2024 年 64 巻 10 号 p. 1517-1527
The increase of hydrogen usage in a blast furnace is expected to affect the reduction degradation of ferrous burden materials and influence the gas permeability inside the furnace. Previous studies show a disagreement on the effect of H2 on reduction degradation, with the extent of degradation depending on the H2 content and type of ferrous burden materials. In this study, the reduction degradation of sinter, lump, and pellet was compared using the reduction degradation test under different gas mixtures containing CO and H2, covering the gas composition of conventional and H2 injection blast furnaces. Lump (Newman Blend Lump NBLL) and pellets show a lower RDI-2.8 than sinter under all the gas compositions tested. Higher RDI-2.8 values were obtained for all burden materials with a reducing gas containing both CO and H2 compared to CO or H2 only. The addition of H2 to CO increases the pore diffusion rate allowing reducing gas to reach the centre part of the particles, leading to the reduction of hematite to magnetite and subsequent crack formation across the whole particles. Compared to the conventional blast furnace case, NBLL lump and sinter show a lower degradation for the H2 injection case while it was the opposite for the pellet, suggesting the necessity of reviewing overall burden materials to optimise the hydrogen injection in the blast furnace.
The steel industry is responsible for around 9% of global carbon dioxide1) emissions with an average of 1.85 tonnes of CO2 produced for every tonne of steel production. Since the Paris Agreement was adopted in 2015, the pressure on the steel industry to reduce its CO2 emissions has increased. According to the International Energy Agency, the reduction of CO2 emissions from the steel industry will reach more than 50% by 2050 through the adoption of a range of potential strategies and technologies.2)
Despite the maturity of its technology and high process efficiency, steel production using the blast furnace–basic oxygen furnace routes emit higher CO2 emissions compared to scrap–electric arc furnace or direct reduction–electric arc furnace routes. Replacing all the blast furnaces with other low-carbon technology will require a lot of time and capital costs, and may be slowed by the fact that the current age of the existing blast furnace assets is relatively young across the world at an average of 13 years.2) The median campaign life of blast furnaces is 17 years,3) and it is normally expected that the blast furnace can run up to 40 years by considering the typical average lifetime of the steel plant. It is therefore estimated that by 2050 between 30–50% of crude steel production will still rely on blast furnace technology including the hydrogen-enriched blast furnace.2)
The injection of hydrogen produced from a renewable electricity into the blast furnace has the potential to reduce CO2 emissions significantly as the reduction by hydrogen generates water instead of CO2 as in reduction by CO. Using an experimental blast furnace, Kamijo et al.4) suggest that the decrease of the CO2 emissions could reach about 16% by injecting H2 while maintaining the constant pig iron temperature and productivity. The limit of H2 injection was investigated numerically with the constraints of minimum top gas temperature and raceway flame temperature, resulting in the maximum hydrogen injection of 19.5 kg-H2/t-HM.5) The full replacement of coke with H2 is also limited as the heat generated from coke combustion is required to heat the furnace. Moreover, coke also provides a gas permeable structure which is crucial for stable gas flow inside the furnace.
The increase of H2 proportion in the reducing gas will not only affect the reducibility but also the reduction degradation of the ferrous burden materials.6,7) Reduction degradation typically occurs in the upper part of the blast furnace where the ferrous burden materials generate fines particles after reduction at low temperatures in the range of 400–700°C. The increase in degradation leads to a decrease in gas permeability8) and lower productivity of blast furnace operation.9) Understanding the degradation behaviour of ferrous burden materials with H2 injection is necessary to prepare optimal burden materials quality, ensuring the stable productivity of the blast furnace.
The fundamental cause of reduction degradation is the volume expansion during the reduction of hematite to magnetite which leads to the generation of stress and the formation of cracks.10) The change in the crystal structure of hematite to magnetite leads to a volume expansion reported to be around 25% during the reduction between 525 and 625°C.11)
Loo and Bristow10) proposed that reduction degradation starts with the reduction of easily accessible hematite leading to crack initiation followed by crack propagation into the surrounding matrix. Further growth and branching of cracks are then driven by the reduction of newly exposed hematite along the walls of the cracks. While the degradation obviously started from the crack initiation, the extent of degradation will be affected by the resistance of the phases to crack propagation as measured by fracture toughness. The ferrous burden materials with a higher composite fracture toughness will likely have a lower reduction degradation.12,13)
Previous investigations on the reduction degradation show that sinter has higher degradation compared to lump and pellet.14,15,16,17) With the addition of H2, the degradation was increased initially and a further increase of H2 decreased the degradation.15,18) Reduction by CO mainly occurred around the surface of particles while H2 reduction proceeded inside the particles due to the higher diffusion rate of H2 compared to CO, leading to the difference in the degradation index.18) The diffusion coefficient of H2 in pellet particles was reported to be higher for H2 than CO, and the diffusion coefficient increases with the increasing of H2 content in the gas mixture.19) A non-topochemical reduction is also suggested to increase the degradation by generating cracks in the radial direction.15) However, other authors found that the addition of H2 decreased the degradation when H2 partially replaced CO and argue that at 500°C the reducing capacity of H2 is lower than CO which leads to lower reduction and degradation.20,21)
A higher increase of degradation with the addition of H2 was observed for the sinter and pellet rather than the lump and it was suggested that this was due to the significant content of silico-ferrite of calcium and aluminium (SFCA) in the sinter and pellet.10) When this SFCA has been reduced in the presence of H2, its fracture toughness value was decreased.10) However, another author found that the lump has a higher increase of degradation with the addition of H2, arguing that the reduction of the lump was accelerated more significantly with H2 as it has lower porosity and smaller pore size compared to sinter and pellet, leading to the higher increase of degradation.16)
Owing to the disagreement in the previous finding, it seems that the effect of H2 on degradation depends on the H2 content and type of ferrous burden materials. Also, limited studies investigate the degradation of different ferrous burden materials in conventional blast furnace and blast furnace with H2 injection. Therefore, evaluating the degradation of sinter, lump, and pellets using various gas mixtures representing the conventional blast furnace and blast furnace with H2 injection is necessary.
In this study, reduction degradation tests were performed to compare the degradation behaviour of sinter, lump, and pellet under different gas mixtures containing CO and H2. The microstructures of reduced samples were compared to observe the characteristics of degradation for different types of ferrous burden materials under different reducing gas mixtures.
A modified reduction degradation test was used in this study to compare the degradation of sinter, Newman Blend Lump (NBLL), and pellet with a set of conditions taken as a compromise between ISO 4696-2 and ISO 4696-1. The samples for the test are basic sinter, lump ore, and fluxed pellet with the composition listed in Table 1, and a particle size between 10 and 12.5 mm. These samples were chosen as typical ferrous burden for blast furnaces in Asia.22) For each test, 600 g of sample was dried in the oven at 105°C for at least 2 hours prior to the test. Samples were then charged into the reduction tube and heated in the electric furnace to 500°C under an N2 atmosphere. The N2 flow rate started at 5 L/min and increased to 20 L/min when approaching 500°C. The samples were then maintained for 30 minutes to stabilize the temperatures before introducing the reducing gas with the same total flow rate of 20 L/min.
| Sample | T.Fe (wt.%) | FeO (wt.%) | CaO (wt.%) | SiO2 (wt.%) | Al2O3 (wt.%) | MgO (wt.%) | Basicity |
|---|---|---|---|---|---|---|---|
| Sinter | 56.70 | 7.60 | 10.04 | 5.43 | 1.87 | 1.76 | 1.85 |
| Lump | 62.50 | 0.49 | 0.01 | 4.35 | 1.28 | 0.04 | 0.00 |
| Pellet | 65.64 | 0.94 | 2.11 | 2.43 | 0.42 | 0.93 | 0.87 |
The isothermal reduction was performed at 500°C for 60 minutes with different reducing gas mixtures as listed in Table 2. The reducing gas mixtures were chosen to include estimates of conventional and hydrogen injection blast furnace operation as well as the common standard tests. Five gas mixtures were selected and named Gas-1 to Gas-5. Gas-1 and Gas-2 adopted the ISO 4696-2 and ISO 4696-1 gas compositions, Gas-3 and Gas-4 were taken from a modelling study simulating blast furnace gas composition by Barrett et al.,5) and Gas-5 represents the reduction only by H2 as an extreme condition for research purposes. The simulated gas composition for Gas-3 and Gas-4 represents the base case (conventional blast furnace operation) and maximum H2 injection case.
| Gas | Gas-1 | Gas-2 | Gas-3 | Gas-4 | Gas-5 |
|---|---|---|---|---|---|
| CO (vol.%) | 30 | 20 | 20 | 17 | – |
| CO2 (vol.%) | – | 20 | 24 | 23 | – |
| H2 (vol.%) | – | 2 | 5 | 13 | 30 |
| N2 (vol.%) | 70 | 58 | 51 | 47 | 70 |
In their modelling study,5) the base case gas already contains H2, as the use of pulverized coal injection (PCI) and the moisture from hot blast are included in the calculation. It was reported that nearly 50% of blast furnaces around the world are using PCI in their operation.23) The top gas composition measurement from actual blast furnaces also contains a H2 content close to 4%, indicating the existence of H2 gas even in the conventional blast furnace.24) In case of maximum H2 injection case, the H2 injection is replacing some portions of the PCI. The H2 injection is limited by maintaining the minimum top gas temperature of 118°C and minimum Raceway Adiabatic Flame Temperature (RAFT) of 2050°C. In both conventional and maximum H2 injection cases, the hot blast temperature and hot metal temperature were maintained constant to ensure the consistency of the blast furnace operation.
After the reduction was completed, the gas was changed to N2 again and the samples were cooled down to room temperature. The whole samples were then removed from the reduction tube and a 50 g sub-sample was taken for additional microstructure and chemical analysis. The remaining sample was then tumbled for 10 minutes using a tumbling drum with an inner diameter of 130 mm at a rotation speed of 30 rpm. Samples were then sieved with a 2.8 mm and 0.5 mm sieve size using a sieve shaker for 1 minute. The mass that remained in each sieve was then measured and used to calculate the Reduction Degradation Index (RDI) as defined by RDI-2.8 and RDI-0.5 using the Eqs. (1) and (2) as below:
| (1) |
| (2) |
where m0 (g) is the mass of sample after reduction and before tumbling, m1 (g) is the mass of the sample retained on the 2.8 mm sieve, and m2 (g) is the mass of the sample retained on the 0.5 mm sieve. The results of RDI-2.8 and RDI-0.5 shown in this study are the average value of duplicate tests.
About 15 g of sinter, lump, and pellet (as-received and reduced samples) were mounted in epoxy resin, ground and polished to a mirror finish to obtain the microstructure using a Zeiss Axioscope reflected light microscope. For the reduced samples, the mounted samples were taken from the material after reduction and before tumbling to preserve the original particle shape. To confirm the effect of reduction on degradation during the test, the RDI of the samples with and without reduction was compared. The RDI without reduction was based on the results of the tumbling test for the as-received samples and after the samples were heated in N2 at 500°C.
The cross-section images and microstructures of sinter, lump, and pellet are shown in Figs. 1 and 2. The samples used in this study are basic sinter, lump ore, and fluxed pellet. Sinter contains silico-ferrite of calcium and aluminium (SFCA) as a main bonding phase and the rest are magnetite, hematite (primary and secondary), dicalcium silicates, and glass. The Newman Blend Lump (NBLL) ore contains various textures with the main phases being hematite and goethite and its structure is relatively dense in comparison to sinter and pellet. The fluxed pellet shows a quite uniform structure of hematite with microporosity distributed across the whole particles.
Image analysis of Fig. 1 was performed to compare the porosity and pore size distribution of sinter, lump, and pellet. ImageJ software was used to perform the segmentation of pore and other solid phases. Pore was segmented by adjusting the grey scale value, and the pore area was measured to calculate the cumulative porosity and equivalent pore diameter. The cumulative porosity is calculated using the Eq. (3) where the pore area and total sample area are measured in mm2. The equivalent pore diameter is calculated under the assumption that the pore is in a circle shape and measured in μm. This method has a limitation to obtain the actual pore volume and pore shape as it only relies on the 2D cross-section of the sample. However, it was reported that the porosity calculated from the microstructure obtained using optical microscopy (2D) has a similar trend with the porosity calculated using a combination of mercury intrusion porosimetry and nitrogen pycnometry (3D).25) Therefore, the use of image analysis from microstructure in this study should be sufficient to compare the porosity and pore size distribution of sinter, lump, and pellet.
| (3) |
Figure 3 shows the cumulative porosity of sinter, lump, and pellet. Sinter has the highest porosity of 43.6%, followed by pellet and lump with 34.5% and 11.6% respectively. A wide range of pore size was observed for sinter with the largest pore size close to 1.3 mm. Up to 50% of pore area in sinter was occupied by pores with a diameter of more than 192 μm, while in case of pellet and lump it was 107 μm and 36 μm. In general, sinter not only has the highest porosity, but also a larger pore size compared to pellet and lump.
Figure 4 shows the comparison of RDI-2.8 and RDI-0.5 of the samples with and without reduction. For the as-received samples (no heating), the degradation was mainly due to surface abrasion. Lump shows a higher RDI-2.8 and RDI-0.5 due to the nature of the lump which contains a proportion of adhering fines particles on the surface. The degradation was increased for lump after heating in N2 at 500°C for 60 minutes due to the thermal shock and transformation of goethite to hematite (which can be associated with decrepitation) while there was no change for sinter and pellet which have already undergone thermal treatment.
It was observed during the test that the sinter and pellet only lost less than 0.1% of mass during heating to 500°C while it was up to 4.6% for lump due to dehydroxylation of goethite. A significant increase of RDI-2.8 and RDI-0.5 after the reduction in CO gas confirmed the degradation during the test was due to the reduction of hematite to magnetite. Cracks can be formed during the cooling step after the reduction; however, they will not significantly affect the degradation index as evidenced by a similar RDI value for as-received samples and samples after heating in N2 for the sinter and pellet.
3.2. Reduction Degradation of Sinter, Lump, and Pellet with Different Gas MixturesThe results of RDI-2.8 for sinter, lump, and pellet with different gas mixtures are shown in Fig. 5. Sinter has a higher RDI-2.8 compared with lump and pellet under all gas mixtures and it was consistent with the previous finding by various authors.10,14,16) In general, sinter has a higher porosity and pore size than lump and pellet which makes it more susceptible to breakage once the cracks are formed and propagated. The pores distributed across sinter particles can act as a channel for reducing gas and enhancing the reduction and crack propagation.
Another reason for the higher degradation of sinter is that sinter structure consists of various phases with different fracture toughness values. Among these phases, glass shows the lowest fracture toughness value.10,13,26) Glass was distributed throughout the sinter particles and cracks might preferentially run through this phase.10)
As the cracks were formed mainly due to the reduction of hematite to magnetite, it can be projected that a higher degradation can be obtained for iron ores with a higher hematite content. However, despite a lower hematite content than lump and pellet, sinter shows a higher degradation which indicates that the degradation is not only dependent on the hematite content but also on overall phase configuration and pore structure. Although some authors highlighted the significance of secondary hematite content with degradation,27,28,29) other authors conclude that the composite property of the overall phases has a better correlation with the degradation.12,13)
With the change in reducing gas composition, all ferrous burden materials show a higher RDI-2.8 in the mixture of CO and H2 compared to only CO and only H2. With the addition of even a small percentage of H2 to CO (2% for Gas-2), the RDI-2.8 increases substantially. This is particularly true for sinter and pellets, where there was a 50% and 137% increase respectively. It was suggested that the SFCA content in the sinter and pellet was responsible for the higher increase in the degradation.10) Further increases in H2 led to smaller changes in degradation. With the addition of H2 to CO, degradation was initially increased but further increase of H2 content decreased the degradation. These results were consistent with previous studies by Murakami et al.18) and Mizutani et al..15) Comparing the Gas-1 and Gas-2 (gas composition from ISO 4696-2 and ISO 4696-1), a small proportion of H2 (2%) in the Gas-2 increased the RDI-2.8. A higher increase of degradation was obtained for sinter and pellet compared to lump which is consistent with the previous study.10)
Sinter and lump show similar behaviour for the RDI-2.8 values under Gas-3 and Gas-4 which represent the gas composition of conventional blast furnaces and H2 injection blast furnaces. The RDI-2.8 of the H2 injection case was lower than the conventional case, suggesting the benefit of increasing H2 content in blast furnaces using sinter and lump. On the other hand, the RDI-2.8 of the pellet for the H2 injection case was higher than the conventional case, showing a similar value with lump under the same gas mixtures. The RDI-2.8 of all burden materials then decreased when the gas composition shifted to only H2 (Gas-5).
Compared to the reduction by CO gas, it was suggested that reduction with the mixtures of CO and H2 can proceed more to the centre of the particles as H2 has a higher pore diffusion rate than CO.18,30) Wu et al.30) proposed that the degradation was due to a combination of the volume expansion during the reduction of hematite to magnetite and the carbon deposition reaction. With the addition of H2, reduction reached the inside part of the particle, producing internal cracks which later become a channel for CO to diffuse into the inside part. The reduction by CO continued to proceed in the inside part and carbon deposition then occurred in the internal cracks, generating more cracks which lead to heavy degradation.
However, with a further increase of H2, the carbon deposition reaction becomes more limited as the proportion of CO and CO2 in the system decreases, leading to the decrease of degradation. Another possibility for the decrease of degradation with a further increase of H2 is the strain level in the magnetite crystals formed. It has been reported that the strain level in magnetite crystal formed from H2 reduction was lower than by CO reduction, resulting in no stress concentration at the surface,31) and lower expansion after reduction with H2–H2O than CO–CO2.32)
Figure 6(a) shows the RDI-0.5 of sinter, lump, and pellet. In general, pellet shows the highest RDI-0.5 (except for Gas-1) followed by lump and sinter. Despite having a lower RDI-2.8 than sinter, a higher RDI-0.5 of pellet means the ultra-fines generation during the degradation was more dominant for pellets. It suggests that the degradation in pellet typically appeared as surface cracking, while it was internal cracking for sinter, and the combination of surface and internal cracking for lump. The ratio of RDI-0.5/RDI-2.8 is in the range of 85–98% for pellet, 41–49% for lump, and 22–27% for sinter (Fig. 6(b)); showing that most of the degradation for pellet is in the form of ultra-fines particles and it was related with the fact that pellet was produced using ultra-fines iron ores feed. On the other hand, sinter was produced using a mixtures of iron ores with a relatively larger size than for pellet feed. After the sintering process some portion of larger iron ore particles (nuclei) maintain their shape and during degradation will break into relatively larger size than in the case of pellet.
The reduction degree of sinter, lump, and pellet under CO (Gas-1), H2 (Gas-5), and CO–H2 (Gas-3) as a reducing gas was shown in Fig. 7. The reduction degree was calculated based on the mass loss data during the reduction period and the total removable oxygen from the samples. Reduction degree under H2 (Gas-5) was higher than CO (Gas-1) and CO–H2 (Gas-3) for all burden materials. Lump shows the highest reduction degree followed by pellet and sinter under the same gas composition. Although sinter has higher porosity than pellet and lump, the lower reduction degree of sinter as shown in Fig. 7 indicated the effect of mineralogy on the reduction. Sinter has a higher proportion of magnetite and SFCA (which are less reducible than hematite), leading to lower reduction than lump and pellet. Metallic iron was found on all the reduced samples, ranging from 1.2% to 11.6% based on chemical analysis of the sub-samples. The metallic iron was formed from the reduction of magnetite instead of wustite as below 570°C the wustite phase is not thermodynamically favourable. Reduction also proceeded as a heterogeneous reaction, leading to the formation of metallic iron and magnetite at a relatively low reduction degree.
The RDI-2.8 and reduction degree were plotted in Fig. 8 to find any relationship between reduction degree and reduction degradation. In general, the correlation between reduction degree and RDI-2.8 was not clear although there was a decreasing tendency of the RDI-2.8 with the increase of reduction degree from ~10% of reduction degree. With the similar value of reduction degree (9–12%), the RDI-2.8 was significantly different, meaning that the degradation was not only dependent on the amount of reduction but also from the characteristics of the materials, including the formation of cracks during the reduction. A similar observation was also obtained from previous studies which indicated that there is no direct correlation between reduction degree and degradation.15,33) A stronger correlation between reduction degree and degradation might occur by increasing the reduction time while maintaining the other parameters constant (same materials, reduction temperature, and gas composition).
The study by Wu et al.30) suggested that the carbon deposition reaction can increase the degradation during low temperature reduction. However, in the current study, it seems that the carbon deposition reaction did not contribute significantly to the increase of degradation. The carbon content from the sub-samples of the reduced samples was measured and presented in Table 3. The carbon content with CO–H2 as a reducing gas was lower than only CO, while the RDI-2.8 was clearly higher. These results suggested that in this study the effect of carbon deposition was not significant to the increase of degradation. Moreover, the amount of carbon was relatively small considering the total mass of the samples, meaning that the addition of mass due to carbon deposition did not significantly affect the calculation of reduction degree. Using the case of lump under Gas-1 as an example, the equivalent value of reduction degree from the carbon mass is only about 1.6% which is much smaller compared to reduction degree of 16%.
| Materials | Gas-1 | Gas-3 |
|---|---|---|
| Sinter | 0.23 | 0.03 |
| Lump | 0.43 | 0.01 |
| Pellet | 0.12 | 0.04 |
Figure 9 shows the microstructure of sinter after reduction with a reducing gas of only CO (Gas-1) and CO–H2 (Gas-3). The variation in particle size and shape were unavoidable as the samples in this study are industrial iron ores. Comparing to the as-received samples in Fig. 1, cracks were clearly shown after the reduction. Cracks were connected between pores and found in a variety of phases. As the cracks were initiated due to the reduction of hematite, the initiation of cracks will likely start from the area near the pores as pores become a channel for the gas to reach the iron oxides. Once the cracks were initiated, they propagated through the surrounding phases until they reached the neighbouring pores. Qualitatively, more cracks were formed after reduction with the CO–H2 than only CO, especially in the centre part of the particles, which supports the previous assumption of the reduction with CO–H2 proceeding more into the centre of particles.
Figure 10 shows the detailed observation of cracks at the surface and centre of the sinter particles. Cracks were found in all the phases and connected to the pores. The branching of cracks and variation in crack thickness were also clearly visible. Most of the hematite has been reduced to magnetite and trace of metallic iron was also observed near the pores, indicating that the reduction of hematite was favourable near the pores. It is likely that pores can not only act as the location for crack initiation as they provide access for reducing gas, but also stopping the propagation of cracks. However, the evaluation of the roles of cracks is not possible using current microstructure analysis and in-situ observation of cracks is necessary.
Comparing the microstructure after reduction with only CO (A and B) and CO–H2 (C and D) in Fig. 10 shows that cracks were not found only on selected phase but across all the phases, including SFCA as main bonding phase, glass, and both of magnetite that formed during sintering and reduction. In case of reduction with only CO, more metallic iron was formed at the surface than at the centre indicating the limited diffusion of CO gas into the centre of particles. This condition may lead to the difference in the extent of the reduction and degradation at the surface and the centre. In case of reduction with CO–H2, the microstructure at the surface and centre was quite similar due to the higher diffusion rate of the gas with the presence of H2. This condition leads to the reduction and formation of cracks across the whole particles as previously described.
The microstructure of lump after reduction with a reducing gas of only CO (Gas-1) and CO–H2 (Gas-3) was shown in Fig. 11 with detail observation of cracks in Fig. 12. Compared to the as-received samples in Fig. 1, it was obvious that cracks were formed after reduction especially at the location where hematite had been reduced to magnetite (pinkish colour). Under reduction with only CO, some portions of hematite (whiteish colour) remained unreduced in the centre part of lumps due to the limited diffusion rate of CO gas whereas under CO–H2 most of hematite was reduced to magnetite. This condition leads to more crack formation throughout the particle under CO–H2 gas and higher degradation.
In Fig. 12, the surface area of lump (E) under CO gas shows most of the hematite was reduced to magnetite and traces of metallic iron were found. Cracks were found on both magnetite formed from hematite or goethite. In the centre area (F), more cracks were found on the reduced area than unreduced area (hematite and goethite). It seems that the cracks were propagated along with the reduction path and similar observation also previously obtained by Adam et al.34) and Hbika et al.35) With the addition of H2 (G and H), most of hematite was reduced to magnetite and the microstructure was quite similar both at the surface and centre part. Similar with the case in sinter, the higher diffusion rate of the gas with H2 addition leads to the reduction and formation of cracks across the whole particles leading to a higher degradation.
Figures 13 and 14 shows the microstructure of and details observation of cracks for pellet after reduction with a reducing gas of only CO (Gas-1) and CO–H2 (Gas-3). In general, pellet shows less cracks than sinter and lump after reduction and it was in good agreement with the RDI-2.8 results. With less cracks in the structure, pellets can maintain their original particle shape during the tumbling process and degradation occurred through surface abrasion and generated mostly fines particles.
Under the reduction with only CO gas, similar finding with lump was observed as some hematite remained unreduced at the centre part (J) and traces of metallic iron were found at the surface (I). Most cracks were formed as a thin crack (fissure) on the magnetite, and no cracks were observed on the unreduced hematite. As shown in Fig. 14, in case of CO–H2 (K and L) the structure at the surface and centre was similar with most of hematite being reduced to magnetite and no trace of metallic iron observed. As already observed in case of sinter and lump, with the higher diffusion rate under CO–H2, reduction and cracks occurred across the whole particles, producing higher degradation.
The microstructure analysis results of sinter, lump, and pellet as shown in Figs. 9, 10, 11, 12, 13, 14 confirm that cracks were formed during the reduction and were responsible for the degradation. Qualitatively, the amount of cracks and degradation index in this study shows a good agreement and this is consistent with previous studies.18,36) Comparing Figs. 9, 11, and 13, it can be seen that the formation of macro-cracks is more prevalent in sinter and lump, leading to a higher degradation than pellet.
The presence of H2 in CO gas increases the diffusion rate resulting in reduction proceed more into the centre of particles along with the cracks formation. Limited diffusion of reduction by only CO gas was clearly shown in case of lump and pellet through the formation of metallic iron near the surface and unreduced hematite at the centre, indicating the topochemical reduction behaviour.
The implication of current finding to industrial practice is the comparison of reduction degradation behaviour using a gas composition close to the actual gas composition is necessary as the presence of H2 clearly changes the degradation index value. Although in this study the degradation of sinter and lump shows decreasing tendency when the gas composition shifted from conventional to H2 injection blast furnace, the increase of degradation for pellet suggests that the overall burden design should be reviewed to optimize the operation with H2 injection. The quality target for sinter and pellet may also change when optimised for the H2 injection blast furnace.
Within the scope of current study, it is not possible to confirm whether the presence of H2 affects the location of crack initiation or the path for crack propagation. Fundamental understanding of the effect of CO and H2 on the crack initiation and crack propagation is still unclear and should be studied in future work. The use of in-situ method like the study by Kim et al.37) to observe crack formation during reduction might give valuable information for the above fundamental study. Another alternative is adopting non-destructive methods such as computed tomography which allow the comparison of the structure before and after reduction for the same sample.
The reduction degradation of lump (Newman Blend Lump NBLL), sinter, and pellet was studied using a modified reduction degradation test with a wide range of reducing gas mixtures containing CO and H2. The reduction degradation index (RDI) was calculated and the RDI-2.8 was used to compare the degradation of different ferrous burden materials. The microstructure of the as-received samples and reduced samples were analysed to observe the formation of cracks and characteristics of degradation. The results of this study can be summarized as follows:
• Lump (NBLL) and pellets show a lower RDI-2.8 than sinter under all the gas compositions tested. The higher porosity of sinter and the existence of a low fracture toughness phase is thought to be responsible for the higher degradation.
• Higher RDI-2.8 values were obtained for all burden materials with reducing gas mixtures containing CO and H2 compared to CO or H2 only. The addition of H2 to CO increases the pore diffusion rate, allowing the reducing gas to reach the centre part of the particles, leading to the reduction and cracks formation across the whole particles.
• The highest degradation was observed for 2% H2 addition in the case of sinter and lump, with further increases in H2 simulating current blast furnace operation or maximum H2 injection cases leading to lower values of RDI-2.8.
• Pellet shows the highest degradation for the maximum H2 injection case, suggesting that moving the blast furnace operation towards higher H2 content can decrease the degradation of lump and sinter while increasing the degradation of pellet. The design of the overall burden materials should be reviewed to optimise the operation with H2 injection.
• The degradation of pellet typically appeared as surface cracking with a high proportion of ultrafine particles after reduction as indicated by a higher value of RDI-0.5, compared to sinter and lump which broke into relatively larger particles.
• There is no clear relationship between reduction degree and RDI-2.8, indicating that the degradation is not only dependent on the amount of reduction but also from the characteristics of cracks during reduction. More fundamental study of crack formation during reduction under different gas composition should be performed in the future.
The authors acknowledge the financial support and permission to publish this paper from BHP. The authors also thank Leanne Matthews and Gareth Penny for their technical assistance, and Dr. Siyu Cheng for the valuable feedback.
