A method for improving the desulfurization capacity of high-titanium slag using artificial intelligence models

Abstract

This invention provides a method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model, applicable to the smelting of high-vanadium-titanium ore in blast furnaces. The method includes: controlling the MgO content in the slag to 8.0%-9.5% by weight, the binary basicity of the slag to 1.10-1.17, and the Al2O3 content ≤14.5%, while maintaining the MgO to Al2O3 ratio in the slag at 0.60-0.65; maintaining the physical heat of molten iron ≥1440℃; controlling the blast furnace inlet blast temperature ≥1200℃; controlling the sulfur load in the blast furnace inlet to 4.5-5.5 kg, and controlling the proportion of pellets in the blast furnace inlet to 46%-50%; and constructing a dynamic adjustment model based on predicted sulfur load increments and predicted titanium reduction increments, so that the oxygen enrichment rate is dynamically adjusted within a first threshold range, and the blast furnace coal ratio is dynamically adjusted within a second threshold range. This invention can dynamically adjust the oxygen enrichment rate and blast furnace coal ratio according to the sulfur load increment and titanium reduction increment, thereby improving the desulfurization capacity of high-titanium slag.

Description

Technical Field

[0001] This invention belongs to the field of blast furnace smelting technology, and in particular relates to a method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model. Background Technology

[0002] Vanadium and titanium resources, as strategic mineral resources of the nation, have irreplaceable application value in fields such as steel, aerospace, and new energy. The Panxi region of my country is the world's largest concentrated production area of ​​vanadium-titanium magnetite, boasting large reserves and high comprehensive utilization potential, and has become an important component of my country's blast furnace smelting raw material system. When blast furnaces employ a high proportion of vanadium-titanium ore for smelting, the titanium dioxide (TiO2) content in the slag will be stably maintained at a level of 21%-22.5%, forming typical high-titanium slag. The physicochemical properties of this type of slag differ significantly from those of conventional blast furnace slag, posing unique challenges to stable blast furnace smelting and hot metal quality control.

[0003] Currently, vanadium-titanium ore resources in the Panxi region generally suffer from high sulfur loads, resulting in a long-term stable sulfur load of around 6.0 kg / TFe in the blast furnace. During blast furnace smelting, sulfur in molten iron is mainly removed by the slag. The desulfurization capacity of the slag directly determines the sulfur content and pass rate of the molten iron. If the slag desulfurization capacity is insufficient, the sulfur content in the molten iron will exceed the standard requirements, leading not only to the scrapping of the molten iron or the need for additional refining, but also to exacerbating the erosion of the blast furnace lining, affecting the quality of subsequent steel grades, and significantly increasing smelting costs.

[0004] However, the existing blast furnace smelting process for high-vanadium-titanium ores has insurmountable technical defects when dealing with the desulfurization requirements of high-titanium slag, specifically in the following four aspects: 1. The slag has high viscosity and poor fluidity, resulting in harsh desulfurization and mass transfer conditions.

[0005] In high-titanium slag, TiO2 is prone to over-reduction under the high-temperature reducing atmosphere of the blast furnace, generating high-melting-point, high-hardness compounds such as titanium carbide (TiC), titanium nitride (TiN), or titanium carbonitride (Ti(C,N)). These compounds are suspended in the slag as solid particles, forming what is commonly known in the industry as foamy slag, which directly leads to a significant increase in slag viscosity and a sharp decrease in fluidity. On the one hand, the increased viscosity worsens the physical mixing effect at the slag-iron interface, greatly increasing the resistance to the diffusion of sulfur from the molten iron to the slag, reducing the mass transfer rate of sulfur, and making it difficult for the desulfurization reaction to proceed fully. On the other hand, poor fluidity makes slag-iron separation difficult, easily causing metallic iron beads to be entrained in the slag, while slag residues also remain in the molten iron, further deteriorating the quality of the molten iron.

[0006] Second, the effective alkalinity is reduced, and the sulfur capacity of the slag is insufficient.

[0007] The desulfurization capacity of blast furnace slag is directly related to its effective basicity. Conventional blast furnaces enhance desulfurization capacity by adjusting the content of basic oxides such as CaO and MgO to increase effective basicity. However, in high-titanium blast furnace slag, TiO2 preferentially reacts with CaO and MgO to form stable composite mineral phases such as perovskite (CaTiO3) and magnesia-titanium (MgTiO3). This reaction consumes a large amount of free CaO and free MgO intended for desulfurization, resulting in a significant decrease in the actual effective basicity of the slag. Insufficient effective basicity directly leads to a decrease in slag sulfur capacity, making it difficult to achieve the expected desulfurization effect even with increased total basicity.

[0008] 3. Fluctuations in slag oxidizability and imbalance in sulfur distribution.

[0009] Titanium exhibits significant valence changes in the reducing atmosphere of a blast furnace (Ti 4+ Ti³ + This valence change process directly affects the oxygen potential of the slag. Fluctuations in the slag oxygen potential disrupt the distribution balance of sulfur between the slag and iron phases. When the oxygen potential is too high, sulfur in the molten iron is more likely to exist in an oxidized state, such as SO2, making it difficult to migrate into the slag. When the oxygen potential is too low, it exacerbates the over-reduction of TiO2 to generate more Ti (C,N), further worsening the slag fluidity. Existing processes lack targeted control measures for the valence change behavior of titanium, making it impossible to stabilize the slag oxygen potential, resulting in large fluctuations in the sulfur distribution coefficient and making it difficult to achieve stable desulfurization.

[0010] Fourth, iron in the slag causes "sulfur reversion," which negates the desulfurization effect.

[0011] As mentioned earlier, the high viscosity and poor fluidity of high-titanium slag lead to incomplete slag-iron separation. Metallic iron beads entrained in the slag dissolve some of the sulfur already present in the slag. This process is known as sulfur reversion. Sulfur reversion reduces the actual amount of sulfur absorbed by the slag, effectively offsetting part of the desulfurization effect. Simultaneously, the entrained iron beads further increase slag viscosity, creating a vicious cycle. Existing furnace-side operating processes have failed to effectively solve the slag-iron separation problem, and sulfur reversion has become one of the key bottlenecks restricting the desulfurization efficiency of high-titanium slag.

[0012] In summary, under the blast furnace smelting scenario of high-TiO2 and high-sulfur-load vanadium-titanium ore, existing processes cannot simultaneously solve problems such as poor slag fluidity, insufficient effective basicity, fluctuating oxidizing properties, and sulfur reversion. This results in slag desulfurization capacity consistently failing to meet actual needs, leading to low hot metal qualification rates, high smelting costs, and limited resource utilization. Therefore, developing a process that can specifically improve the physicochemical properties of high-titanium slag and enhance its desulfurization capacity has become an urgent need for the efficient utilization of vanadium-titanium ore and stable blast furnace smelting in the Panzhihua-Xichang region.

[0013] Chinese patent application CN102978312B discloses a novel blast furnace smelting method for high-vanadium-titanium and low-MgO slag, belonging to the field of blast furnace steelmaking technology. This method includes the steps of feeding vanadium-titanium sinter, acidic lump ore, vanadium-titanium pellets, and coke into a blast furnace for smelting. The vanadium-titanium sinter contains 2.2–2.6% MgO and 2.4–2.8% Al2O3 by weight. During smelting, the blast furnace slag basicity (CaO / SiO2) is controlled at 1.28–1.33, the physical heat of molten iron is ≥1450℃, and the MgO content, Al2O3 content, and TiO2 content in the blast furnace slag are controlled at 6.8–8.5%, 11–13%, and 18–21%.

[0014] The aforementioned existing technologies have a narrow applicable range of titanium content, failing to cover the critical range where titanium over-reduction exacerbates the problem. They lack operational control measures specifically for titanium over-reduction and also lack synergistic regulation of slag fluidity, further amplifying the impact of titanium over-reduction. Summary of the Invention

[0015] The purpose of this invention is to provide a method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model, which partially solves or alleviates the above-mentioned deficiencies in the prior art. It can dynamically adjust the oxygen enrichment rate and blast furnace coal ratio according to the sulfur load increment and titanium reduction degree increment, thereby improving the desulfurization capacity of high-titanium slag.

[0016] To solve the aforementioned technical problems, the present invention specifically adopts the following technical solution: A first aspect of the present invention is to provide a method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model, for use in blast furnace smelting of high-vanadium-titanium ore, comprising: By weight, the MgO content in the slag should be controlled at 8.0%-9.5%, the binary basicity of the slag should be controlled at 1.10-1.17, the Al2O3 content should be ≤14.5%, and the ratio of MgO to Al2O3 in the slag should be maintained at 0.60-0.65. Maintain the physical heat of molten iron ≥1440℃; control the blast furnace inlet temperature ≥1200℃; The sulfur load in the blast furnace should be controlled at 4.5-5.5 kg, and the proportion of pellets in the blast furnace should be 46%-50%. A dynamic adjustment model is constructed based on the predicted sulfur load increment and the predicted titanium reduction increment, so that the oxygen enrichment rate is dynamically adjusted within the first threshold range and the blast furnace coal ratio is dynamically adjusted within the second threshold range.

[0017] Furthermore, the dynamic adjustment model is as follows: O%=O_base+α1×(ΔS_pred-ε1)+β1×(ΔR_Ti_pred-ε2); C_ratio= C_base +α2×(ΔS_pred- ε3)-β2×(ΔR_Ti_pred-ε4); Where O% is the oxygen enrichment rate, C_ratio is the blast furnace coal ratio, O_base is the basic oxygen enrichment rate, C_base is the basic coal ratio, ΔS_pred is the predicted sulfur load increment, ΔR_Ti_pred is the predicted titanium reduction increment, α1, β1, α2 and β2 are adjustment coefficients, and ε1~ε4 are deviations.

[0018] Furthermore, the methods for obtaining the predicted sulfur load increment and the predicted titanium reduction increment include: A dataset is constructed, wherein each sample in the dataset includes input data and output data; the input data includes sulfur load-related parameters, titanium reduction degree-related parameters, and operational intervention parameters over N hours; the output data includes the corresponding sulfur load increment and titanium reduction degree increment over M hours after N hours. Train a prediction model using the dataset; The collected N hours of input data are fed into the trained model to obtain the sulfur load increment and titanium reduction increment after N hours, M hours later.

[0019] Furthermore, the sulfur load-related parameters include the total sulfur content of the ore fed into the furnace, the pellet ratio, the sulfur content of the sinter, the sulfur content of the coke, and the total amount of ore fed into the furnace; the titanium reduction degree-related parameters include the TiO2 content of the slag, the physical heat of the molten iron, the hearth temperature, the blast kinetic energy, and the air volume; the operational intervention parameters include the blast furnace coal ratio and the oxygen enrichment rate.

[0020] Furthermore, the prediction model is an LSTM, comprising: The input layer has a dimension of number of features × number of time steps; The hidden layer consists of two LSTM network layers: the first LSTM network has 64 neurons and the second LSTM network has 32 neurons; each LSTM network is followed by a Dropout layer. The fully connected layer is a Dense layer with 16 neurons, and non-linear features are extracted using the ReLU activation function; The output layer consists of two neurons that output the predicted sulfur load increment and the predicted titanium reduction increment, respectively.

[0021] Furthermore, the first threshold range is 4.5%-6.5%, and the second threshold range is 155-170 kg / ton of iron.

[0022] Furthermore, when the predicted sulfur load increment is greater than or equal to the first threshold, or / and the predicted titanium reduction increment is greater than or equal to the second threshold, the emergency adjustment model is activated to adjust the oxygen enrichment rate and blast furnace coal ratio.

[0023] Furthermore, the emergency adjustment model is as follows: O%_pre=O_base_pre+α3×(ΔS_pred-ε5)+β3×MAX(0, ΔR_Ti_pred-ε6); C_ratio_pre= C_base_pre +α4×(ΔS_pred- ε7)-β4×MAX(0, ΔR_Ti_pred-ε8); Where O%_pre is the pre-adjusted value of oxygen enrichment rate, C_ratio_pre is the pre-adjusted value of blast furnace coal ratio, O_base is the pre-adjusted value of basic oxygen enrichment rate, C_base is the pre-adjusted value of basic coal ratio, ΔS_pred is the predicted sulfur load increment, ΔR_Ti_pred is the predicted titanium reduction increment, α3, β3, α4 and β4 are adjustment coefficients, and ε5~ε8 are deviations.

[0024] Furthermore, the vanadium-titanium ore is pretreated by roasting it at 800-900℃ with oxygen-enriched air for 1.5-2.0 hours.

[0025] Furthermore, during the roasting process, 0.8-1.2% of CaO-Al2O3 composite desulfurizing agent is added to the vanadium-titanium ore.

[0026] Beneficial effects: The pretreatment stage of this invention uses oxygen-enriched roasting at 800-900℃ and 0.8-1.2% CaO-Al2O3 composite desulfurizing agent to pre-fix free TiO2 to generate MgTiO3 and CaTiO3 and pre-desulfurize, thereby reducing titanium over-reduction reactants and sulfur load in the furnace from the source, and reducing the burden on subsequent in-furnace desulfurization.

[0027] In terms of core furnace control, the synergistic control of slag composition, thermal parameters, and raw material ratio precisely matches the needs of high-titanium slag to inhibit titanium over-reduction, ensure fluidity, and enhance desulfurization activity, forming a three-dimensional support of composition, temperature, and raw materials. In the intelligent response phase, the dynamic and emergency models adjust the oxygen enrichment rate and coal ratio in real time based on predicted data, transforming the static parameters of pretreatment and in-furnace control into a dynamically adaptable flexible system, thereby achieving self-optimization of parameters throughout the entire process. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0031] In this document, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" may be used interchangeably.

[0032] In this document, the terms "upper," "lower," "inner," "outer," "front," "rear," "one end," and "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the present invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0033] In this document, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0034] In this document, "and / or" includes any and all combinations of one or more of the listed related items.

[0035] In this article, "multiple" means two or more, that is, it includes two, three, four, five, etc.

[0036] As used in this specification, the term "about" typically means + / -5% of the value, more typically + / -4% of the value, more typically + / -3% of the value, more typically + / -2% of the value, even more typically + / -1% of the value, and even more typically + / -0.5% of the value.

[0037] In this specification, certain embodiments may be disclosed in a range-bound format. It should be understood that this "range-bound" description is merely for convenience and brevity and should not be construed as a rigid limitation on the disclosed range. Therefore, the description of a range should be considered as having specifically disclosed all possible subranges and the individual numerical values ​​within those ranges. For example, a description of the range 1-6 should be considered as having specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6, etc., and the individual numbers within those ranges, such as 1, 2, 3, 4, 5, and 6. This rule applies regardless of the breadth of the range.

[0038] Example 1: like Figure 1 As shown, this invention provides a method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model, the specific steps of which include: S1 involves pretreating vanadium-titanium ore by roasting it at 800-900℃ with oxygen-enriched air for 1.5-2.0 hours; during the roasting process, 0.8-1.2% of CaO-Al2O3 composite desulfurizing agent is added to the vanadium-titanium ore.

[0039] In vanadium-titanium ore, 30-40% of TiO2 exists in a free state. After being fed into the furnace, it is easy to generate Ti (C,N) under a high-temperature reducing atmosphere, which leads to a sharp increase in slag viscosity and deterioration in fluidity. In addition, the sulfur content of vanadium-titanium ore raw materials in the Panxi region is generally 0.6-0.8%, which directly results in a sulfur load of more than 6.0 kg / TFe in the furnace, exceeding the conventional desulfurization capacity of the slag.

[0040] The purpose of the pretreatment in this step is to achieve the directional fixation of free TiO2 through roasting, reduce the over-reduction reaction products of titanium in the furnace and reduce the sulfur load in the furnace by pre-desulfurizing the raw materials, thereby reducing the desulfurization burden in the furnace from the source and avoiding excessive reliance on parameters such as high MgO and high alkalinity in the furnace, thus balancing the desulfurization effect and smelting cost.

[0041] Specifically, the first step involves oxygen-enriched roasting to directionally fix free TiO2 and suppress over-reduction of titanium in the furnace.

[0042] During roasting, the free TiO2 in the vanadium-titanium ore reacts with the CaO and MgO contained in the raw material to directionally generate stable magnesium-titanium ore (MgTiO3) and perovskite (CaTiO3) under a weak oxidizing atmosphere at 800-900℃. The reaction formula is as follows: TiO2 + MgO → MgTiO3; TiO2 + CaO → CaTiO3; These two titanates are extremely stable and do not easily decompose in the reducing atmosphere of blast furnace smelting. They can permanently fix free TiO2 and prevent it from being reduced to Ti (C,N) in the furnace.

[0043] The purpose of introducing oxygen-enriched air with an oxygen concentration of 25-30% is to maintain a weak oxidizing atmosphere and avoid the strong reducing atmosphere from causing FeO to be over-reduced to metallic Fe, which would increase the risk of over-reduction of titanium. The weak oxidizing atmosphere can accelerate the reaction rate of TiO2 with CaO / MgO, so that the conversion rate of free TiO2 can reach 60-70% within 1.5-2.0h.

[0044] The second step involves using a CaO-Al2O3 composite desulfurizing agent to pre-desulfurize the raw materials.

[0045] Single CaO exhibits low desulfurization efficiency at 800-900℃, necessitating the use of Al2O3 as a flux to form a low-melting-point desulfurization system for efficient pre-desulfurization. The preferred mass ratio of CaO to Al2O3 in the composite desulfurizing agent is 3:1. CaO serves as the primary desulfurizing agent, reacting with sulfur in the raw materials (mainly in the form of FeS and MnS) to generate stable CaS, with the reaction formula: FeS + CaO → CaS + FeO.

[0046] Al2O3 acts as a flux, forming a low-melting-point calcium aluminate with CaO. This molten state occurs at 800-900℃, significantly improving the fluidity and dispersibility of CaO, increasing the contact area between CaO and FeS, and thus enhancing desulfurization efficiency.

[0047] The amount of desulfurizing agent added should be controlled between 0.8% and 1.2%. Too low an amount will result in insufficient liquid phase of calcium aluminate, poor CaO dispersion, low desulfurization rate, and inability to effectively reduce sulfur load; excessive Al2O3 will remain in the raw materials, leading to excessive Al2O3 content in the subsequent slag, which will increase the viscosity of the slag.

[0048] S2, by weight, controls the MgO content in the slag at 8.0%-9.5%, the binary basicity of the slag at 1.10-1.17, and the Al2O3 content at ≤14.5%, while maintaining the ratio of MgO to Al2O3 in the slag at 0.60-0.65.

[0049] In high-titanium slag, TiO2 is easily over-reduced to high-melting-point compounds such as TiC and TiN under high-temperature reducing atmosphere, leading to a sharp increase in slag viscosity and hindering desulfurization mass transfer. The role of MgO is to preferentially react with free TiO2 to form stable magnesia-titanium ore (MgTiO3), with the reaction formula being TiO2 + MgO → MgTiO3. Magnesia-titanium ore has extremely high stability and does not decompose under the reducing atmosphere of the blast furnace, permanently fixing free TiO2 and preventing its reduction to Ti (C,N).

[0050] Binary basicity directly determines the content of free CaO in the slag, with the reaction formula being (CaO) + [S] + [C] → (CaS) + CO↑. However, in high-titanium slag, TiO2 preferentially reacts with CaO to form perovskite: TiO2 + CaO → CaTiO3, consuming a large amount of CaO. If the basicity is too low, the free CaO will be insufficient after CaO is consumed by TiO2, resulting in a sharp drop in desulfurization capacity; if the basicity is too high, excess CaO will react with SiO2 to form high-melting-point calcium silicate, leading to an increase in the slag melting temperature and deterioration in fluidity, which in turn inhibits desulfurization mass transfer.

[0051] Al₂O₃ and MgO / Al₂O₃ are viscosity modifiers used to control slag flowability. Al₂O₃ is a typical impurity in high-vanadium-titanium ore, and it reacts with CaO to form high-viscosity calcium aluminate, exacerbating slag flowability problems. When Al₂O₃ exceeds 14.5%, even with the addition of MgO, the amount of calcium aluminate formed will surge, causing the viscosity at 1450℃ to exceed 13 Pa. s, exceeding the viscosity required for desulfurization ≤11Pa s-critical value.

[0052] MgO can react with Al2O3 to form low-viscosity magnesium aluminum spinel, and the optimal range for MgO to neutralize Al2O3 is 0.60-0.65. If the ratio is <0.60, there is insufficient MgO, which cannot effectively counteract the thickening effect of Al2O3; if the ratio is >0.65, excess MgO will react with SiO2 to form high-melting-point magnesium silicate, which will instead increase the melting point of the slag.

[0053] After MgO fixes free TiO2, it reduces the consumption of CaO by TiO2, ensuring that a binary basicity of 1.10-1.17 is sufficient to maintain a free CaO content of 3.5%-4.5%, thus avoiding the deterioration of fluidity caused by high basicity. When Al2O3 = 12%-14.5%, MgO 8.0%-9.5% precisely maintains the MgO / Al2O3 ratio at 0.60-0.65, ensuring that the proportion of magnesium aluminum spinel in the slag reaches 20%-25%, and the viscosity remains stable at 9-10 Pa at 1450℃. The fluidity of s provides optimal kinetic conditions for the desulfurization reaction. Low alkalinity reduces the reaction between CaO and TiO2, while high MgO preferentially fixes free TiO2. Together, these factors keep the titanium reduction degree below 0.3, fundamentally avoiding the vicious cycle of titanium over-reduction → increased viscosity → deterioration of desulfurization.

[0054] S3 maintains the physical heat of molten iron ≥1440℃; controls the blast furnace inlet blast temperature ≥1200℃.

[0055] The physical heat of molten iron is the actual temperature of the molten iron inside the hearth, which can be monitored using thermocouples or infrared thermometers in the molten iron trough. Since the desulfurization reaction of high-titanium slag is a strongly endothermic reaction, temperature is crucial for driving the reaction forward. According to thermodynamic calculations, the equilibrium constant K for the desulfurization reaction at 1440℃ is 1.2 × 10³, a significant increase compared to K = 8 × 10² at 1420℃. Simultaneously, high temperature increases the diffusion coefficient of sulfur between slag and iron, reaching 6.0 × 10³ at 1440℃. -9 m² / s, compared to 4.5 × 10 at 1400℃ -9 The m² / s is increased by 33%, ensuring that sulfur can quickly migrate from molten iron to slag and avoid sulfur being retained in molten iron.

[0056] High-titanium slag, due to the presence of high-melting-point particles such as Ti (C, N), exhibits a much higher viscosity sensitivity to temperature compared to conventional slag. When the physical heat of molten iron is ≥1440℃, the slag temperature is simultaneously maintained at 1450-1470℃, at which point the slag viscosity can be stabilized at 9-11 Pa. If the physical heat is less than 1440℃, the slag viscosity will increase sharply, the slag-iron interface will be poorly mixed, desulfurization and mass transfer will be hindered, and the sulfur content of the molten iron will easily exceed the standard.

[0057] Titanium over-reduction tends to occur in localized low-temperature zones within the hearth. A physical heat of molten iron ≥1440℃ indicates sufficient heat reserve in the hearth and a uniform temperature field, which can prevent the formation of a central dead stock or localized low-temperature zones, reducing the amount of Ti (C,N) generated and fundamentally reducing the factors that cause slag viscosity deterioration.

[0058] The blast furnace inlet temperature is the temperature at which hot air is blown into the blast furnace from the hot blast stove. It can be monitored by thermocouples in the hot blast pipeline and is the most important external heat source for the blast furnace.

[0059] The heat in the blast furnace hearth mainly comes from the combustion of coke and injected fuel, and the blast temperature directly determines the theoretical combustion temperature (Tf) of the combustion reaction. According to empirical formulas, for every 100°C increase in blast temperature, Tf increases by approximately 80-100°C. When the blast temperature is ≥1200°C, Tf can be maintained at 2150-2250°C, ensuring complete fuel combustion and continuous heat generation in the hearth, providing a heat reserve for the physical heat of molten iron ≥1440°C. If the blast temperature is <1150°C, Tf will drop below 2050°C, resulting in incomplete fuel combustion, reduced heat output from the hearth, and the physical heat of molten iron easily falling below 1440°C.

[0060] High blast temperature can replace part of the heating effect of coke, avoiding the increased costs caused by simply relying on increasing coke and pulverized coal injection to maintain the hot state. The blast furnace desulfurization reaction mainly occurs in the high-temperature zone from below the softening zone to the hearth. A stable blast temperature of ≥1200℃ can fix the position of the high-temperature zone in the blast furnace, ensuring that the desulfurization reaction can be fully carried out in the ideal high-temperature and high-alkalinity region.

[0061] S4 controls the sulfur load in the blast furnace to be 4.5-5.5 kg, and controls the proportion of pellets in the blast furnace to be 46%-50%; The sulfur load in the blast furnace is the total amount of sulfur introduced by all raw materials fed into the blast furnace, including sinter, pellets, coke, and lump ore.

[0062] According to the desulfurization distribution coefficient formula LS = (sulfur content in slag) / (sulfur content in molten iron), and considering the characteristic of high-titanium slag LS≥5.5, when the sulfur load at the furnace is 4.5-5.5 kg / TFe, the slag can remove more than 90% of the sulfur. If the sulfur load exceeds 5.5 kg, even if LS=5.5, the sulfur content in the molten iron will rise to more than 0.145%, exceeding the qualified standard. If the sulfur load is less than 4.5 kg, although the sulfur content in the molten iron can be reduced to below 0.12%, low-sulfur raw materials must be used, resulting in a sharp increase in cost and an economic imbalance.

[0063] Compared to sinter and lump ore, vanadium-titanium pellets, especially acidic oxidized pellets, have unique advantages such as low sulfur content, high permeability, and high strength. The sulfur content of vanadium-titanium pellets in the Panxi region is generally ≤0.05%, far lower than that of vanadium-titanium sinter and acidic lump ore. Increasing the proportion of pellets can directly reduce the overall sulfur content of the raw materials fed into the furnace.

[0064] S5 constructs a dynamic adjustment model based on the predicted sulfur load increment and the predicted titanium reduction increment, so that the oxygen enrichment rate is dynamically adjusted within the first threshold range and the blast furnace coal ratio is dynamically adjusted within the second threshold range.

[0065] Specifically, the dynamic adjustment model is as follows: O%=O_base+α1×(ΔS_pred-ε1)+β1×(ΔR_Ti_pred-ε2); C_ratio= C_base +α2×(ΔS_pred- ε3)-β2×(ΔR_Ti_pred-ε4); Where O% is the oxygen enrichment rate, C_ratio is the blast furnace coal ratio, O_base is the basic oxygen enrichment rate, C_base is the basic coal ratio, ΔS_pred is the predicted sulfur load increment, ΔR_Ti_pred is the predicted titanium reduction increment, α1, β1, α2 and β2 are adjustment coefficients, and ε1~ε4 are deviations.

[0066] In the smelting of high-titanium slag, fluctuations in sulfur load directly increase the demand for desulfurization, while fluctuations in titanium reduction indirectly interfere with desulfurization by affecting slag viscosity. Since there is a 30-60 minute lag between parameter adjustments and the observed effects in the blast furnace, traditional real-time feedback control often leads to uncontrolled desulfurization.

[0067] In this embodiment, a neural network is used to predict ΔS_pred (the future increase in sulfur load relative to the current value) and ΔR_Ti_pred (the future increase in titanium reduction relative to the current value) 30 minutes in advance. The oxygen enrichment rate and coal ratio are adjusted in advance through quantitative formulas, so that the desulfurization reaction can be initiated before the sulfur load increases and the titanium reduction deteriorates. The lag is transformed into a control advance, so as to prevent problems before they occur.

[0068] The role of oxygen enrichment rate is to increase the hearth temperature and slag oxygen potential. Its adjustment direction is positively correlated with ΔS_pred and ΔR_Ti_pred. When ΔS_pred > ε1, the sulfur load will increase, and 0% will increase. For every 0.1% increase in oxygen enrichment rate, the theoretical combustion temperature (Tf) increases by 8-10℃, the slag temperature increases simultaneously, the sulfur diffusion coefficient increases by 5-8%, and desulfurization kinetics are enhanced. At the same time, oxygen enrichment increases the oxygen potential in the furnace, promoting the forward desulfurization reaction.

[0069] When ΔR_Ti_pred > ε2, the titanium reduction degree will increase, with an additional 0% increase. High oxygen potential can inhibit the reduction reaction of titanium and reduce Ti³. + The conversion to Ti (C,N) reduces the risk of slag viscosity deterioration from the source.

[0070] In addition, the method for obtaining the predicted sulfur load increment and the predicted titanium reduction increment in this embodiment includes the following steps: S51 constructs a dataset, in which each sample includes input data and output data; the input data includes sulfur load-related parameters, titanium reduction-related parameters, and operational intervention parameters over N hours; the output data includes the corresponding sulfur load increment and titanium reduction-related increment over M hours after N hours.

[0071] More specifically, the sulfur load-related parameters include the total sulfur content of the ore fed into the furnace (%), the pellet ratio (%), the sulfur content of the sinter (%), the sulfur content of the coke (%), and the total amount of ore fed into the furnace (t / h). These parameters directly affect the trend of sulfur load changes.

[0072] Parameters related to titanium reduction include slag TiO2 content (%), physical heat of molten iron (°C), hearth temperature (°C), and blast energy (kg). The air volume (m / s) and airflow (m³ / min) affect the reduction reaction of titanium, thus determining the degree of titanium reduction. Operational intervention parameters include coal ratio (kg / TFe) and oxygen enrichment rate (%). Historical adjustments to these operational parameters will have a lag effect on sulfur load and titanium reduction, and need to be included in the model to capture the intervention effect.

[0073] For the 12 parameters mentioned above, the sampling frequency is 1 time / minute, and N is set to 180 data points / parameters over three hours. The data from the previous 3 hours is used as an input window, and the sulfur load and titanium reduction data for the next 0.5 hours are used as output labels to form a sample set.

[0074] S52 uses a dataset to train a prediction model.

[0075] The prediction model selected in this embodiment is LSTM. The reason is that the changes in sulfur load and titanium reduction in blast furnace smelting are affected by the cumulative effect of long-term historical parameters, and the gating mechanism of LSTM can effectively capture this long-term dependence.

[0076] Specifically, the prediction model is LSTM, including: The input layer has a dimension of number of features × number of time steps, i.e., 12 × 180.

[0077] The hidden layer consists of two LSTM network layers: the first LSTM network has 64 neurons and the second LSTM network has 32 neurons. Each LSTM network is followed by a Dropout layer to prevent overfitting.

[0078] The fully connected layer is a Dense layer with 16 neurons, and non-linear features are extracted using the ReLU activation function; The output layer consists of two neurons that output the predicted sulfur load and the predicted titanium reduction degree, respectively (when substituted into the dynamic adjustment mode, they need to be processed into the predicted sulfur load increment and the predicted titanium reduction degree increment).

[0079] When training the model, 70% of the samples are used as the training set, 20% as the validation set, and 10% as the test set.

[0080] The loss function is the mean squared error (MSE), which is given by the formula MSE = (1 / n)Σ[(y_pred - y_true)²].

[0081] The optimizer and training strategy use the Adam optimizer with early stopping to avoid overfitting; The accuracy target is achieved by training for 50-100 rounds until the prediction error of S_load on the test set is ≤5% and the prediction error of R_Ti is ≤8%.

[0082] S53 inputs the collected N hours of input data into the trained model to obtain the sulfur load increment and titanium reduction increment after N hours, M hours later.

[0083] Every 5 minutes, the system automatically collects the latest 12 parameters from the blast furnace control system and updates the data window for the previous 3 hours. Every 5 minutes, the system initiates LSTM model inference and outputs the predicted curves for sulfur load and titanium reduction degree for the next 30 minutes. The system uses a weighted average for the overlapping intervals of two consecutive predictions to eliminate high-frequency fluctuations and ensure curve smoothness.

[0084] In addition, this embodiment also includes an emergency adjustment mechanism to cope with sudden changes in sulfur load and titanium reduction.

[0085] Specifically, when the predicted sulfur load increment is greater than or equal to the first threshold, or / and the predicted titanium reduction increment is greater than or equal to the second threshold, the emergency adjustment model is activated to adjust the oxygen enrichment rate and blast furnace coal ratio.

[0086] More specifically, the emergency adjustment model is as follows: O%_pre=O_base_pre+α3×(ΔS_pred-ε5)+β3×MAX(0, ΔR_Ti_pred-ε6); C_ratio_pre= C_base_pre +α4×(ΔS_pred- ε7)-β4×MAX(0, ΔR_Ti_pred-ε8); Where O%_pre is the pre-adjusted value of oxygen enrichment rate, C_ratio_pre is the pre-adjusted value of blast furnace coal ratio, O_base is the pre-adjusted value of basic oxygen enrichment rate, C_base is the pre-adjusted value of basic coal ratio, ΔS_pred is the predicted sulfur load increment, ΔR_Ti_pred is the predicted titanium reduction increment, α3, β3, α4 and β4 are adjustment coefficients, and ε5~ε8 are deviations.

[0087] If the predicted increase in sulfur load is greater than or equal to the first threshold (e.g., ΔS_pred ≥ 0.5 kg / TFe), it means that the sulfur load will exceed the upper limit of 5.5 kg / TFe in the next 30 minutes. Conventional adjustments cannot offset the pressure of the increased sulfur load, and there is a risk that the sulfur content in the molten iron will exceed 0.14%. If the predicted increase in titanium reduction is greater than or equal to the second threshold (e.g., ΔR_Ti_pred ≥ 0.08), it means that the titanium reduction degree will exceed the critical value of 0.35 in the next 30 minutes, and the slag viscosity may increase from 10 Pa. s rose to 13 Pa Above s, desulfurization mass transfer is hindered.

[0088] To address the above situation, the emergency adjustment model introduces MAX(0, ΔR_Ti_pred-ε6) and MAX(0, ΔR_Ti_pred-ε8), which only initiate additional adjustments when the increase in titanium reduction truly exceeds the limit, thus avoiding excessive intervention.

[0089] When ΔR_Ti_pred≤ε6, such as when ε6=0.06, the MAX term is 0, and the titanium reduction degree does not contribute to the adjustment of the oxygen enrichment rate, thus avoiding unnecessary oxygen potential increases triggered by small fluctuations; when ΔR_Ti_pred>ε6, the MAX term=ΔR_Ti_pred-ε6, and the adjustment intensity increases linearly with the extent of exceeding the standard, achieving a precise response where the more severe the exceedance, the stronger the intervention.

[0090] The adjustment coefficients (α3, β3, α4, β4) of the emergency model are significantly larger than those of the conventional model (α1, β1, α2, β2). For example, α3≈0.5% / (kg / TFe) (conventional α1=0.3%), for every 0.5kg increase in sulfur load, the oxygen enrichment rate increases by an additional 0.1%, accelerating the increase in hearth temperature; β3≈0.4% / 0.1R_Ti (conventional β1=0.2%), for every 0.1% increase in titanium reduction, the oxygen enrichment rate increases by an additional 0.2%, more strongly inhibiting titanium reduction. The amplified coefficients make the emergency adjustment response faster than the conventional adjustment, ensuring that the deterioration trend is contained within 30 minutes.

[0091] To address the issue that sulfur load will severely exceed the limit if ΔS_pred ≥ ε5, O%_pre is significantly increased by α3 × (ΔS_pred - ε5). For every 0.5% increase in oxygen enrichment, the theoretical combustion temperature (Tf) rises by 40-50℃, the slag temperature rises simultaneously to 1470-1480℃, the sulfur diffusion coefficient increases by 20-25%, and the desulfurization reaction rate accelerates, offsetting the pressure of increased sulfur load.

[0092] To address the issue that titanium reduction will severely exceed the limit if ΔR_Ti_pred ≥ ε6, O%_pre is additionally increased by β3×MAX(0, ΔR_Ti_pred-ε6). This increases the oxygen potential and Ti³ + To Ti 4+The oxidation reaction rate increases, the amount of Ti (C,N) generated decreases, and the increase in slag viscosity is controlled.

[0093] Similarly, for cases where the sulfur load will be severely exceeded if ΔS_pred≥ε7, C_ratio_pre is significantly increased by α4×(ΔS_pred-ε7). The increased coal ratio releases additional heat, ensuring that the physical heat of molten iron is maintained above 1450℃, which is higher than the conventional 1440℃, thus providing thermodynamic support for high-intensity desulfurization.

[0094] To address the issue that titanium reduction will severely exceed the limit if ΔR_Ti_pred≥ε8, C_ratio_pre is forcibly reduced by -β4×MAX (0, ΔR_Ti_pred-ε8). This reduces the coal ratio, which in turn reduces the fixed carbon content in the furnace, weakens the reducing atmosphere, and decreases the titanium reduction reaction rate. Combined with the increase in oxygen enrichment rate, this forms a synergistic inhibition of low reduction and high oxygen potential.

[0095] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0096] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model, used in blast furnace smelting of high-vanadium-titanium ore, characterized in that... include: By weight, the MgO content in the slag should be controlled at 8.0%-9.5%, the binary basicity of the slag should be controlled at 1.10-1.17, the Al2O3 content should be ≤14.5%, and the ratio of MgO to Al2O3 in the slag should be maintained at 0.60-0.

65. Maintain the physical heat of molten iron ≥1440℃; control the blast furnace inlet temperature ≥1200℃; The sulfur load in the blast furnace should be controlled at 4.5-5.5 kg, and the proportion of pellets in the blast furnace should be controlled at 46%-50%. A dynamic adjustment model is constructed based on the predicted sulfur load increment and the predicted titanium reduction increment, so that the oxygen enrichment rate is dynamically adjusted within the first threshold range and the blast furnace coal ratio is dynamically adjusted within the second threshold range.

2. The method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 1, characterized in that... The dynamic adjustment model is as follows: O%=O_base+α1×(ΔS_pred-ε1)+β1×(ΔR_Ti_pred-ε2); C_ratio= C_base +α2×(ΔS_pred- ε3)-β2×(ΔR_Ti_pred-ε4); Where O% is the oxygen enrichment rate, C_ratio is the blast furnace coal ratio, O_base is the basic oxygen enrichment rate, C_base is the basic coal ratio, ΔS_pred is the predicted sulfur load increment, ΔR_Ti_pred is the predicted titanium reduction increment, α1, β1, α2 and β2 are adjustment coefficients, and ε1~ε4 are deviations.

3. The method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 2, characterized in that... The methods for obtaining the predicted sulfur load increment and the predicted titanium reduction increment include: A dataset is constructed, wherein each sample in the dataset includes input data and output data; the input data includes sulfur load-related parameters, titanium reduction degree-related parameters, and operational intervention parameters over N hours; the output data includes the corresponding sulfur load increment and titanium reduction degree increment over M hours after N hours. Train a prediction model using the dataset; The collected N hours of input data are fed into the trained model to obtain the sulfur load increment and titanium reduction increment after N hours (M hours).

4. The method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 2, characterized in that: The sulfur load-related parameters include the total sulfur content of the ore fed into the furnace, the pellet ratio, the sulfur content of the sinter, the sulfur content of the coke, and the total amount of ore fed into the furnace; the titanium reduction degree-related parameters include the TiO2 content of the slag, the physical heat of the molten iron, the hearth temperature, the blast kinetic energy, and the air volume; the operational intervention parameters include the blast furnace coal ratio and the oxygen enrichment rate.

5. The method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 3, characterized in that: The prediction model is LSTM, including: The input layer has a dimension of number of features × number of time steps; The hidden layer consists of two LSTM network layers: the first LSTM network has 64 neurons and the second LSTM network has 32 neurons; each LSTM network is followed by a Dropout layer. The fully connected layer is a Dense layer with 16 neurons, and non-linear features are extracted using the ReLU activation function; The output layer consists of two neurons that output the predicted sulfur load increment and the predicted titanium reduction increment, respectively.

6. The method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 1, characterized in that: The first threshold range is 4.5%-6.5%, and the second threshold range is 155-170 kg / ton of iron.

7. The method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 2, characterized in that: When the predicted sulfur load increment is greater than or equal to the first threshold, or / and the predicted titanium reduction increment is greater than or equal to the second threshold, the emergency adjustment model is activated to adjust the oxygen enrichment rate and blast furnace coal ratio.

8. The method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 7, characterized in that... The emergency adjustment model is as follows: O%_pre=O_base_pre+α3×(ΔS_pred-ε5)+β3×MAX(0, ΔR_Ti_pred-ε6); C_ratio_pre= C_base_pre +α4×(ΔS_pred- ε7)-β4×MAX(0, ΔR_Ti_pred-ε8); Where O%_pre is the pre-adjusted value of oxygen enrichment rate, C_ratio_pre is the pre-adjusted value of blast furnace coal ratio, O_base is the pre-adjusted value of basic oxygen enrichment rate, C_base is the pre-adjusted value of basic coal ratio, ΔS_pred is the predicted sulfur load increment, ΔR_Ti_pred is the predicted titanium reduction increment, α3, β3, α4 and β4 are adjustment coefficients, and ε5~ε8 are deviations.

9. The method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 1, characterized in that: Vanadium-titanium ore is pretreated by roasting it at 800-900℃ with oxygen-enriched air for 1.5-2.0 hours.

10. A method for improving the desulfurization capacity of high-titanium slag using an artificial intelligence model according to claim 2, characterized in that: During the roasting process, 0.8-1.2% of CaO-Al2O3 composite desulfurizing agent is added to the vanadium-titanium ore.

Images