Multi-modal information fusion converter intelligent steelmaking method

Abstract

The invention relates to a multi-modal information fusion converter intelligent steelmaking method, belongs to the technical field of converter steelmaking, and solves the problem of low parameter prediction precision in the prior art. Screening the heat data with the historical slag thickness value and the production time meeting the requirements from the historical heat data to obtain a self-learning library of the corresponding steel grade; based on the data of each heat in the self-learning library and the historical mean value and the required value of each parameter of the to-be-produced steel grade, the modulus increment of each heat in the self-learning library is obtained; obtaining the weight of each heat in the self-learning library based on the modulus increment of each heat; on the basis of the weight of each heat and data of each heat in a self-learning library, obtaining standard heat data of the steel grade to be produced; and on the basis of the standard heat data and the technological requirements of the current production heat, the predicted values of the parameters in the current production heat are obtained, and the steelmaking process is conducted on the basis of the predicted values of the parameters. And accurate parameter prediction is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of converter steelmaking, and in particular to a converter intelligent steelmaking method based on multimodal information fusion. Background Art

[0002] As a core link in iron and steel production, converter steelmaking attaches great importance to the accurate diagnosis of molten steel quality and the precise control of key process parameters. For example, the prediction accuracy of the carbon content, temperature, and phosphorus content at the end of converter steelmaking is directly related to the quality and production efficiency of molten steel. Among them, improper control of the carbon content will lead to a decline in the quality of molten steel, while inaccurate temperature control will increase unnecessary energy consumption. Therefore, accurately predicting key process parameters is of great significance for reducing production costs and shortening the steelmaking cycle.

[0003] The traditional gas-phase carbon determination method analyzes the flue gas discharged from the converter through a mass spectrometer, and based on the carbon content in the flue gas, obtains the carbon content in the molten steel to achieve the control of the carbon content in the molten steel. However, the traditional gas-phase carbon determination method has high equipment investment, high maintenance costs, and high technical requirements for operators. Moreover, in actual applications, due to the complex environment in the converter, such as slag erosion and gas flow scouring, the measurement accuracy is relatively low, which in turn affects the accuracy of process parameter prediction. In addition, the traditional method has obvious deficiencies in dealing with the strong coupling relationship between time-series and scalar parameters, resulting in unreliable collaborative analysis of multivariable parameters and difficulty in meeting the complex requirements of actual industrial processes. Summary of the Invention

[0004] In view of the above analysis, the embodiments of the present invention aim to provide a converter intelligent steelmaking method based on multimodal information fusion to solve the problem of low accuracy in predicting converter process parameters in the prior art.

[0005] The embodiments of the present invention provide a converter intelligent steelmaking method based on multimodal information fusion, including:

[0006] Based on the required slag thickness of the steel grade to be produced, screening out the furnace data with historical slag thickness values and production times meeting the requirements from the historical furnace data to obtain a self-learning library for the corresponding steel grade; the furnace data is the numerical values corresponding to the parameters of each furnace.

[0007] Based on the furnace data of each furnace in the self-learning library and the historical mean values and required values of the parameters of the steel grade to be produced, obtaining the model increment of each furnace in the self-learning library.

[0008] Based on the model increment of each furnace, obtaining the weight of each furnace in the self-learning library.

[0009] Based on the weights of each furnace and the furnace data in the self-learning library, obtaining the standard furnace data of the steel grade to be produced.

[0010] Based on the standard heat data and the process requirements of the current production heat, the predicted values of the respective parameters in the current production heat are obtained, and the steelmaking process is carried out based on the predicted values of the respective parameters.

[0011] Based on a further improvement of the above method, the predicted values of the respective parameters in the current production heat are obtained by the following formula based on the standard heat data and the process requirements of the current production heat:

[0012]

[0013] Wherein, W q represents the value of the q-th parameter in the standard heat data, and W q′ represents the predicted value of the q-th parameter in the current production heat, W0 q represents the process requirement value of the q-th parameter in the current production heat, and α q represents the influence parameter of the q-th parameter.

[0014] Based on a further improvement of the above method, the standard heat data of the steel grade to be produced is obtained by the following formula based on the weights of the respective heats and the heat data in the self-learning library:

[0015]

[0016] Wherein, W q represents the value of the q-th parameter in the standard heat data, represents the q-th parameter of the i-th heat in the self-learning library, and γ i represents the weight of the i-th heat in the self-learning library, and N is a preset quantity, representing the number of heats in the self-learning library.

[0017] Based on a further improvement of the above method, the weights of the respective heats in the self-learning library are obtained by the following formula based on the die increments of the respective heats:

[0018]

[0019] Wherein, γ i represents the weight of the i-th heat in the second self-learning library, and M i represents the die increment of the i-th heat in the self-learning library.

[0020] Based on a further improvement of the above method, the die increments of the respective heats in the self-learning library are obtained by the following formula based on the heat data in the self-learning library and the historical average values and requirement values of the respective parameters of the steel grade to be produced:

[0021]

[0022] Wherein, M iDenote the mold increment of the \(i\)-th heat in the self-learning database, where \(i\) ranges from 1 to \(N\), \(H_0\) represents the average thickness of the converters in the historical heats of the steel grade to be produced, \(H\) i Denote the thickness of the converter in the \(i\)-th heat in the self-learning database, \(T_0\) represents the average smelting cycle of the historical heats of the steel grade to be produced, \(T\) i Denote the smelting cycle of the \(i\)-th heat in the self-learning database, \(X\) COAim Denote the required carbon-oxygen product of the steel grade to be produced, \(X\) iCO Denote the carbon-oxygen product of the \(i\)-th heat in the self-learning database, \(F\) 0k Denote the required tonnage of the \(k\)-th type of scrap of the steel grade to be produced, \(n\) represents the total categories of scrap, \(F\) ki Denote the tonnage of the \(k\)-th type of scrap in the \(i\)-th heat in the self-learning database, \(W\) 0liq Denote the required hot metal volume of the steel grade to be produced, \(W\) iliq Denote the hot metal volume of the \(i\)-th heat in the self-learning database, \(Q\) 0T Denote the required oxygen volume of the steel grade to be produced, \(Q\) iT Denote the oxygen volume of the \(i\)-th heat in the self-learning database, \(x\) 0I Denote the required hot metal composition of the steel grade to be produced, \(x\) Ii Denote the hot metal composition of the \(i\)-th heat in the self-learning database, \(x\) 0AC Denote the end-point carbon content of the steel grade to be produced, \(x\) iA0 Denote the end-point carbon content of the \(i\)-th heat in the self-learning database, \(x\) 0A0 Denote the end-point oxygen content of the steel grade to be produced, \(x\) iA0 Denote the end-point oxygen content of the \(i\)-th heat in the self-learning database, \(T\) 0A Denote the end-point temperature of the steel grade to be produced, \(T\) iA Denote the end-point temperature of the \(i\)-th heat in the self-learning database, \(\alpha\) represents the influence factor of the corresponding parameter.

[0023] Based on the further improvement of the above method, based on the required slag thickness of the steel grade to be produced, screening the furnace data with historical slag thickness values and production times meeting the requirements from the historical furnace data, the self-learning database for the corresponding steel grade includes:

[0024] Using the following formula, based on the required slag thickness of the steel grade to be produced, screening the furnaces with historical slag thickness values meeting the requirements from the historical furnace data:

[0025] S aim -\(\Delta S\leq S_0\leq S\) aim +\(\Delta S\)

[0026] where \(S_0\) represents the historical slag thickness value of a furnace in the historical furnace data, \(S\) aim represents the required slag thickness of the steel grade to be produced, \(\Delta S\) represents the allowable slag thickness deviation of the steel grade to be produced;

[0027] Sort the heats with historical slag thickness values meeting the requirements according to the generation time, and select the preset number of heats with the latest time to obtain the self-learning library.

[0028] Based on further improvement of the above method, the obtained historical slag thickness value includes:

[0029] Calculate the coupled slag thickness value for the corresponding heat based on video, audio, and vibration slag thickness values;

[0030] Based on the coupled slag thickness value, calculate the historical slag thickness value for the corresponding heat.

[0031] Based on further improvement of the above method, use the following formula to calculate the historical slag thickness value for the corresponding heat based on the coupled slag thickness value:

[0032]

[0033] Among them, S0 represents the historical slag thickness value of a heat, and S v represents the coupled slag thickness value of the corresponding heat, and Q represents the total oxygen supply.

[0034] Based on further improvement of the above method, use the following formula to calculate the coupled slag thickness value for the corresponding heat based on video, audio, and vibration slag thickness values:

[0035]

[0036] Among them, S v represents the coupled slag thickness value of a heat, S H represents the vibration slag thickness value of the corresponding heat, S V represents the video slag thickness value of the corresponding heat, S Y represents the audio slag thickness value of the corresponding heat, S AIM represents the standard slag thickness value of the steel grade to be produced, γ Y1 , γ V1 and γ H1 respectively represent the weight coefficients of audio, video, and vibration slag thickness values at low slag thickness, and γ Y2 , γ V2 and γ H2 respectively represent the weight coefficients of audio, video, and vibration slag thickness values at high slag thickness.

[0037] Based on further improvement of the above method, the acquisition device for slag thickness data includes: an audio slag sensor, a flame analysis camera, and an oxygen lance vibration sensor; among them,

[0038] Calculate the audio slag thickness value using the noise intensity during splashing collected by the audio slag sensor;

[0039] Calculate the video slag thickness value using the number of furnace mouth flame slags collected by the flame analysis camera;

[0040] The vibration slag thickness value is obtained by calculating the acceleration collected by the oxygen lance vibration sensor.

[0041] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0042] 1. The historical data of converter steelmaking is fully utilized. The parameters used involve raw material parameters, molten steel quality parameters, efficiency parameters, process control parameters and equipment parameters. The collaborative analysis of multivariate parameters is adopted for each parameter. The relationship between parameters is fully utilized to obtain the modulus increment of each furnace in the self-learning library, and the weight of each furnace in the self-learning library is further obtained to obtain the standard furnace data of the steel type to be produced; combined with the process requirements of the current production furnace, the predicted values of each parameter in the current production furnace are finally obtained; the prediction accuracy of the parameters is improved, and accurate parameter prediction is achieved;

[0043] 2. By applying multimodal computing and comprehensively considering the relationship and influence between audio, video and vibration data, multiple process parameters of the converter are applied to self-learning, which optimizes the self-learning data and improves the learning quality, thereby increasing the converter endpoint hit rate, reducing the production cost caused by trial and error, and shortening the converter steelmaking cycle;

[0044] 3. The process requirements of the current furnace comprehensively consider new process requirements, production conditions, and advances in smelting technology to ensure the representativeness, advancement, and applicability of standard furnace data. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components;

[0046] Figure 1 The present invention is a flowchart of a converter intelligent steelmaking method with multimodal information fusion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0048] A specific embodiment of the present invention discloses a converter intelligent steelmaking method with multi-modal information fusion, such as Figure 1 As shown. Includes:

[0049] S1: Based on the required slag thickness of the steel grade to be produced, the historical heat data with the historical slag thickness value and production time meeting the requirements are selected to obtain the self-learning library of the corresponding steel grade; the heat data is the value corresponding to each parameter of the corresponding heat;

[0050] S2: Obtain the die increment of each heat in the self-learning library based on the data of each heat in the self-learning library and the historical mean and required values of each parameter of the steel grade to be produced.

[0051] S3: Obtain the weight of each heat in the self-learning library based on the die increment of each heat.

[0052] S4: Obtain the standard heat data of the steel grade to be produced based on the weights of each heat and the data of each heat in the self-learning library.

[0053] S5: Obtain the predicted values of each parameter in the current production heat based on the standard heat data and the process requirements of the current production heat, and carry out the steelmaking process based on the predicted values of each parameter.

[0054] Historical heat data refers to the data of heats that have been smelted, including hot metal temperature, hot metal composition, scrap steel information, heat condition information of the heat, smelting process audio information, etc. In an optional implementation manner, the historical heat data can be preprocessed first, and the preprocessing includes data elimination and data normalization; wherein,

[0055] Data elimination includes: eliminating the heats with missing values in the historical heats; eliminating the heats with abnormal zero values of variables; eliminating the heats with abnormally large and small discrete values. Data elimination is used to detect and eliminate outliers in the historical heat data, ensure the consistency and reliability of the data, and provide a high-quality data basis for subsequent steps.

[0056] Data normalization is used to extract key features from the eliminated data, such as the frequency components of the historical slag thickness value audio, the image features of the video, and the spectral characteristics of the vibration. For the data collected by sensors, there will be extremely large or extremely small values, that is, there will be cases where the values are greater than the maximum value of the range and less than the minimum value of the range. When these situations occur, the normalization process can unify the scale of the data, make the distribution of the data more consistent, and make the key features more obvious.

[0057] Specifically, S1 includes steps S11 and S12:

[0058] S11: Use the following formula to screen out the heats with historical slag thickness values meeting the requirements from the historical heat data of the steel grade to be produced:

[0059] S aim -ΔS≤S0≤S aim +ΔS

[0060] wherein, S0 represents the historical slag thickness value of a heat in the historical heat data, S aimS represents the required slag thickness of the steel grade to be produced, and ΔS represents the allowable slag thickness deviation of the steel grade to be produced. Exemplarily, ΔS is taken as 0.15, S of ordinary steel aim is taken as 0.4, S of low-phosphorus steel aim is taken as 0.5. When the converter capacity is 120t, ΔS is taken as 0.15, and when the converter capacity is 200t, ΔS is taken as 0.18.

[0061] It should be noted that calculating the historical slag thickness value includes steps S111 and S112:

[0062] S111: Calculate the coupled slag thickness value of the corresponding heat based on the video, audio, and vibration slag thickness values using the following formula:

[0063]

[0064] where S v represents the coupled slag thickness value of a heat, S H represents the vibration slag thickness value of the corresponding heat, S V represents the video slag thickness value of the corresponding heat, S Y represents the audio slag thickness value of the corresponding heat, S AIM represents the standard slag thickness value of the steel grade to be produced, γ Y1 , γ V1 and γ H1 represent the weight coefficients of the audio, video, and vibration slag thickness values at low slag thickness respectively, γ Y2 , γ V2 and γ H2 represent the weight coefficients of the audio, video, and vibration slag thickness values at high slag thickness respectively. Exemplarily, γ Y1 , γ V1 and γ H1 are taken as 0.8, 0.05, and 0.15 respectively; γ Y2 , γ V2 and γ H2 are taken as 0.05, 0.8, and 0.15 respectively; when the converter capacity is 120t, S AIM is 0.3 meters.

[0065] Calculate the audio slag thickness value S Y using the following formula:

[0066]

[0067] where I represents the noise intensity collected in real time during steelmaking, I0 represents the noise intensity during splashing, S YI represents the audio slag thickness value, ΔH represents the height of the oxygen lance relative to the molten steel surface, R represents the diameter of the converter mouth, L represents the real-time oxygen flow rate during blowing, L0 represents the initial set oxygen flow rate during blowing, Y represents the real-time height of the hood, and Y0 represents the maximum opening height of the hood. It should be noted that I0 is the noise intensity during splashing, which is a fixed value determined by the performance of each converter; the audio system will collect a sound intensity during splashing at the site in the early stage of operation and use this value as I0, and I is the real-time collected value during the smelting process of each heat of steel.

[0068] The video slag thickness value S is obtained by the following formula V :

[0069]

[0070] else S V = S Y

[0071] Among them, S V represents the video slag thickness value, G represents the number of slag in the flame at the converter mouth collected by the video, G0 represents the critical number of slag in the flame at the converter mouth collected by the video, and Y0 represents the critical thickness of the converter slag. Exemplarily, G0 is 300, Y0 is a constant, and the reference value is 600 mm.

[0072] The vibration slag thickness value S is obtained by the following formula H :

[0073]

[0074] Among them, G represents the acceleration collected by the oxygen lance vibration sensor, a and b represent constants, F represents the oxygen pressure of the oxygen lance, L H represents the height of the oxygen lance, B H represents the correction term for the height of the furnace top. Exemplarily, a is 3, b is 5, and the correction term for the height of the furnace top is 50 cm; the specific parameter values need to be obtained through the experimental summary of the converter system.

[0075] In order to obtain the audio, video and vibration slag thickness values, the acquisition device for slag thickness data includes: an audio slag sensor, a flame analysis camera and an oxygen lance vibration sensor; among them,

[0076] The audio slag thickness value is calculated by using the noise intensity during splashing collected by the audio slag sensor;

[0077] The video slag thickness value is calculated by using the number of slag in the flame at the converter mouth collected by the flame analysis camera;

[0078] The vibration slag thickness value is calculated by using the acceleration collected by the oxygen lance vibration sensor.

[0079] In an alternative embodiment, the converter body is equipped with a total of 24 high-precision sensing devices, including 8 audio slag-making sensors evenly distributed on the furnace wall of the converter to monitor the audio data of the slag layer in the furnace in real time and obtain the real-time noise intensity during the steelmaking process; 4 flame analysis cameras installed on the top of the converter and aimed at the furnace mouth to capture and analyze the shape, color, and temperature of the flame, thereby collecting video data of the slag layer and obtaining the quantity of the slag at the furnace mouth flame; 3 oxygen lance vibration sensors located at the root of the oxygen lance to monitor the vibration of the oxygen lance during the blowing process and obtain acceleration information to ensure the stability and safety of the blowing; 9 thickness measurement sensors distributed on the inner wall of the converter to monitor the thickness change of the converter in real time, collect slag thickness data, and prevent potential safety hazards caused by the over-thinning of the converter. The installation positions of these devices are carefully designed to ensure the accuracy and real-time nature of the data, providing the converter steelmaking with all-round sensing capabilities to achieve comprehensive monitoring of different production stages and different production states; and optimizing the effect of data collection, further improving the accuracy of data collection and the efficiency and quality of steelmaking.

[0080] S112: Based on the coupled slag thickness value, calculate the historical slag thickness value corresponding to the furnace number using the following formula:

[0081]

[0082] Wherein, S0 represents the historical slag thickness value of a furnace number, S v represents the coupled slag thickness value corresponding to the furnace number, and Q represents the total oxygen supply.

[0083] S12: Sort the furnace numbers with historical slag thickness values meeting the requirements according to the generation time, and screen out the data of the preset number of the latest furnace numbers to obtain the self-learning library. Exemplarily, the preset number is 6.

[0084] Furthermore, use the following formula to obtain the mold increment of each furnace number in S2:

[0085]

[0086] Wherein, M i represents the mold increment of the i-th furnace number in the self-learning library, the value of i ranges from 1 to N, H0 represents the average thickness of the converter of the historical furnace numbers of the steel type to be produced, H i represents the thickness of the converter of the i-th furnace number in the self-learning library, T0 represents the average smelting cycle of the historical furnace numbers of the steel type to be produced, T i represents the smelting cycle of the i-th furnace number in the self-learning library, X COAim represents the required carbon-oxygen product of the steel type to be produced, X iCO represents the carbon-oxygen product of the i-th furnace number in the self-learning library, F 0k represents the required tonnage of the k-th type of scrap steel of the steel type to be produced, n represents the total categories of scrap steel, Fki Denotes the tonnage of the k-th type of scrap steel in the i-th heat in the self-learning database, W 0liq Denotes the required hot metal volume for the steel grade to be produced, W iliq Denotes the hot metal volume in the i-th heat in the self-learning database, Q 0T Denotes the required oxygen volume for the steel grade to be produced, Q iT Denotes the oxygen volume in the i-th heat in the self-learning database, x 0I Denotes the required hot metal composition for the steel grade to be produced, x Ii Denotes the hot metal composition in the i-th heat in the self-learning database, x 0A Denotes the final carbon content of the steel grade to be produced, x iA0 Denotes the final carbon content in the i-th heat in the self-learning database, x 0A0 Denotes the final oxygen content of the steel grade to be produced, x iA0 Denotes the final oxygen content in the i-th heat in the self-learning database, T 0A Denotes the final temperature of the steel grade to be produced, T iA Denotes the final temperature in the i-th heat in the self-learning database, α represents the influence factor of the corresponding parameter, which is a fixed value in the industry.

[0087] It should be noted that when calculating the mold increment of each heat, the parameters used involve raw material parameters, including the type and tonnage of scrap steel, hot metal volume, and hot metal composition; molten steel quality parameters, including oxygen volume, final carbon content, final oxygen content, and final temperature; efficiency parameters, including smelting cycle; equipment parameters, including the average thickness of the converter; process control parameters, including carbon-oxygen product; by deeply analyzing the historical heat data, the compliance between the actual smelting results and the process requirements is evaluated, and the mold increment of each heat is obtained.

[0088] Furthermore, the weight of each heat in S3 is obtained using the following formula:

[0089]

[0090] Among them, γ i Denotes the weight of the i-th heat in the second self-learning database, M i Denotes the mold increment of the i-th heat in the self-learning database, N is a preset quantity, representing the number of heats in the self-learning database. Exemplarily, N is 6.

[0091] Furthermore, the standard heat data of the steel grade to be produced in S4 is obtained using the following formula:

[0092]

[0093] Among them, W q Denotes the value of the q-th parameter in the standard heat data, Denotes the q-th parameter of the i-th heat in the self-learning database, γ iIt represents the weight of the i-th heat in the self-learning library.

[0094] It should be noted that in the process of generating the standard heat in this embodiment, multi-modal information fusion is carried out, comprehensively considering the mutual relationship and influence among audio, video and vibration data, and performing fusion analysis on real-time production data and historical data, which improves the real-time performance and adaptability of the generation of standard heat data; comprehensively considering factors such as new process requirements, production conditions and the progress of smelting technology, etc., to ensure the representativeness and advancement of the standard heat data.

[0095] Further, the predicted values of each parameter in S5 for the current production heat are obtained by the following formula:

[0096]

[0097] where, W q represents the value of the q-th parameter in the standard heat data, W q′ represents the predicted value of the q-th parameter in the current production heat, W0 q represents the process requirement value of the q-th parameter in the current production heat, α q represents the influence parameter of the q-th parameter.

[0098] When calculating the predicted value of the current production heat, considering the changes in real-time production conditions, such as raw material composition, temperature, charging amount, etc., the static calculation result is adjusted in real time; ensuring a high degree of matching between the static calculation result and the actual production conditions, and providing accurate parameter prediction for the converter steelmaking process.

[0099] Further, based on the predicted values of each parameter, an accurate initial setting is provided for the converter steelmaking process, and the steelmaking process of the current heat is started. Exemplarily, the parameters include: raw material parameters, including the type and tonnage of scrap steel, hot metal volume, hot metal composition; liquid steel quality parameters, including oxygen content, end carbon content, end oxygen content, end temperature; efficiency parameters, including smelting cycle; equipment parameters, including the average thickness of the converter; process control parameters, including carbon-oxygen product.

[0100] The heat data collected each time steel is made is added to the historical data of the corresponding steel grade to complete the update of the historical data.

[0101] The present invention makes full use of the historical data of converter steelmaking, and the parameters used involve raw material parameters, molten steel quality parameters, efficiency parameters, process control parameters and equipment parameters. The collaborative analysis of multivariable parameters is adopted for each parameter, and the relationship between the parameters is fully utilized to obtain the modulus increment of each furnace in the self-learning library, and further obtain the weight of each furnace in the self-learning library, and obtain the standard furnace data of the steel type to be produced; then combined with the process requirements of the current production furnace, the predicted value of each parameter in the current production furnace is finally obtained; the prediction accuracy of the parameters is improved, and accurate parameter prediction is achieved; by applying multimodal calculation, the mutual relationship and influence between audio, video and vibration data are comprehensively considered, and multiple process parameters of the converter are applied to self-learning, the self-learning data is optimized, the learning quality is improved, and then the converter endpoint hit rate is improved, the production cost caused by trial and error is reduced, and the converter steelmaking cycle is shortened; the process requirements of the current furnace comprehensively consider new process requirements, production conditions and metallurgical technology progress factors, to ensure the representativeness, advancement and applicability of the standard furnace data.

[0102] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0103] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A converter intelligent steelmaking method for multimodal information fusion, characterized in that, Including: Based on the required slag thickness of the steel grade to be produced, screen the furnace data with historical slag thickness values and production times that meet the requirements from the historical furnace data to obtain the self-learning library for the corresponding steel grade; the furnace data are the values corresponding to the parameters of each furnace. Based on the furnace data of each furnace in the self-learning library and the historical average and required values of the parameters of the steel grade to be produced, obtain the die increment of each furnace in the self-learning library. Based on the die increment of each furnace, obtain the weight of each furnace in the self-learning library. Based on the weights of each furnace and the furnace data in the self-learning library, obtain the standard furnace data of the steel grade to be produced. Based on the standard furnace data and the process requirements of the current production furnace, obtain the predicted values of the parameters in the current production furnace, and carry out the steelmaking process based on the predicted values of the parameters.

2. The converter intelligent steelmaking method for multi-modal information fusion according to claim 1, characterized in that, Using the following formula, based on the standard furnace data and the process requirements of the current production furnace, obtain the predicted values of the parameters in the current production furnace: Among them, W q represents the value of the q-th parameter in the standard furnace data, and W q′ represents the predicted value of the q-th parameter of the current production furnace, and W0 q represents the process requirement value of the q-th parameter of the current production furnace, and α q represents the influencing parameter of the q-th parameter.

3. The converter intelligent steelmaking method for multimodal information fusion according to claim 2, characterized in that Using the following formula, based on the weights of each furnace and the furnace data in the self-learning library, obtain the standard furnace data of the steel grade to be produced: Among them, W q represents the value of the q-th parameter in the standard heat data, represents the q-th parameter of the i-th heat in the self-learning library, and γ i represents the weight of the i-th heat in the self-learning library. N is a preset quantity, representing the number of heats in the self-learning library.

4. The converter intelligent steelmaking method for multi-modal information fusion according to claim 3, characterized in that Using the following formula, based on the die increment of each furnace, obtain the weight of each furnace in the self-learning library: Among them, γ i represents the weight of the i-th heat in the second self-learning library, and M i represents the mold increment of the i-th heat in the self-learning library.

5. The converter intelligent steelmaking method for multi-modal information fusion according to claim 4, characterized in that Using the following formula, based on the furnace data of each furnace in the self-learning library and the historical average and required values of the parameters of the steel grade to be produced, obtain the die increment of each furnace in the self-learning library: Among them, M i represents the mold increment of the i-th heat in the self-learning library, where the value of i ranges from 1 to N, H0 represents the average thickness of the converters in the historical heats of the steel grade to be produced, H i represents the thickness of the converter in the i-th heat in the self-learning library, T0 represents the average smelting cycle of the historical heats of the steel grade to be produced, T i represents the smelting cycle of the i-th heat in the self-learning library, X COAim represents the required carbon-oxygen product of the steel grade to be produced, X iCO represents the carbon-oxygen product of the i-th heat in the self-learning library, F 0k represents the required tonnage of the k-th type of scrap for the steel grade to be produced, n represents the total categories of scrap, F ki represents the tonnage of the k-th type of scrap in the i-th heat in the self-learning library, W 0liq represents the required hot metal amount of the steel grade to be produced, W ili represents the hot metal amount of the i-th heat in the self-learning library, Q 0T represents the required oxygen amount of the steel grade to be produced, Q iT represents the oxygen amount of the i-th heat in the self-learning library, x 0I represents the required hot metal composition of the steel grade to be produced, x Ii represents the hot metal composition of the i-th heat in the self-learning library, x 0AC represents the final carbon content of the steel grade to be produced, x iA0 represents the final carbon content of the i-th heat in the self-learning library, x 0A0 represents the final oxygen content of the steel grade to be produced, x iA0 represents the final oxygen content of the i-th heat in the self-learning library, T 0A represents the final temperature of the steel grade to be produced, T iA represents the final temperature of the i-th heat in the self-learning library, and α represents the influence factor of the corresponding parameter.

6. The converter intelligent steelmaking method for multi-modal information fusion according to claim 1, characterized in that, Based on the required slag thickness of the steel grade to be produced, screening the furnace data with historical slag thickness values and production times that meet the requirements from the historical furnace data to obtain the self-learning library for the corresponding steel grade includes: Using the following formula, based on the required slag thickness of the steel grade to be produced, screen the furnaces with historical slag thickness values that meet the requirements from the historical furnace data: S aim -ΔS ≤ S0 ≤ S aim +ΔS Among them, S0 represents the historical slag thickness value of a furnace in the historical furnace data, and S aim represents the required slag thickness of the steel grade to be produced, and ΔS represents the allowable slag thickness deviation of the steel grade to be produced; Sort the furnaces with historical slag thickness values that meet the requirements according to the production time, and screen out the furnace data of the preset number of the latest times to obtain the self-learning library.

7. The converter intelligent steelmaking method for multimodal information fusion according to claim 1 or 6, characterized in that Obtaining the historical slag thickness value includes: Based on the video, audio and vibration slag thickness values, calculate the coupled slag thickness value corresponding to the furnace. Based on the coupled slag thickness value, calculate the historical slag thickness value corresponding to the furnace.

8. The converter intelligent steelmaking method for multimodal information fusion according to claim 7, characterized in that Using the following formula, based on the coupled slag thickness value, calculate the historical slag thickness value corresponding to the furnace: Among them, S0 represents the historical slag thickness value of a heat, and S v represents the coupled slag thickness value corresponding to the heat, and Q represents the total oxygen supply volume.

9. The converter intelligent steelmaking method for multi-modal information fusion according to claim 8, characterized in that, Using the following formula, based on the video, audio and vibration slag thickness values, calculate the coupled slag thickness value corresponding to the furnace: Among them, S v represents the coupled slag thickness value of a heat, S H represents the vibrating slag thickness value corresponding to the heat, S V represents the video slag thickness value corresponding to the heat, S Y represents the audio slag thickness value corresponding to the heat, S AIM represents the standard slag thickness value of the steel grade to be produced, γ Y1 , γ V1 and γ H1 respectively represent the weight coefficients of the audio, video and vibrating slag thickness values at low slag thickness, γ Y2 , γ V2 and γ H2 respectively represent the weight coefficients of the audio, video and vibrating slag thickness values at high slag thickness.

10. The converter intelligent steelmaking method for multimodal information fusion according to claim 1 or 9, characterized in that The acquisition device for slag thickness data includes: an audio slag sensor, a flame analysis camera and an oxygen lance vibration sensor; wherein, Calculate the audio slag thickness value using the noise intensity during splashing collected by the audio slag sensor. Calculate the video slag thickness value using the number of furnace mouth flame slags collected by the video collected by the flame analysis camera. Calculate the vibration slag thickness value using the acceleration collected by the oxygen lance vibration sensor.

Images