future Work Use URANS to simulate the transient case proposed in previous slides, to find: meniscus level based on dynamic pressure from the CFD models compare this simulated meniscus level with real measurement to further validate and calibrate the stopper-based model the most accurate curve of flow rate vs. time, via this iterative procedure Carry out LES to study the transient flow behavior during the multiple stopper rod movements during this process, to find out: how the flow pattern in the mold is changed by the varying inlet velocity (flow rate) University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 49 Future Work Perform particle transport and entrapment simulations during LES run, to find: how the particles are entrapped due to the flow pattern change preferential locations for the inclusions to get entrapped Perform this modeling process for different casting conditions, to find the critical stopper rod moving velocity that can cause sliver defects University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 50

26 Acknowledgment Continuous Casting Consortium Members (ABB, ArcelorMittal, Baosteel, Corus, LWB Refractories, Nucor Steel, Nippon Steel, Postech, Posco, ANSYS-Fluent) M. Yavuz in ArcelorMittal Global R&D at East Chicago Rajneesh and other graduate students and visiting scholars at Metals Processing Simulation Lab, UIUC University of Illinois at Urbana-Champaign Metals Processing Simulation Lab Rui Liu 5