Dynamic proportioning method for recovering tungsten from alkali roasting of smelting furnace slag based on on-line analysis of slag composition

Abstract

The application discloses a dynamic dosing method for recovering tungsten in a fuming furnace based on online analysis of slag components, which comprises the following steps: S1. At the end of sulfuration and volatilization, the composition information of the current furnace slag is obtained through online analysis; S2. Based on the obtained slag composition information, the optimal amount of Na2CO3 for the current furnace is calculated by using a preset calculation model; S3. The optimal amount of Na2CO3 for the current furnace is calculated, and the amount of Na2CO3 added to the fuming furnace is automatically controlled. The application realizes real-time and accurate regulation and control of alkali fusion dosing, reduces the tungsten conversion rate fluctuation from ±15% of fixed dosing to within ±3%, and at the same time, guarantees the stability of slag fluidity and water quenching performance, and significantly improves the process stability and alkali utilization rate of the tungsten recovery process.

Description

Technical Field

[0001] This invention belongs to the field of non-ferrous metal smelting process control technology, specifically relating to a dynamic batching method for tungsten recovery by alkaline melting in a fuming furnace based on online analysis of slag composition. It is particularly suitable for real-time and precise control of the amount of Na2CO3 added during the alkaline melting recovery of tungsten in the slag of a tin smelting furnace. Background Technology

[0002] In tin smelting, the fuming furnace is a key piece of equipment for processing tin-containing materials such as tin ore and tin slag. The fuming furnace slag contains a considerable amount of tungsten (mainly scheelite CaWO4). By adding Na2CO3 to the molten slag at the end of the sulfidation and volatilization phase in the fuming furnace for alkaline fusion conversion, the CaWO4 in the slag can be converted into soluble Na2WO4. Subsequently, water quenching allows the Na2WO4 to enter the aqueous solution, from which the tungsten can be recovered. This technology utilizes the existing high-temperature heat source of the fuming furnace, eliminating the need for a separate sintering facility, and offers advantages such as low investment and low energy consumption.

[0003] However, significant inter-furnace fluctuations exist in the fuming furnace production process. Due to variations in the source and proportion of raw materials (tin middlings, tin slag, return materials, etc.), the slag quantity, tungsten content, and gangue composition (CaO, SiO2, Al2O3, FeO, etc.) of the slag produced in each furnace exhibit substantial variations. Taking a tin smelter as an example, the W content in the water-quenched slag varies from 0.35% to 0.70%, the Ca content from 4.5% to 8.0%, and the SiO2 content from 10% to 16% across different furnace batches. This compositional fluctuation leads to the following problems when using a fixed Na2CO3 addition amount: (1) When the W content in the slag of a certain furnace is too high and the amount of Na2CO3 added is insufficient, CaWO4 cannot be completely converted into Na2WO4, the tungsten recovery rate drops significantly, resulting in resource waste; (2) When the W content in the slag of a certain furnace is too low and the amount of Na2CO3 added is too high, the excessive Na2CO3 entering the slag will cause changes in slag shape, increase basicity, and worsen the fluidity and water quenching performance of the slag. In severe cases, it may cause production accidents such as furnace condition deterioration and tuyer blockage; (3) When the CaO content in the gangue is high, additional Na2CO3 is consumed. When the SiO2 content is high, it participates in the alkali fusion reaction (generating CaSiO3). When the Al2O3 content is high, sodium aluminate is generated, consuming alkali. These fluctuations in gangue composition further exacerbate the complexity of alkali matching.

[0004] Currently, the batching operation for tungsten recovery through alkaline fusion in fuming furnaces mainly relies on the experience and judgment of operators, using a fixed Na2CO3 addition amount or manual adjustments based on pre-furnace sampling and analysis results. Using a fixed Na2CO3 addition amount cannot accommodate compositional fluctuations between furnace runs, and pre-furnace sampling and analysis suffers from analytical lag (it typically takes 2-4 hours from sampling to results), failing to provide real-time guidance for adjusting the batching for each furnace run. Existing technologies lack methods and systems for online analysis of fuming furnace slag composition and dynamic calculation of the optimal Na2CO3 addition amount. Summary of the Invention

[0005] This invention addresses the technical challenge of balancing tungsten conversion rate and slag stability when the slag composition fluctuates during alkali fusion tungsten recovery in fuming furnaces. It provides a dynamic batching method for alkali fusion tungsten recovery in fuming furnaces based on online slag composition analysis.

[0006] The technical solution adopted in this invention is as follows: A dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis is described. This method involves adding Na₂CO₃ to the molten slag in the furnace at the end of the sulfidation and volatilization phase of tin smelting in the fuming furnace to perform alkaline fusion conversion and recover tungsten from the slag. The method includes the following steps: S1. Online acquisition of slag composition information: During the final stage of sulfidation and volatilization in tin smelting in a fuming furnace, the composition information of the slag for the current furnace batch is acquired through an online analysis system. The slag composition information includes at least W content and Ca content. The online analysis system is one or a combination of XRF spectroscopy online analysis system and LIBS laser-induced breakdown spectroscopy online analysis system. S2. Based on the acquired slag composition information, calculate the optimal amount of Na2CO3 added for the current furnace using a preset calculation model; the calculation model is: Theoretical Na2CO3 addition amount = k × m 渣 × w 钨 × M Na2CO3 / M CaWO4 Where k is a correction factor, ranging from 1.2 to 2.0; m 渣 The current slag quantity; w 钨 M represents the mass fraction of tungsten in the slag. Na2CO3 M is the molar mass of Na₂CO₃. CaWO4 Here is the molar mass of CaWO4; S3. Based on the calculated optimal amount of Na2CO3, the amount of Na2CO3 added to the fuming furnace is automatically controlled by the automatic control system.

[0007] Furthermore, the slag composition information mentioned in step S1 also includes SiO2 content, Fe content, and Al2O3 content.

[0008] Furthermore, in the calculation model of step S2, the correction coefficient k is dynamically adjusted according to the Ca / Si / Al gangue content in the slag. The normal content is defined as SiO2 10%–16%, Al2O3 2.5%–4.5%, and CaO 6%–8% by mass. When the Ca content is high, k is 1.6–2.0; when the Ca content is normal, k is 1.3–1.6; and when the Ca content is low, k is 1.2–1.4. When the SiO2 content is high, k is decreased by 0.1–0.2; and when the Al2O3 content is high, k is increased by 0.1–0.2.

[0009] Further, in step S2, slag stability is verified. The verification method is to first calculate the basicity of the slag after adding Na2CO3, A / C = m. Al2O3 / m CaO m Al2O3 m is the mass of Al2O3. CaO The quality of CaO is considered; when the A / C ratio exceeds the set threshold of 0.6, reduce the amount of Na2CO3 added or add CaO to adjust the slag type.

[0010] Furthermore, the automatic control system is configured with an online analysis layer, a data acquisition layer, a batching calculation layer, and an execution control layer. The online analysis layer includes an online analysis system. The data acquisition layer includes a data acquisition and calculation module consisting of a data acquisition unit and a slag type verification submodule. The batching calculation layer includes a batching calculation module consisting of an alkali quantity calculation submodule, a Na2CO3 solution feeding device, and a slag type verification submodule. The data acquisition layer and the batching calculation layer share a single slag type verification submodule, which receives slag composition parameters and slag quantity input data from the data processing unit and performs slag type stability verification. The execution control layer and the fuming furnace control system PLC / DCS achieve bidirectional data interaction via an industrial communication protocol. After obtaining slag composition information, the online analysis layer inputs it to the data acquisition layer. The data acquisition and processing unit filters and reduces noise from the input data, and then, combined with manually input or automatically estimated slag quantity, the alkali quantity calculation submodule of the batching calculation module first calculates the slag quantity according to formula m. Na2CO3 =k×m 渣 ×w 钨 Calculate the theoretical alkali content by multiplying by 0.3681, then use A / C=m Al2O3 / m CaO For values ​​≤0.6, perform slag type verification. If the A / C value exceeds the limit, reduce the amount of alkali or add CaO. The final output is the optimized amount of Na2CO3 added, m. Na2CO3 m is the theoretical alkali content. 渣 w represents the current slag quantity for the furnace run. 钨 m represents the mass fraction of tungsten in the slag. Al2O3m is the mass of Al2O3. CaO To ensure the quality of CaO, the execution control layer receives three signals from the fuming furnace control system PLC / DCS: furnace temperature, slag level height, and tuyere pressure. It then comprehensively judges the timing and rate of addition and issues a feeding command to the fuming furnace control system PLC / DCS. The fuming furnace control system PLC / DCS continuously adds the alkali material into the fuming furnace in batches through a weighing or volume system with a metering accuracy of ±5%, thus achieving closed-loop automatic control.

[0011] Furthermore, in step S1, when the online analysis method is XRF spectroscopy, the XRF probe is installed on the furnace wall of the fuming furnace to measure the composition of the slag in real time, with measurements taken every 1 to 5 minutes. When the online analysis method is LIBS laser-induced breakdown spectroscopy, the probe of the LIBS laser-induced breakdown spectroscopy online analysis system is aimed at the surface of the molten pool inside the furnace for in-situ measurement to obtain slag composition information, with measurements taken every 10 to 60 seconds.

[0012] Furthermore, the amount of Na2CO3 added in step S3 is controlled with an accuracy of ±5%, and precise measurement is achieved through a weighing and / or volumetric metering system.

[0013] Furthermore, the automatic control system is also equipped with a historical data self-learning module connected to the PLC / DCS control system of the fuming furnace. This module performs self-learning optimization on historical data. The optimization method involves storing the slag composition, Na2CO3 addition amount, tungsten conversion rate, and slag type parameters for each furnace in a database. Data mining and regression analysis are used to optimize the value strategy of the correction coefficient k' and the slag type stability verification threshold. The optimization strategy for the correction coefficient k' is a continuous function: k' = a + b×(Ca%-Ca0) - c×max(SiO2%-SiO2). 20 , 0) +d×max(Al2O3%-Al2O 30 , 0), where a, b, c, d are regression coefficients, CaO, SiO 20 Al2O 30 The baseline contents of Ca, SiO2, and Al2O3 are respectively, and k' is determined by multiple regression analysis; Ca%, SiO2%, and Al2O3% are the mass contents of Ca, SiO2, and Al2O3 in the slag obtained from the online analysis of the current furnace.

[0014] Compared with the prior art, the present invention has the following beneficial effects: (1) Achieve precise batching: The fluctuation range of tungsten conversion rate is reduced from ±15% of fixed batching to within ±3%.

[0015] (2) Maintaining slag stability: Slag stability verification effectively avoids problems such as slag damage and decreased fluidity caused by excessive alkali.

[0016] (3) Online analysis and real-time response: The measurement cycle is 10s to 5min, which is much faster than the 2 to 4h lag time of traditional sampling chemical analysis, thus enabling real-time monitoring of slag composition.

[0017] (4) Historical data self-learning optimization: Na2CO3 utilization rate increased from about 75% initially to over 90%. Attached Figure Description

[0018] Figure 1 This is a flowchart of the dynamic batching method for tungsten recovery by alkaline melting in a fuming furnace based on online analysis of slag composition, as described in this invention. Figure 2 This is a schematic diagram of the architecture linking the online analysis system with the flue gas furnace control system. Detailed Implementation

[0019] The present invention will be further explained below with reference to the embodiments. The following embodiments do not limit the scope of protection of the present invention in any way. All technical solutions obtained by equivalent substitution are within the scope of protection of the present invention. Specific conditions not specified in the embodiments are performed according to conventional conditions. Example 1

[0020] A tin smelter's tin fuming furnace produces approximately 20 tons of slag per furnace. The slag composition varies as follows: W 0.35%~0.70%, Ca 4.5%~8.0%, SiO2 10%~16%, Al2O3 2.5%~5.0%.

[0021] See Figure 1 The illustrated process describes a dynamic batching method for tungsten recovery via alkaline fusion in a fuming furnace, based on online slag composition analysis. This method involves adding Na₂CO₃ to the molten slag at the end of the sulfidation and volatilization phase in the fuming furnace to perform alkaline fusion conversion and recover tungsten from the slag. The method includes the following steps: S1. Online acquisition of slag composition information: An XRF online analysis probe is installed on the furnace wall of the fuming furnace to measure the slag composition in real time, with a measurement cycle of 3 minutes / time. The slag composition obtained by XRF online analysis at the end of the sulfidation and volatilization of a certain batch is W 0.55%, Ca 6.3%, SiO2 12.0%, and Al2O3 3.8%.

[0022] S2. Dynamic Calculation of Optimal Na2CO3 Addition Amount: Based on the acquired slag composition information, the optimal Na2CO3 addition amount for the current furnace is calculated using a preset calculation model. The slag volume for this furnace is m. 渣=20t. Ca content 6.3%, within the normal range, k = 1.4; SiO2 12.0%, normal, no adjustment needed; Al2O3 3.8%, normal, no adjustment needed. m Na2CO3 =1.4×20×0.0055×105.99 / 287.93=56.8kg. CaO content is about 8.9%, Al2O3 content is about 3.8%, A / C=3.8 / 8.9=0.43, which is lower than the threshold of 0.6, and the slag type is stable.

[0023] S3. The automatic control system adds Na2CO3 in three batches of 56.8 kg each, with a feeding accuracy of ±5%. After addition, Na2CO3 undergoes an alkali-fusion reaction with the slag, converting into added Na2WO4. The tungsten conversion rate for this furnace was 72.5%, the slag shape was normal, water quenching was smooth, and alkali consumption was reasonable. Statistics from 30 consecutive furnaces show that the tungsten conversion rate fluctuated between 68.5% and 73.8%, with a fluctuation range of only ±2.7%.

[0024] The automatic control system is described in [reference]. Figure 2 It adopts a four-layer architecture, consisting of an online analysis layer, a data acquisition layer, a batching calculation layer, and an execution control layer, from top to bottom. Data communication between the layers is achieved via industrial Ethernet or fieldbus. Specifically, it includes the following components and their connections: (1) Online analysis layer, including: XRF Online Analysis System: The XRF probe is installed on the furnace wall of the fuming furnace to measure the slag flow in real time. The measurement cycle is 1 to 5 minutes / time. The elements analyzed are W, Ca, Si, Fe and Al. The output slag composition data (mass percentage) is output. LIBS Online Analysis System: The probe of the LIBS laser-induced breakdown spectroscopy online analysis system is installed on the furnace wall (directly opposite the XRF probe installation position), and is aimed at the surface of the molten pool inside the furnace to perform in-situ measurements. The measurement cycle is 10-60s / time. The elements analyzed are W, Ca, Si, Fe, and Al, and the output slag composition data (mass percentage) is provided. The two systems can operate independently or in parallel. When operating in parallel, the analysis data undergoes weighted averaging processing via the data fusion module (data weight of 0.6 for both the XRF online analysis system and the LIBS laser-induced breakdown spectroscopy online analysis system), and the fused component data is output to the data acquisition layer.

[0025] (2) Data acquisition layer, including data acquisition and processing unit and slag type verification submodule: Data acquisition and processing unit: Receives the component data fused by the online analysis layer, performs signal filtering and noise reduction processing, and completes the component data fusion calculation; Slag type verification submodule: Receives the output of the data processing unit and outputs slag composition parameters such as W%, Ca%, SiO2%, and Al2O3%, while receiving slag quantity input data (manual input or automatic estimation). Feedback signals are received from the batching calculation layer through the "model optimization feedback" connection, which are used for the optimization of model parameters in the next batch.

[0026] (3) Batching calculation layer, which is set up with a batching calculation module consisting of an alkali calculation submodule, a Na2CO3 solution feeding device and a slag type verification submodule (within the dashed box).

[0027] Alkali content calculation submodule: Based on formula m Na2CO3 =k×m 渣 ×w 钨 The theoretical alkali content is calculated using ×0.3681, and the k value is dynamically adjusted based on the contents of Ca, SiO2, and Al2O3. Slag type verification submodule: Performs slag type stability verification. The verification method is to first calculate the basicity of the slag after adding Na2CO3, A / C = m. Al2O3 / m CaO m Al2O3 m is the mass of Al2O3. CaO The mass of CaO is considered; when the A / C ratio exceeds the set threshold of 0.6, the amount of Na2CO3 added is reduced or CaO is added to adjust the slag type. The verified amount of Na2CO3 added is output to the Na2CO3 solution feeding device.

[0028] The Na2CO3 solution feeding device can use an existing automatic feeding device. After receiving the optimal Na2CO3 addition instruction from the batching calculation module, the Na2CO3 solution stored in the storage tank is accurately measured by the volume metering system (metering accuracy ±5%) of the feeding device and sent to the slag in the fuming furnace through a metering pump / pneumatic conveying device to achieve continuous feeding (feeding time is about 20 minutes).

[0029] The data acquisition layer and the batching calculation layer share a slag type verification submodule, which receives the slag composition parameters and slag quantity input data output by the data processing unit and performs slag type stability verification.

[0030] (4) Execution control layer: The execution control layer and the fuming furnace control system PLC / DCS achieve bidirectional data interaction via an industrial communication protocol. The execution control layer also includes a historical data self-learning module connected to the fuming furnace control system PLC / DCS, which performs self-learning optimization on historical data. Its built-in database stores data such as slag composition, regression analysis optimization correction coefficient k' value, tungsten conversion rate, and slag type parameters for each furnace. The optimization method involves storing the slag composition, Na2CO3 addition amount, tungsten conversion rate, and slag type parameters for each furnace in the database, and then optimizing the value strategy of the correction coefficient k' and the slag type stability verification threshold through data mining and regression analysis. The optimization strategy for the correction coefficient k' is a continuous function: k' = a + b×(Ca%-Ca0) - c×max(SiO2%-SiO2). 20 , 0) +d×max(Al2O3%-Al2O 30 , 0), where a, b, c, d are regression coefficients, CaO, SiO 20 Al2O 30 The baseline contents of Ca, SiO2, and Al2O3 are respectively, and k' is determined through multiple regression analysis; Ca%, SiO2%, and Al2O3% are the mass contents of Ca, SiO2, and Al2O3 in the slag obtained from the online analysis of the current furnace. The k-value function and A / C threshold are optimized through regression analysis, and the optimization results are fed back to the batching calculation layer.

[0031] The online analysis layer acquires slag composition information, which is then fused and transmitted to the data acquisition layer. The data acquisition and processing unit filters and reduces noise from the input data, and combines it with manually input or automatically estimated slag quantity. The alkali quantity calculation submodule of the batching calculation module first calculates the alkali quantity according to formula m. Na2CO3 =k×m 渣 ×w 钨 Calculate the theoretical alkali content by multiplying by 0.3681, then use A / C=m Al2O3 / m CaO For values ​​≤0.6, perform slag type verification. If the A / C value exceeds the limit, reduce the amount of alkali or add CaO. The final output is the optimized amount of Na2CO3 added, m. Na2CO3 m is the theoretical alkali content. 渣 w represents the current slag quantity for the furnace run. 钨 m represents the mass fraction of tungsten in the slag. Al2O3 m is the mass of Al2O3. CaOFor the quality of CaO. After the batching calculation layer completes the alkali dosage calculation and slag type verification, it outputs the optimal Na2CO3 addition command to the execution control layer, and simultaneously stores the operating data in the historical database. The historical data self-learning module optimizes the k' value function and A / C threshold through regression analysis, and feeds back to the batching calculation layer through the "model optimization feedback" connection to form a closed-loop optimization. After receiving three signals from the fuming furnace control system PLC / DCS—furnace temperature, slag surface height, and tuyere pressure—the execution control layer comprehensively judges the timing and rate of addition, and then issues a feeding command to the fuming furnace control system PLC / DCS. The fuming furnace control system PLC / DCS continuously adds the alkali into the fuming furnace in batches through a weighing or volume system with a metering accuracy of ±5%, realizing closed-loop automatic control. Example 2

[0032] Similar to the fuming furnace in Example 1, but using LIBS laser-induced breakdown spectroscopy for online analysis. The probe of the LIBS laser-induced breakdown spectroscopy online analysis system is installed on the furnace wall (directly opposite the XRF installation position), and is aimed at the surface of the molten pool inside the furnace to perform in-situ measurements, with a measurement cycle of 30 seconds per measurement.

[0033] The slag composition of a certain furnace batch, analyzed by LIBS, contained 0.68% W, 7.8% Ca, 14.5% SiO2, and 4.6% Al2O3, with a slag volume of 22t. The Ca content of 7.8% was too high, so k was set to 1.7; the SiO2 content of 14.5% was too high, so k was adjusted down by 0.15 to 1.55; and the Al2O3 content of 4.6% was too high, so k was adjusted up by 0.1 to 1.65. Na2CO3 =1.65×22×0.0068×0.3681=90.7kg. CaO content is approximately 10.9%, Al2O3 content is approximately 4.6%, A / C=4.6 / 10.9=0.42, which is lower than the threshold of 0.6. Na2CO3 was added to the molten slag in the furnace in four batches according to 90.7kg, and the tungsten conversion rate was 70.8%, with normal slag shape. The W content and gangue composition of this furnace were both high, and the dynamic batching automatically adjusted the alkali amount to ensure the conversion rate. If a fixed batch of 60kg were used, the tungsten conversion rate would only be about 48%. Example 3

[0034] Similar to the fuming furnace in Example 1, a dual system is also installed, consisting of an XRF online analysis probe (located on the furnace wall) and a LIBS laser-induced breakdown spectroscopy online analysis system (located on the furnace wall directly opposite the XRF installation location). See also Figure 2 XRF weight 0.4, LIBS weight 0.6.

[0035] XRF analysis of the slag from a certain furnace batch showed a W content of 0.42% and a Ca content of 5.0%; LIBS analysis showed a W content of 0.44% and a Ca content of 5.2%. The weighted average was: W = 0.6 × 0.42% + 0.4 × 0.44% = 0.428%, Ca = 0.6 × 5.0% + 0.4 × 5.2% = 5.08%. The slag volume was 18 tons. The Ca content of 5.08% was slightly low, so k was set to 1.3; SiO2 and Al2O3 were normal and no adjustment was needed. Na2CO3 =1.3×18×0.00428×0.3681=36.8kg. A / C=0.42, lower than the threshold of 0.6. Adding 36.8kg results in a tungsten conversion rate of 71.5%, and the alkali consumption is reasonable. The weighted average of the dual analysis system improves the accuracy and reliability of the analysis, avoiding measurement biases that may occur with a single system. If a fixed addition of 70kg (set for high W content) is used, the alkali content is severely excessive, the A / C rises to 0.58, close to the threshold, and the slag fluidity deteriorates. Example 4

[0036] Using the fuming furnace from Example 1 as an example, after 200 heat runs, the database accumulated data on slag composition, Na₂CO₃ addition, tungsten conversion rate, and slag type parameters for each heat run. Multiple regression analysis was performed on the 200 heat run data to establish regression models between the tungsten conversion rate and each factor. (See [link to relevant documentation]). Figure 1 The system performs self-learning optimization on historical data, storing the slag composition, Na2CO3 addition amount, tungsten conversion rate, and slag type parameters for each furnace in a database. Through data mining and regression analysis, it optimizes the value strategy of the correction coefficient k' and the slag type stability verification threshold. The regression analysis optimizes the value strategy of the correction coefficient k' into a continuous function: k' = a + b×(Ca%-Ca0) - c×max(SiO2%-SiO2). 20 , 0) +d×max(Al2O3%-Al2O 30 , 0), where a, b, c, d are regression coefficients, CaO, SiO 20 Al2O 30 The baseline contents of Ca, SiO2, and Al2O3 were determined through multiple regression analysis. Ca%, SiO2%, and Al2O3% represent the mass contents of Ca, SiO2, and Al2O3 in the slag obtained from the online analysis of the current furnace.

[0037] Analysis results show that the optimal value of the correction coefficient k' in regression analysis has an approximately linear relationship with the Ca content; the alkali saving effect is significant when SiO2 > 13%; the original threshold A / C = 0.6 is too conservative and can be relaxed to 0.65 in practice. After optimization, the strategy for selecting the value of the correction coefficient k' is optimized from three discrete values ​​to a continuous function: k' = 1.15 + 0.12 × (Ca% - 4.0) - 0.05 × max(SiO2% - 13, 0) + 0.08 × max(Al2O3% - 4.0, 0). The slag type threshold is adjusted from 0.6 to 0.65. After optimization, the average tungsten conversion rate of 100 heats increased from 69.8% to 72.3%, the Na2CO3 utilization rate increased from 76% to 91%, and the fluctuation range of tungsten conversion rate was further narrowed to within ±2.5%. Example 5

[0038] The same fuming furnace as in Example 1, the online analysis results for a certain batch were W 0.62%, Ca 4.8%, SiO2 10.2%, Al2O3 5.8%, and slag volume 19t. The Ca content of 4.8% was too low, so k was set to 1.25; the Al2O3 content of 5.8% was too high, so k was adjusted upward by 0.15 to 1.4. Na2CO3 =1.4×19×0.0062×0.3681=60.4kg. CaO content is approximately 6.7%, Al2O3 5.8%, A / C=5.8 / 6.7=0.87, exceeding the threshold of 0.6. The system triggers a slag type warning. Maintain the Na2CO3 addition at 60.4kg, while adding 15kg of CaO to reduce the A / C value to 0.60. After adding CaO, the A / C drops to 0.60, the slag type stabilizes, and the tungsten conversion rate is 71.2%. If 60.4kg of Na2CO3 is added directly without adjustment, the A / C=0.87 will cause severe vitrification of the slag, making the slag blocks difficult to break after water quenching, hindering tungsten dissolution, and resulting in an actual conversion rate of only about 55%. Comparative Example 1

[0039] The same fuming furnace and slag as in Example 1 were used, without any online analysis equipment. Based on operational experience, 55 kg of Na2CO3 was added per furnace (approximately corresponding to a median W of 0.50%), without considering compositional fluctuations between furnace cycles.

[0040] The production data for 10 consecutive furnaces is as follows: Furnace 1: W 0.35%, excess alkali, tungsten conversion rate 69.5%, slag A / C ratio increased to 0.56. Furnace 2: W 0.70%, insufficient alkali, tungsten conversion rate 52.1%, a large amount of tungsten not converted. Furnace 3: W 0.45%, slightly higher alkali content, tungsten conversion rate 70.2%. Furnace 4: W 0.68%, insufficient alkali, tungsten conversion rate 54.8%. The tungsten conversion rates of the 10 furnaces ranged from 52.1% to 74.2%, with a fluctuation range of ±11.1%. Two of the furnaces had an A / C ratio greater than 0.6 due to excessive alkali, resulting in poor slag fluidity and difficulty in water quenching. Compared to the fluctuation range of ±2.7% in Example 1, this comparative example showed greater conversion rate fluctuations and higher slag type risk, verifying the inherent defects of the fixed feed method.

[0041] Comparative Example 2

[0042] The same fuming furnace as in Example 1 was used. No online analysis equipment was installed; the amount of Na2CO3 added was adjusted based on the results of pre-furnace sampling and testing. It took 3 hours from sampling to obtaining results, and the batching for the current furnace could only be adjusted based on the test results of the previous furnace.

[0043] The actual composition of a certain batch was 0.68% W and 7.8% Ca (the same as in Example 2), but the test results of the previous batch were 0.45% W and 5.5% Ca. Based on the results of the previous batch, the operators calculated that approximately 47 kg of Na2CO3 should be added, with k=1.4 and W=0.45%.

[0044] The actual alkali requirement for this furnace was approximately 90.7 kg (calculated according to Example 2), but only 47 kg was added, resulting in a severe deficiency and a tungsten conversion rate of only 43.5%. By the time the test results came out, the furnace had already been slag discharged, and the tungsten had been lost along with the slag.

[0045] This comparative example shows that even with testing methods, the 2-4 hour analytical lag cannot provide real-time guidance, and ingredient adjustments are always "a step behind," which can easily lead to serious errors in alkali dosage when there are large fluctuations in composition.

[0046] Comparative Example 3

[0047] The same fuming furnace and online analysis system as in Example 1 were used, but the correction coefficient k in the calculation model was fixed at 1.4 and was not adjusted according to the contents of Ca, SiO2, and Al2O3.

[0048] The online analysis results for a certain furnace batch were: W 0.65%, Ca 7.8% (high), SiO2 14.5% (high), and Al2O3 4.8% (high). With k=1.4 fixed, m was calculated. Na2CO3 =1.4×20×0.0065×0.3681=67.0kg.

[0049] According to the dynamic adjustment rules, if Ca is too high, k should be set to 1.7; if SiO2 is too high, it should be adjusted to 1.55; and if Al2O3 is too high, it should be adjusted to 1.65. The actual optimal k is 1.65, which corresponds to an alkali amount of 79.0 kg.

[0050] When k=1.4, 67.0 kg of alkali was added, which is 12.0 kg less than the optimal amount of 79.0 kg. This resulted in insufficient alkali, leading to a tungsten conversion rate of only 61.2%, approximately 10 percentage points lower than the dynamically adjusted 71.0%. This comparative example verifies the necessity of dynamically adjusting the k value according to the gangue composition—ignoring fluctuations in gangue composition will lead to significant deviations in alkali content.

[0051] Comparative Example 4

[0052] The same fuming furnace and online analysis system as in Example 1 were used. The calculation model was running normally, but the slag stability verification function was turned off. Na2CO3 was added only according to the calculation results, and no A / C verification or adjustment was performed.

[0053] The composition of a certain batch was W 0.62%, Ca 4.8%, and Al2O3 5.8% (same as in Example 5). Dynamic calculation yielded m. Na2CO3 =60.4kg. Because the slag type verification was turned off, it was added directly at 60.4kg, without adding CaO.

[0054] After addition, the slag A / C ratio reached 0.87, far exceeding the safety threshold. The slag was viscous and had extremely poor fluidity during discharge. Water quenching resulted in a large number of glassy slag lumps, from which tungsten could not be dissolved, resulting in an actual tungsten conversion rate of only 55.1%. Simultaneously, a large amount of insufficiently water-quenched slag accumulated at the bottom of the slag pool, requiring shutdown for cleaning and disrupting production continuity.

[0055] This comparative example shows that slag stability verification is a key step in ensuring process safety. Ignoring verification may not only reduce tungsten recovery rate but also cause production accidents.

[0056] Comparative Example 5

[0057] The same fuming furnace as in Example 1, with the online analysis system and dynamic batching model operating normally, but the historical data self-learning optimization function is turned off, and the value strategy of the regression analysis optimization correction coefficient k' and the slag type verification threshold remain unchanged within 500 furnace runs.

[0058] In the initial stage of operation (first 100 furnaces): the k' value strategy was set based on theoretical analysis, and the average tungsten conversion rate was 68.5%, which was acceptable.

[0059] Mid-term operation (200-300 furnaces): As the source of raw materials for the furnace changes, the Al2O3 content in the slag gradually increases (from an average of 3.5% to 4.8%), while the fixed k' value strategy fails to adapt in time, resulting in a systematically low alkali content and an average tungsten conversion rate of 64.3%.

[0060] In the later stages of operation (400-500 furnaces): Operators discovered the problem and manually adjusted the k' value, but lacking system data support, the adjustment range relied on trial and error based on experience, and the average tungsten conversion rate only recovered to 66.8%.

[0061] In contrast, after enabling self-learning optimization in Example 4, the model automatically adjusted the k' value function and slag type threshold based on accumulated data, increasing the average conversion rate from 69.8% to 72.3% and the alkali utilization rate to 91%.

[0062] This invention enables real-time and precise control of alkali fusion batching, reducing the fluctuation of tungsten conversion rate from ±15% of fixed batching to within ±3%, while ensuring the stability of slag fluidity and water quenching performance, significantly improving the process stability and alkali utilization rate of tungsten recovery.

Claims

1. A dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis, characterized in that, The method involves adding Na2CO3 to the molten slag in the furnace at the end of the sulfidation and volatilization stage of tin smelting in a fumigation furnace to carry out alkaline fusion conversion and recover tungsten from the slag. The method includes the following steps: S1. Online acquisition of slag composition information: During the final stage of sulfidation and volatilization in tin smelting in a fuming furnace, the composition information of the slag for the current furnace batch is acquired through an online analysis system. The slag composition information includes at least W content and Ca content. The online analysis system is one or a combination of XRF spectroscopy online analysis system and LIBS laser-induced breakdown spectroscopy online analysis system. S2. Based on the acquired slag composition information, calculate the optimal amount of Na2CO3 added for the current furnace using a preset calculation model; the calculation model is: Theoretical Na2CO3 addition amount = k × m 渣 × w 钨 × M Na2CO3 / M CaWO4 Where k is a correction factor, ranging from 1.2 to 2.0; m 渣 The current slag quantity; w 钨 M represents the mass fraction of tungsten in the slag. Na2CO3 M is the molar mass of Na₂CO₃. CaWO4 Here is the molar mass of CaWO4; S3. Based on the calculated optimal amount of Na2CO3 added, the amount of Na2CO3 added to the fuming furnace is automatically controlled by the automatic control system.

2. The dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis according to claim 1, characterized in that, The slag composition information mentioned in step S1 also includes SiO2 content, Fe content, and Al2O3 content.

3. The dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis according to claim 1, characterized in that, In the calculation model of step S2, the correction coefficient k is dynamically adjusted according to the Ca / Si / Al gangue content in the slag. The normal content is SiO2 10%~16%, Al2O3 2.5%~4.5%, and CaO 6%~8% by mass. When the Ca content is high, k is 1.6~2.0; when the Ca content is normal, k is 1.3~1.6; when the Ca content is low, k is 1.2~1.

4. When the SiO2 content is high, k is reduced by 0.1~0.2; when the Al2O3 content is high, k is increased by 0.1~0.

2.

4. The dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis according to claim 1, characterized in that, In step S2, slag stability is verified by first calculating the basicity of the slag after adding Na2CO3, A / C = m. Al2O3 / m CaO m Al2O3 m is the mass of Al2O3. CaO The quality of CaO is considered; when the A / C ratio exceeds the set threshold of 0.6, the amount of Na2CO3 added is reduced or CaO is added to adjust the slag type.

5. The dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis according to claim 1, 2, 3, or 4, is characterized in that... The automatic control system comprises an online analysis layer, a data acquisition layer, a batching calculation layer, and an execution control layer. The online analysis layer houses an online analysis system. The data acquisition layer consists of a data acquisition and calculation module comprised of a data acquisition unit and a slag type verification submodule. The batching calculation layer comprises an alkali quantity calculation submodule, a Na2CO3 solution feeding device, and a slag type verification submodule. The data acquisition and batching calculation layers share a single slag type verification submodule, which receives slag composition parameters and slag quantity input data from the data processing unit and performs slag type stability verification. The execution control layer and the fuming furnace control system PLC / DCS achieve bidirectional data interaction via an industrial communication protocol. After obtaining slag composition information, the online analysis layer inputs it to the data acquisition layer. The data acquisition and processing unit filters and reduces noise from the input data, and then, combined with manually input or automatically estimated slag quantity, the alkali quantity calculation submodule of the batching calculation module first calculates the slag quantity according to formula m. Na2CO3 =k×m 渣 ×w 钨 Calculate the theoretical alkali content by multiplying by 0.3681, then use A / C=m Al2O3 / m CaO For values ​​≤0.6, perform slag type verification. If the A / C value exceeds the limit, reduce the amount of alkali or add CaO. The final output is the optimized amount of Na2CO3 added, m. Na2CO3 m is the theoretical alkali content. 渣 w represents the current slag quantity for the furnace run. 钨 m represents the mass fraction of tungsten in the slag. Al2O3 m is the mass of Al2O3. CaO To ensure the quality of CaO, the execution control layer receives three signals from the fuming furnace control system PLC / DCS: furnace temperature, slag level height, and tuyere pressure. It then comprehensively judges the timing and rate of addition and issues a feeding command to the fuming furnace control system PLC / DCS. The fuming furnace control system PLC / DCS continuously adds the alkali material into the fuming furnace in batches through a weighing or volume system with a metering accuracy of ±5%, thus achieving closed-loop automatic control.

6. The dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis according to claim 5, characterized in that, Step S1: When the online analysis method is XRF spectroscopy, the XRF probe is installed on the furnace wall of the fuming furnace to measure the composition of the slag in real time, with measurements taken every 1-5 minutes. When the online analysis method is LIBS laser-induced breakdown spectroscopy, the probe of the LIBS laser-induced breakdown spectroscopy online analysis system is aimed at the surface of the molten pool inside the furnace for in-situ measurement to obtain slag composition information, with measurements taken every 10-60 seconds.

7. The dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis according to claim 5, characterized in that, The accuracy of the amount of Na2CO3 added in step S3 is ±5%.

8. The dynamic batching method for tungsten recovery by alkaline fusion in a fuming furnace based on online slag composition analysis according to claim 5, characterized in that, The automatic control system is also equipped with a historical data self-learning module connected to the PLC / DCS control system of the fuming furnace. This module performs self-learning optimization on historical data. The optimization method involves storing the slag composition, Na2CO3 addition amount, tungsten conversion rate, and slag type parameters for each furnace in a database. Data mining and regression analysis are used to optimize the value strategy of the correction coefficient k' and the slag type stability verification threshold. The optimization strategy for the correction coefficient k' is a continuous function: k' = a + b×(Ca%-Ca0) -c×max(SiO2%-SiO2). 20 , 0) +d×max(Al2O3%-Al2O 30 , 0), where a, b, c, d are regression coefficients, CaO, SiO 20 Al2O 30 The baseline contents of Ca, SiO2, and Al2O3 are respectively, and k' is determined by multiple regression analysis; Ca%, SiO2%, and Al2O3% are the mass contents of Ca, SiO2, and Al2O3 in the slag obtained from the online analysis of the current furnace.

Images