Effect of Selective Pellet Loading on Burden Distribution and Blast Furnace Operations

Abstract

Good permeability of the blast furnace bed is of paramount importance for stable operations of the same and to achieve it, controlled burden distribution is the key. While coke is primarily used to maintain permeability of the furnace bed (with the ferrous burden particles being considerably smaller), global calls for reducing CO₂ emissions substantially are pushing blast furnaces to consume lower amounts of the same. Considering this, other burden distribution methods must be explored to maximize permeability of the burden layer. To do so, efforts are made to control pellet dumping in the burden layer in such a way that it occupies more volume in the mid-radial portions away from the walls. Pellets being spherical in shape and having narrower size distributions than sinter offer more inter particle porosity. Discrete element method (DEM) simulations are used to study the effect of the delay in pellet loading time in the ferrous burden (composed of pellets and sinter) on the positioning in the hopper. This positioning is found to subsequently affect the hopper output constitution, which then affects the radial constitution of the ferrous burden layer. Delaying the pellet loading on the ferrous burden is found to delay emptying of the same from the hopper, resulting in more pellets occupying the mid-radial region of ferrous burden layer in the blast furnace. Based on the observations from simulations, trials are taken in blast furnace and relevant parameters are monitored to capture the effect of such a charging practice on operations.

1. Introduction

The blast furnace is the most popular ironmaking process in the world. It takes solid raw materials which are sources of iron ore and carbon and hot blast gas and discharges molten iron alongwith slag from the bottom as well as CO and CO₂ rich gas from the top. The major portion of the iron ore is fed in the form of agglomerates known as pellets and sinter, both of which vary significantly in terms of physical and chemical properties. To utilize the different advantages offered by both, their mixtures are charged inside the furnace as part of the iron-ore carrying burden (which will be hereby referred to as the ferrous burden). The ferrous burden also contains some amount of lump iron ore as obtained from mines, fluxes in the form of limestone, dolomite, etc. Alternate layers of ferrous burden and carbon containing coke are charged into the blast furnace for stable operations.

Continuous descent (and subsequent softening, melting and reduction) of the ferrous burden and flow of hot blast gas upwards is integral to stable blast furnace operations. For this, it is important that the burden is permeable enough for the gas and molten metal to pass through easily, but not too permeable to allow the constituents inside sufficient reaction time. This makes burden distribution very important in the context of achieving stable blast furnace operations, and hence is a huge interest of study among researchers. It is desired that the flow of gas upwards be concentrated more towards the centre of blast furnace to reduce heat load on the walls. Due to this reason, centre coke is also charged in the centre of the blast furnace to maintain good permeability of that region. In the other regions along the radius, alternate layers of coke and ferrous burden are present. Since coke particles are larger than pellets and sinter in the ferrous burden, a thick enough layer of the same will ensure good permeability through that region. However, with increasing focus towards reducing CO₂ emissions, blast furnaces are proactively working towards operating with lower coke rates than before. Consuming lesser coke means coke layer thickness decreases, thus making the region less permeable. If the permeability decreases below the acceptable limit, gas flow and burden descent can be profoundly affected which is detrimental to stability of operations. Hence, it is imperative that alternative methods of improving permeability be explored and implemented.

Most of the modern blast furnaces are equipped with a rotating chute that distributes material circumferentially at desired radial positions with material input from a hopper above. However, material inside a hopper is subject to segregation, which affects the uniformity of the material output. Studies have shown that segregation in the hopper is affected by many parameters like the size distribution of constituent material particles,^{1,2,3,4,5,6,7,8)} hopper geometry,^7,8,9,10,11) shape of constituent particles,^7,12,13,14) etc. Discrete element method (DEM) simulations have shown that interaction parameters between material particles and wall^7,15,16,17) can affect hopper flow as well. These factors contribute to non-uniform material output with time of hopper emptying. If there are more than one type of particles in the mixture, the constitution of the same at hopper output is liable to change with time.^18,19) As a result, the rotating chute input is always different and thus circumferential distribution of material in the blast furnace is non-uniform. Different particles of different sizes will constitute different positions along the radius of the blast furnace, thus dividing the entire layer in zones with differing permeability.

By utilizing the knowledge of varying hopper output with time, the charging of material from hopper to blast furnace can be controlled. In this study, this idea is used to identify suitable charging schemes for maximisation of the pellet concentration in the blast furnace burden layer in the mid-radial position. This is done to improve the permeability of the mid-radial region of the burden, as the centre is already dominated by centre coke and gas flow at the walls needs to be controlled to protect them from large heat losses. Since pellets are nearly spherical in shape, they provide more inter-particle porosity than sinter particles which are more irregularly shaped. There are studies in literature which deal with custom charging of certain burden materials on belt and its effect on hopper loading and burden distribution,^{18,20,21,22,23)} but such studies for pellets are very limited. Effect of selective pellet loading on the hoppers by controlling its charging time w.r.t. the entire ferrous burden batch (from the raw material stock house) and subsequent burden layer formation in the blast furnace is evaluated using DEM simulations and the optimum charging schemes are identified. The implications of selected charging schemes on blast furnace operations are also presented in detail.

2. Materials and Methods

2.1. Discrete Element Method

The discrete element method is a popular tool used to simulate the flow of particulate solids in a given geometry. It is essentially a numerical method which is based on Newton’s second law of motion to calculate accelerations, velocities, and positions of the concerned particles from a starting condition for a desired simulation time in very small steps of the order of 10⁻⁵ s. The motion of particles in the simulation is mainly affected by collisions with other particles or geometries and presence of other forces in the domain, particularly gravity which is the most dominant one. At small size ranges, electrostatic, hydrostatic, magnetic, etc. forces can also profoundly affect particle motion, whose effect can be incorporated in simulations if needed. By tracking the motion and behaviour of every individual particle, the flow behaviour of the entire bulk solid can be studied. The commercial software Altair EDEM is used for simulations, and in that the Hertz-Mindlin no-slip contact model is employed to model the collisions between the particles and particles/geometries.

At any instant, the force balance and torque balance on an i^th particle which is assumed to be in contact with k number of particles (all particles considered as spherical) can be depicted as   

$$m_i\frac{d\overrightarrow{v_i}}{dt}=m_i\overrightarrow{g}+{\displaystyle \sum }_{j=1}^k\left(F_{cn,ij}+F_{dn,ij}\right)\hat{n}+{\displaystyle \sum }_{j=1}^k\left(F_{ct,ij}+F_{dt,ij}\right)\hat{t}$$(1)

  

$$I_i\frac{d\overrightarrow{{\omega }_i}}{dt}={\displaystyle \sum }_{j=1}^k\left(\overrightarrow{T_{t,ij}}+\overrightarrow{T_{r,ij}}\right)$$(2)

$\hat{n}$ and $\hat{t}$ are the unit vectors accounting for contact between the i^th and j^th particles in the normal and tangential directions respectively. F_cn,ij and F_ct,ij represent the elastic components of the normal and tangential forces respectively, whereas F_dn,ij and F_dt,ij represent the damping components of the normal and tangential forces respectively. $m_i\overrightarrow{g}$ is the gravitational force on the i^th particle. $\overrightarrow{T_{t,ij}}$ and $\overrightarrow{T_{r,ij}}$ represent the torques generated due to tangential force and rolling friction respectively. The detailed expressions of all forces and dependent parameters are given in Tables 1 and 2. For more information, the Altair EDEM reference guide can be used.²⁴⁾

Table 1. Detailed equations of different force and torque components used during collisions.

Symbol	Equation
Fcn,ij	$\frac{4}{3}{E}^{*}\sqrt{R^{*}}{\delta }_n{}^{\frac{3}{2}}$
Fdn,ij	$-2\sqrt{\frac{5}{6}}\beta \sqrt{S_n{m}^{*}}{v}_n^{rel}$
Fct,ij	Fct,ij = −Stδt
Fdt,ij	$F_{dt,ij}=-2\sqrt{\frac{5}{6}}\beta \sqrt{S_t{m}^{*}}{v}_t^{rel}$
$\overrightarrow{T_{t,ij}}$	$R_i\hat{n}\times \left(F_{ct,ij}+F_{dt,ij}\right)\hat{t}$
$\overrightarrow{T_{r,ij}}$	$-{\mu }_r{F}_{cn,ij}{R}_i\widehat{{\omega }_i}$

Table 2. Detailed description of parameters used for calculations.

Symbol	Parameter	Equation (if any)
mi	Mass of ith particle
vi	Velocity of ith particle
Ii	Moment of inertia of ith particle
ωi	Angular velocity of ith particle
E*	Equivalent Young’s modulus	${\left(\frac{1-{\upsilon }_i{}^2}{E_i}+\frac{1-{\upsilon }_j{}^2}{E_j}\right)}^{-1}$
Ei	Young’s modulus of ith particle
υi	Poisson’s ratio of ith particle
R*	Equivalent radius	${\left(\frac{1}{R_i}+\frac{1}{R_j}\right)}^{-1}$
Ri	Radius of ith particle
δn	Normal deformation
β	Function of coefficient of restitution	$\frac{lne}{\sqrt{{(lne)}^2+{\pi }^2}}$
e	Coefficient of restitution
m*	Equivalent mass	${\left(\frac{1}{m_i}+\frac{1}{m_j}\right)}^{-1}$
$v_n^{rel}$	Normal component of relative velocity
G*	Equivalent shear modulus	${\left(\frac{2-{\upsilon }_i}{G_i}+\frac{2-{\upsilon }_j}{G_j}\right)}^{-1}$
Gi	Shear modulus of ith particle	$\frac{E_i}{2\left(1+{\upsilon }_i\right)}$
δt	Tangential deformation
$v_t^{rel}$	Tangential component of relative velocity
μr	Coefficient of rolling friction
$\widehat{{\omega }_i}$	Unit angular velocity at point of contact

The tangential force must always be less than μ_s times the normal force, where μ_s is the coefficient of static friction. Material specific parameters like solid density ρ_s, Young’s modulus E, Poisson’s ratio υ, coefficient of restitution e, coefficient of static friction μ_s and coefficient of rolling friction μ_r are given as inputs to Altair EDEM. The latter 3 must be given separate values for each separate contact between constituent materials. Values of these parameters are either determined from experiments or from calibration exercises.²⁵⁾ Size distribution and shape of particles can also be defined according to user. For particles that deviate from the spherical shape, they can be represented using clusters of multiple spheres. In their case, the relevant constituent sphere of a multispherical particle will be considered for contact detection and calculations will take place accordingly.

2.2. Materials

For this study, we consider four types of material, namely sinter, pellet, coke, and centre coke. Sinter and pellet constitute the ferrous burden while coke is charged across the radius of the furnace and centre coke is charged in the centre of the furnace. Coke and centre coke are essentially the same material physically and chemically, but different sized fractions taken from the same coke batch produced in coke plants. The ferrous burden dumped in the blast furnace also contains some amount of lump ore and fluxes, but their proportion in the burden is significantly less compared to pellets and sinter, hence for the sake of simplicity, we consider only sinter and pellets as part of the ferrous burden in our simulations. Size distributions taken for the materials considered are mentioned in the Table 3 below.

Table 3. Size distribution of materials used in simulations.

Material	Size	Mass Percentage
Pellet	6–9 mm	5.02
	9–12.5 mm	68.73
	12.5–16 mm	26.25
Sinter	5–10 mm	27.97
	10–15 mm	28.45
	15–20 mm	29.91
	20–40 mm	13.67
Coke	35–40 mm	10.78
	40–45 mm	16.45
	45–50 mm	72.77
Centre Coke	50–60 mm	77.26
	60–70 mm	13.78
	70–80 mm	8.96

The material specific parameters and interaction parameters are taken from previous studies^26,27) where detailed explanation for determining the same have been given. Pellets are assumed to be spheres, whereas multispherical representations are taken for sinter and coke. Parameters for coke and centre coke are the same. For interactions between coke and other particles, values of interaction parameters are taken as the average of interaction parameters of the individual materials in question (Table 5). It is assumed that entire geometry against which burden material would flow is lined with the same material (ceramic-rubber liner) which is hereby referred to as the wall. Interactions of the solid materials with walls are also listed in Table 4.

Table 4. DEM input parameters for different materials involved in simulations.

Input Parameter	Sinter	Pellet	Coke	Wall
Solids Density ρs (kg/m3)	3050	3700	985	7900
Young’s modulus E (MPa)	250	250	25	25000
Poisson’s ratio υ	0.25	0.25	0.25	0.25
Particle-Particle Coefficient of Restitution e	0.4	0.7	0.6	–
Particle-Particle Coefficient of Static Friction μs	0.7	0.49	0.55	–
Particle-Particle Coefficient of Rolling Friction μr	0.08	0.06	0.1	–

Table 5. DEM input interaction parameters for interactions between different materials involved in simulations.

Input Parameter	Sinter-Pellet	Pellet-Coke	Coke-Sinter	Sinter-Wall	Pellet-Wall	Coke-Wall
Particle-Particle Coefficient of Restitution e	0.55	0.65	0.5	0.3	0.5	0.4
Particle-Particle Coefficient of Static Friction μs	0.6	0.52	0.64	0.72	0.9	0.9
Particle-Particle Coefficient of Rolling Friction μr	0.07	0.08	0.09	0.08	0.06	0.2

2.3. Simulation Setup

All solid burden materials for blast furnace come from upstream units and are first collected in individual material bunkers in the stock house area. Whenever required, the discharge gate of a particular bunker is opened and fixed weight of material is discharged onto a conveyor belt, which transports the material to the blast furnace hoppers where entire batch of material is first stored and then discharged onto the rotating chute and into the blast furnace bed top. The path between stock house bunkers and hopper is usually laden with a few transfer points where possible direction change of material stream takes place. Coke, surface coke, sinter, and pellets each have their own storage bunkers in the stock house. Coke and centre coke both are charged separately. To prepare ferrous burden, the discharge gate of both pellets and sinter bunkers are opened within the same time frame. Since in this study, sinter percentage in burden is more than pellet percentage, the time between opening and closing the sinter bunker gate is essentially the total time of ferrous burden charging, as the pellets are charged within that time only on the sinter layer (or vice versa, depending on how the bunkers are placed along the direction of conveyor belt motion).

For our simulations, we assume that the total time of ferrous burden charging on belt (or time to charge ferrous burden in hopper) is 115 s. In between, pellets are charged for 75 s. Total mass of the ferrous burden taken is 106 tons, out of which 61.48 tons is sinter and 44.52 tons is pellets (pellet percentage in ferrous burden = 42%). The variable is the start of charging time of pellets w.r.t. start of charging time of sinter, which will be henceforth referred to as the delay. A delay of 0 s means pellets and sinter both are present in the ferrous burden charge head. The delay values are chosen in such a way that there is a reasonable buffer between pellet end charging time and sinter end charging time to prevent pure pellet loading on the belt which makes it prone to spillage. Also, delaying pellet charging further would increase the total time of ferrous burden charging in the hoppers, reducing their daily availability and affecting production. A schematic of the ferrous burden charge on the feeding conveyor belt is shown in Fig. 1.

Fig. 1.

Schematic of delayed pellet charging in the ferrous burden with sinter. Delay of 10 s is illustrated here. (Online version in color.)

Assuming that the material transport on conveyor belt is plug flow, and that material would get decently mixed at the transfer points before entering the hopper, the simulations are started from just above the burden hopper before entering the blast furnace. A particle factory is placed above the hopper entry to generate particles according to a given delay configuration and calculated mass flow rates. The total simulation is done in three parts, i) filling in of hopper, ii) emptying of hopper and iii) formation of burden layer in blast furnace (Fig. 2). Geometries are taken as per scale. All sets of simulations are done for all delays studied. For feasible simulation times and memory savings, particle sizes are scaled up 2 times. This leads to having approximately 8 million particles constituting the ferrous burden for the given tonnage in the simulation domain, whereas the same amount for coke and centre coke is restricted to a few hundred thousand. The entire height of the domain from hopper entry to blast furnace top is about 24 m, and the diameter at blast furnace top is 9.5 m.

Fig. 2.

Stages of burden material flow simulations in blast furnace. (Online version in color.)

For hopper filling in simulations, there is a gate at the hopper exit which is kept closed while the factory fills in particles at the specified rates for a fixed amount of time. Once the hopper is filled and material inside is stable, the gate is opened such that the entire ferrous burden material is emptied within 77–78 s of opening. The time taken to empty the hopper is roughly the time for which material will be fed to rotating chute accounting for little to no accumulation in between. The chute rotates at a speed of 8 rpm about the blast furnace centre axis (meaning it makes one material ring in 7.5 s). 10 rings each of coke and ferrous burden are made while 3 rings of centre coke are made. For forming the burden layer, first coke layer is dumped, followed by centre coke and then ferrous burden. Charging sequence is kept the same for all delays studied (details in Table 6). For ferrous burden and coke, charging starts from the walls and gradually moves towards the mid-radial region, the charging positions being specified by ring numbers which are indicative of the chute angles (ring 11 being the most outward ring near the wall and ring 1 being the innermost circle, reserved for centre coke). For ferrous burden, to save further simulation time, charging in the BF layer is done in a 90⁰ sector (accounting for 1/4th of the blast furnace area) (Fig. 3). The area is marked by two walls placed along the radius of the blast furnace, both of which are perpendicular to each other. Particle information at the hopper outlet during emptying is captured and stored and used to create another custom particle factory to feed the rotating chute. A custom 2D profile of actual blast furnace burden layer top is taken and rotated in one full circle around the centre axis to create a base layer on which the burden materials are charged and settled in the order mentioned before (Fig. 4).

Table 6. Charging sequence details with number of rings for each burden material. Chute angle values are w.r.t. the vertical.

Charging Material	Number of rings formed in designated ring number (chute angle in degrees in bracket)
	11 (50.2)	10 (48.2)	9 (45.7)	8 (43.1)	7 (40)	6 (37.1)	5 (34.1)	4 (30.9)	3 (27.8)	2 (23.7)	1 (12)
Coke	2	2	2	2	1	1	0	0	0	0	0
Centre Coke	0	0	0	0	0	0	0	0	0	0	3
Ferrous Burden	0	0	0	3	3	2	2	0	0	0	0

Fig. 3.

Simulation of ferrous burden charging in the blast furnace to form burden layer. (Online version in color.)

Fig. 4.

Base layer created on which burden material is charged. (Online version in color.)

3. Results and Discussion

Observations regarding the effect of the delay of pellet charging on ferrous burden loading inside the hopper and the subsequent hopper output constitution and deposition in the blast furnace burden layer are recorded in detail. Simulations from hopper filling to burden layer formation for coke and centre coke have been done, but any observations regarding them shall not be discussed in this study.

3.1. Hopper Loading

Effect of delaying pellet charging is clearly observed on the hopper loading. Figure 5 shows the inner material distribution of the ferrous burden after it has been filled in the hopper for various delay times given. For visualisation purposes, pellet particles are coloured blue and sinter particles are coloured orange. For a delay of 0 s, there is a mixed layer of pellets and sinter upto a certain height after which there is only sinter which is due to the pure sinter layer which remains as pellet charging is finished earlier. However, after introducing delays, the pellets are sandwiched (alongwith some sinter) between layers of pure sinter, as per the configuration that was shown in Fig. 1. With increasing delays, the lowermost pure sinter layer size increases and the topmost pure sinter layer size decreases. Pure pellet layers are not formed due to reasons mentioned before.

Fig. 5.

Inner material distribution of ferrous burden (blue particles representing pellets and orange particles representing sinter) for delays of (a) 0 s, (b) 5 s, (c) 10 s, (d) 15 s. (Online version in color.)

Apart from individual material positioning, it is important to see the arrangement of larger and smaller particles depositing in the hopper. Segregation of pellets from sinter is not found to be significant. As shown in Fig. 6, we consider a part of the hopper spanning 3 m in height for analysing radial and depth-wise size segregation within materials. This part holds up most of the material. Geometrically, the cross-section at each depth is a circle. In the part considered, the centre of the circle varies linearly with depth. The hopper is divided into 12 equally thick parts in the z-direction (each part being 250 mm thick) and one such part is then divided into 5 equally thick parts in the radial direction (each part spanning 0.2R mm thick, where R is the local radius determined from the relationship of the circle centre locus with depth). Hence particles constituting the first region spanning from 0 to 0.2R are nearer to the centre and those located within a distance of 0.8R to R are nearer to the walls. The weighted average diameter trends for both sinter and pellet in each region is shown in Fig. 6. For both materials, the weighted average diameter is the largest near the walls at all depths. This shows that after landing, larger (and thus heavier) particles tend to roll away as much as possible and deposit more at the walls. Segregation in case of pellets is rather subdued than in that for sinter, the primary reason being the narrow size distribution used for the same in the simulations.

Fig. 6.

Radial and depth-wise size segregation analysis in (a) marked region of hopper spanning 3 m in height, where centre of circular cross-section (and its radius) at each depth varies linearly with the depth, radial trends of weighted average diameter of (b) pellets and (c) sinter for regions at different depths. (Online version in color.)

3.2. Hopper Discharging

Any level of control on the hopper output material depends on the type of flow that happens within the hopper. Material discharge flow from hopper can be divided into mainly two types of flow, namely mass flow where first material filled is the first material to get out, and funnel flow, where material above the opening is discharged first followed by material near the walls. For our purpose, mass flow would be the most ideal outcome, but studies have shown that the type of hopper used in simulation mainly results in funnel flow.^28,29,30) We use the profile shown in Fig. 5(a) and colour the particles (irrespective of whether they are pellets or sinter) according to the time in which they exit the hopper. From the profile shown in Fig. 7, it is clear that the material directly above the hopper exit is the first to be discharged followed by material near the walls, indicating strong funnel flow. It is expected that for all burden materials in consideration, the flow nature will be the same. Figure 8 shows a virtual sensor kept just outside the hopper exit which can record all particle level information at different times, and it is used to measure the weighted average diameter of pellets and sinter output from hopper with time. Since particle sizes are doubled in the simulations, the sizes reported are divided by 2. For both materials, average diameter remains unchanged and increases towards the later stages of emptying.^3,29) This stems from our observations during hopper filling in where larger sized particles roll outwards and occupy the walls. Since material near the walls is the last to exit the hopper, the particle weighted average diameter at the hopper exit increases with discharge time. When larger sized particles are discharged towards the end of hopper emptying, correspondingly the rotating chute charges material at that time in the mid-radial region, thus it may aid improving permeability in that region.

Fig. 7.

Hopper emptying profile for a ferrous burden batch. (Online version in color.)

Fig. 8.

(a) Virtual sensor at hopper exit. (b) Pellet and (c) sinter weighted average diameter trends during hopper emptying. (Online version in color.)

For periods of hopper emptying when pellet concentration in the output is maximum, the chute can be correspondingly placed over mid-radial positions to maximise pellet deposition in those areas. Figure 9 shows the hopper output trends for pellet and sinter mass flow rates. For all delays given, there are times before t = 31 s of hopper discharge gate opening where continuously the pellet concentration in the hopper output is more than that of sinter. These periods shrink further with increasing delays, as pellet exit is also delayed, thus pushing further the time required for pellets to dominate hopper output. However, irrespective of the delay given, pellet domination in the hopper output ends around t = 31 s. As shown in Fig. 5, effect of delay on the lower part of hopper where initial layer of pure sinter is present is more evident than in the upper part of the hopper. Since cross-sectional area of lower part of hopper is quite less than that of the upper part, any shift in the interface between pure sinter layer and mixed layer will be more prominent there over that in the upper part of the layer. Since material follows funnel flow, the time by which material at the top interface starts moving is almost the same for all delays as there is very little difference in the top interface position for all cases.

Fig. 9.

Pellet and sinter mass flow trends during hopper emptying for different delays. Blue regions signify periods when pellet concentration in output is more than that of sinter. (Online version in color.)

For the periods of pellet domination at hopper output (till t = 31 s), correspondingly the chute makes the first 4 rings of ferrous burden which are near the walls. If more pellets are present in the ferrous burden during those times, there will be less of them left to occupy positions in the mid-radial region during later stages of charging. Hence, variation in pellet amount inside hopper with emptying time needs to be looked at. Figure 10 shows clear impact of delay on the pellet amount left inside hopper with time of emptying. Delaying the pellet filling in the hopper is seen to delay the pellet loading in the output as well, but the magnitudes of those delays is not like that of delays during hopper filling. For a delay of 0 s, pellets are first observed in the hopper output within 0.2 s of opening the discharge gate, but for a delay of 15 s, pellets take about 3.7 s from gate opening to appear in the output. Trends for all delays are visibly different but they overlap around t = 69 s. This shows that delaying pellet charging during filling in can push pellets in the output further back, thus allowing more of them to go in the mid-radial region while charging in the blast furnace.

Fig. 10.

Hopper pellet concentration trends during hopper emptying for different delays. (Online version in color.)

3.3. Burden Layer Formation

As discussed earlier, ferrous burden is charged in the blast furnace after charging required amount of coke and centre coke in the same. Ferrous burden is charged in the form of 10 rings at different radial positions, which accounts for a total charging time of 75 s. There is a little amount of material that leaves the hopper after 75 s which is not considered in the simulation. Number of rings formed for each material is provided in table no. 6. To view the progression of ferrous burden layer formation on the coke layer and centre coke layer, we consider a radial slice of the system such that the target sector of 90⁰ is divided in two equal halves. For the case of 0 s, it is observed that ferrous burden presence in the mid-radial region is felt from the 3rd ring onwards (after 15 s of hopper gate opening), following which material goes on depositing in that region only. During the 4th and 5th ring charging, some coke is also displaced from the coke layer and transported towards the centre, displaying the phenomena of coke push (Fig. 11). For increasing delays, a small pellet free layer is formed at the walls during the 1st ring charging, which is applicable only for our chosen geometry. For some delay values, pellet supply during the first 7.5 s (time for 1st ring formation) is non-uniform, which means non-uniform circumferential distribution of ferrous burden is a possibility.

Fig. 11.

Radial profile of ferrous burden layer formation (a) before starting charging (b) after 1st ring (c) after 2nd ring (d) after 3rd ring (e) after 4th ring (f) after 5th ring (g) after 6th ring (h) after 7th ring (i) after 8th ring (j) after 9th ring (k) after 10th ring. Black particles represent coke, grey particles centre coke, and maroon particles ferrous burden (sinter and pellets). (Online version in color.)

The distribution of sinter and pellets radially in the layer is determined by considering only those particles in the middle 70⁰ of the 90⁰ sector in which burden is charged. The particles in the remaining two sectors of 10⁰ are not taken as wall effects may be prevalent there. The domain is divided in 11 regions in the radial direction in such a way that all regions have the same cross-sectional area at the same depth (Table 7). It is very well seen that the pellet concentration in the regions between r = 2025 mm and r = 3508 mm increases with increase in delay. These regions correspond to the mid-radial region. Conversely, pellet concentrations at the walls (r = 3789 mm to r = 4750 mm) decreases with increase in delay (Fig. 12(a)). For sinter, with increasing delays, concentrations at walls and mid-radial region increase and decrease respectively (Fig. 12(b)). The effect of increasing pellet concentration in the mid-radial region is also visible on the porosity in that region. Sample space taken to calculate the porosity of layer is shown in Fig. 13. Care is taken not to include coke particles in the same. Increase in pellet amount does increase the porosity of the region, while wall porosity conversely decreases as it is less populated with pellets, as seen from Table 8. This suggests that permeability of blast furnace gases through the mid-radial region can be eased, while the walls become less permeable and might see reduction in heat losses through the same.

Table 7. Region-wise demarcation in the radial direction to determine distribution of pellets and sinter.

Region	Inner Radius (mm)	Outer Radius (mm)
R1	0	1432
R2	1432	2025
R3	2025	2480
R4	2480	2864
R5	2864	3202
R6	3202	3508
R7	3508	3789
R8	3789	4051
R9	4051	4296
R10	4296	4529
R11	4529	4750

Fig. 12.

Percentage mass distribution within regions in the burden layer described in Table 7 at different delays for (a) pellets and (b) sinter. (Online version in color.)

Fig. 13.

Sample space considered for computing porosity for (a) mid-radial region and (b) wall. (Online version in color.)

Table 8. Porosity of ferrous burden at mid-radial and wall regions for different delays.

Delay (s)	Mid-radial Porosity (%)	Wall Porosity (%)
0	40.36	37.96
5	40.75	37.43
10	40.99	37.09
15	41.12	36.82

3.4. Trial in Blast Furnace and Effect on Operations

A continuous trial at a blast furnace was taken while keeping 10 s of delay in pellet charging of ferrous burden. The observations were compared later with periods where identical furnace conditions were maintained before the trial. In those periods, there was no delay given (i.e. 0 s delay). For isolating effects of delay discharge, charging sequence of burden materials was kept same and it was ensured that there was not much difference in the burden batch weights used. Pellet percentages in the ferrous burden varied between 40–43%. Furnace was operated with minimal deviations in coke rate as well as coal rate and hot blast gas volume injected. This was important as significant changes in the coke rate and blowing parameters profoundly affect the permeability of the furnace, which would hamper determining effects of delay of pellet charging.

Effect of delaying pellet discharge is clear on the upper blast furnace wall temperatures, heat losses from the walls and central working index (CWI) of the blast furnace. All data shown is the daily averaged data and it has been normalised such that the lowest point in the respective data set corresponds to 0 while the highest to 1. Increasing delay of pellet charging in ferrous burden has led to reduced wall temperatures in the upper part of blast furnace (Fig. 14(a)) which can be attributed to increased sinter accumulation at the walls resulting in decreased permeability in those regions. Due to this, overall heat losses from the blast furnace walls have also gone down (Fig. 14(b)). Decreased wall permeability has seemingly corresponded to improvement in the permeability near the centre of the blast furnace which is reflected in the central working index (CWI), an indicator of the central working nature of the blast furnace (Fig. 14(c)). There are functional above burden probes on which temperature detecting sensors are strategically placed from furnace wall to centre. According to their locations, the sensors are divided into central, mid-radial and peripheral regions, with the central region being above the BF centre. The average temperature of the sensors in the central region divided by the average temperature of all sensors is the central working index (CWI). This shows that hot blast gas is now flowing better from the centre rather than the walls. The trial has shown that pellet delayed discharge can be used as a method to control the wall temperatures.

Fig. 14.

Effect of pellet delay discharge on (a) blast furnace upper wall temperature (b) heat loss through walls and (c) central working index (CWI). Dashed line represents time when delay was increased from 0 to 10 s. (Online version in color.)

While the trial did throw up very encouraging results, the delay values need to be chosen very carefully depending on the degree of control needed in the wall temperatures. Very high temperatures are obviously to be avoided to protect the walls from thermally induced damages, but very low temperatures are also not desired as that may result in low utilization of materials near the walls and loss of production. Multiple factors contribute to changes in wall temperatures, which are normally controlled by changing burden charging schemes, coke rates, etc. Pellet delay discharge can be used in conjunction with these methods to achieve required wall temperatures.

4. Conclusion

Permeability of the burden bed inside the blast furnace is important to maintain stable operations, and to achieve that, distribution of the burden materials must be controlled. The iron ore carrying ferrous burden is usually a mixture of different types of materials, pellets and sinter being the major ones, and with major impetus on reducing greenhouse gas emissions from the blast furnace, it falls upon the existing burden materials to be utilised efficiently. In this study, effect of selective loading practices of pellets in the ferrous burden and the subsequent process of material filling in and discharging from the hopper to settling in the burden layer has been discussed in detail. Keeping the same batch weights of materials, effect of the delay time of pellet discharge from initial sinter discharge is seen. Owing to the shape of the hopper, strong funnel flow of material is observed and segregation in the material both in terms of size (for both pellets and sinter) and constitution in the output is seen. Delaying the pellet discharge w.r.t. the ferrous burden is also shown to affect hopper output constitution of the ferrous burden, with more pellets being pushed back during discharging time at higher delays. As ferrous burden charging begins from the wall and then concludes in the mid-radial region, the pushed back pellets eventually occupy more space in the mid-radial regions of the blast furnace with the wall relatively getting relieved of pellets, resulting in increasing and decreasing trends in porosity for the mid-radial region and wall respectively with increasing delays. Trial at higher delay has shown that wall temperatures reduce and central working capability of the furnace improves, indicating increased resistance at the walls possibly due to more pellets being present in the mid-radial region of the furnace which improves permeability there. This practice along with other procedures shows good promise to help achieve an additional degree of control over furnace wall heat losses.

References

• 1)   Y.  Kajiwara,  T.  Jimbo,  T.  Joko,  Y.  Aminaga and  T.  Inada: Trans. Iron Steel Inst. Jpn., 24 (1984), 799. https://doi.org/10.2355/isijinternational1966.24.799
• 2)   N.  Standish: Powder Technol., 45 (1985), 43. https://doi.org/10.1016/0032-5910(85)85059-2
• 3)   Y.  Aminaga,  Y.  Kajiwara,  T.  Inada,  T.  Tanaka and  K. I.  Sato: Trans. Iron Steel Inst. Jpn., 27 (1987), 851. https://doi.org/10.2355/isijinternational1966.27.851
• 4)   Y.  Yu and  H.  Saxén: Chem. Eng. Sci., 65 (2010), 5237. https://doi.org/10.1016/j.ces.2010.06.025
• 5)   M.  Kondo,  H.  Kosakabashi,  K.  Okabe,  H.  Marushima,  H.  Takahashi and  J.  Kurihara: Tetsu-to-Hagané, 65 (1979), S593 (in Japanese). https://doi.org/10.2355/tetsutohagane1955.65.11_S588
• 6)   S.  Matsuzaki and  Y.  Taguchi: Tetsu-to-Hagané, 88 (2002), 823 (in Japanese). https://doi.org/10.2355/tetsutohagane1955.88.12_823
• 7)   W. R.  Ketterhagen,  J. S.  Curtis,  C. R.  Wassgren and  B. C.  Hancock: Powder Technol., 179 (2008), 126. https://doi.org/10.1016/j.powtec.2007.06.023
• 8)   P.  Artega and  U.  Tüzün: Chem. Eng. Sci., 45 (1990), 205. https://doi.org/10.1016/0009-2509(90)87093-8
• 9)   T. J.  Goda and  F.  Ebert: Powder Technol., 158 (2005), 58. https://doi.org/10.1016/j.powtec.2005.04.019
• 10)   J.  Wan,  F.  Wang,  G.  Yang,  S.  Zhang,  M.  Wang,  P.  Lin and  L.  Yang: Powder Technol., 335 (2018), 147. https://doi.org/10.1016/j.powtec.2018.03.041
• 11)   S.  Jung and  W.  Chung: ISIJ Int., 41 (2001), 1324. https://doi.org/10.2355/isijinternational.41.1324
• 12)   P. W.  Cleary and  M. L.  Sawley: Appl. Math. Model., 26 (2002), 89. https://doi.org/10.1016/S0307-904X(01)00050-6
• 13)   C. M.  Wensrich,  A.  Katterfeld and  D.  Sugo: Granul. Matter, 16 (2014), 327. https://doi.org/10.1007/s10035-013-0465-1
• 14)   D.  Höhner,  S.  Wirtz and  V.  Scherer: Powder Technol., 226 (2012), 16. https://doi.org/10.1016/j.powtec.2012.03.041
• 15)   C.  González-Montellano,  Á.  Ramírez,  E.  Gallego and  F.  Ayuga: Chem. Eng. Sci., 66 (2011), 5116. https://doi.org/10.1016/j.ces.2011.07.009
• 16)   S. M.  Derakhshani,  D. L.  Schott and  G.  Lodewijks: Powder Technol., 269 (2015), 127. https://doi.org/10.1016/j.powtec.2014.08.072
• 17)   D.  Markauskas and  R.  Kačianauskas: Granul. Matter, 13 (2011), 143. https://doi.org/10.1007/s10035-010-0196-5
• 18)   Y.  Matsui,  A.  Sato,  T.  Oyama,  T.  Matsuo,  S.  Kitayama and  R.  Ono: ISIJ Int., 43 (2003), 166. https://doi.org/10.2355/isijinternational.43.166
• 19)   W. X.  Xu,  S. S.  Cheng,  Q.  Niu,  W.  Hu and  J. W.  Bang: Metall. Res. Technol., 116 (2019), 314. https://doi.org/10.1051/metal/2018099
• 20)   Y.  Kashihara,  Y.  Iwai,  N.  Ishiwata,  N.  Oyama,  H.  Matsuno,  H.  Horikoshi,  K.  Yamamoto and  M.  Kuwabara: ISIJ Int., 57 (2017), 665. https://doi.org/10.2355/isijinternational.ISIJINT-2016-613
• 21)   A.  Murao,  Y.  Kashihara,  S.  Watakabe and  M.  Sato: ISIJ Int., 51 (2011), 1360. https://doi.org/10.2355/isijinternational.51.1360
• 22)   S.  Watakabe,  K.  Takeda,  H.  Nishimura,  S.  Goto,  N.  Nishimura,  T.  Uchida and  M.  Kiguchi: ISIJ Int., 46 (2006), 513. https://doi.org/10.2355/isijinternational.46.513
• 23)   A.  Murao,  Y.  Kashihara,  N.  Oyama,  M.  Sato,  S.  Watakabe,  K.  Yamamoto and  Y.  Fukumoto: ISIJ Int., 55 (2015), 1172. https://doi.org/10.2355/isijinternational.55.1172
• 24)  Altair: Introduction to Altair EDEM and Physics models, https://www.altair.com/resource/introduction-to-altair-edem-and-physics-models, (accessed 2022-07-02).
• 25)   C. J.  Coetzee: Powder Technol., 310 (2017), 104. https://doi.org/10.1016/j.powtec.2017.01.015
• 26)   A.  Chakrabarty,  R.  Biswas,  S.  Basu and  S.  Nag: Adv. Powder Technol., 33 (2022), 103358. https://doi.org/10.1016/j.apt.2021.11.010
• 27)   A.  Chakrabarty,  S.  Basu,  S.  Nag,  U.  Ghosh and  M.  Patra: ISIJ Int., 61 (2021), 782. https://doi.org/10.2355/isijinternational.ISIJINT-2020-498
• 28)   W.  Xu,  S.  Cheng,  Q.  Niu and  G.  Zhao: ISIJ Int., 57 (2017), 1173. https://doi.org/10.2355/isijinternational.ISIJINT-2017-003
• 29)   J.  Qiu,  D.  Ju,  J.  Zhang and  Y.  Xu: Powder Technol., 314 (2017), 218. https://doi.org/10.1016/j.powtec.2017.02.020
• 30)   Y.  Kashihara,  Y.  Morikawa,  T.  Sato,  N.  Ishiwata and  M.  Sato: ISIJ Int., 55 (2015), 1165. https://doi.org/10.2355/isijinternational.55.1165