The Effect of Electromagnetic Stirring and Flow Control Devices on Eight-Strand Tundish Performance

DEGREE PROJECT IN MATERIALS SCIENCE AND ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM, SWEDEN 2018

The Effect of Electromagnetic

Stirring and Flow Control Devices on Eight-Strand Tundish

Performance

BINTANG BERGAS CENDEKIA

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT

i

ABSTRACT

The strand similarity and inclusion removal capability are two critical parameters to measure the performance of multi-strand tundish in clean steel production. In this work, the effect of two flow regulators, i.e., Flow Control Devices (FCD) and Electromagnetic Stirring (EMS) on eight-strand tundish performance have been investigated by establishing a water model and conducting numerical simulations of water model. The water model was focused on revealing the effect of stirring while the simulation was employed to investigate the effect of two FCDs, namely baffle wall and turbo-stopper. The analysis of strand similarity and inclusion removal were conducted by analyzing flow characteristics derived from Combined Model of Residence Time Distribution (RTD) curve and observing the flow movement in the tundish model. In addition, the tundish capability to remove inclusions was also studied by injecting inclusion particles using Discrete Phase Model (DPM) in ANSYS Fluent. Experiment results cause the Combined Model needs to be modified. This modification was employed when analyzing tundish configuration involving stirring. By using the modified Combined Model, the stirring can significantly increase the well-mix volume to almost 100% as it annihilates dead zone. The stirring also increases the similarity between strands and makes the RTD curve more similar to ideal mixing curve. However, the problem of short-circuiting flow need to be solved and care should be taken into consideration regarding the selection of stirring direction as well as bath surface condition when implementing EMS in reality. The simulation results show that the addition of baffle wall and turbo-stopper are beneficial to improve mixing as well as to avoid the short-circuiting flow. Furthermore, compared to individual FCD, the combination of baffle wall and turbo-stopper results in the best performance to remove inclusions by providing surface-directed flow and generating a higher plug flow.

Keywords: multi-strand tundish, strand similarity, flow characteristic, EMS, flow control

device

ii

PREFACE

I would like to express my gratitude to my supervisor at KTH, Mikael Ersson for all his

help and guidance throughout this project. I also want to thank Lidong Teng and Hongliang

Yang as my supervisor in ABB Metallurgy for all their support and advice during my time in

ABB Metallurgy. Finally, I want to thank LPDP (Indonesia Endowment Fund for Education)

for giving me scholarship during my master degree education in KTH.

iii

TABLE OF CONTENT

1 I

... 1

2 T

M

... 3

2.1 Tundish Role in Clean Steel Production ... 3

2.2 Non-Metallic Inclusion Removal in the tundish ... 4

2.3 Current Strategies on Inclusion Removal in the tundish ... 4

2.4 Tundish Flow... 5

2.4.1 Tundish Flow Characterization ... 5

2.4.2 Tundish Flow Problem ... 6

2.5 Tundish Performance Measurement... 6

2.6 Residence Time Distribution Experiment ... 7

2.6.1 Tracer Injection Method ... 7

2.6.2 Dimensionless C-Curve ... 8

2.7 Flow Characterization Calculation ... 9

2.7.1 RTD curve of Plug flow ... 9

2.7.2 RTD curve of Well-Mixed Volume ... 10

2.7.3 Combined Model ... 10

2.7.4 Plug Volume Fraction ... 11

2.7.5 Dead Volume Definition ... 11

2.7.6 Dead Volume Fraction ... 12

2.7.7 Well-Mixed Volume Fraction ... 13

2.7.8 Multi-strand Tundish Model ... 13

2.7.9 Strand Similarity ... 14

2.8 Tundish Water Model Criteria ... 14

2.8.1 Geometric Similarity ... 14

2.8.2 Dynamic Similarity Consideration ... 15

2.9 Multi-strand tundish problem ... 16

2.10 Electromagnetic Stirring Technology ... 17

3 B

T

N

M

... 18

3.1 Governing Equations ... 18

3.1.1 Conservation of Mass ... 18

3.1.2 Conservation of Momentum ... 18

3.1.3 Species Transport Conservation ... 18

3.1.4 Discrete Phase Model (DPM) ... 19

3.2 Turbulence Model ... 19

3.2.1 The Reynold Average Navier Stokes (RANS) model ... 19

3.2.2 The realizable 𝒌−∈ method ... 20

3.2.3 Law of the wall and Near Wall-Treatment ... 20

iv

4 E

S

... 21

4.1 Model selection ... 21

4.2 Similarity Consideration ... 22

4.3 Experimental Tools design ... 22

4.3.1 General Setup : Material and – Scale consideration ... 22

4.3.2 Water Tank Requirement ... 23

4.3.3 Inlet and Outlet Configuration ... 23

4.3.4 Turbo-stopper and Baffle wall attachment ... 23

4.4 Water Pump for Stirring Position ... 26

4.4.1 Pump Location-1 ... 26

4.4.2 Pump Location-2 ... 27

4.4.3 Pump Location-3 ... 27

4.5 Experimental Method and Setup ... 28

4.5.1 Velocity Mapping ... 28

4.5.2 Flow Behavior Observation ... 29

4.5.3 RTD Experiment ... 29

4.5.4 Dye-Color Injection ... 30

4.6 Data Processing Steps of RTD Experiment ... 30

4.7 Experiment variables ... 31

5 N

S

S

... 33

5.1 Computer specifications ... 33

5.2 Geometry Domain ... 33

5.3 Simulation Steps ... 33

5.3.1 Tracer Injection Simulation ... 34

5.3.2 Inclusion Injection Simulation ... 34

5.4 Simulation List ... 35

5.5 Assumption and Model Setup ... 36

5.6 Meshing ... 36

5.7 Convergence Criteria... 37

5.8 Mesh Sensitivity Study ... 38

6 E

R

... 40

6.1 Velocity Mapping ... 40

6.2 Flow Behavior Observation ... 40

6.3 RTD Experiment ... 42

6.3.1 Repeatability ... 42

6.3.2 RTD Curve ... 43

6.3.3 Flow Characteristics... 46

6.4 Dye-Color Injection... 46

7 N

S

R

... 50

v

7.1 RTD Curve ... 50

7.2 Flow Characteristics Calculation ... 51

7.3 Flow-related variable Comparison ... 51

7.4 Results of Inclusion Injection Simulation ... 52

8 EXPERIMENT ANALYSIS ... 53

8.1 Velocity Mapping ... 53

8.2 Analysis of individual strand... 54

8.3 Analysis of Overall Performance of tundish ... 54

8.3.1 Plug Flow ... 55

8.3.2 Dead and Well-Mixed Volume ... 56

8.3.3 Plug to Dead Zone ratio ... 56

8.4 New Approach of Dead Zone calculation in Stirring Case ... 58

8.4.1 Reason 1: Fast Moving Fluid ... 58

8.4.2 Reason 2: Quick Mixing time ... 58

8.4.3 Reason 3: Similarity with RTD Curve of Ideal Mix Flow... 59

8.4.4 Modified Combined Model... 60

8.5 Type of RTD curve in Stirring Case ... 61

8.6 Analysis of Overall Performance Using the New Approach ... 63

8.7 Strand Similarity ... 64

8.8 The Similarity with Ideal Mix Curve ... 64

8.8.1 Derivation formula ... 67

8.8.2 Analysis of deviation from ideal mix... 68

8.9 Plug to Well-Mix Volume ratio ... 69

8.10 Particle Collision ... 70

8.11 Surface Condition ... 71

8.12 Ethical and Social Aspect Consideration... 72

9 N

S

A

... 73

9.1 Wall Y+ Problem ... 73

9.2 Validation of CFD Simulation ... 73

9.2.1 Validation of Color Tracer Injection ... 73

9.2.2 Validation of Prediction Regarding Dead Zone Location ... 74

9.2.3 Validation of RTD Curve and Flow Characteristics ... 75

9.3 Effect of Baffle wall ... 77

9.4 Effect of Turbo-stopper ... 79

9.5 Dead Zone Comparison ... 81

9.6 Inclusion Injection Simulation ... 81

10 C

... 83

11 F

S

... 84

12 REFERENCES ... 85

1

1 INTRODUCTION

Tundish has become a fundamental component in clean steel production because it acts as a metallurgical reactor to remove non-metallic inclusions by allowing floatation of non- metallic inclusion towards the slag on the surface. Besides, tundish can also homogenize the steel composition and temperature in such a way by providing a proper flow that promotes floatation of inclusion and sufficient residence time so that the heat and concentration can be distributed homogeneously to the whole tundish. Thus, the melt flow has become an interesting parameter for researchers as it is an influencing factor to improve tundish performance [1].

Currently, several tools to adjust the melt flow within the tundish has been developed as standard practices in casting foundries such as baffles, weird, dam, or turbo-stopper. Those techniques, which commonly known as flow control devices (FCD), aim to optimize the melt flow by creating optimum turbulence and mixing within the melt. However, the consideration of FCD installation becomes more complicated in a multi-strand tundish. The vast amount of strands lead to increase the tendency of non-homogeneous melt [2]. The melt flowed to the strand closer to the ladle shroud only needs a shorter time to reach the strand so that it tends to have a higher temperature and contains more inclusions compared to the melt which flows to the further strand [3]. The uneven temperature of each strand can also lead to the different cast steel microstructure [4].

The particular combination of FCD could still be used to overcome the similarity problem in the multi-strand tundish. Several studies related to the use of FCD in multi-strand tundish have been conducted by researchers. Zheng and Zhu conducted an experimental study using a physical model of the ten-strand tundish and they stated that the combination of the specific design of turbulence inhibitor and baffles could reduce the inclusion as 42% and lead an evener distribution of inclusion among the strands [5]. This result has a good agreement with Tomasz and Marek who did the similar study for six-strand tundish by numerical modelling and industrial measurements. It is concluded that specific design of baffle could reduce the transient zone in the tundish [6].

Nevertheless, there are some limitations when using FCD in the multi-strand tundish.

Firstly, the multi-strand tundish usually has insufficient working spaces which make challenging to install several FCD. In addition, the shape and design of FCD are also very case sensitive [2, 5, 7, 8] which means the selection of flow control devices must involve both of tundish design and casting parameters. Moreover, the FCD refractory material is also susceptible to wear during the long practice so that it affects the productivity, quality as well as the total cost of its implementation. The last, FCD cannot provide the adjustment of flow for the whole process time.

One of the solutions that could overcome those problems is to use electromagnetic stirring

(EMS) installed on the tundish to control the melt flow. This technology has been developed

by ABB Metallurgy in Västerås, Sweden. One or two EMS which is installed on the side wall

of tundish can provide stirring during the whole casting process thus there is a possibility to

replace or simplify the FCDs using EMS technology. Theoretically, a horizontal stirring

generated in the melt will mix the melt and adjust the melt flow so that the composition and

temperature could be homogenized. From the flexibility point of view, it can be adjusted easily

2 based on the tundish design and the casting process. Furthermore, although the EMS equipment is more expensive than FCD, the longer lifetime and lower running cost will make the total cost becomes comparable. All those advantages make this technology seems to be promising in the future.

However, a comprehensive analysis of the melt flow generated from EMS has not well- understood due to the lack of studies related to the application of this technology in the tundish application. Hence, the focus of this work is to investigate the possibility to replace FCD with EMS in multi-strand tundish application by comparing the flow pattern generated in the tundish for both cases. The investigation was conducted via physical modelling of eight strand tundish and numerical modelling of the experiment. The water model experiment was focused on revealing the effect of stirring while the numerical modelling was employed to investigate the effect of such FCD. The analysis of flow movement, Residence Time Distribution (RTD) curve and dye-color injection were employed to measure two parameters of tundish performance, i.e., flow characteristics and strand similarity in different tundish configurations.

In addition, a simulation of inclusion injection was also conducted to get more understanding

regarding the effect of baffle wall and turbo-stopper on inclusion removal.

3

2 TUNDISH METALLURGY AND PHYSICAL MODELLING 2.1 Tundish Role in Clean Steel Production

In a high-quality steel production route, the removal of non-metallic inclusion becomes a necessary step to gain high cleanliness of steel. It can be accomplished in several stages before casting process such as in hot metal treatment and ladle treatment as shown in Figure 1.

However, in the recent years the removal of inclusion is also conducted in the tundish and even in the mold of continuous casting.

Figure 1 Steelmaking stages and important stage to reduce inclusion [9]

The tundish is an intermediate vessel between casting ladle and the mold. In continuous casting, the molten steel from ladle will flow down into a tundish as can been in Figure 2. The steel melt then continues to flow through a copper mold, where the solidification occurs.

Furthermore, the further cooling applied to ensure the strands are fully solidified before the solidified steel is shaped by roller and cut into a certain dimension.

Figure 2 Schematic illustration of continuous casting process [10]

Tundish has several functions in continuous casting. The primary role is to be a vessel

where the steel melt is distributed to different strands. In addition, tundish is necessary to

provide a constant casting rate by controlling the melt level. Tundish can also serve

homogenization of steel composition and temperature if the melt has proper flow and sufficient

time to stay in the tundish so that the heat and concentration can be distributed homogeneously

to the whole tundish. However, over the last decades, the tundish has a new important role

4 which influences the quality of steel. Since it is the last vessel before mold in continuous casting, the inclusion removal in the tundish is essential for clean steel production.

2.2 Non-Metallic Inclusion Removal in the tundish

The existence of non-metallic inclusion in steel leads to a detrimental effect on the mechanical properties of steel. Besides, non-metallic inclusion can also cause clogging problems in the nozzle of continuous casting. Furthermore, the non-metallic inclusion in clogging may also be carried by the stream during the process, which results in non- homogeneous mechanical properties in steel product. Therefore, the amount and size of non- metallic inclusion are necessary to be limited.

One of the fundamental mechanism of inclusion removal in the tundish is by promoting the floatation of non-metallic inclusion to the tundish slag. The inclusion then can be trapped on the surface so that it can be removed together with the slag. The floatation of inclusion spontaneously happens, since the density of inclusion is lighter than hot liquid steel. However, it is more problematic for small inclusions as the buoyancy force is highly dependent on the diameter of inclusion as can be seen in Equation 1.

𝑉

=

2 𝑔 (𝜌𝑓−𝜌𝑖𝑛𝑐)

9𝜇𝑓

Equation 11

𝑉

is the Stoke rise velocity, 𝑅

is the radius of inclusion, g is the gravity, 𝜇 is the laminar viscosity of the steel, 𝜌

, and 𝜌

are the density of steel and inclusion respectively. Thus, it is evident that inclusions with the larger size could float to the surface more quickly than the smaller inclusions.

2.3 Current Strategies on Inclusion Removal in the tundish

There are different mechanisms of inclusion removal in the tundish and one of the dominant mechanism is by the Stokes floatation [11]. Therefore, the inclusion removal in the tundish is highly dependent on how long the melt can stay in the tundish so that the inclusion has enough time to float towards the slag at the surface before it goes to the outlet of tundish [12]. The time, which is called as residence time, could be affected by different parameters of the tundish, whether it is related with the tundish design or the process parameter [11]. One of the parameters related to the design is the size of the tundish. A larger tundish results in a longer residence time since the melt has to travel with a longer distance to the outlet [11]. However, this option is not preferable due to the limitation of cost and space in reality.

There is, however another way to promote the flotation of inclusion by adjusting the melt

flow in the tundish using a flow control devices (FCD) as illustrated in Figure 3. The addition

of this FCD could change the melt pathway towards the surface as well as prevent any short-

circuiting flow. However, it should be mentioned that the low turbulence at the wall and the

surface should be maintained otherwise the exogenous inclusion due to refractory wear or slag

entrainment appears.

5

Figure 3 Different type of flow control devices (FCD) in the tundish [13]

Nevertheless, the addition of FCD is limited by the available space inside the tundish.

In addition, it is also highly dependent on tundish design. The parameter, such as the location or the height of turbo-stopper, influences the mixing phenomena significantly [8]. The different design of baffles can also create different phenomena which are very specific for every tundish [5, 7]. Moreover, the position of strand and casting parameter such as shroud immersion depth can also affect the performance of FCD [2, 8]. Therefore the consideration when selecting a type and design of FCD must involve both tundish and casting parameters. Another disadvantage is the FCD is susceptible to wear for long time usage. This can affect the productivity, quality and as well as the total cost of its implementation. The last, the FCD cannot provide the adjustment of flow for the whole process time.

The other method to adjust the melt flow inside tundish is by performing a stirring by argon gas or electromagnetic stirring (EMS). Theoretically, the stirring enhances the probability of collision and agglomeration between inclusions resulting in a larger size and floatation velocity [11]. In addition, it also helps the temperature and composition homogenization due to increasing of mixing intensity. However, the gas stirring is limited by injection rate due to a possibility of slag entrainment may occur. The limitation of using FCD and argon stirring could be partly solved by using EMS, which is investigated in this project.

2.4 Tundish Flow

The main objective when designing a tundish is to provide a melt flow that promotes floatation of inclusions and a homogeneous composition in the whole tundish [14]. On the other hand, the flow should be carefully controlled to prevent over-stirring that may induce an exogenous inclusion in the melt. In this chapter, the model of flow characteristic in the tundish was discussed and the preferable flow was also elaborated.

2.4.1 Tundish Flow Characterization

Since the melt flow affects the performance of tundish, a Combined Models of flow characterization has been developed by Sahai and Emi [11]. In this model, the flow inside tundish consists of three types: Plug flow, well-mixed, and dead volumes.

Figure 4 illustrates the difference between those flows in the tundish. The plug volumes

6 represent a ‘direct channel flow’ where the longitudinal mixing has no effect on this flow, but the transverse mixing may be present to some extent. It means the residence time for every volume element in the plug flow region is identical. The well-mixed volume is the volume where the optimum turbulence create the perfect mixing so that the homogeneous concentration achieved at any location. The combination of plug flow and well-mixed flow in the tundish is termed as an active region. Meanwhile, in the dead volumes, the fluid movement is very slow and even tend to be stagnant. Due to this definition, the dead volume is calculated as a portion of the fluid which stays more than two times of theoretical residence time in the tundish [15].

Figure 4 Illustration of different type of flow in the tundish: (a) Plug Flow, (b) Well-Mixed Flow, (c) Dead Volume [11, 16]

This model is developed to analyze the behavior of melt flow in the tundish by analyzing the fraction of each flow-type from a Residence Time Distribution (RTD) curve obtained from tracer injection experiment. The details of the tracer experiment and the formula to calculate each flow-types is explained in section 2.7.

2.4.2 Tundish Flow Problem

Besides the flow should be adjusted to promote floatation of inclusion, problems related to the flow also have to be avoided during the continuous casting process, namely surface turbulence, short-circuiting flow, dead zone and vortexing [14]. High surface turbulence exposed the melt to the air which leads to the re-oxidation reaction as well as entrains the slag to the melt. Meanwhile, the short-circuiting flow is undesired as it contains more inclusions due to insufficient time for floatation of inclusion. The dead zone has a colder temperature which can lead to different steel temperature of different strands. Thus the dead zone could greatly affect the steel quality due to the different microstructure of each strand [17].

Meanwhile, vortexing can entrain the slag to the melt when the melt level is too shallow.

2.5 Tundish Performance Measurement

The investigation of tundish performance is necessary before implementing tundish and FCD design to the real process. In multi-strand tundish, the tundish performance is measured by determining the inclusion removal capability and strand similarity. The tundish capability to remove inclusion can be observed from flow characteristics generated. In this project, the eight-strand tundish performance was measured using a physical model and numerical model.

The physical model or water model is utilized to perform the experiment on a reduced

scale of the real industrial tundish. It can be a reliable approach if the geometry and dynamic

similarity between the model and real tundish are maintained. By performing a tracer injection

experiment in this model, an RTD curve can be obtained. In the end, the inclusion removal

7 capability can be determined by calculating the fraction of flow-type from the RTD curve.

Meanwhile, the numerical model is carried out by simulating the fluid flow in CFD software. Besides analyzing the RTD Curve, the numerical model is also useful to predict the behavior of melt flow in the tundish. The relevant variables in fluid flow such as turbulence energy in any location of tundish can also be observed. Another advantage is the quick result may be obtained for every configuration change in the tundish. However, the appropriate assumption and correct input of boundary condition is necessary to obtain a reliable result.

For both methods, following variables are determined in order to quantify the performance of tundish as a metallurgical reactor to remove inclusions:

1. Minimum dead volume;

2. High Mixing Volume;

3. The melt flow with less turbulence near the wall and slag-metal interface;

4. High Plug Volume or no short-circuiting melt flow from inlet to the outlet of tundish;

5. The Long residence time of melt in the tundish;

6. Large Plug to dead or mixing ratio [11].

2.6 Residence Time Distribution Experiment

The residence time distribution (RTD) curve has been obtained by performing the experiment in water model. A certain amount of soluble and non-reactive chemical is injected through the inlet so that the tracer flows together with the water stream. The tracer should be detectable so that the important phenomena such as short-circuiting stream could be easily observed. Commonly, the tracer used in such experiment is a dye, acid or salt. The time in which the tracer injected is defined as t =0 and then the concentration of tracer will be measured at the outlet of the model. The results are plotted as an RTD curve where the dimensionless time and the dimensionless concentration becomes x- and y-axes respectively.

2.6.1 Tracer Injection Method

Two methods commonly used to inject the tracer during the RTD experiment, i.e., the step input and the pulse input. The difference between the injection method and the RTD curve result is summarized in Figure 5. In the pulse input, a certain amount of tracer is suddenly injected into the water stream in a short time. It is indicated as a peak Figure 5 and the result is named as a C-Curve. The concentration gradually increases up to the maximum peak before it gradually decreases to the initial level. This method is employed to model the continuous casting during steady condition.

In the step input, a certain amount of tracer is injected at the beginning of the experiment

until reach a certain level of concentration. Then the concentration is maintained during the

duration of the experiment. Thus, the concentration at the outlet increases until reaching the

same level of concentration at the inlet. The result of this method is called as F-Curve. This

method is more useful when investigating the transient state such as in grade changing during

casting. In this work, a salt solution of 20%NaCl was used as a tracer and the pulse method

was chosen since the steady state in casting process is the interesting condition.

8

Figure 5 the schematic RTD curve of pulse and step injection method [18]

2.6.2 Dimensionless C-Curve

The response of tracer injection by pulse method is plotted as dimensionless C-curve as can be seen in Figure 6. This curve shows a changing of tracer concentration detected at the outlet with the increasing of flow time. In order to obtain the dimensionless time (𝜃), the flow time is divided by the theoretical average residence time (𝑡

) which can be calculated using Equation 2. The theoretical average residence time is defined as the volume of tundish (V) per volumetric flow rate of water input (Q) as expressed in Equation 3.

𝜃 =

𝑡𝑎𝑣𝑒𝑟𝑎𝑔𝑒

Equation 2

𝑡

=

𝑄

Equation 3

Figure 6 A typical dimensionless C-Curve or Residence Time Distribution (RTD) curve

The area under C-Curve should be equal to unity. Therefore, the dimensionless concentration (C), can be determined by the concentration observed at the outlets (c) divided by the total area under c vs. 𝜃 curve as expressed in Equation 4.

𝐶 =

∫ 𝑐 𝑑𝜃₀^∞

Equation 4

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0.0 0.2 0.4 0.5 0.7 0.9 1.1 1.3 1.4 1.6 1.8 2.0 2.1 2.3 2.5 2.7 2.9 3.0 3.2 3.4 3.6 3.8 3.9

Dimensionless Concentration

Dimensionless time

9 When investigating the similarity between strands, the mean of dimensionless residence time (𝜃

) is necessary to describe the difference of RTD curve and to calculate the dead volume fraction. The mean residence time can be determined by Equation 5.

𝜃

=

∑^∞_𝑖=0𝑐_𝑖

Equation 5

2.7 Flow Characterization Calculation

The dimensionless C-Curve was used to study the similarity between strands in the model as well as the tundish performance by determining the flow characteristics. The explanation of each flow characteristics and its formula is explained in this section.

2.7.1 RTD curve of Plug flow

Due to the absence of longitudinal mixing in the ideal plug flow, the liquid melt will have the same residence time in the tundish. This will produce a maximum peak at the theoretical residence time (𝜃 = 1 ) as shown in Figure 7. However, in the actual process, a horizontal mixing always presence due to the turbulence or the molecular diffusion [11]. This lead to a dispersed plug flow which has a broader peak than the ideal condition as illustrated in Figure 8.

Figure 7 RTD curve of ideal plug flow

Figure 8 RTD curves of Non-ideal plug flow with various dispersion number constants [16]

0.00 0.20 0.40 0.60 0.80 1.00 1.20

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9

Dimensionless Concentration

Dimensionless time

10

2.7.2 RTD curve of Well-Mixed Volume

In an ideal well-mixed flow, the extreme mixing causes an identical tracer concentration at the outlet and at any position in the tundish. The residence curve of this flow is derived from the mass balance equation of tracer injection as written in Equation 6 to Equation 9 [11].

𝑅𝑎𝑡𝑒 𝑜𝑓 𝑡𝑟𝑎𝑐𝑒𝑟 𝑖𝑛𝑝𝑢𝑡 − 𝑅𝑎𝑡𝑒 𝑜𝑓 𝑇𝑟𝑎𝑐𝑒 𝑂𝑢𝑡𝑝𝑢𝑡 = 𝑅𝑎𝑡𝑒 𝑜𝑓 𝑎𝑐𝑐𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 Equation 6 𝑄(0) − 𝑄(𝑐) =

𝑑𝑡

𝑉

Equation 7

𝑑𝑐

𝐶

= −

𝑉

𝑑𝑡 Equation 8 The integration of Equation 8 leads to the solution as stated in Equation 9. From this equation,

the RTD curve is plotted and presented in Figure 9.

𝑐 = 𝑒

Equation 9 In an ideal mix curve, the dimensionless concentration at 𝜃 = 0 equals the average concentration of tracer injected (𝐶 = 1). Then the concentration at any point of tundish decreases exponentially with the increasing of time.

Figure 9 RTD curve of ideal mix volume

2.7.3 Combined Model

Sahai and Emi proposed a new model for analyzing flow in the tundish based on an assumption that the tundish contains three different type of flow: plug flow volume, well-mix flow volume and the dead volume [11]. The model is derived based on a configuration where the plug flow and well-mixed flow presence in a series reactor. The result of RTD curve of this configuration is presented in Figure 10. In this model, the minimum residence time is considered as the plug volume fraction. Meanwhile, the dead zone is defined as a melt that moves very slowly or tends to be stagnant so that it has a long residence time in the tundish.

The further explanation regarding the calculation of those three flow-type fractions are discussed in the next chapter.

0.00 0.20 0.40 0.60 0.80 1.00 1.20

0.0 0.2 0.3 0.5 0.6 0.8 0.9 1.1 1.3 1.4 1.6 1.7 1.9 2.0 2.2 2.4 2.5 2.7 2.8 3.0 3.1 3.3 3.5 3.6

Dimensionless Concentration

Dimensionless time

11

Figure 10 RTD curves of reactor containing ideal plug flow and ideal well-mixed flow in series [11]

2.7.4 Plug Volume Fraction

There are two different approaches used by researchers to calculate the fraction of plug flow. The first way is mentioned in Equation 10, where it is derived from an RTD curve as shown in Figure 10. The plug volume fraction is considered the first appearance of tracer at the outlet indicated by a vertical peak line.

𝑉𝑝

𝑉

= 𝜃

Equation 10 However, the different approaches should be used if there is a deviation of C-Curve from the ideal Combined Model curve due to the presence of longitudinal mixing in the system.

Typically the peak of C-curve becomes wider as illustrated previously in Figure 8 where there is a discrepancy between 𝜃

and 𝜃

[19, 20]. This phenomenon typically happened in the tundish. This second approach, which is named as a dispersed plug flow, was used in this work.

This approach was formulated by Ahuja and Sahai [21] and was used in several studies [20, 22, 23]. By using this method, the plug flow volume is determined by the average of minimum dimensionless time and the time when the peak is observed as expressed in Equation 11.

𝑉𝑝

𝑉

=

2

Equation 11

2.7.5 Dead Volume Definition

According to the Combined Model, the dead volume is defined either as stagnant fluid or slowly moving fluid. From the first definition, it can be concluded that the flow does not flow through the dead regions so that the total flow rate (Q) is always the same as the flow rate through active regions (𝑄

), as illustrated in Figure 11(a).

Meanwhile, from the second definition, the dead volume is determined by calculating the portion of the fluid which stays more than two times of theoretical residence time in the tundish as it moves very slowly [19]. Therefore there is a possibility where the fluid can enter the dead zones or even continually exchange to the active region as illustrated in Figure 11(b).

The second definition was used in this work since it suits the reality in the configuration of bare

tundish with and without FCD where the fluid with a long residence time is characterized by

the long tail in the RTD curve.

12

Figure 11 Schematic Model of Dead Volume as a stagnant fluid [19]

2.7.6 Dead Volume Fraction

According to the two definitions of dead volume, there are two different formula to calculate this flow in the tundish. First, the fraction of the dead volume can be calculated based on the difference of theoretical residence time and the mean residence time as expressed in Equation 12.

𝑉𝑑

𝑉

= 1 − 𝜃

Equation 12 It should be stated that the 𝜃

used in calculating the fraction of dead volume applies within the range from 𝜃 = 0 to 𝜃 = 2 and this formula is shown in Equation 13.

𝜃

=

∑2 𝑐𝑖 𝑖=0

𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 13 However, according to the second definition, some portion of fluid moves very slowly

so that the flow rate through the active region must be considered in the calculation. Therefore Equation 12 is modified into Equation 14.

𝑉𝑑

𝑉

= 1 −

𝑄

𝑥 𝜃

Equation 14 The term

𝑄

is the flow rate through the active region and it represents the area under the RTD curve up to 𝜃 = 2 as shown in Figure 12. Meanwhile, the flow rate through the dead region (

𝑄

) is defined as the area under the curve after 𝜃 = 2. Thus, if the fluid in dead zones is completely stagnant, the fraction of

𝑄

becomes 0 so that the formula in Equation 12 is valid.

The formula also indicates that the smaller flow rate through the dead region is preferable because it can reduce the fraction of dead zone in the tundish.

The total fraction of tracer stays in the tundish from t=0 to ∞ should be equal to 1. This represents the total area under the curve which is always unity [11, 18] and this relationship is expressed in Equation 15.

∫ 𝐶(𝑡)𝑑𝑡

= 1 Equation 15

13 Therefore, the term

𝑄

can be determined based on Equation 16.

𝑄_𝑎

𝑄 = 𝑡𝑜𝑡𝑎𝑙 𝑎𝑟𝑒𝑎 𝑢𝑛𝑑𝑒𝑟 𝑡ℎ𝑒 𝑐𝑢𝑟𝑣𝑒 𝑢𝑝 𝑡𝑜 (𝜃 = 2) 𝑄_𝑎

𝑄

=

∑^∞_𝑖=0𝑐_𝑖∆𝜃

Equation 16

Figure 12 Representation of flow rate through active and dead region in RTD curve [19]

2.7.7 Well-Mixed Volume Fraction

From the Combined Model theory, the well-mixed volume can be determined using Equation 17.

𝑉𝑚 𝑉

=

𝐶𝑚𝑎𝑥

Equation 17 However, this formula cannot be used directly in the current case as the actual C-curve is not ideal as in the Combined Model. It is indicated by the presence of the dead volume and a broader peak which represents a dispersed plug flow in the tundish. Therefore the fraction of well-mixed volume is simply defined as the remaining fluid beside the dead and dispersed plug flow as shown in Equation 18.

𝑉𝑚

𝑉

= 1 −

𝑉

−

𝑉

Equation 18

2.7.8 Multi-strand Tundish Model

All equations of flow characteristic mentioned previously were developed for analyzing the performance of single strand tundish. Hence, another approach must be employed to calculate the flow characteristics in a multi-strand tundish. The formula proposed by Zheng and Zhu [3] was used in this work. In this approach, the flow fraction of each strand initially was calculated using Combined Model formula and then its value was combined with other strands as summarized in Equation 19 to Equation 21. This value is called the overall flow characteristics and the tundish performance between different configurations is evaluated by these variables.

𝑉𝑝 𝑉

=

𝑁

[(

2

) + (

2

) + ⋯ . + (

2

)] Equation 19

14

𝑉_𝑑

𝑉

= 1 −

𝑁

(

𝑄1

𝑥 𝜃

+

𝑄2

𝑥 𝜃

+ ⋯ +

𝑄𝑁

𝑥 𝜃

) Equation 20

𝑉𝑚

𝑉

= 1 −

𝑉

−

𝑉

Equation 21

2.7.9 Strand Similarity

Besides the flow characteristics which become a reference to measure tundish capability to remove inclusion, a similarity among strands also become an essential parameter to measure the performance of multi-strands tundish. It is because the excellent similarity among strands is essential to guarantee an uniform temperature and steel cleanliness.

Unfortunately, researchers have a different way to measure this parameter, such as by comparing the maximum concentration of the inner and outer strand tundish [24] or by calculating the standard deviation of some important variables, i.e., the first time tracer appears at the outlets [25, 26]. However, an approach developed by Zheng and Zhu was used in this work and the formula is presented by Equation 22 [5]. The N in the equation represents the number of strands, z is the number of instantaneous time 𝑡

and 𝐶 ̅ (𝑡

) is the average of dimensionless concentration at time 𝑡

.

𝑆

̅̅̅ =

∑ {

𝑁 2

𝑗=1 ]

𝑁−1

}

1/2

𝑍𝑗=1

Equation 22

In the other word, 𝑆 ̅̅̅ is the strand similarity from strand 1 to N which defined as the

average of total standard deviation of the dimensionless concentration from time j=1 to t=z.

Therefore, the lower this value means the deviation of concentration in all strands at every time is small so that it leads to the better strand similarity. This approach is better than others since it considers all of the concentration of every strand in every recorded time in the experiment [5].

2.8 Tundish Water Model Criteria

The utilization of water model to perform RTD experiment is useful due to its reliability to represent the actual industrial tundish. However, it is obvious that several parameters are different and it is impossible to create similar condition as the reality. Hence, the model has to fulfill a number of similarity criteria, i.e., geometric, dynamic, kinetic, thermal, and chemical similarity. Since the steel melt is assumed as non-reactive in isothermal condition, the thermal and chemical similarity consideration is excluded in this study.

2.8.1 Geometric Similarity

The tundish size has a great effect on the capability of tundish to remove inclusions

since the melt can stay longer in a bigger tundish [1]. Besides, the design parameter such as

FCD location in the tundish also influences the tundish performance [27]. Thus, the changing

of geometry may lead to a different result in the physical modelling. In order to minimize this

effect, each linear dimensions in the tundish have to be reduced by the same scale factor 𝜆 as

shown in Equation 23. Lm is the dimension of model and Lp is the prototype or real industrial

15 tundish dimension, so that λ may vary from 0 to 1. A reduced scale model has a benefit concerning space and for experiment setting. Nevertheless, a reduced scale model has the drawback as it is impossible to achieve a similar Froude and Reynold number in dynamic similarity consideration [15]. The more detail explanation of the problem to achieve a similar Froude and Reynold Number in dynamic similarity is explained in section 2.8.2.

𝜆 =

𝐿𝑝

Equation 23

2.8.2 Dynamic Similarity Consideration

The dynamic similarity between industrial tundish and water model entails that the presence of all forces act on steel melt must be changed by a certain scale in the water model.

In the tundish, the considered forces are the forces related to fluid flow behavior such as the gravity, the viscosity and the inertial forces. In fluid mechanics, the flow behavior could be expressed in the form of a ratio between the two forces which is known as the Reynold and Froude number. The Reynold number is a ratio between the inertial to the viscous forces whereas the Froude number represents the ratio of viscous to the gravitational forces. Formulas of those two similarity considerations were shown in Equation 24 and Equation 25 where the Re and Fr is the symbol of Reynold and Froude number respectively, 𝜐 is the velocity of the fluid, 𝜌 is the density of the fluid, 𝐿 is the characteristic length and 𝜇 is the dynamic viscosity of the melt.

𝑅𝑒 =

𝜇

Equation 24 𝐹𝑟 =

𝑔𝐿

Equation 25 In dynamic similarity consideration, the Reynold and Froude numbers are desired to be similar between model and the industrial tundish. For Reynold number similarity, this leads to Equation 26.

(

𝜇

)

𝑚𝑜𝑑𝑒𝑙

= (

𝜇

)

𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑖𝑎𝑙 𝑡𝑢𝑛𝑑𝑖𝑠ℎ

Equation 26 Since the difference between the kinematic viscosity of steel at 1600℃ and water at room temperature as shown in Table 1 is insignificant, the Equation 26 can be simplified into Equation 27. By using a similar step in Froude number similarity consideration, the relationship between model and industrial tundish is expressed in Equation 28.

Table 1 Comparison of kinematic viscosity of steel and water [28]

𝜐

≈

𝜆

𝜐

Equation 27 𝜐

≈ √𝜆 𝜐

Equation 28 Based on Equation 27 and Equation 28, it is clear that the similarity of Froude and Reynold number can only be achieved in a full-scale model as 𝜆 = 1. It is impossible to satisfy

Property Steel at 1600 ℃ Water at 20 ℃

Dynamic Viscosity, 𝜇 (kg m^-1 s^-1) 0.0064 0.001

Density,𝜌 (kg m^-3) 7014 1000

Kinematic Viscosity, ^𝜇

𝜌 (m² s^-1) 1. 10^-6 0.913. 10^-6

16 both number in a reduced scale model, because when the Reynold number similarity is achieved, the velocity becomes higher in the model and the opposite result occurs if the Froude number similarity becomes the reference. That is the reason why researchers have a different opinion regarding which number should be used when developing a water model.

In relation with this problem, a comparison study of the importance of Reynold and Froude number on fluid flow behavior in the tundish has been carried out by Sahai and Emi.

They stated that actually neither of the Reynold number nor Froude number similarity is necessary to predict the fluid flow behavior in water model experiment. However, the Froude number similarity is useful for the prediction of the inclusion removal in the tundish [15].

The reason of why Reynold number similarity is not important in water model experiment is because the turbulent flow becomes the dominant fluid flow in the tundish. In turbulent flow, a combination of molecular and turbulent viscosity is used to describe the diffusive momentum transfer capability. The turbulent viscosity is much important due to the exchange happened by eddies length are much dominant than by the molecular exchange.

Therefore, a Reynold number similarity is not so important to be fulfilled as in turbulent flow the inertia flow is much higher than the viscous forces. The Reynold number similarity is only important to be considered in a laminar flow where the viscous layer becomes the only mechanism of diffusive momentum transfer [15].

The Froude number similarity criteria are also not important for modelling fluid flow since the isothermal condition in water model causes the gravity forces do not affect the melt flow. Thus it is unnecessary to achieve a Froude number similarity for modelling fluid flow.

However, the Froude number similarity can provide the similarity of important aspect related with inclusion removal, such as the inclusion size and inclusion density as can be seen in Figure 13 [15]. Hence, the water model with Froude Number similarity can describe the similar behavior of floatation of inclusion or agglomeration. This is the reason of why the Froude number was chosen as the dynamic similarity consideration in this study.

Figure 13 Relationship between inclusion size (left) and density (right) in industrial tundish and water model [15]

2.9 Multi-strand tundish problem

The multi-strand tundish is commonly used in a billet steel production, where the number of strands varies from three to ten. This high number of strands may lead to a significant quality dissimilarity among strands. Since the inclusion removal depends on the residence times, the further strand from the ladle shroud tend to has more inclusions and lower temperature than the strand near to the ladle shroud. Thus, the inclusion removal capability as well as the strand similarity are essential to be investigated.

To overcome the problem, the melt flow inside tundish is usually be adjusted by installing

17 several flow control devices (FCD) such as dams, weirds, or baffle wall. However, the proper FCD varies for different tundish geometry and design. This means the solution of optimum FCD design and its position in the tundish is very specific for each case. In addition, the multi- strand tundish typically has a limited working space and the refractory material may contaminate the steel. Another weakness of FCD is it is vulnerable to wear for long-term usage.

It affects the productivity, quality and also the total cost of its implementation. The last, the FCD cannot provide the adjustment of melt flow for the whole process time.

One solution that could overcome those problems is using electromagnetic stirring (EMS) for stirring the melt. This technology has been developed by ABB Metallurgy in Västerås, Sweden. The horizontal stirring created by EMS will mix and adjust the melt flow then it homogenize the composition and temperature. Thus, there is a possibility to replace FCD with EMS. In addition, EMS is also more flexible to be controlled based on the tundish design and the casting process. Despite being more expensive than FCD, EMS has a much longer lifetime that the total cost becomes comparable. However, the comprehensive behavior of the flow generated by EMS has not been understood due to the lack of studies related with EMS in the tundish.

2.10 Electromagnetic Stirring Technology

EMS is a set of tools which can provide mixing in the melt with the help of electromagnetic force generated by coil induction. This technology can be used as a method to adjust flow behavior inside the tundish so that better steel cleanliness can be obtained. The EMS unit consists of four components namely an electromagnetic coil, frequency converter, transformer and a water station as can be seen in Figure 14.

Figure 14 Typical Electromagnetic Stirring System and Components [29]

The basic principle is the current from the converter will flow into the coil. This current then will produce a strengthened electromagnetic field which penetrates into the steel melt.

Furthermore, the electromagnetic field induces a current in the steel melt. As a result, the Lorentz Force will be generated in the steel melt and this becomes the source of stirring force.

Meanwhile, the water station act as a cooling unit for reducing the temperature of iron ore

inside the coil. In this work, the EMS will be replaced by the water pumps in the water model

experiment. This method has not been used previously in the tundish stirring. Even though it

is definitely not similar, however, it can be the good way to model the EMS in water.

18

3 BASIC THEORY OF NUMERICAL MODELLING

A Computational Fluid Dynamics (CFD) can be a method to predict the fluid behavior in the tundish model by solving governing equations related to fluid flow. It can also complement the result of physical experiments since it is allowed us to observe more variables related to fluid flow. However, correct settings such as meshing or turbulence model are required in order to have reliable results,. In this project, a CFD Software ANSYS Fluent 18.2 was used.

Theories behind some setup in this software are elaborated in this section.

3.1 Governing Equations

ANSYS Fluent simulates the fluid flow behavior based on the conservation of mass and momentum equation. Since the tracer as a chemical species is injected during the water model experiment, the conservation of species equation was activated. For simulating particle trap during particle injection, a Discrete Phase Model (DPM) was selected. Meanwhile, the equation of heat transfer was turned due to isothermal assumption. Those equations are solved together during simulation in this work. Firstly, the mass and momentum are iterated until reach a steady state condition. The conservation of species and discrete phase model equation are solved after the tracer or inclusion particle injection in a transient simulation.

3.1.1 Conservation of Mass

The conservation of mass is expressed in Equation 29:

𝜕𝜌

𝜕𝑡

+ 𝛻 ∙ 𝜌𝑣⃗ = 𝑆

Equation 29 The 𝐯 ⃗⃗⃗⃗ represents the fluid velocity, ρ is the density and S

is the mass added to the system from a dispersed phase, which equals to 0 in this work.

3.1.2 Conservation of Momentum

The equation of momentum conservation is expressed in Equation 30.

𝜕

𝜕𝑡

(𝜌𝑣⃗) + 𝛻 ∙ (𝜌𝑣⃗𝑣⃗) = −𝛻𝑝 + 𝜌𝑔⃗ + 𝛻 ∙ 𝜏̿ + 𝐹⃗ Equation 30 This equation is derived from the Newton’s second law where p represents the static pressure, g is the gravity and 𝜏̿ is the stress tensor acts on fluid and 𝐅 ⃗ is the external force. The left side in the equation is the initial force and the right side represents pressure forces, viscous forces and external forces respectively. This equation, together with the mass continuity equation is known as Navier-Stokes equation [30].

3.1.3 Species Transport Conservation

Fluent predicts the mass fraction changing of the species Yi, by solving the equation of convection-diffusion of the solution as shown in Equation 31.

𝜕

𝜕𝑡

(𝜌𝑌

) + 𝛻 ∙ (𝜌𝑣⃗𝑌

) = −𝛻 ∙ 𝐽 ⃗⃗⃗ + 𝑅

+ 𝑆

Equation 31

𝑅

and 𝑆

represent the net rate of production of species i by chemical reaction and by addition

from any other sources respectively. Meanwhile, 𝑱 ⃗⃗⃗ represents the diffusive flux which occurs

19 due to the difference between temperature and concentration in the system.

3.1.4 Discrete Phase Model (DPM)

The Discrete Phase Model (DPM) was used to study the particle trajectory from an injection. In this work, the particle injected represents the alumina inclusion which is dispersed due to turbulent flow. The particle trajectory is calculated by solving the particle force balance equation as expressed in Equation 32.

𝑑𝑈_𝑖^𝑃

𝑑𝑡

= 𝐹

(𝑢

− 𝑢

) + 𝑔

(

𝜌𝑝

) +

𝜌𝑝

Equation 32 The first part in the right side of Equation 32 represents the drag force, while the second and the last are the gravity and external forces [31]. The DPM also consider the particle-wall interactions such as escape, reflect or trap as shown in Figure 15.

Figure 15 Particle Interaction in DPM [31]

3.2 Turbulence Model

The turbulence flow is characterized by a high Reynold number, high degree of instability as well as irregular movements where the mass, momentum and species transport is always changing every time [32]. Because the flow inside tundish always involves the turbulence flow, it needs to be modelled aiming to get more accurate results as in reality.

Figure 16 Turbulence flow structure [32]

There are small and large structures in a turbulence flow as shown in Figure 16. In order to describe the behavior of turbulence in those structures, several mathematical models which solves the energy and dissipation rate in those structures has been developed. In CFD, three approaches can be used to solve the turbulence flow: Reynold Average Navier Stokes (RANS), Large Eddy Simulation (LES) or Direct Numerical Simulation (DNS). The illustration of those approaches is shown in Figure 17. The RANS method, specifically a Realizable 𝑘−∈ model was chosen due to its simplicity.

3.2.1 The Reynold Average Navier Stokes (RANS) model

The basic principle of RANs model is expressed in Equation 34.

𝑈

= 𝑈 ̅ + 𝑈

Equation 33

20

Figure 17 Three difference method of turbulence model [33]

Where the velocity at certain time (𝑼

) is a sum of the average velocity (𝑼 ̅) and a fluctuating component (𝑼

). The RANS Equation then can be obtained by substituting Equation 34 into mass and continuity equation in Navier-Stokes Equation. The results are basically similar as Navier-Stokes equation, but the velocity contains the ensemble time averaging of velocity and the fluctuating component.

3.2.2 The realizable 𝒌−∈ method

The realizable 𝑘−∈ method is an improvement from the standard 𝑘−∈ method, where it satisfies certain mathematical constraint on the Reynold stress tensor. This model was used in this simulation since it is more suitable and more accurate for predicting the flow behavior which involves rotation and recirculation and spreading from inlet jets as occur in this problem [34].

3.2.3 Law of the wall and Near Wall-Treatment

´ The flow near wall experiences differences behavior depending on the distance from the wall. The behavior of fluid flow near wall can be explained by the law of the wall. This law divided fluid near wall into three regions: viscous sublayer, a buffer layer and fully turbulent region. The viscous sublayer is the region closest to the wall where the fluid flow is almost laminar as the viscosity plays an important role in momentum transfer. Meanwhile, the fully turbulent region is the region where the turbulent plays a dominant role in momentum transfer.

The buffer layer is the layer between a viscous and turbulent layer where the fluid flow as a mix between laminar and turbulent flow. Due to this phenomena, a grid setup near the wall needs to be setup properly so that the accurate results are obtained.

There are two ways to setup the mesh around the wall. The first way is by resolving the

viscous sublayer and the second is by using a logarithmic-based wall function. The first method

required a high mesh resolution near the wall whereas the second method used a wall function

which relates the flow variables near the wall to the cell at the wall. In this work, the wall

inflation used for generating a thin mesh near the wall. In addition, the enhanced wall treatment

was used in the turbulence model setting window. This treatment was chosen because it is a

preferred method for predicting the mixing phenomenon [32].

21

4 EXPERIMENT SETUP 4.1 Model selection

An eight-strand industrial tundish was selected as the prototype in this project. Due to the symmetrical consideration, only a half of tundish was manufactured. The model was completed with flow control devices namely baffle walls and turbo-stopper. The tundish model and baffle wall were respectively made from Plexiglas 20 mm and 10 mm of thickness while the turbo-stopper was made from polypropylene. In order to have a flexible experiment setting, the baffle wall and turbo-stopper have a detachable design. The dimension of tundish model and flow control devices are shown in Figure 18 and Figure 19. The water was used to replace 1600℃ of hot steel melt due to a comparable kinematic viscosity as discussed section 2.8.2.

Figure 18 Drawing of tundish water model

Figure 19 drawing of turbo-stopper (left) and baffle wall (right)

22

4.2 Similarity Consideration

The Froude number similarity was employed as a similarity criterion with a scale factor of 40%. Comparison between important properties and parameters in industrial tundish and water model was summarized in Table 2.

Table 2. Comparison of properties and process parameter in industrial tundish and water model

4.3 Experimental Tools design

All experiment tools were designed considering the process parameter mentioned in Table 2. The schematic layout and photo of the whole experiment tools are shown in Figure 20 while the actual photo is presented in Figure 21. The lower water tank and big water tank are put on the floor, while the tundish water model as well as upper water tank are held by tables as the supporting structures.

In order to circulate the water during experiments, all water tanks and tundish model are connected by pipes. The upper and big water tank as well as tundish model are connected to the water tap source by P-11, P-12 and P-13 for a faster water filling. Meanwhile, the water sink route is provided by P-9 and P-10. All crucial connections between tundish and tanks are completed with valve and flow meter so that the net balance of flow rate can be adjusted. The four outlets which represent SEN outlet in industrial tundish completed with conductivity meters. The data transmission system is provided by connecting conductivity meter to the laptop. The water conductivity as the transmitted data is converted into the salt concentration.

Details of some important configurations are discussed in next section.

4.3.1 General Setup : Material and – Scale consideration

The Plexiglas was selected as the material of tundish model because of its transparent property which is useful for the observation of the flow behavior during experiments. The same material also used for the upper water tank as the water level during experiment needs to be controlled at a constant level in order to provide a constant velocity at the inlet. Meanwhile, the other water tanks are made from stainless steel to prevent corrosion. The 40% scale was used because this is the maximum scale that fits in the available room space. The model geometry was kept same as the real industrial tundish to obtain an accurate result.

Parameter Unit Industrial Tundish Water Model

Scale Factor 1 0.4

Density kg/m

7038 998

Kinematic Viscosity m2/s 9.13 E-7 1.006 E-06

Length at base mm 6062.4 2438.8

Bath level mm 850 340

Volume of steel/water at max bath level m

5.94 0.3734

Volumetric Flow rate at the inlet l/h 13648 1380

Mass Flow Rate kg/min 1592.3 23

Theoretical Mean residence time min 26.12 16.23

Froude Number 3.285 E-05 3.29 E-05