Converter molten steel carbon content and temperature real-time judgment method

Abstract

The invention relates to a converter molten steel carbon content and temperature real-time judgment method, and belongs to the technical field of molten steel quality monitoring methods in the converter steelmaking production process. According to the technical scheme, the method comprises the following steps: installing an infrared camera, and carrying out system calibration; an image data processing center is constructed, and automatic processing of infrared image acquisition-preprocessing-feature extraction is realized; carbon content and temperature of molten steel are calculated and judged, a carbon temperature coupling calculation-dynamic threshold judgment secondary model is constructed, and accurate carbon temperature calculation and abnormal early warning are achieved; outputting a judgment result and responding to execution. The method has the beneficial effects that the carbon content and the temperature of the molten steel in the converter are judged on line in real time through an infrared camera and an image analysis algorithm, and the method is suitable for molten steel state monitoring in the middle and later periods of converter steelmaking in iron and steel enterprises; and data support is provided for converter blowing end point control, steel grade component accurate regulation and control and production efficiency improvement.

Description

Technical Field

[0001] This invention relates to a method for real-time determination of carbon content and temperature in molten steel from a converter, belonging to the technical field of steel quality monitoring methods in the converter steelmaking process. Background Technology

[0002] In converter steelmaking, the carbon content (final carbon) and temperature of molten steel are core indicators determining steel quality, blowing time, and energy consumption. Precise control of carbon content can prevent excessive decarburization or carbon replenishment in subsequent refining processes, while temperature control can prevent problems such as steel leakage and nozzle blockage during casting. Currently, methods for detecting carbon content and temperature in molten steel are mainly divided into offline sampling and online indirect detection, both of which have significant limitations.

[0003] 1. Offline sampling and testing method: The traditional method involves manually scooping molten steel samples with a sample spoon during the blowing process, cooling them, and then using a carbon-sulfur analyzer to detect the carbon content and a thermocouple thermometer to measure the temperature. This method has three major drawbacks:

[0004] (1) Prominent lag: It takes 8-15 minutes from sampling to obtaining results, during which the converter continues to blow, which may cause the carbon content or temperature to exceed the target range, resulting in "scrap steel" or increased refining costs.

[0005] (2) Insufficient representativeness: A single sampling can only reflect the local state of molten steel. When the molten steel is unevenly mixed in the converter (such as insufficient bottom blowing), the sample cannot represent the overall molten steel situation, and the misjudgment rate is >15%.

[0006] (3) Low safety: Manual sampling requires proximity to the high-temperature converter (furnace mouth temperature reaches 1200℃~1500℃), which is prone to molten steel splashing and burn accidents. In addition, the blowing process is interrupted during sampling, which affects production efficiency.

[0007] 2. Existing online indirect detection methods: Some enterprises use secondary nozzle detection or furnace gas analysis technology, but these methods have inherent shortcomings:

[0008] (1) Sub-lance inspection: The inspection probe needs to be inserted into the molten steel. The probe wears out quickly (cost per use > 500 yuan, lifespan only 1 time), and the inspection interval is > 5 minutes. It cannot track carbon content and temperature changes in real time. When the molten steel surface fluctuates, the probe is easily damaged, causing the inspection to be interrupted.

[0009] (2) Furnace gas analysis: Carbon content is estimated by analyzing the composition of converter flue gas (e.g., CO and CO2 concentrations), but it is affected by the flue gas dust removal system (e.g., dust adsorption gas), resulting in large detection errors (carbon content error ±0.03%), and the temperature of molten steel cannot be obtained directly.

[0010] (3) Poor adaptability: Existing methods are mostly for fixed steel grades (such as plain carbon steel). When switching to low alloy steel or high carbon steel, the detection model needs to be recalibrated (time > 4 hours), which cannot meet the flexible production needs of multiple steel grades.

[0011] With the increasing demand for "high-efficiency, precise, and low-consumption" production in the steel industry (e.g., high-end bearing steel requires a final carbon content tolerance of ±0.01% and a temperature tolerance of ±5℃), traditional testing methods are no longer adequate. The industry urgently needs a non-contact, real-time, high-precision, multi-steel-grade-compatible online technology for determining the carbon content and temperature of molten steel. Specific requirements are as follows:

[0012] 1. Real-time requirement: Changes in carbon content and temperature in molten steel must be identified within 2 seconds to prevent loss of control at the blowing endpoint;

[0013] 2. Non-contact requirement: The testing equipment does not need to come into direct contact with molten steel or high-temperature furnace, ensuring equipment lifespan and production safety;

[0014] 3. Environmental adaptability requirements: Can withstand high temperatures at the converter site (ambient temperature ≤ 80℃) and dust (concentration ≤ 20mg / m³). 3 ), smoke obstruction and interference;

[0015] 4. Flexible adaptation requirements: When switching steel grades (carbon content 0.03% to 1.2%), there is no need to stop the machine to calibrate the model.

[0016] Furthermore, while the "method for detecting carbon content in converter steel based on laser spectroscopy" proposed in patent 202010689231.5 achieves non-contact detection, the laser equipment is expensive (a single unit > 800,000 yuan) and is affected by the absorption of flue gas at the furnace mouth (flue gas concentration > 50 mg / m³). 3 (Time accuracy drops by 40%), making large-scale promotion difficult.

[0017] Patent 202110754126.3, proposing an "online monitoring method for converter molten steel temperature," revolves around pre-embedding multiple thermocouples in the converter's sidewall. These thermocouples collect the furnace temperature in real time, establishing a mapping relationship between the furnace temperature and the molten steel temperature to indirectly calculate the molten steel temperature. However, this method has significant limitations in practical applications. First, the thermocouples need to operate in a high-temperature environment (furnace temperature 800-1000℃) for extended periods, resulting in a short service life (less than one month), frequent replacements, and high costs. Second, the furnace temperature is greatly affected by the furnace lining thickness and cooling water volume, causing the mapping relationship with the molten steel temperature to drift, leading to a temperature calculation error >15℃. Furthermore, this method only monitors temperature and cannot obtain carbon content data, requiring additional detection methods and increasing system complexity. Summary of the Invention

[0018] The purpose of this invention is to provide a method for real-time judgment of carbon content and temperature in molten steel in a converter. By relying on an infrared camera and image analysis algorithm, this method enables online real-time judgment of carbon content and temperature in molten steel in a converter. It is applicable to the monitoring of molten steel status in the middle and late stages of converter steelmaking in steel enterprises, providing data support for the endpoint control of converter blowing, precise regulation of steel composition, and improvement of production efficiency, and effectively solving the above-mentioned problems existing in the background technology.

[0019] The technical solution of this invention is: a method for real-time determination of carbon content and temperature in converter steel, comprising the following steps:

[0020] (1) Deployment and calibration of infrared camera system: Install infrared cameras and perform system calibration;

[0021] (2) Construct an image data processing center. The on-site data processing center adopts an industrial control computer and is equipped with image processing software developed based on infrared image modules to realize automated processing of "infrared image acquisition-preprocessing-feature extraction".

[0022] (3) Calculation and determination of carbon content and temperature in molten steel: Based on the preprocessed infrared image and calibration relationship, a two-level model of "carbon temperature coupling calculation - dynamic threshold determination" is constructed to realize accurate carbon temperature calculation and abnormal early warning.

[0023] (4) Judgment result output and execution response.

[0024] The specific steps in step (1) are as follows:

[0025] (11) Camera installation location and layout

[0026] a) Installation location: Build high-temperature protection supports on both sides of the same horizontal plane at the converter opening, 10-15m away from the edge of the opening, avoiding the splash zone and the flue gas concentration zone.

[0027] b) Dual-view layout: Two infrared cameras are symmetrically distributed with a horizontal angle controlled between 30° and 60° to ensure coverage of the core area of ​​the furnace flame without any blind spots.

[0028] c) Installation requirements: The infrared camera is equipped with a high-temperature resistant protective cover and a nitrogen purging device. The bracket adopts an anti-vibration design, and the lens axis is kept parallel to the furnace opening plane.

[0029] (12) Camera selection parameters

[0030] A high-sensitivity infrared camera must be selected, and it must meet the following requirements:

[0031] a) Infrared resolution ≥ 640×512 pixels to ensure accurate identification of subtle temperature differences in flames at long distances;

[0032] b) Frame rate ≥ 25fps, one frame is captured every 0.04 seconds to meet real-time requirements;

[0033] c) Spectral range of 8-14μm, suitable for long-distance detection of 10-15m, reducing dust interference;

[0034] d) Temperature measurement range 200-2000℃, covering the converter flame temperature range, with a measurement accuracy of ±2℃;

[0035] e) Environmental adaptability: Operating temperature -30-80℃, protection level IP68, electromagnetic interference resistance;

[0036] f) Data interface: Gigabit Ethernet, supporting synchronous transmission of infrared images and temperature data;

[0037] (13) System calibration

[0038] a) Cold calibration: During shutdown, a standard temperature source and a carbon content standard sample are placed in the furnace mouth simulation device, and flame simulation images from two infrared cameras are acquired simultaneously; the infrared characteristic parameters of the flame are extracted, and an initial mapping model between the characteristic parameters and the standard carbon temperature is established:

[0039] Carbon content C (%) = a1 × G + b1 × S + c1

[0040] Temperature T (°C) = a² × P + b² × G + c²

[0041] Where: G is the average ash value of the flame, S is the area ratio of the high-temperature region, P is the peak value of infrared radiation intensity; a1, b1, c1, a2, b2, c2 are calibration coefficients, which are obtained by fitting the experimental data of the standard sample.

[0042] b) Flue gas interference compensation: Real-time acquisition of flue gas concentration at the furnace inlet, obtained through a dust sensor and denoted as ρ, is used to correct the infrared characteristic parameters based on the flue gas concentration.

[0043] The corrected grayscale value G' = G × [1 - 0.02 × ρ]

[0044] Step (2) comprises three sub-models:

[0045] (21) Image synchronous acquisition model

[0046] a) Acquisition frequency: 25fps, synchronously acquire image data from two infrared cameras, denoted as I1(t) and I2(t), where t is time, and simultaneously read the blowing parameters from the converter PLC via the OPC protocol;

[0047] b) Data frame format: timestamp + camera number + infrared image data + flame temperature matrix + blowing parameters + check digit, to ensure data traceability;

[0048] c) Storage strategy: Real-time infrared images and temperature data are cached in memory, and the processed carbon temperature data and key images are compressed and stored in the industrial database every minute;

[0049] (22) Image preprocessing model

[0050] a) Noise reduction: A bilateral filtering algorithm is used to eliminate image noise caused by long-distance transmission and smoke and dust, while preserving flame edge details;

[0051] b) Smoke obstruction correction: Identify and mark smoke obstruction areas using an infrared image segmentation algorithm, and fill them with the average infrared temperature of adjacent unobstructed areas;

[0052] c) Flame region extraction: An adaptive threshold segmentation algorithm is used to extract the core flame region at the furnace opening, eliminating background interference from the furnace body refractory material and the furnace opening frame.

[0053] (23) Abnormal Image Processing Model

[0054] a) Image blur determination: Calculate the temperature contrast of each frame of infrared image. If the contrast is <50℃, it is determined to be blurry. If 6 consecutive frames of images are blurry, it is determined to be lens contamination. Trigger the nitrogen purging device to purge strongly, and replace it with the average carbon temperature of the first 4 valid images.

[0055] b) Camera fault diagnosis: If a camera has no data output for 15 consecutive seconds, or the data transmission error rate is >8%, a fault alarm is triggered, the data of that camera is marked as invalid, and single-view compensation calculation is enabled. That is, based on the infrared characteristic parameters of another camera, combined with the blowing time and oxygen flow process parameters, the calibration coefficient is corrected to ensure continuous carbon temperature calculation.

[0056] In step (3),

[0057] (31) Carbon-temperature coupling calculation sub-model

[0058] a) Core parameter calculation: The average gray value G of the flame, the area ratio of the high-temperature region S and the peak value of the infrared radiation intensity P are extracted from the infrared images of two infrared cameras. Combined with the smoke interference compensation and calibration model, the carbon temperature data of a single camera is obtained.

[0059] b) Carbon temperature verification: The average value of the calculation results from two infrared cameras is taken as the final carbon temperature to reduce single-view error;

[0060] c) Correction during the blowing stage: Adjust the calibration coefficient according to the blowing time to improve the accuracy of carbon content calculation;

[0061] (32) Dynamic threshold determination sub-model

[0062] a) Threshold setting: Based on historical production data, the relationship between the threshold and oxygen supply time is established through linear regression.

[0063] C_upper limit (t_supply) = d1 × t_supply + e1, C_lower limit (t_supply) = d2 × t_supply + e2

[0064] T_upper limit (t_supply) = d3 × t_supply + e3, T_lower limit (t_supply) = d4 × t_supply + e4

[0065] In the formula, t is the oxygen supply time, and d1, e1, etc. are fitting coefficients, which are determined by process experiments;

[0066] b) Exception detection logic:

[0067] Level 1 warning: When C(t) and T(t) exceed the qualified threshold range, a Level 1 warning is triggered, indicating abnormal carbon temperature, which needs to be monitored;

[0068] Level 2 warning: When C(t) exceeds [C_lower limit(t_supply)-0.005, C_upper limit(t_supply)+0.005] and T(t) exceeds [T_lower limit(t_supply)-3, T_upper limit(t_supply)+3], and the duration is ≥2 seconds, a level 2 warning is triggered, which is judged as a serious abnormality, and the blowing parameters need to be adjusted immediately.

[0069] Trend prediction: Based on the carbon temperature change trend in the first 4 seconds, predict the carbon temperature C_pred(t+1.5s) and T_pred(t+1.5s) for the next 1.5 seconds. If the predicted value exceeds the level 2 warning threshold, trigger the warning 1.5 seconds in advance.

[0070] The specific steps in step (4) are as follows:

[0071] (41) Real-time display: The central control room HMI dynamically displays the furnace flame infrared thermal image, real-time carbon temperature curve, dynamic threshold range, camera status, and key blowing parameters.

[0072] (42) Alarm and adjustment response: Level 1 warning triggers yellow indicator light flashing + low frequency buzzer, prompting the operator to check parameters; Level 2 warning triggers red indicator light constantly on + high frequency buzzer, and at the same time sends a signal to converter PLC to automatically adjust the blowing parameters;

[0073] (43) Data traceability: Automatically generate daily / weekly carbon temperature judgment reports, including average carbon temperature value, number of abnormalities and early warning handling records, and support export to Excel / PDF format for quality traceability and process optimization.

[0074] The beneficial effects of this invention are: by relying on infrared cameras and image analysis algorithms to realize online real-time judgment of carbon content and temperature of molten steel in converters, this method is applicable to the monitoring of molten steel status in the middle and late stages of converter steelmaking in steel enterprises, and provides data support for converter blowing endpoint control, precise control of steel composition and improvement of production efficiency. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0076] Figure 2 This is a flowchart of the method of the present invention;

[0077] In the picture: converter opening 1, infrared camera 1 2, infrared camera 2 3. Detailed Implementation

[0078] To make the purpose, technical solutions, and advantages of the invention's embodiments clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only a small part of the embodiments of the present invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0079] A method for real-time determination of carbon content and temperature in converter steel includes the following steps:

[0080] (1) Deployment and calibration of infrared camera system: Install infrared cameras and perform system calibration;

[0081] (2) Construct an image data processing center. The on-site data processing center adopts an industrial control computer and is equipped with image processing software developed based on infrared image modules to realize automated processing of "infrared image acquisition-preprocessing-feature extraction".

[0082] (3) Calculation and determination of carbon content and temperature in molten steel: Based on the preprocessed infrared image and calibration relationship, a two-level model of "carbon temperature coupling calculation - dynamic threshold determination" is constructed to realize accurate carbon temperature calculation and abnormal early warning.

[0083] (4) Judgment result output and execution response.

[0084] The specific steps in step (1) are as follows:

[0085] (11) Camera installation location and layout

[0086] a) Installation location: Build high-temperature protection supports on both sides of the same horizontal plane at the converter opening, 10-15m away from the edge of the opening, avoiding the splash zone and the flue gas concentration zone.

[0087] b) Dual-view layout: Two infrared cameras are symmetrically distributed with a horizontal angle controlled between 30° and 60° to ensure coverage of the core area of ​​the furnace flame without any blind spots.

[0088] c) Installation requirements: The infrared camera is equipped with a high-temperature resistant protective cover and a nitrogen purging device. The bracket adopts an anti-vibration design, and the lens axis is kept parallel to the furnace opening plane.

[0089] (12) Camera selection parameters

[0090] A high-sensitivity infrared camera must be selected, and it must meet the following requirements:

[0091] a) Infrared resolution ≥ 640×512 pixels to ensure accurate identification of subtle temperature differences in flames at long distances;

[0092] b) Frame rate ≥ 25fps, one frame is captured every 0.04 seconds to meet real-time requirements;

[0093] c) Spectral range of 8-14μm, suitable for long-distance detection of 10-15m, reducing dust interference;

[0094] d) Temperature measurement range 200-2000℃, covering the converter flame temperature range, with a measurement accuracy of ±2℃;

[0095] e) Environmental adaptability: Operating temperature -30-80℃, protection level IP68, electromagnetic interference resistance;

[0096] f) Data interface: Gigabit Ethernet, supporting synchronous transmission of infrared images and temperature data;

[0097] (13) System calibration

[0098] a) Cold calibration: During shutdown, a standard temperature source and a carbon content standard sample are placed in the furnace mouth simulation device, and flame simulation images from two infrared cameras are acquired simultaneously; the infrared characteristic parameters of the flame are extracted, and an initial mapping model between the characteristic parameters and the standard carbon temperature is established:

[0099] Carbon content C (%) = a1 × G + b1 × S + c1

[0100] Temperature T (°C) = a² × P + b² × G + c²

[0101] Where: G is the average ash value of the flame, S is the area ratio of the high-temperature region, P is the peak value of infrared radiation intensity; a1, b1, c1, a2, b2, c2 are calibration coefficients, which are obtained by fitting the experimental data of the standard sample.

[0102] b) Flue gas interference compensation: Real-time acquisition of flue gas concentration at the furnace inlet, obtained through a dust sensor and denoted as ρ, is used to correct the infrared characteristic parameters based on the flue gas concentration.

[0103] The corrected grayscale value G' = G × [1 - 0.02 × ρ]

[0104] Step (2) comprises three sub-models:

[0105] (21) Image synchronous acquisition model

[0106] a) Acquisition frequency: 25fps, synchronously acquire image data from two infrared cameras, denoted as I1(t) and I2(t), where t is time, and simultaneously read the blowing parameters from the converter PLC via the OPC protocol;

[0107] b) Data frame format: timestamp + camera number + infrared image data + flame temperature matrix + blowing parameters + check digit, to ensure data traceability;

[0108] c) Storage strategy: Real-time infrared images and temperature data are cached in memory, and the processed carbon temperature data and key images are compressed and stored in the industrial database every minute;

[0109] (22) Image preprocessing model

[0110] a) Noise reduction: A bilateral filtering algorithm is used to eliminate image noise caused by long-distance transmission and smoke and dust, while preserving flame edge details;

[0111] b) Smoke obstruction correction: Identify and mark smoke obstruction areas using an infrared image segmentation algorithm, and fill them with the average infrared temperature of adjacent unobstructed areas;

[0112] c) Flame region extraction: An adaptive threshold segmentation algorithm is used to extract the core flame region at the furnace opening, eliminating background interference from the furnace body refractory material and the furnace opening frame.

[0113] (23) Abnormal Image Processing Model

[0114] a) Image blur determination: Calculate the temperature contrast of each frame of infrared image. If the contrast is <50℃, it is determined to be blurry. If 6 consecutive frames of images are blurry, it is determined to be lens contamination. Trigger the nitrogen purging device to purge strongly, and replace it with the average carbon temperature of the first 4 valid images.

[0115] b) Camera fault diagnosis: If a camera has no data output for 15 consecutive seconds, or the data transmission error rate is >8%, a fault alarm is triggered, the data of that camera is marked as invalid, and single-view compensation calculation is enabled. That is, based on the infrared characteristic parameters of another camera, combined with the blowing time and oxygen flow process parameters, the calibration coefficient is corrected to ensure continuous carbon temperature calculation.

[0116] In step (3),

[0117] (31) Carbon-temperature coupling calculation sub-model

[0118] a) Core parameter calculation: The average gray value G of the flame, the area ratio of the high-temperature region S and the peak value of the infrared radiation intensity P are extracted from the infrared images of two infrared cameras. Combined with the smoke interference compensation and calibration model, the carbon temperature data of a single camera is obtained.

[0119] b) Carbon temperature verification: The average value of the calculation results from two infrared cameras is taken as the final carbon temperature to reduce single-view error;

[0120] c) Correction during the blowing stage: Adjust the calibration coefficient according to the blowing time to improve the accuracy of carbon content calculation;

[0121] (32) Dynamic threshold determination sub-model

[0122] a) Threshold setting: Based on historical production data, the relationship between the threshold and oxygen supply time is established through linear regression.

[0123] C_upper limit (t_supply) = d1 × t_supply + e1, C_lower limit (t_supply) = d2 × t_supply + e2

[0124] T_upper limit (t_supply) = d3 × t_supply + e3, T_lower limit (t_supply) = d4 × t_supply + e4

[0125] In the formula, t is the oxygen supply time, and d1, e1, etc. are fitting coefficients, which are determined by process experiments;

[0126] b) Exception detection logic:

[0127] Level 1 warning: When C(t) and T(t) exceed the qualified threshold range, a Level 1 warning is triggered, indicating abnormal carbon temperature, which needs to be monitored;

[0128] Level 2 warning: When C(t) exceeds [C_lower limit(t_supply)-0.005, C_upper limit(t_supply)+0.005] and T(t) exceeds [T_lower limit(t_supply)-3, T_upper limit(t_supply)+3], and the duration is ≥2 seconds, a level 2 warning is triggered, which is judged as a serious abnormality, and the blowing parameters need to be adjusted immediately.

[0129] Trend prediction: Based on the carbon temperature change trend in the first 4 seconds, predict the carbon temperature C_pred(t+1.5s) and T_pred(t+1.5s) for the next 1.5 seconds. If the predicted value exceeds the level 2 warning threshold, trigger the warning 1.5 seconds in advance.

[0130] The specific steps in step (4) are as follows:

[0131] (41) Real-time display: The central control room HMI dynamically displays the furnace flame infrared thermal image, real-time carbon temperature curve, dynamic threshold range, camera status, and key blowing parameters.

[0132] (42) Alarm and adjustment response: Level 1 warning triggers yellow indicator light flashing + low frequency buzzer, prompting the operator to check parameters; Level 2 warning triggers red indicator light constantly on + high frequency buzzer, and at the same time sends a signal to converter PLC to automatically adjust the blowing parameters;

[0133] (43) Data traceability: Automatically generate daily / weekly carbon temperature judgment reports, including average carbon temperature value, number of abnormalities and early warning handling records, and support export to Excel / PDF format for quality traceability and process optimization.

[0134] In practical applications, this invention involves erecting fixed supports on both sides of the converter opening 1 at the same horizontal plane, and arranging two infrared cameras, namely infrared camera 2 and infrared camera 3, to simultaneously acquire infrared images of the flame at the furnace opening during the blowing process. Through preprocessing such as image denoising, segmentation, and feature extraction, combined with the calibration relationship between "flame infrared characteristics - molten steel carbon temperature" and flue gas interference compensation, the real-time carbon content and temperature of the molten steel are calculated. Finally, a dynamic threshold is used to determine whether the carbon temperature is within the target range, triggering an early warning and blowing parameter adjustment response. The specific steps are as follows:

[0135] Step 1: Deployment and Calibration of Infrared Camera System

[0136] (1) Camera installation location and layout

[0137] 1) Installation location: Build high-temperature protection supports on both sides of the same horizontal plane at the converter opening 1, with a horizontal distance of 10-15m from the edge of the opening (avoiding the splash zone and the flue gas concentration zone);

[0138] 2) Dual-view layout: Two infrared cameras (infrared camera 1 2 and infrared camera 2 3) are symmetrically distributed, with the horizontal angle controlled between 30° and 60° (to ensure coverage of the core area of ​​the furnace flame and eliminate blind spots);

[0139] 3) Installation Requirements: Infrared Camera 1 (2) and Infrared Camera 2 (3) must be equipped with high-temperature resistant protective covers (temperature resistance ≥120℃) and nitrogen purging devices (pressure 0.5-0.7MPa, flow rate 4-6m³ / h). 3 / h, to prevent dust from adhering to the lens at long distances), the bracket adopts an anti-vibration design (vibration control ≤0.08g), and the lens axis is kept parallel to the furnace mouth plane.

[0140] (2) Infrared camera selection parameters

[0141] A high-sensitivity infrared camera must be selected, and it must meet the following requirements:

[0142] 1) Infrared resolution: ≥640×512 pixels (ensuring accurate identification of subtle temperature differences in flames at long distances);

[0143] 2) Frame rate: ≥25fps (1 frame is captured every 0.04 seconds to meet real-time requirements);

[0144] 3) Spectral range: 8-14μm (This band has strong penetrating power in flue gas, suitable for long-distance detection of 10-15m, and reduces dust interference);

[0145] 4) Temperature measurement range: 200-2000℃ (covering the converter flame temperature range), temperature measurement accuracy ±2℃;

[0146] 5) Environmental adaptability: Operating temperature -30-80℃, protection level IP68 (dustproof and waterproof), electromagnetic interference resistance (compliant with industrial EMC standards);

[0147] 6) Data interface: Gigabit Ethernet (GigE Vision protocol), supporting synchronous transmission of infrared images and temperature data.

[0148] (3) System calibration

[0149] 1) Cold calibration: When the machine is shut down, place a standard temperature source (temperature range 800-1800℃, accuracy ±1℃) and a carbon content standard sample (carbon content 0.03%~1.2%) in the furnace mouth simulation device, and simultaneously acquire flame simulation images from infrared camera 2 and infrared camera 3; extract the flame infrared characteristic parameters (average gray value, high temperature area ratio, peak infrared radiation intensity), and establish an initial mapping model between the characteristic parameters and the standard carbon temperature:

[0150] Carbon content C (%) = a1 × G + b1 × S + c1

[0151] Temperature T (°C) = a² × P + b² × G + c²

[0152] (Where: G is the average ash value of the flame, S is the area ratio of the high-temperature region, P is the peak value of infrared radiation intensity; a1, b1, c1, a2, b2, c2 are calibration coefficients, obtained by fitting experimental data of standard samples).

[0153] 2) Flue gas interference compensation: Real-time acquisition of flue gas concentration at the furnace inlet (obtained through a dust sensor, denoted as ρ), and correction of infrared characteristic parameters based on flue gas concentration:

[0154] The corrected grayscale value G' = G × [1 - 0.02 × ρ]

[0155] Step 2: Construction of the Image Data Processing Center

[0156] The on-site data processing center uses an industrial control computer (IPC) equipped with image processing software developed based on Python + OpenCV (infrared image module) to automate the "infrared image acquisition-preprocessing-feature extraction" process, comprising three sub-models:

[0157] (1) Image synchronous acquisition model

[0158] 1) Acquisition frequency: 25fps, simultaneously acquiring image data from two infrared cameras (denoted as I1(t) and I2(t), where t is time), and reading blowing parameters (oxygen flow rate, oxygen supply time, bottom blowing intensity) from the converter PLC via the OPC protocol;

[0159] 2) Data frame format: "timestamp + camera number + infrared image data (compressed TIFF format) + flame temperature matrix + blowing parameters + check digit" to ensure data traceability;

[0160] 3) Storage strategy: Real-time infrared images and temperature data are cached in memory (retaining the most recent 15 seconds of data), and the processed carbon temperature data and key images (abnormal moments) are compressed and stored in an industrial database (such as PostgreSQL) every minute.

[0161] (2) Image preprocessing model

[0162] 1) Noise reduction: The "bilateral filtering algorithm" (filter kernel size 5×5) is used to eliminate image noise caused by long-distance transmission and smoke and dust, while preserving the details of the flame edge;

[0163] 2) Smoke obstruction correction: The smoke obstruction area (grayscale value < 50) is identified and marked by the "infrared image segmentation algorithm", and filled with the average infrared temperature of the adjacent non-obstruction area;

[0164] 3) Flame region extraction: The core flame region of the furnace mouth is extracted by the "adaptive threshold segmentation algorithm" (lower threshold 100, upper threshold 200) to eliminate background interference from furnace body refractory materials, furnace mouth frame and other factors.

[0165] (3) Abnormal Image Processing Model

[0166] 1) Image blur determination: Calculate the "temperature contrast" of each frame of infrared image (the temperature difference between high temperature area and low temperature area, the contrast <50℃ is judged as blur). If 6 consecutive frames of images are blurry, it is judged as lens contamination, triggering the nitrogen purging device to purge strongly (pressure increased to 0.7MPa, lasting for 15 seconds), and at the same time, the average carbon temperature of the previous 4 valid images is used to replace it.

[0167] 2) Camera fault diagnosis: If a camera has no data output for 15 consecutive seconds, or the data transmission error rate is >8%, a fault alarm is triggered, the data of that camera is marked as invalid, and "single-view compensation calculation" is enabled (based on the infrared characteristic parameters of another camera, combined with process parameters such as blowing time and oxygen flow rate to correct the calibration coefficient and ensure continuous carbon temperature calculation).

[0168] Step 3: Calculation and Judgment Model for Carbon Content and Temperature in Molten Steel

[0169] Based on the preprocessed infrared image and calibration relationship, a two-level model of "carbon temperature coupled calculation - dynamic threshold determination" is constructed to achieve accurate carbon temperature calculation and anomaly early warning.

[0170] (1) Carbon-temperature coupling calculation sub-model

[0171] 1) Core parameter calculation: The average gray value G of the flame, the area ratio of the high temperature region S, and the peak value of the infrared radiation intensity P are extracted from the infrared images of infrared camera 12 and infrared camera 23. Combined with the smoke interference compensation and calibration model, the carbon temperature data of a single camera is obtained.

[0172] 2) Carbon temperature verification: The average value of the calculation results of infrared camera 12 and infrared camera 23 is taken as the final carbon temperature (e.g., camera 1: C1 = 0.06%, T1 = 1650℃, camera 2: C2 = 0.05%, T2 = 1648℃, final C(t) = 0.055%, T(t) = 1649℃) to reduce single-view error;

[0173] 3) Correction during the blowing stage: Adjust the calibration coefficient according to the blowing time (early stage, middle stage, and late stage). For example, in the late stage of blowing (oxygen supply time > 12 minutes), a1 is adjusted to 1.05 times the original coefficient to improve the accuracy of carbon content calculation.

[0174] (2) Dynamic threshold determination sub-model

[0175] 1) Threshold setting: Based on historical production data, the relationship between the threshold and oxygen supply time is established through linear regression:

[0176] C_upper limit (t_supply) = d1 × t_supply + e1, C_lower limit (t_supply) = d2 × t_supply + e2

[0177] T_upper limit (t_supply) = d3 × t_supply + e3, T_lower limit (t_supply) = d4 × t_supply + e4

[0178] (where t is the oxygen supply time, and d1, e1, etc. are fitting coefficients, which are determined by process experiments);

[0179] 2) Exception detection logic:

[0180] Level 1 warning: When C(t) and T(t) exceed the qualified threshold range, a Level 1 warning is triggered (indicating abnormal carbon temperature, requiring attention).

[0181] Level 2 warning: When C(t) exceeds [C_lower limit(t_supply)-0.005, C_upper limit(t_supply)+0.005] and T(t) exceeds [T_lower limit(t_supply)-3, T_upper limit(t_supply)+3], and the duration is ≥2 seconds, a Level 2 warning is triggered (judged as a serious abnormality, and the blowing parameters need to be adjusted immediately).

[0182] Trend prediction: Based on the carbon temperature change trend in the first 4 seconds (such as the slope of carbon content change k_C=[C(t)-C(t-4s)] / 4s), predict the carbon temperature C_pred(t+1.5s) and T_pred(t+1.5s) in the next 1.5 seconds. If the predicted value will exceed the level 2 warning threshold, trigger the warning 1.5 seconds in advance.

[0183] Step 4: Determine the output of the judgment result and execute the response.

[0184] (1) Real-time display: The central control room HMI dynamically displays the furnace mouth flame infrared thermal imaging image (labeled with carbon content, temperature value and qualified status), real-time carbon temperature curve and dynamic threshold range, camera status and key blowing parameters.

[0185] (2) Alarm and adjustment response: Level 1 warning triggers a flashing yellow indicator light + low-frequency buzzer to prompt the operator to check parameters; Level 2 warning triggers a solid red indicator light + high-frequency buzzer, and at the same time sends a signal to the converter PLC to automatically adjust the blowing parameters (such as reducing oxygen flow).

[0186] (3) Data traceability: Automatically generate daily / weekly carbon temperature judgment reports, including average carbon temperature value, number of abnormalities, early warning handling records, etc., and support export to Excel / PDF format for quality traceability and process optimization.

[0187] Example:

[0188] Based on an application case study of a 120t converter in a steel plant (producing bearing steel GCr15), the implementation steps are explained in detail:

[0189] Step 1: Deployment and Calibration of Infrared Camera System

[0190] 1. Infrared camera installation: Frames are erected on both sides of the converter opening at the same horizontal plane, 12m away from the opening. Two FLIRA655sc infrared cameras (640×512 resolution, 25fps frame rate) are arranged at a 45° horizontal angle. The protective cover is made of 316 stainless steel and the nitrogen purging pressure is 0.6MPa.

[0191] 2. Calibration: Place a standard temperature source (1580℃, accuracy ±1℃) and a carbon content standard sample (0.08%), and acquire images from two infrared cameras: average flame grayscale value G = 180, high-temperature area ratio S = 35%, peak infrared radiation intensity P = 2200; fitting calibration coefficients: a1 = 0.0005, b1 = 0.0001, c1 = 0.02; a2 = 0.5, b2 = 1.2, c2 = 1300;

[0192] 3. Flue gas interference compensation: During production, the flue gas concentration at the furnace inlet ρ = 8 mg / m³ 3 The corrected grayscale value G' = 180 × [1 - 0.02 × 8] = 177.12.

[0193] Step 2: Construction of the Image Data Processing Center

[0194] 1. IPC configuration: Intel Core i9 processor, 32GB RAM, 2TB SSD, Windows Server 2022 operating system installed, image processing software developed based on Python + OpenCV;

[0195] 2. Image preprocessing: A frame of the image contains smoke obstruction (grayscale value 45). After bilateral filtering for noise reduction, the obstruction area is filled by a segmentation algorithm, and the core flame area (grayscale value 100-200) is extracted.

[0196] 3. Anomaly Handling: If the temperature contrast of 6 consecutive frames of images from the infrared camera is 45℃ (<50℃), it is determined to be lens contamination, triggering a strong purge (0.7MPa, 15 seconds), during which the average carbon temperature of the previous 4 frames (C=0.078%, T=1578℃) is used as the replacement.

[0197] Step 3: Carbon Temperature Calculation and Determination

[0198] 1. Process parameters: bearing steel GCr15, oxygen supply time t_supply = 14 minutes in the later stage of blowing, carbon content within acceptable range of 0.08±0.01%, temperature within acceptable range of 1580±5℃;

[0199] 2. Carbon temperature calculation: At t = 16:45:00, C1 of infrared camera one is 0.092% and T1 is 1586℃; C2 of infrared camera two is 0.091% and T2 is 1585℃; finally, C(t) = (0.092 + 0.091) / 2 = 0.0915%, T(t) = (1586 + 1585) / 2 = 1585.5℃;

[0200] 3. Anomaly detection: C(t) = 0.0915% exceeds the upper limit by 0.09%, and T(t) = 1585.5℃ exceeds the upper limit of 1585℃, triggering a level one warning; if it does not drop after 2 seconds, a level two warning is triggered.

[0201] Step 4: Execute the response

[0202] 1. HMI display: The flame infrared thermal imaging icon indicates a carbon content of 0.0915% (red) and a temperature of 1585.5℃ (red). The real-time curve exceeds the threshold range.

[0203] 2. Automatic Adjustment: The system sends a signal to the PLC to adjust the oxygen flow rate from 2800 m³ / h. 3 / h dropped to 2600m 3 / h, after 1.5 seconds, C(t) drops to 0.088%, T(t) drops to 1583℃, returning to the acceptable range;

[0204] Data logging: The system automatically records abnormal events and handling measures, and generates a daily report.

[0205] Compared with the prior art, the present invention has the following significant advantages:

[0206] 1. Real-time and high-precision judgment: Through 25fps high-frequency acquisition and carbon temperature coupling model, carbon temperature anomalies can be identified within 2 seconds, with carbon content accuracy of ±0.005% and temperature accuracy of ±2℃, accurately capturing subtle changes in the later stage of blowing.

[0207] 2. Non-contact and strong environmental adaptability: The dual cameras are 10-15m away from the furnace opening, with no direct contact and a service life of 24-36 months; equipped with protective and purging devices, they effectively resist high temperature and dust interference, with an effective data rate of >99%;

[0208] 3. Flexible adaptation to multiple steel grades: When switching steel grades, only the dynamic threshold library needs to be updated (time < 30 minutes), without the need for machine shutdown for calibration, adapting to the production of multiple specifications with carbon content from 0.03% to 1.2%;

[0209] 4. Low cost and easy deployment: The cost of two infrared cameras is less than 50,000 yuan per unit, and the total investment is only 12.5% ​​of that of laser spectroscopy equipment; the dual-view horizontal layout (10-15m distance + 30-60° angle) is easy to install and maintain, and easy to promote on a large scale, which can reduce the unqualified rate of carbon temperature of molten steel by 50% to 70%.

Claims

1. A method for real-time determination of carbon content and temperature of molten converter steel, characterized by Comprise the following steps: (1) infrared camera system deployment and calibration, install infrared camera, system calibration; (2) build image data processing center, field data processing center uses industrial control computer, carries image processing software based on infrared image module development, realizes infrared image acquisition-preprocessing-feature extraction automation processing; (3) molten steel carbon content and temperature calculation and determination, based on the preprocessed infrared image and calibration relation, build carbon temperature coupling calculation-dynamic threshold determination two level model, realize carbon temperature accurate calculation and abnormal early warning; (4) the output and execution response of the judgment result.

2. The method for real-time judging the carbon content and temperature of converter steel according to claim 1, characterized in that: In the step (1), the specific steps are as follows: (11) camera installation position and layout a) installation position, build high temperature support on both sides of the same horizontal plane of the converter mouth, horizontal distance from the edge of the mouth 10-15 m, avoid spatter area and smoke concentration area; b) double view layout, two infrared cameras are symmetrically distributed, the horizontal angle is controlled in 30°-60°, ensure to cover the core area of the flame, no visual blind area; c) installation requirements, infrared camera is externally installed with high temperature resistant protective cover and nitrogen purging device, support adopts anti-vibration design, lens axis is parallel to the plane of the mouth; (12) camera selection parameters Select high sensitivity infrared camera, need to meet: a) infrared resolution ≥640×512 pixels to ensure the identification accuracy of the fine temperature difference of the flame at a long distance; b) frame rate ≥25fps, collect 1 frame of image every 0.04 seconds, meet the real-time requirement; c) spectral range 8-14 μm, adapt to 10-15 m long distance detection, reduce dust interference; d) temperature measurement range 200-2000 ℃, cover the temperature interval of the converter flame, temperature measurement accuracy ±2 ℃; e) environmental adaptability: working temperature-30-80 ℃, protection level IP68, anti-electromagnetic interference; f) data interface: gigabit Ethernet, support infrared image and temperature data synchronous transmission; (13) system calibration a) cold state calibration: when stopping, place standard temperature source and carbon content standard sample in the mouth simulation device, simultaneously collect flame simulation images of two infrared cameras; extract infrared feature parameters of the flame, establish initial mapping model of feature parameters and standard carbon temperature: Carbon content C (%) = a1×G + b1×S + c1 Temperature T (℃) = a2×P + b2×G + c2 Wherein: G is the average gray value of the flame, S is the area ratio of high temperature area, P is the peak value of infrared radiation intensity; a1, b1, c1, a2, b2, c2 are calibration coefficients, which are obtained by fitting experimental data of standard sample; b) smoke interference compensation: real-time acquisition of the concentration of the smoke at the mouth, the concentration of the smoke at the mouth is obtained by the dust sensor, recorded as ρ, the infrared feature parameters are corrected according to the concentration of the smoke: Corrected gray value G' = G × [1-0.02×ρ].

3. The method for real-time judging the carbon content and temperature of converter steel according to claim 1, characterized in that: The step (2) comprises three submodels: (21) image synchronous acquisition model a) acquisition frequency: 25fps, synchronous acquisition of image data of two infrared cameras, recorded as I1(t), I2(t), t is time, at the same time, read blowing parameters from converter PLC through OPC protocol; b) Data frame format: timestamp + camera number + infrared image data + flame temperature matrix + blowing parameters + check digit, to ensure data traceability; c) Storage strategy: Real-time infrared images and temperature data are cached in memory, and the processed carbon temperature data and key images are compressed and stored in the industrial database every minute; (22) Image preprocessing model a) Noise reduction: A bilateral filtering algorithm is used to eliminate image noise caused by long-distance transmission and smoke and dust, while preserving flame edge details; b) Smoke obstruction correction: Identify and mark smoke obstruction areas using an infrared image segmentation algorithm, and fill them with the average infrared temperature of adjacent unobstructed areas; c) Flame region extraction: An adaptive threshold segmentation algorithm is used to extract the core flame region at the furnace opening, eliminating background interference from the furnace body refractory material and the furnace opening frame. (23) Abnormal Image Processing Model a) Image blur determination: Calculate the temperature contrast of each frame of infrared image. If the contrast is <50℃, it is determined to be blurry. If 6 consecutive frames of images are blurry, it is determined to be lens contamination. Trigger the nitrogen purging device to purge strongly, and replace it with the average carbon temperature of the first 4 valid images. b) Camera fault diagnosis: If a camera has no data output for 15 consecutive seconds, or the data transmission error rate is >8%, a fault alarm is triggered, the data of that camera is marked as invalid, and single-view compensation calculation is enabled. That is, based on the infrared characteristic parameters of another camera, combined with the blowing time and oxygen flow process parameters, the calibration coefficient is corrected to ensure continuous carbon temperature calculation.

4. The method for real-time judging the carbon content and temperature of converter steel according to claim 1, characterized in that: In step (3), (31) Carbon-temperature coupling calculation sub-model a) Core parameter calculation: The average gray value G of the flame, the area ratio of the high-temperature region S and the peak value of the infrared radiation intensity P are extracted from the infrared images of two infrared cameras. Combined with the smoke interference compensation and calibration model, the carbon temperature data of a single camera is obtained. b) Carbon temperature verification: The average value of the calculation results from two infrared cameras is taken as the final carbon temperature to reduce single-view error; c) Correction during the blowing stage: Adjust the calibration coefficient according to the blowing time to improve the accuracy of carbon content calculation; (32) Dynamic threshold determination sub-model a) Threshold setting: Based on historical production data, the relationship between the threshold and oxygen supply time is established through linear regression. C_upper limit (t_supply) = d1 × t_supply + e1, C_lower limit (t_supply) = d2 × t_supply + e2 T_upper limit (t_supply) = d3 × t_supply + e3, T_lower limit (t_supply) = d4 × t_supply + e4 In the formula, t is the oxygen supply time, and d1, e1, etc. are fitting coefficients, which are determined by process experiments; b) Exception detection logic: Level 1 warning: When C(t) and T(t) exceed the qualified threshold range, a Level 1 warning is triggered, indicating abnormal carbon temperature, which needs to be monitored; Level 2 warning: When C(t) exceeds [C_lower limit(t_supply)-0.005, C_upper limit(t_supply)+0.005] and T(t) exceeds [T_lower limit(t_supply)-3, T_upper limit(t_supply)+3], and the duration is ≥2 seconds, a level 2 warning is triggered, which is judged as a serious abnormality, and the blowing parameters need to be adjusted immediately. Trend prediction: Through the trend of carbon temperature change in the first 4 seconds, the carbon temperature C_pred(t+1.5s) and T_pred(t+1.5s) in the next 1.5 seconds are predicted. If the predicted value exceeds the secondary warning threshold, the warning will be triggered 1.5 seconds in advance.

5. The method for real-time judging the carbon content and temperature of converter steel according to claim 1, characterized in that: In the step (4), the specific steps are as follows: (41) Real-time display: The HMI in the central control room dynamically displays the infrared thermal imaging of the furnace mouth flame, the real-time carbon temperature curve, the dynamic threshold range, the camera status, and the key blowing parameters. (42) Alarm and adjustment response: The first warning triggers the yellow indicator light to flash and the low-frequency buzzer to sound, prompting the operator to check the parameters; the second warning triggers the red indicator light to always light and the high-frequency buzzer to sound, and sends a signal to the converter PLC to automatically adjust the blowing parameters. (43) Data tracing: Daily / weekly carbon temperature judgment reports are automatically generated, including average carbon temperature value, number of abnormalities, and warning processing records, which can be exported in Excel / PDF format for quality tracing and process optimization.

Images