Electric arc furnace foamed slag splashing control early warning method, system, device and storage medium

Abstract

This invention discloses a method, system, device, and storage medium for controlling and warning of foamed slag splashing in electric arc furnaces, relating to the field of electric arc furnace foamed slag splashing control and warning technology. The method includes: acquiring infrared images and high-temperature visible light images of the smelting area inside the electric arc furnace; obtaining dynamic visual features and thermal distribution change features based on the infrared and high-temperature visible light images; inputting the dynamic visual features and thermal distribution change features into a preset splashing risk identification model to obtain an initial splashing risk prediction result; determining the current electric arc furnace submerged arc state and the real-time height of the foamed slag based on the infrared and high-temperature visible light images; determining the target splashing risk judgment result based on the initial splashing risk prediction result, the current electric arc furnace submerged arc state, and the real-time height of the foamed slag; and performing electric arc furnace foamed slag splashing control and warning based on the target splashing risk judgment result. This improves the accuracy of electric arc furnace foamed slag splashing control and warning.

Description

Technical Field

[0001] This invention relates to the field of electric arc furnace foam slag splash control and early warning technology, and in particular to a method, system, equipment and storage medium for electric arc furnace foam slag splash control and early warning. Background Technology

[0002] Existing early warning methods for electric arc furnace steelmaking splashing are insufficient in their ability to identify samples in the early stages of splashing, their risk output is prone to fluctuations, and there is a lack of sufficient consistency constraints between the prediction results and the actual changes in the electric arc furnace steelmaking process, resulting in high false alarm and false negative rates. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to overcome the shortcomings of the prior art and provide a method, system, device and storage medium for controlling and warning of foam slag splashing in electric arc furnaces.

[0004] This invention provides the following technical solution: In a first aspect, the present invention provides a method for controlling and warning of foamy slag splashing in an electric arc furnace, the method comprising: Infrared and high-temperature visible light images of the smelting area inside the electric arc furnace were acquired. Dynamic visual features and thermal distribution change features are obtained from the infrared image and the high-temperature visible light image; The dynamic visual features and the thermal distribution change features are input into a preset splash risk identification model to obtain an initial splash risk prediction result. The current submerged arc status of the electric arc furnace and the real-time height of the foamy slag are determined based on the infrared image and the high-temperature visible light image. The target splash risk assessment result is determined based on the initial splash risk prediction result, the current electric arc furnace submerged arc status, and the real-time height of the foam slag. The electric arc furnace foam slag splash control and early warning are then performed based on the target splash risk assessment result.

[0005] In an optional implementation, the method further includes: Acquire training samples, which include normal samples, splash precursor samples, and splash samples; The training samples are input into the initial splash risk identification model to obtain the risk prediction value; Calculate the precursor enhancement loss, risk increment constraint loss, and temperature-brightness motion consistency loss based on the training samples and the risk prediction value; calculate the target loss value based on the precursor enhancement loss, the risk increment constraint loss, and the temperature-brightness motion consistency loss. The model parameter update direction of the initial splash risk identification model is optimized based on the target loss value until the trained preset splash risk identification model is obtained.

[0006] In an optional implementation, the dynamic visual features include the intensity of foam movement, the thermal distribution change features include local brightness change and local temperature change, and the formula for calculating the target loss value is: In the formula, For the aforementioned precursor enhancement loss, For the aforementioned risk-increasing constraint loss, For the aforementioned temperature and brightness motion consistency loss, , All are loss weighting coefficients; The formula for calculating the precursor enhancement loss is as follows: In the formula, N is the number of training samples, and t is the time index of the sample. Let be the weight of the t-th sample. This represents the true label of the t-th sample. Let ln be the risk prediction value corresponding to the t-th sample, and ln be the natural logarithm function. The formula for calculating the loss under the increasing risk constraint is as follows: In the formula, M is the number of sample points within the precursor time window; The formula for calculating the temperature-brightness motion consistency loss is as follows: In the formula, This represents the splash risk probability output by the model at time time . The comprehensive precursor index at time t; The formula for calculating the comprehensive precursor index is as follows: In the formula, The normalized foam motion intensity, This represents the normalized rate of change of local average brightness. This represents the normalized local average rate of change of temperature. , , These are the feature weight coefficients.

[0007] In an optional implementation, the precursor enhancement loss is used to increase the training weights of samples within a preset time window before the splash occurs. The risk-increasing constraint loss is used to penalize risk outputs that do not satisfy the increasing relationship within the time interval before the splash occurs. The temperature-brightness motion consistency loss is used to ensure that the risk prediction value output by the preset splash risk identification model remains consistent with the comprehensive precursor index composed of foam motion intensity, brightness change, and temperature change.

[0008] In an optional implementation, obtaining dynamic visual features and thermal distribution change features based on the infrared image and the high-temperature visible light image includes: The infrared image and the high-temperature visible light image are preprocessed to obtain a fused image sequence; The fused image sequence is segmented to obtain a foam region image sequence; The pixel motion vector of the foam region image sequence is calculated based on the optical flow algorithm, the foam motion intensity is determined based on the pixel motion vector, and the foam motion intensity is used as the dynamic visual feature. In the fused image sequence, a first region of interest is obtained from the high-temperature visible light image that corresponds to the area directly below the electrode or the area with high sputtering incidence, and the local average brightness change rate of the first region of interest is calculated. In the fused image sequence, a second region of interest is obtained corresponding to the high-temperature activity area directly below the electrode in the infrared thermal image, and the local average temperature change rate of the second region of interest is calculated. The average brightness change rate and the average temperature change rate are used as the characteristics of the thermal distribution change.

[0009] In an optional implementation, determining the current submerged arc state of the electric arc furnace and the real-time height of the foamed slag based on the infrared image and the high-temperature visible light image includes: Image morphological features and temperature distribution features are obtained from the fused image sequence; The image morphological features and the temperature distribution features are input into the submerged arc state recognition model to obtain the current submerged arc state of the electric arc furnace. The image sequence of the foam region is mapped to obtain the spatial coordinates of the foam. The real-time height of the foam slag is obtained by calculating the spatial coordinates of the foam based on the spatial relationship between the surface boundary of the foam slag and the preset reference position.

[0010] In an optional implementation, determining the target splash risk assessment result based on the initial splash risk prediction result, the current electric arc furnace submerged arc state, and the real-time height of the foamed slag includes: If the current electric arc furnace is fully submerged and the real-time height of the foam slag is within the preset target range, then the target splash risk judgment result is the initial splash risk judgment result. If the current electric arc furnace is partially submerged or not submerged, the initial splash risk assessment result is enhanced to obtain the target splash risk assessment result. If the real-time height of the foam residue is not within the preset target range, the initial splash risk assessment result is enhanced to obtain the target splash risk assessment result. If the current electric arc furnace is partially or not submerged, and the real-time height of the foam slag is not within the preset target range, then a high-level splash warning will be used as the target splash risk assessment result.

[0011] In a second aspect, the present invention provides an electric arc furnace foam slag splash control and early warning system, the system comprising: The acquisition module is used to acquire infrared images and high-temperature visible light images of the smelting area inside the electric arc furnace; The acquisition module is used to acquire dynamic visual features and thermal distribution change features based on the infrared image and the high-temperature visible light image; The prediction module is used to input the dynamic visual features and the heat distribution change features into a preset splash risk identification model to obtain an initial splash risk prediction result; The determination module is used to determine the current submerged arc state of the electric arc furnace and the real-time height of the foamed slag based on the infrared image and the high-temperature visible light image; The early warning module is used to determine the target splash risk judgment result based on the initial splash risk prediction result, the current electric arc furnace submerged arc state, and the real-time height of the foam slag, and to perform early warning for electric arc furnace foam slag splash control based on the target splash risk judgment result.

[0012] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the electric arc furnace foam slag splash control and early warning method as described in any of the foregoing embodiments.

[0013] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the electric arc furnace foam slag splash control and early warning method as described in any of the foregoing embodiments.

[0014] This invention discloses a method, system, equipment, and storage medium for controlling and warning of foamy slag splashing in electric arc furnaces. Based on infrared and high-temperature visible light images, it extracts dynamic visual features and thermal distribution change characteristics of the foamy slag region. These features are then used to identify splashing precursors, obtaining initial splashing risk prediction results. Model prediction ensures the ability to identify samples during the splashing precursor stage, improving the robustness of warnings under complex operating conditions. Simultaneously, it achieves synchronous non-contact monitoring of the submerged arc state and the real-time height of the foamy slag based on infrared and high-temperature visible light images. Combined with the initial splashing risk prediction results, a joint judgment is made, further reducing false alarm and false negative rates, thus enabling timely warnings. This improves smelting stability, reduces operational risks and smelting costs, and enhances smelting efficiency and steel quality. Attached Figure Description

[0015] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope of protection of the present invention. In the various drawings, similar components are numbered similarly.

[0016] Figure 1 A flowchart of the electric arc furnace foam slag splash control and early warning method proposed in this embodiment is shown. Figure 2 Another flowchart of the electric arc furnace foam slag splash control and early warning method proposed in this embodiment is shown; Figure 3 This illustration shows another flowchart of the electric arc furnace foam slag splash control and early warning method proposed in this embodiment; Figure 4 A schematic diagram of another process for the electric arc furnace foam slag splash control and early warning method proposed in this embodiment is shown; Figure 5 A schematic diagram of the structure of the electric arc furnace foam slag splash control and early warning system proposed in this embodiment is shown.

[0017] Explanation of reference numerals in the attached diagram: 500 - Electric arc furnace foam slag splash control and early warning system; 501 - Acquisition module; 502 - Acquisition module; 503 - Prediction module; 504 - Determination module; 505 - Early warning module. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] The components of the embodiments of the invention described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] In the following, the terms “comprising,” “having,” and their cognates, which may be used in various embodiments of the invention, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as excluding, firstly, the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more features, numbers, steps, operations, elements, components, or combinations thereof.

[0021] Furthermore, the terms "first," "second," and "third" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0022] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the invention pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of the invention.

[0023] Example 1 This disclosure provides a method for controlling and warning of foam slag splashing in an electric arc furnace.

[0024] Please see Figure 1 The method for controlling and warning of foam slag splashing in electric arc furnace includes steps S101 to S106, and each step is described in detail below.

[0025] Step S101: Acquire infrared images and high-temperature visible light images of the smelting area inside the electric arc furnace.

[0026] In this embodiment, infrared images and high-temperature visible light images of the smelting area inside the electric arc furnace are simultaneously acquired by an infrared camera and a high-temperature industrial camera set at the observation position of the electric arc furnace.

[0027] Step S102: Obtain dynamic visual features and thermal distribution change features based on the fused image sequence.

[0028] In this embodiment, dynamic visual features and thermal distribution change features are extracted from the fused image sequence. The dynamic visual features reflect the mechanical instability process of the foam slag interface, while the thermal distribution change features reflect the thermodynamic imbalance process of the energy field inside the furnace. Both can provide characteristics before splashing occurs.

[0029] Please see Figure 2 In one specific embodiment, step S102 includes steps S1021 to S1026, and each step is described in detail below.

[0030] Step S1021: Perform image preprocessing on the infrared image and the high-temperature visible light image to obtain a fused image sequence.

[0031] In this embodiment, at least one of the following image preprocessing methods is applied to the infrared image and the high-temperature visible light image: time synchronization, spatial registration, defogging, noise reduction, brightness correction, distortion correction, and temperature calibration, to obtain a fused image sequence.

[0032] Step S1022: Segment the fused image sequence to obtain a foam region image sequence.

[0033] In this embodiment, a semantic segmentation model is used to segment the fused image sequence to obtain a bubble region image sequence.

[0034] Understandably, accurately separating the main foam slag region and eliminating irrelevant background interference from furnace walls, electrodes, and molten steel surfaces allows subsequent dynamic feature extraction to focus on the direct carrier of splashing (the foam layer), significantly improving the purity of the physical meaning of features and the efficiency of model training. Simultaneously, the foam region is the core interface for splash energy release; traditional global image analysis introduces a large amount of noise features (such as furnace wall thermal radiation fluctuations), and this step achieves controllable focus through segmentation.

[0035] Step S1023: Calculate the pixel motion vector of the foam region image sequence based on the optical flow algorithm, determine the foam motion intensity based on the pixel motion vector, and use the foam motion intensity as the dynamic visual feature.

[0036] In this embodiment, pixel motion vectors of the foam region image sequence are calculated based on an optical flow algorithm (e.g., the Farneback dense optical flow algorithm), and the foam motion intensity, i.e., dynamic visual features, is determined according to the average amplitude, peak value, variance, or directional dispersion of the pixel motion vectors. For example, the 90th percentile of the pixel motion vector amplitude is used as the foam motion intensity.

[0037] Understandably, by using dynamic visual features to quantify the macroscopic fluidity and internal energy disturbance of foam slag, the invisible latent variables of the metallurgical process, such as slag turbulence and gas escape rate, are transformed into calculable, traceable, and strongly correlated visual dynamic indicators with splash risk, thus making up for the lack of mechanism in the single-dimensional observation of temperature / brightness.

[0038] Step S1024: Obtain the first region of interest in the high-temperature visible light image corresponding to the region directly below the electrode or the region with high sputtering incidence in the fused image sequence, and calculate the local average brightness change rate of the first region of interest.

[0039] In this embodiment, a first region of interest (ROI) corresponding to the area directly below the electrode or a high-incidence area of ​​sputtering is obtained from the fused image sequence in a high-temperature visible light image. The average brightness change rate of the first ROI within a continuous time window is calculated to obtain the local average brightness change rate, thereby capturing the transient response of energy input at the electrode-slag interface and reflecting the local arc stability and slag layer breakdown risk. A sudden increase in brightness often indicates arc exposure, excessively thin slag layer, or local boiling, which is a differentiated precursor signal from slowly heating sputtering. When the average brightness change rate exceeds a preset brightness threshold, it is marked as a brightness burst event.

[0040] Step S1025: Obtain the second region of interest in the infrared thermal image corresponding to the high-temperature activity area directly below the electrode in the fused image sequence, and calculate the local average temperature change rate of the second region of interest.

[0041] In this embodiment, a second region of interest (ROI) corresponding to the high-temperature activity area directly below the electrode is obtained from the fused image sequence in the infrared thermal image. The average temperature change rate of the second ROI within a continuous time window is calculated to obtain the local average temperature change rate. This allows for precise monitoring of the heat accumulation rate at the electrode tip and identification of local overheating phenomena caused by slag thinning or arc penetration. The temperature change rate reflects the dynamic process of thermal equilibrium disruption better than absolute temperature, reducing false triggering under steady-state high-temperature conditions. Furthermore, infrared light avoids visible light blind spots (such as behind smoke and dust) and is specifically used to detect hidden thermal runaway. When the average temperature change rate exceeds a preset threshold, it is marked as a temperature burst event.

[0042] Step S1026: The average brightness change rate and the average temperature change rate are used as the thermal distribution change characteristics.

[0043] In this embodiment, the average brightness change rate and the average temperature change rate are integrated into a thermal distribution change feature, enabling risk identification to have multiple verifications based on energy (brightness) and thermodynamics (temperature).

[0044] Step S103: Input the dynamic visual features and the heat distribution change features into the preset splash risk identification model to obtain the initial splash risk prediction result.

[0045] In this embodiment, dynamic visual features and thermal distribution change features are input into a preset splash risk identification model, and an initial splash risk prediction result is output, thereby realizing an end-to-end mapping from multi-source heterogeneous features to splash probability. At the same time, the model prediction ensures the ability to identify samples in the splash precursor stage.

[0046] The preset splash risk identification model is a neural network model based on time series feature input. The neural network model includes any one or a combination of one-dimensional convolutional neural network, recurrent neural network, long short-term memory network, gated recurrent unit network, and Transformer network.

[0047] Please see Figure 3 In one specific embodiment, the method further includes steps S301 to S304, which are described in detail below.

[0048] Step S301: Obtain training samples, which include normal samples, splash precursor samples, and splash samples.

[0049] In this embodiment, a combination of historical production videos, operation records, and manual verification is used to determine the time of the splashing event, and training samples are constructed according to the time series.

[0050] The training samples include normal samples, splash precursor samples, and splash samples. Normal samples are image sequences that have not occurred and have no obvious precursor features; splash precursor samples are image sequences within a preset time window before the splash occurs; splash samples are image sequences within a short time interval after the splash occurs.

[0051] Step S302: Input the training samples into the initial splash risk identification model to obtain the risk prediction value.

[0052] In this embodiment, the dynamic visual features and thermal distribution change features extracted from the training samples are input into the initial splash risk identification model to obtain the risk prediction value. A time alignment benchmark between the model output and the actual splash process is established, providing a differentiable target signal for the subsequent gradient backpropagation of the triple loss function. This ensures that the model parameter update direction always points to the optimization goal of improving precursor sensitivity, strengthening trend consistency, and aligning with physical mechanisms.

[0053] Step S303: Calculate the precursor enhancement loss, risk increment constraint loss, and temperature-brightness motion consistency loss based on the training samples and the risk prediction value; calculate the target loss value based on the precursor enhancement loss, the risk increment constraint loss, and the temperature-brightness motion consistency loss.

[0054] In this embodiment, the precursor enhancement loss, risk increment constraint loss, and temperature-brightness motion consistency loss are calculated based on the training samples and risk prediction values. Furthermore, the target loss value is calculated based on the precursor enhancement loss, risk increment constraint loss, and temperature-brightness motion consistency loss to guide the model optimization direction, thereby improving the model prediction accuracy and thus enhancing the ability to identify samples in the precursor stage of splashing.

[0055] The formula for calculating the target loss value is as follows: In the formula, As a precursor to amplify the loss, To constrain losses due to increasing risk, For Wen Liang's motion consistency loss, , All of these are loss weighting coefficients.

[0056] The formula for calculating the precursor enhancement loss is: In the formula, N is the number of training samples, and t is the time index of the sample. Let be the weight of the t-th sample. This represents the true label of the t-th sample. Let be the risk prediction value corresponding to the t-th sample, and ln be the natural logarithm function.

[0057] The precursor enhancement loss is used to increase the training weights of samples within a preset time window before the splash occurs, thereby enhancing the model's ability to identify the precursor stages of splash. To improve the model's ability to identify samples in the key precursor stages before splash occurs, training samples within the preset time window before splash occurrence are assigned higher loss weights than ordinary samples, constituting the precursor enhancement loss. The preset time window can be set to 1 to 10 seconds before splash occurrence, preferably 2 to 5 seconds. The closer the sample is to the splash occurrence time, the higher its loss weight.

[0058] The formula for calculating the loss under the increasing risk constraint is: In the formula, M is the number of sample points within the precursor time window.

[0059] Understandably, the risk of splashing typically increases gradually during its actual evolution, accompanied by enhanced slag surface movement, abnormal brightness, and abrupt temperature changes. To make the model output more consistent with the temporal evolution of splash precursors, a risk-increasing constraint loss is applied during training to penalize risk outputs that do not satisfy the increasing relationship within the interval before the splashing occurs. This reduces the irregular fluctuations in risk probability between adjacent time points and enhances the rationality of its temporal evolution.

[0060] The formula for calculating the motion consistency loss of Wenliang is: In the formula, This represents the splash risk probability output by the model at time time . Let be the comprehensive precursor index at time t.

[0061] The formula for calculating the comprehensive precursor index is: In the formula, The normalized foam motion intensity, This represents the normalized rate of change of local average brightness. This represents the normalized local average rate of change of temperature. , , These are the feature weight coefficients.

[0062] The comprehensive precursor index is used to fuse and characterize multi-source precursor information before splashing occurs, mapping the intensity of foam slag movement, local brightness changes, and local temperature changes into a single precursor index. By introducing a temperature-brightness-movement consistency loss, the risk probability output by the splashing risk identification model is constrained to be consistent with the comprehensive precursor index, thereby making the model output more consistent with the actual evolution of splashing precursors and improving the interpretability and anti-interference capability of the early warning results.

[0063] Step S304: Optimize the model parameter update direction of the initial splash risk identification model according to the target loss value until the trained preset splash risk identification model is obtained.

[0064] In this embodiment, the model parameter update direction of the initial splash risk identification model is optimized according to the target loss value until the model is stable, thus obtaining the trained preset splash risk identification model.

[0065] Step S104: Determine the current submerged arc state of the electric arc furnace and the real-time height of the foamed slag based on the fused image sequence.

[0066] In this embodiment, the current submerged arc state of the electric arc furnace and the real-time height of the foamed slag are determined based on the fused image sequence. The submerged arc state reflects the degree of arc exposure and the energy release state within the furnace, while the real-time height of the foamed slag reflects the slag layer coverage and slag surface fluctuation.

[0067] Please see Figure 4 In one specific embodiment, step S104 includes steps S1041 to S1044, and each step is described in detail below.

[0068] Step S1041: Obtain image morphological features and temperature distribution features based on the fused image sequence.

[0069] In this embodiment, image morphological features and temperature distribution features are extracted from the fused image sequence. For example, the boundary of the foam slag region is extracted from the foam region image to use the boundary contour features of the foam slag region as image morphological features; at the same time, image brightness uniformity features, high-temperature region area ratio features, and temperature gradient distribution features are extracted from the fused image sequence as temperature distribution features.

[0070] As an example, the fused image sequence is preprocessed to determine the target analysis region, i.e., the smelting region inside the furnace, in the preprocessed fused image sequence. The distribution of pixel brightness values ​​within the target analysis region is statistically analyzed, and the image brightness uniformity characteristics are determined based on the variance, standard deviation, or coefficient of variation of the brightness values. High-temperature pixels within the target analysis region are segmented according to a preset temperature threshold or brightness temperature threshold, and the area ratio of high-temperature pixels to the total area of ​​the target analysis region is determined. The spatial temperature gradient is calculated based on the temperature values ​​of each pixel within the target analysis region, and the temperature gradient distribution characteristics are determined based on the average value, maximum value, or distribution dispersion of the temperature gradient.

[0071] Understandably, by characterizing the spatial geometry and thermodynamic state of foam slag through image morphology features and temperature distribution features, a continuous spectrum description beyond binary (buried / unburied) can be provided for the identification of buried arc state.

[0072] Step S1042: Input the image morphological features and the temperature distribution features into the submerged arc state recognition model to obtain the current submerged arc state of the electric arc furnace.

[0073] In this embodiment, image morphological features and temperature distribution features are input into the submerged arc state recognition model to obtain the current submerged arc state of the electric arc furnace. This enables real-time autonomous perception of the core process state of the electric arc furnace, eliminating reliance on manual observation or auxiliary sensors (such as electrode current sudden changes), and ensuring the objectivity and continuity of state judgment in unattended or high-risk environments.

[0074] The arc burial status identification results include three states: fully buried arc, partially buried arc, and no buried arc. Fully buried arc means that the arc is fully covered by foam slag, partially buried arc means that the arc is partially exposed, and no buried arc means that the arc is clearly exposed.

[0075] Step S1043: Perform coordinate mapping on the image sequence of the foam region to obtain the spatial coordinates of the foam.

[0076] In this embodiment, by combining camera calibration parameters (camera intrinsic and extrinsic parameters) and online calibration results, the image pixel coordinates in the foam region image sequence are mapped to the actual foam space coordinates.

[0077] Understandably, establishing a quantitative mapping relationship between two-dimensional image pixels and three-dimensional physical space inside the furnace transforms visual measurements into engineering parameters that can participate in automatic control, providing a geometric benchmark for calculating the height of foamed slag and supporting subsequent height feedback control.

[0078] Step S1044: Calculate the spatial coordinates of the foam based on the spatial relationship between the surface boundary of the foam slag and the preset reference position to obtain the real-time height of the foam slag.

[0079] In this embodiment, based on the spatial relationship between the surface boundary of the foam slag and the preset reference position, the height component in the foam space coordinates corresponding to the surface boundary pixel of the foam slag is extracted, and the difference between the height component and the preset reference position is calculated to obtain the real-time height of the foam slag, thereby realizing non-contact, real-time, online measurement of the thickness of the foam slag layer.

[0080] It should be noted that in other embodiments, a checkerboard calibration plate can be used for offline calibration to obtain the camera intrinsic and extrinsic parameter matrices and establish a conversion model between pixel coordinates and furnace physical coordinates; the furnace wall refractory brick joints of known height in the image are used as a reference benchmark to perform online calibration of the conversion model; based on the pixel distance between the upper surface of the foam slag and the surface of the molten steel, the real-time height H of the foam slag is calculated in combination with the conversion model, and the calculation formula is: H=Δp×k+b, where Δp is the pixel distance, k is the pixel ratio coefficient, and b is the calibration offset.

[0081] Step S105: Determine the target splash risk judgment result based on the initial splash risk prediction result, the current electric arc furnace submerged arc state, and the real-time height of the foam slag, and perform electric arc furnace foam slag splash control and early warning based on the target splash risk judgment result.

[0082] In this embodiment, based on the initial splash risk prediction result, the current electric arc furnace submerged arc status, and the real-time height of foam slag, the target splash risk judgment result is generated and an early warning is issued according to the preset logic, thereby constructing a three-dimensional decision matrix of "risk-status-height", upgrading the early warning behavior from passive response to active working condition adaptation.

[0083] In one specific embodiment, step S105 includes: if the current electric arc furnace is fully submerged and the real-time height of the foam slag is within a preset target range, then the target splash risk judgment result is the initial splash risk judgment result; if the current electric arc furnace is partially submerged or not submerged, then the initial splash risk judgment result is enhanced to obtain the target splash risk judgment result; if the real-time height of the foam slag is not within the preset target range, then the initial splash risk judgment result is enhanced to obtain the target splash risk judgment result; if the current electric arc furnace is partially submerged or not submerged and the real-time height of the foam slag is not within the preset target range, then a high-level splash warning is used as the target splash risk judgment result.

[0084] In this embodiment, the ideal foam slag height range is determined based on the arc length, molten steel temperature, and smelting stage, and is used as a preset target range.

[0085] If the current electric arc furnace is fully submerged and the real-time height of the foam slag is within the preset target range, then the target splash risk judgment result is the initial splash risk judgment result, that is, the model's initial judgment is maintained under safe operating conditions.

[0086] If the current arc burial status of the electric arc furnace is partially or not burial, the initial splash risk assessment result is enhanced to obtain the target splash risk assessment result, that is, the early warning level is automatically enhanced under abnormal arc burial status.

[0087] If the real-time height of the foam residue is not within the preset target range, the initial splash risk assessment result is enhanced to obtain the target splash risk assessment result, i.e., automatic enhanced early warning when the foam height exceeds the limit.

[0088] If the current electric arc furnace is partially or not submerged, and the real-time height of the foamy slag is not within the preset target range, then the high-level splash warning will be used as the result of the target splash risk assessment. That is, when the double risks are superimposed, the highest level intervention command will be triggered, realizing the deep coupling of the warning strategy and the metallurgical process, and significantly improving the effectiveness of operation guidance and the efficiency of human-machine collaboration.

[0089] It should be noted that a comprehensive splashing risk score can be calculated based on the initial splashing risk prediction results, the current submerged arc status of the electric arc furnace, and the real-time height of the foamed slag. The target splashing risk assessment result can then be determined based on this comprehensive splashing risk score. The comprehensive splashing risk score can be expressed as follows: In the formula, P is the state risk component determined based on the initial splash risk prediction result, A is the state risk component determined based on the current submerged arc state of the electric arc furnace, and H is the height risk component determined based on whether the real-time height of the foamed slag is abnormal. , , The feature weight coefficient is used. Based on the comparison between the comprehensive splash risk score and the preset threshold, splash risk is divided into low risk, medium risk, and high risk.

[0090] It should be further noted that when the target splash risk assessment result reaches a preset threshold, a splash warning signal is output, and a preventive adjustment command is sent to the electric arc furnace control system. The adjustment command includes at least one of the following: adjusting the carbon injection quantity, adjusting the oxygen blowing quantity, adjusting the power supply intensity, and adjusting the electrode power.

[0091] When the probability of splash risk exceeds the preset threshold for multiple consecutive detection cycles, a splash warning signal is triggered, and the main oxygen lance flow rate and electrode power are reduced within a preset time to achieve preventive control.

[0092] This embodiment can also be configured with a data storage module and a model update module. The data storage module is used to store original images, recognition results, risk scores, control records, and event retrospective information. The model update module can periodically retrain or incrementally update the splash risk identification model based on actual working condition data to continuously improve the model's adaptability to specific furnace types and on-site working conditions.

[0093] The electric arc furnace foam slag splash control and early warning method proposed in this embodiment extracts dynamic visual features and thermal distribution change features of the foam slag area based on infrared images and high-temperature visible light images. It then uses these dynamic visual features and thermal distribution change features to identify splash precursors and obtain initial splash risk prediction results. The model prediction can ensure the sample identification capability in the splash precursor stage and improve the early warning robustness under complex operating conditions. At the same time, it achieves synchronous non-contact monitoring of the submerged arc status and the real-time height of the foam slag based on infrared images and high-temperature visible light images. Combined with the initial splash risk prediction results for joint judgment, it can further reduce the false alarm rate and the missed alarm rate, thereby providing timely early warning. This can improve smelting stability, reduce operating risks and smelting costs, and improve smelting efficiency and steel quality.

[0094] Example 2 Furthermore, this disclosure provides an electric arc furnace foam slag splash control and early warning system 500, please refer to [link to relevant documentation]. Figure 5 ,include: The acquisition module 501 is used to acquire infrared images and high-temperature visible light images of the smelting area inside the electric arc furnace; The acquisition module 502 is used to acquire dynamic visual features and heat distribution change features based on the fused image sequence; Prediction module 503 is used to input the dynamic visual features and the heat distribution change features into a preset splash risk identification model to obtain an initial splash risk prediction result; The determination module 504 is used to determine the current submerged arc state of the electric arc furnace and the real-time height of the foamed slag based on the fused image sequence; The early warning module 505 is used to determine the target splash risk judgment result based on the initial splash risk prediction result, the current electric arc furnace submerged arc state, and the real-time height of the foam slag, and to perform early warning for electric arc furnace foam slag splash control based on the target splash risk judgment result.

[0095] In an optional implementation, the system further includes: A training module is used to acquire training samples, including normal samples, splash precursor samples, and splash samples; input the training samples into an initial splash risk identification model to obtain risk prediction values; calculate precursor enhancement loss, risk increment constraint loss, and temperature-brightness motion consistency loss based on the training samples and the risk prediction values; calculate a target loss value based on the precursor enhancement loss, the risk increment constraint loss, and the temperature-brightness motion consistency loss; optimize the model parameter update direction of the initial splash risk identification model based on the target loss value until a pre-trained splash risk identification model is obtained.

[0096] In an optional implementation, the dynamic visual features include the intensity of foam movement, the thermal distribution change features include local brightness change and local temperature change, and the formula for calculating the target loss value is: In the formula, For the aforementioned precursor enhancement loss, For the aforementioned risk-increasing constraint loss, For the aforementioned temperature and brightness motion consistency loss, , All are loss weighting coefficients; The formula for calculating the precursor enhancement loss is as follows: In the formula, N is the number of training samples, and t is the time index of the sample. Let be the weight of the t-th sample. This represents the true label of the t-th sample. Let ln be the risk prediction value corresponding to the t-th sample, and ln be the natural logarithm function. The formula for calculating the loss under the increasing risk constraint is as follows: In the formula, M is the number of sample points within the precursor time window; The formula for calculating the temperature-brightness motion consistency loss is as follows: In the formula, This represents the splash risk probability output by the model at time time . The comprehensive precursor index at time t; The formula for calculating the comprehensive precursor index is as follows: In the formula, The normalized foam motion intensity, This represents the normalized rate of change of local average brightness. This represents the normalized local average rate of change of temperature. , , These are the weighting coefficients.

[0097] In an optional implementation, the precursor enhancement loss is used to increase the training weights of samples within a preset time window before the splash occurs. The risk-increasing constraint loss is used to penalize risk outputs that do not satisfy the increasing relationship within the time interval before the splash occurs. The temperature-brightness motion consistency loss is used to ensure that the risk prediction value output by the preset splash risk identification model remains consistent with the comprehensive precursor index composed of foam motion intensity, brightness change, and temperature change.

[0098] In an optional implementation, the acquisition module 502 is further configured to perform image preprocessing on the infrared image and the high-temperature visible light image to obtain a fused image sequence; segment the fused image sequence to obtain a foam region image sequence; calculate the pixel motion vector of the foam region image sequence based on an optical flow algorithm, determine the foam motion intensity based on the pixel motion vector, and use the foam motion intensity as the dynamic visual feature; acquire a first region of interest in the high-temperature visible light image corresponding to the region directly below the electrode or the high-incidence area of ​​sputtering in the fused image sequence, and calculate the local average brightness change rate of the first region of interest; acquire a second region of interest in the infrared thermal image corresponding to the high-temperature activity area directly below the electrode, and calculate the local average temperature change rate of the second region of interest; and use the average brightness change rate and the average temperature change rate as the heat distribution change feature.

[0099] In an optional implementation, the determining module 504 is further configured to: acquire image morphological features and temperature distribution features based on the fused image sequence; input the image morphological features and temperature distribution features into the submerged arc state recognition model to obtain the current submerged arc state of the electric arc furnace; perform coordinate mapping on the foam region image sequence to obtain foam space coordinates; and calculate the foam space coordinates based on the spatial relationship between the foam slag surface boundary and a preset reference position to obtain the real-time height of the foam slag.

[0100] In an optional implementation, the early warning module 505 is further configured to: if the current electric arc furnace is fully submerged and the real-time height of the foam slag is within a preset target range, then the target splash risk judgment result is the initial splash risk judgment result; if the current electric arc furnace is partially submerged or not submerged, then the initial splash risk judgment result is enhanced to obtain the target splash risk judgment result; if the real-time height of the foam slag is not within the preset target range, then the initial splash risk judgment result is enhanced to obtain the target splash risk judgment result; if the current electric arc furnace is partially submerged or not submerged and the real-time height of the foam slag is not within the preset target range, then a high-level splash warning is used as the target splash risk judgment result.

[0101] The system provided in this embodiment can execute the steps of the electric arc furnace foam slag splash control and early warning method provided in Embodiment 1. To avoid repetition, the steps will not be repeated.

[0102] The electric arc furnace foam slag splash control and early warning system proposed in this embodiment extracts dynamic visual features and thermal distribution change features of the foam slag area based on infrared and high-temperature visible light images. These features are then used to identify splash precursors and obtain initial splash risk prediction results. Model prediction ensures the ability to identify samples in the splash precursor stage, improving the robustness of early warning under complex operating conditions. Simultaneously, synchronous non-contact monitoring of the submerged arc state and real-time foam slag height is achieved based on infrared and high-temperature visible light images. Combined with the initial splash risk prediction results, joint judgment can further reduce false alarm and false alarm rates, enabling timely early warning. This improves smelting stability, reduces operational risks and smelting costs, and enhances smelting efficiency and steel quality.

[0103] Example 3 Furthermore, this disclosure provides a computer device including a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the electric arc furnace foam slag splash control and early warning method described in Embodiment 1.

[0104] The device provided in this embodiment can execute the steps of the electric arc furnace foam slag splash control and early warning method provided in Embodiment 1. To avoid repetition, it will not be described again.

[0105] Example 4 This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the electric arc furnace foam slag splash control and early warning method described in Embodiment 1.

[0106] In this embodiment, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0107] The computer-readable storage medium provided in this embodiment can implement the electric arc furnace foam slag splashing control and early warning method provided in Embodiment 1. To avoid repetition, it will not be described again here.

[0108] In all examples shown and described herein, any specific values ​​should be interpreted as merely exemplary and not as limitations; therefore, other examples of exemplary embodiments may have different values.

[0109] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0110] The above-described embodiments are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for controlling and warning of foamy slag splashing in an electric arc furnace, characterized in that, The method includes: Infrared and high-temperature visible light images of the smelting area inside the electric arc furnace were acquired. Dynamic visual features and thermal distribution change features are obtained from the infrared image and the high-temperature visible light image; The dynamic visual features and the thermal distribution change features are input into a preset splash risk identification model to obtain an initial splash risk prediction result. The current submerged arc status of the electric arc furnace and the real-time height of the foamy slag are determined based on the infrared image and the high-temperature visible light image. The target splash risk assessment result is determined based on the initial splash risk prediction result, the current electric arc furnace submerged arc status, and the real-time height of the foam slag. The electric arc furnace foam slag splash control and early warning are then performed based on the target splash risk assessment result.

2. The method for controlling and warning of foamy slag splashing in an electric arc furnace according to claim 1, characterized in that, The method further includes: Acquire training samples, which include normal samples, splash precursor samples, and splash samples; The training samples are input into the initial splash risk identification model to obtain the risk prediction value; Calculate the precursor enhancement loss, risk increment constraint loss, and temperature-brightness motion consistency loss based on the training samples and the risk prediction value; calculate the target loss value based on the precursor enhancement loss, the risk increment constraint loss, and the temperature-brightness motion consistency loss. The model parameter update direction of the initial splash risk identification model is optimized based on the target loss value until the trained preset splash risk identification model is obtained.

3. The method for controlling and warning of foamy slag splashing in an electric arc furnace according to claim 2, characterized in that, The dynamic visual features include the intensity of foam movement, the thermal distribution change features include the local average brightness change rate and the local average temperature change rate, and the formula for calculating the target loss value is: In the formula, For the aforementioned precursor enhancement loss, For the aforementioned risk-increasing constraint loss, For the aforementioned temperature and brightness motion consistency loss, , All are loss weighting coefficients; The formula for calculating the precursor enhancement loss is as follows: In the formula, N is the number of training samples, and t is the time index of the sample. Let be the weight of the t-th sample. This represents the true label of the t-th sample. Let ln be the risk prediction value corresponding to the t-th sample, and ln be the natural logarithm function. The formula for calculating the loss under the increasing risk constraint is as follows: In the formula, M is the number of sample points within the precursor time window; The formula for calculating the temperature-brightness motion consistency loss is as follows: In the formula, This represents the splash risk probability output by the model at time time . The comprehensive precursor index at time t; The formula for calculating the comprehensive precursor index is as follows: In the formula, The normalized foam motion intensity, This represents the normalized rate of change of local average brightness. This represents the normalized local average rate of change of temperature. , , These are the feature weight coefficients.

4. The method for controlling and warning of foamy slag splashing in an electric arc furnace according to claim 3, characterized in that, The precursor enhancement loss is used to increase the training weights of samples within a preset time window before the splash occurs. The risk-increasing constraint loss is used to penalize risk outputs that do not satisfy the increasing relationship within the time interval before the splash occurs. The temperature-brightness motion consistency loss is used to ensure that the risk prediction value output by the preset splash risk identification model remains consistent with the comprehensive precursor index composed of foam motion intensity, brightness change, and temperature change.

5. The method for controlling and warning of foamy slag splashing in an electric arc furnace according to claim 1, characterized in that, The step of obtaining dynamic visual features and thermal distribution change features based on the infrared image and the high-temperature visible light image includes: The infrared image and the high-temperature visible light image are preprocessed to obtain a fused image sequence; The fused image sequence is segmented to obtain a foam region image sequence; The pixel motion vector of the foam region image sequence is calculated based on the optical flow algorithm, the foam motion intensity is determined based on the pixel motion vector, and the foam motion intensity is used as the dynamic visual feature. In the fused image sequence, a first region of interest is obtained from the high-temperature visible light image that corresponds to the area directly below the electrode or the area with high sputtering incidence, and the local average brightness change rate of the first region of interest is calculated. In the fused image sequence, a second region of interest is obtained corresponding to the high-temperature activity area directly below the electrode in the infrared thermal image, and the local average temperature change rate of the second region of interest is calculated. The average brightness change rate and the average temperature change rate are used as the characteristics of the thermal distribution change.

6. The method for controlling and warning of foamy slag splashing in an electric arc furnace according to claim 5, characterized in that, The step of determining the current submerged arc state of the electric arc furnace and the real-time height of the foamy slag based on the infrared image and the high-temperature visible light image includes: Image morphological features and temperature distribution features are obtained from the fused image sequence; The image morphological features and the temperature distribution features are input into the submerged arc state recognition model to obtain the current submerged arc state of the electric arc furnace. The image sequence of the foam region is mapped to obtain the spatial coordinates of the foam. The real-time height of the foam slag is obtained by calculating the spatial coordinates of the foam based on the spatial relationship between the surface boundary of the foam slag and the preset reference position.

7. The method for controlling and warning of foamy slag splashing in an electric arc furnace according to claim 1, characterized in that, The determination of the target splash risk assessment result based on the initial splash risk prediction result, the current electric arc furnace submerged arc state, and the real-time height of the foamed slag includes: If the current electric arc furnace is fully submerged and the real-time height of the foam slag is within the preset target range, then the target splash risk judgment result is the initial splash risk judgment result. If the current electric arc furnace is partially submerged or not submerged, the initial splash risk assessment result is enhanced to obtain the target splash risk assessment result. If the real-time height of the foam residue is not within the preset target range, the initial splash risk assessment result is enhanced to obtain the target splash risk assessment result. If the current electric arc furnace is partially or not submerged, and the real-time height of the foam slag is not within the preset target range, then a high-level splash warning will be used as the target splash risk assessment result.

8. A foam slag splash control and early warning system for electric arc furnaces, characterized in that, The system includes: The acquisition module is used to acquire infrared images and high-temperature visible light images of the smelting area inside the electric arc furnace; The acquisition module is used to acquire dynamic visual features and thermal distribution change features based on the infrared image and the high-temperature visible light image; The prediction module is used to input the dynamic visual features and the heat distribution change features into a preset splash risk identification model to obtain an initial splash risk prediction result; The determination module is used to determine the current submerged arc state of the electric arc furnace and the real-time height of the foamed slag based on the infrared image and the high-temperature visible light image; The early warning module is used to determine the target splash risk judgment result based on the initial splash risk prediction result, the current electric arc furnace submerged arc state, and the real-time height of the foam slag, and to perform early warning for electric arc furnace foam slag splash control based on the target splash risk judgment result.

9. A computer device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the electric arc furnace foam slag splash control and early warning method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the electric arc furnace foam slag splash control and early warning method as described in any one of claims 1 to 7.

Images