A method and device for dynamic control of a slag splashing furnace protection, an electronic device and a storage medium

Abstract

This application provides a method, device, electronic device, and storage medium for dynamic control of slag splashing protection in a converter. The method acquires a target slag splashing image at the converter's opening and the current slag splashing mode of the converter. The target slag splashing image is input into a preset slag splashing level recognition model to obtain a target slag splashing level characterizing the intensity of the current slag splashing. The slag splashing level recognition model is trained based on image samples labeled with slag splashing levels. Based on a slag splashing lance position parameter configuration table corresponding to the target slag splashing level and the current slag splashing mode, a lance position control command is generated. The oxygen lance is then controlled to perform the corresponding lance position action according to the lance position control command to complete the dynamic control of slag splashing protection in the converter. This method provides a more practical and accurate dynamic control method for slag splashing protection in a converter.

Description

Technical Field

[0001] This application relates to the field of metallurgical technology, and in particular to a dynamic control method, device, electronic equipment and storage medium for slag splashing furnace protection. Background Technology

[0002] As a widely used furnace lining maintenance technology, the basic principle of converter slag splashing protection is to use high-pressure nitrogen gas to forcefully splash and coat the molten slag, which is still in a high-temperature molten state after converter smelting, onto the inner wall of the furnace lining. After cooling, a dense solid protective slag layer is formed. This protective slag layer can effectively slow down the erosion of the furnace lining refractory material by high-temperature molten slag and steel through physical isolation and buffering, thereby significantly extending the furnace life and greatly reducing the consumption cost of refractory materials. This has important practical significance for cost reduction and efficiency improvement in metallurgical enterprises.

[0003] Currently, relevant converter slag splashing control technologies mainly fall into two categories: manual slag splashing methods and static slag splashing methods. Manual slag splashing requires operators to observe the flame characteristics at the furnace opening and the slag condition within the furnace throughout the process, and to manually adjust the lance position in real time based on subjective experience. This process is highly dependent on the operator's skill and on-site judgment, is highly subjective, exhibits significant individual differences, makes it difficult to guarantee the stability of slag splashing quality, and is labor-intensive, making standardized control difficult. To overcome the limitations of manual operation, static slag splashing methods have emerged. These methods typically pre-determine operating parameter tables for lance position changes over time based on process parameters such as steel composition, final temperature, and furnace type, combined with manual slag splashing experience. During the slag splashing process, the system executes the set lance positions in a fixed sequence, thereby achieving a certain degree of automation. However, this static method is essentially an open-loop control, unable to provide real-time sensing and dynamic feedback on the slag layer distribution and slag splashing effect during the slag splashing process. Therefore, when furnace conditions fluctuate or anomalies occur, it is difficult to adaptively adjust the lance position and purging parameters, resulting in poor slag splashing uniformity and furnace protection effect, failing to fully meet the refined control requirements of the slag splashing furnace protection process under complex smelting conditions. Clearly, a new dynamic control method for slag splashing furnace protection is urgently needed to solve at least one of the above problems.

[0004] It should be noted that the above content only provides background information related to this application and does not necessarily constitute prior art. Summary of the Invention

[0005] This application provides a dynamic control method, device, electronic equipment, and storage medium for slag splashing furnace protection, in order to solve the technical bottleneck of existing slag splashing control modes in terms of real-time status perception and dynamic adjustment, and the technical problem that they cannot make intelligent decisions and adaptive controls based on the actual slag splashing conditions in the furnace.

[0006] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0007] This application provides a dynamic control method for slag splashing protection of a converter, comprising: acquiring a target slag splashing image at the furnace opening of a target converter and acquiring the current slag splashing mode of the target converter; inputting the target slag splashing image into a preset slag splashing level recognition model to obtain a target slag splashing level for characterizing the intensity of the current slag splashing, wherein the slag splashing level recognition model is trained based on image samples with slag splashing level labels; generating a lance position control command according to a slag splashing lance position parameter configuration table corresponding to the target slag splashing level and the current slag splashing mode, and controlling the oxygen lance to perform the corresponding lance position action according to the lance position control command to complete the dynamic control of slag splashing protection of the converter.

[0008] In one embodiment of this application, the method for obtaining the slag splashing level recognition model based on the foregoing scheme includes: acquiring a slag splashing level label and a sample slag splashing image, wherein the slag splashing level label is determined according to the density of slag particle splashing; labeling the sample slag splashing image according to the slag splashing level label to generate an image sample with a slag splashing level label; and training a preset slag splashing level recognition model based on the image sample with the slag splashing level label to obtain the slag splashing level recognition model.

[0009] In one embodiment of this application, based on the aforementioned scheme, after inputting the target slag splash image into a preset slag splash level recognition model, the method further includes: performing flame tracking on the target slag splash image to locate the core region of the slag splash, and adjusting the region of interest according to the location result; selecting a preset number of consecutive frames from multiple frames of the target slag splash images, and determining the analysis region of each frame image according to the region of interest; inputting the analysis region of each frame image into the feature extraction network of the slag splash level recognition model to extract the high-dimensional feature sequence of each frame image; integrating the high-dimensional feature sequence of the preset number of frames of images through the temporal feature fusion module of the slag splash level recognition model to generate a comprehensive feature representation that integrates the spatiotemporal context; and sequentially passing the comprehensive feature representation through the global average pooling layer for dimensionality reduction, the random deactivation layer for overfitting suppression, and the fully connected layer for mapping in the slag splash level recognition model to obtain the original slag splash level.

[0010] In one embodiment of this application, after obtaining the original slag splash level based on the aforementioned scheme, the method further includes: performing statistical analysis on the original slag splash level within a preset time window to obtain the current slag splash level in the current time window; calculating the level change range based on the current slag splash level and the previous slag splash level in the previous time window; if the level change range is greater than a preset change range, replacing the current slag splash level with a preset transition value to obtain a constrained slag splash level after jump constraint; and normalizing and outputting the constrained slag splash level according to a preset frequency to obtain the target slag splash level after rectification optimization.

[0011] In one embodiment of this application, the method for obtaining the slag-splashing lance position parameter configuration table based on the aforementioned scheme includes: determining the slag-splashing furnace protection position and slag-splashing target according to the current slag-splashing mode; dividing the slag-splashing process into a static slag adjustment stage, a dynamic slag-splashing stage, and an endpoint determination stage according to a preset stage division rule; determining the lance position height and lance position holding time of the static slag adjustment stage, the lance lifting rules and lance lowering rules of the dynamic slag-splashing stage, and the endpoint standard level and endpoint lance position of the endpoint determination stage, based on the slag-splashing furnace protection position and the slag-splashing target, respectively, thereby obtaining the slag-splashing lance position parameter configuration table, wherein the lance lifting rules include the lance lifting standard level and the corresponding lance lifting height and lance lifting holding time, and the lance lowering rules include the lance lowering standard level and the corresponding lance lowering height and lance lowering holding time.

[0012] In one embodiment of this application, based on the aforementioned scheme, the slag-splashing lance position parameter configuration table includes the lance position height and lance position holding time during the static slag adjustment stage, the lance lifting standard level and corresponding lifting height and lifting holding time during the dynamic slag-splashing stage, the lance lowering standard level and corresponding lowering height and lowering holding time, and the endpoint standard level and endpoint lance position during the endpoint determination stage; according to the target slag-splashing level and the slag-splashing lance position parameter configuration table corresponding to the current slag-splashing mode, a lance position control command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the lance position control command, including: generating a lance position holding command according to the lance position height and the lance position holding time, and controlling the oxygen lance to maintain a fixed lance position during the static slag adjustment stage according to the lance position holding command until the static slag adjustment stage ends; the static After the slag adjustment stage, the dynamic slag splashing stage begins. Based on the comparison between the target slag splashing level and the standard levels for raising and lowering the lance, a lance position action command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the command. After the lance position action command is executed, the target slag splashing level is compared again with the standard levels for raising and lowering the lance. During the dynamic slag splashing stage, the target slag splashing level and the current actual lance position are monitored in real time. If the target slag splashing level is less than or equal to the endpoint standard level, and the current actual lance position is less than or equal to the endpoint lance position, a termination command is generated to end the dynamic control of slag splashing for furnace protection. The lance position control command includes the lance position holding command, the lance position action command, and the termination command.

[0013] In one embodiment of this application, based on the aforementioned scheme, a gun position action command is generated according to the comparison result between the target slag splash level and the standard level for raising and lowering the lance, and the oxygen lance is controlled to perform the corresponding gun position action according to the gun position action command, including: if the target slag splash level is greater than the standard level for raising the lance, a raising action command is generated, and the oxygen lance is controlled to be raised to the corresponding raising height and held for the raising height for the specified duration according to the raising action command; if the target slag splash level is less than or equal to the standard level for lowering the lance, a lowering action command is generated, and the oxygen lance is controlled to be lowered to the corresponding lowering height and held for the lowering height for the specified duration according to the lowering action command.

[0014] This application also provides a dynamic control device for slag splashing protection, comprising: a data acquisition module for acquiring a target slag splashing image at the target converter opening and acquiring the current slag splashing mode of the target converter; a level recognition module for inputting the target slag splashing image into a preset slag splashing level recognition model to obtain a target slag splashing level characterizing the intensity of the current slag splashing, wherein the slag splashing level recognition model is trained based on image samples with slag splashing level labels; and a furnace protection control module for generating lance position control commands according to a slag splashing lance position parameter configuration table corresponding to the target slag splashing level and the current slag splashing mode, and controlling the oxygen lance to perform corresponding lance position actions according to the lance position control commands to complete the dynamic control of slag splashing protection.

[0015] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the dynamic control method for slag splashing and furnace protection as described in any of the above embodiments.

[0016] This application also provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the dynamic control method for slag splashing furnace protection as described in any of the above embodiments.

[0017] The beneficial effects of this application are as follows: By acquiring the target slag splashing image of the target converter mouth and the current slag splashing mode of the target converter, and using a preset slag splashing level recognition model to infer and recognize the target slag splashing image, a quantitative target slag splashing level is quickly output, thereby transforming the traditional subjective judgment that relies on manual visual inspection into an objective and standardized real-time perception signal. On this basis, lance position control commands are automatically generated according to the slag splashing level and the slag splashing mode corresponding to the lance position parameter configuration table, and the oxygen lance is driven to perform the corresponding lance position action, thus constructing a complete closed-loop control link without manual intervention. By solidifying the experience of excellent operators into a reusable algorithm model and parameter configuration table, the digital accumulation and standardized inheritance of core process knowledge are realized. At the same time, it can effectively avoid quality fluctuations caused by human fatigue and differences in operating level, significantly improving the stability, uniformity and consistency of the slag splashing protection effect, and truly realizing the intelligent upgrade of converter slag splashing protection from experience-dependent to data-driven.

[0018] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0020] In the attached diagram: Figure 1 This is a schematic diagram illustrating an exemplary system architecture as shown in an exemplary embodiment of this application; Figure 2 This is a schematic flowchart illustrating a dynamic control method for slag splashing furnace protection, as shown in an exemplary embodiment of this application. Figure 3 This is a schematic diagram of different levels of splatter images illustrating an exemplary embodiment of this application; Figure 4 This is a block diagram illustrating a dynamic control device for slag splashing furnace protection, as shown in an exemplary embodiment of this application; Figure 5 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation

[0021] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0022] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. The drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the shape, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0023] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the present application. However, it will be apparent to those skilled in the art that embodiments of the present application may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the present application.

[0024] First, it should be noted that slag splashing for furnace protection refers to a smelting technique that utilizes the magnesium oxide-containing terminal slag from the converter after tapping to form a high-melting-point protective layer through high-pressure nitrogen splashing, thereby extending the life of the furnace lining. This protective layer effectively reduces refractory material loss by inhibiting oxidation and decarburization on the surface of the furnace lining bricks and mitigating the erosive effect of high-temperature molten slag.

[0025] Currently, relevant converter slag splashing control technologies are mainly divided into two categories: manual slag splashing methods and static slag splashing methods. Manual slag splashing requires operators to monitor the furnace flame and slag splashing conditions in real time throughout the process, and manually adjust the spray gun position based on personal experience. The slag splashing quality is highly dependent on the operator's skill level and on-site judgment, making it subjective and difficult to guarantee the stability and consistency of slag splashing effects between different furnace runs. Furthermore, while static slag splashing methods can automatically control the gun position sequentially according to a preset static parameter table, achieving basic automation, it is essentially an open-loop control mode. It cannot perceive the slag layer distribution and slag splashing effect in the furnace in real time during the slag splashing process, nor can it dynamically adjust the spray gun parameters according to changes in actual operating conditions. Therefore, it is difficult to guarantee the uniformity of slag coverage and the furnace protection effect when facing furnace condition fluctuations or abnormal situations. At the same time, the slag splashing experience accumulated by skilled operators in existing methods is difficult to quantify into standardized control rules, which is not conducive to the technical accumulation and widespread application of operational experience.

[0026] To address these issues, embodiments of the present invention provide a dynamic control method for slag splashing furnace protection, a dynamic control device for slag splashing furnace protection, an electronic device, and a computer-readable storage medium. These embodiments will be described in detail below.

[0027] Figure 1 This is a schematic diagram illustrating an exemplary system architecture as shown in an exemplary embodiment of this application.

[0028] Reference Figure 1As shown, the system architecture may include a data acquisition device 101 and a computer device 102. The computer device 102 may be at least one of a desktop graphics processing unit (GPU) computer, a GPU computing cluster, or a neural network computer. The data acquisition device 101 is used to acquire target slag splashing images at the target converter mouth and to acquire the current slag splashing mode of the target converter. In this embodiment, after acquiring the above data, the data acquisition device 101 provides it to the computer device 102 for processing. Those skilled in the art can use the computer device 102 to input the target slag splashing image into a preset slag splashing level recognition model to obtain a target slag splashing level characterizing the current intensity of slag splashing. The slag splashing level recognition model is trained based on image samples with slag splashing level labels. Based on the slag splashing level and the slag splashing lance position parameter configuration table corresponding to the current slag splashing mode, it generates lance position control commands and controls the oxygen lance to perform corresponding lance position actions according to the lance position control commands to complete the dynamic control of slag splashing protection. It should be noted that the data acquisition device 101 and computer device 102 provided in this embodiment are merely examples and should not impose any limitations on the functions and scope of use of the embodiments of this application.

[0029] It should be noted that the dynamic control method for slag splashing and furnace protection provided in this application embodiment is generally executed by computer equipment 102, and correspondingly, the dynamic control device for slag splashing and furnace protection is generally installed in computer equipment 102.

[0030] Figure 2 This is a schematic flowchart illustrating an exemplary embodiment of the dynamic control method for slag splashing furnace protection, which can be executed by a computing processing device. Figure 1 The computer device 102 shown is illustrated. (Refer to...) Figure 2 As shown, the dynamic control method for slag splashing protection includes at least steps S210 to S240, which are described in detail below: In step S210, a target slag splashing image of the target converter opening is obtained, and the current slag splashing mode of the target converter is obtained.

[0031] In this embodiment, a video stream of the slag splashing process is acquired in real time by image acquisition devices such as high-temperature resistant vision sensors deployed at the target converter mouth. The target slag splashing image at the target converter mouth is then obtained from the video stream. Alternatively, the target slag splashing image at the target converter mouth can be directly acquired using image acquisition devices. It is understood that the target slag splashing image consists of multiple frames of the current slag splashing image at the target converter mouth. Simultaneously, the operator selects the target slag splashing mode through a human-machine interface. The system obtains the current slag splashing mode of the target converter. Slag splashing modes include various modes such as bottom splashing mode and body splashing mode, with different slag splashing modes corresponding to different slag splashing gun position parameter configuration tables. The operator can select the corresponding slag splashing mode based on the current furnace lining erosion status (e.g., the furnace bottom or body requires priority maintenance). The system loads the parameter configuration table corresponding to that mode as the basis for subsequent gun position control.

[0032] In step S220, the target slag splashing image is input into a preset slag splashing level recognition model to obtain the target slag splashing level used to characterize the intensity of the current slag splashing.

[0033] The slag splashing level recognition model is trained based on image samples with slag splashing level labels.

[0034] In one embodiment of this application, the method for obtaining the slag splashing level recognition model includes: acquiring a slag splashing level label and a sample slag splashing image, wherein the slag splashing level label is determined according to the density of slag particle splashing; labeling the sample slag splashing image according to the slag splashing level label to generate an image sample with a slag splashing level label; and training a preset slag splashing level recognition model based on the image sample with the slag splashing level label to obtain the slag splashing level recognition model.

[0035] In this embodiment, firstly, a slag splashing level classification standard is established. Based on the density and intensity of slag splashing, slag splashing images are divided into multiple levels; the higher the level, the denser the slag particles in the scene and the more intense the slag splashing; the lower the level, the sparser the slag particles in the scene and the less intense the slag splashing.

[0036] To improve the accuracy, stability, and adaptability of the model, a dataset was created using slag splash images from different furnace opening brightness areas and different ROI (Region of Interest) locations, considering various lighting conditions, furnace opening positions, and height variations. The constructed dataset includes standard legends for lance raising, lance lowering, and endpoint levels.

[0037] Then, based on the above-mentioned slag splashing level classification standard, the collected slag splashing images are labeled with their respective levels to generate image samples with slag splashing level labels.

[0038] A pre-defined slag level recognition model is trained based on image samples labeled with slag level. In this embodiment, ResNet50 (a classic deep convolutional neural network model, mainly used for image classification tasks) is selected as the core feature extraction network. Since the features of industrial slag images differ significantly from those of natural images, this embodiment does not use pre-trained weights on a general dataset, but instead adopts a strategy of training from scratch, allowing the model (i.e., the slag level recognition model) to focus more on learning feature representations specific to the slag scene domain. End-to-end training is performed on the labeled slag level dataset using the cross-entropy loss function and the Adam (Adaptive Moment Estimation) optimizer. The dataset is divided into training, validation, and test sets in chronological order to ensure the temporal validity of the evaluation. After the above training converges, the slag level recognition model is obtained.

[0039] In one embodiment of this application, after inputting the target slag splash image into a preset slag splash level recognition model, the method further includes: performing flame tracking on the target slag splash image to locate the core region of the slag splash, and adjusting the region of interest based on the location result; selecting a preset number of consecutive frames from multiple frames of target slag splash images, and determining the analysis region of each frame image based on the region of interest; inputting the analysis region of each frame image into the feature extraction network of the slag splash level recognition model to extract the high-dimensional feature sequence of each frame image; integrating the high-dimensional feature sequences of the preset number of frames images through the temporal feature fusion module of the slag splash level recognition model to generate a comprehensive feature representation that integrates the spatiotemporal context; and sequentially passing the comprehensive feature representation through the global average pooling layer for dimensionality reduction, the random deactivation layer for overfitting suppression, and the fully connected layer for mapping in the slag splash level recognition model to obtain the original slag splash level.

[0040] In one embodiment of this application, after obtaining the original slag splashing level, the method further includes: performing statistical analysis on the original slag splashing level within a preset time window to obtain the current slag splashing level of the current time window; calculating the level change range based on the current slag splashing level and the previous slag splashing level of the previous time window; if the level change range is greater than the preset change range, replacing the current slag splashing level with a preset transition value to obtain the constrained slag splashing level after jump constraint; and normalizing the constrained slag splashing level according to a preset frequency to obtain the target slag splashing level after rectification optimization.

[0041] In this embodiment, during the actual slag splashing control process, after the target slag splashing image is input into the pre-trained slag splashing level recognition model, the model performs online inference according to the following steps.

[0042] First, because the slag splashing level identification model integrates dynamic flame tracking function, it performs flame tracking on the target slag splashing image based on real-time image processing algorithm, locates the core area of ​​slag splashing activity, and automatically adjusts the position and range of the region of interest (ROI) according to the location result, ensuring that subsequent analysis always focuses on the most representative slag splashing state, effectively overcoming the interference caused by furnace mouth flame jumping, external obstruction and camera vibration, etc., laying a stable and reliable perception foundation for subsequent feature extraction.

[0043] After locating the region of interest through dynamic flame tracking, the slag splashing level identification model selects a predetermined number of consecutive frames from the target slag splashing image and determines the analysis region of each frame based on the region of interest. The analysis regions of each frame are then normalized in size and input into the feature extraction network of the slag splashing level identification model.

[0044] The ResNet50 network was selected as the core feature extraction network. Since the features of industrial slag splash images differ significantly from those of natural images, this embodiment does not use pre-trained weights on general datasets such as ImageNet (a large image database), but instead adopts a strategy of training from scratch, allowing the model to focus more on learning feature representations specific to the slag splash scene. The analysis regions of each frame image are input into this feature extraction network to extract the high-dimensional feature sequence of each frame.

[0045] To enhance the model's ability to perceive dynamic processes such as slag splashing and morphological changes, a multi-frame temporal detection mechanism is introduced. After extracting multiple frames of high-dimensional feature sequences, the temporal feature fusion module of the slag splashing level recognition model integrates the high-dimensional feature sequences of a preset number of frames to capture the temporal evolution of the slag splashing state and generate a comprehensive feature representation that integrates spatiotemporal context. This comprehensive feature representation, which incorporates spatiotemporal context information from multiple frames, can more accurately determine the current dynamic state of the slag splashing compared to single-frame image classification.

[0046] The combined feature representation after temporal fusion is input into the classifier at the end of the model. This classifier is specifically optimized: first, it is reduced in dimensionality by a global average pooling layer, then a Dropout layer (i.e., a random deactivation layer) is added to suppress overfitting, and finally, it is mapped to the original logistic values ​​of multiple slag levels (i.e., the original slag levels) through a fully connected layer. These original logistic values ​​are the instantaneous level determination results directly output by the model without post-processing.

[0047] To address the issue of drastic output level fluctuations caused by transient occlusion, image noise, or momentary model misjudgment, the model also integrates an intelligent rectification module for recognition results. This module post-processes the original logic values, ensuring the smoothness, stability, and rationality of the output signal. Specifically, it includes the following three mechanisms: Time Window Statistics and Filtering: The module (i.e., the intelligent rectification module for recognition results) maintains a configurable time window and performs statistical analysis on the original slag splash level within the time window. For example, it uses moving average or median filtering to smooth instantaneous noise and obtain the current slag splash level for the current time window. The original slag splash level is the instantaneous value directly output by the model at a certain moment, while the current slag splash level is the smoothed value output at that moment after time window filtering. Essentially, they are the state values ​​of the same parameter at different processing stages.

[0048] Output jump constraint: The module calculates the level change range based on the current slag splash level and the previous slag splash level in the previous time window. The user can preset the maximum allowed level jump value (i.e., preset change range). When the level change range exceeds the preset change range, the module activates the constraint logic, replacing the current slag splash level with a smooth transition value, making the change trend of the output level more consistent with the continuous and gradual characteristics of the slag splash physical process, thus obtaining the constrained slag splash level after jump constraint.

[0049] Adjustable output frequency: The module normalizes the constrained slag splash level after the jump constraint according to a preset frequency, which reduces the receiving pressure on the downstream gun position control system and further smooths the final signal, ultimately obtaining the rectified and optimized target slag splash level signal. The original slag splash level, current slag splash level, constrained slag splash level, and target slag splash level are essentially the same parameter (slag splash level) at different processing stages, representing the step-by-step optimization process from instantaneous model output to time window smoothing, then to jump constraint, and finally frequency normalized output.

[0050] The target slag splash level signal is stable, smooth, and reliable, conforming to the gradual change law of the slag splash physical process, which greatly enhances the reliability of the output signal. It can be directly used in the downstream gun position closed-loop control logic as the basis for generating gun position action commands.

[0051] Furthermore, to meet the stringent real-time requirements of industrial environments, the slag splash level recognition model, after training convergence and before deployment to edge devices, employs INT8 (an 8-bit signed integer data type) quantization technology to perform post-training quantization compression on the model weights and activation values, converting the data precision from 32-bit floating-point numbers to 8-bit integers. This significantly reduces computational load and memory consumption while maintaining basic recognition accuracy, enabling the model to meet the real-time processing needs of industrial environments.

[0052] Understandably, the slag splash level identification model is not a simple static image classifier, but a complete system that integrates dynamic flame tracking, multi-frame time-series analysis, core network optimization, and intelligent rectification of identification results, which can meet the core requirements of accuracy, real-time performance, and robustness in industrial settings.

[0053] In step S230, a lance position control command is generated based on the lance position parameter configuration table corresponding to the target slag splashing level and the current slag splashing mode. The oxygen lance is then controlled to perform the corresponding lance position action according to the lance position control command to complete the dynamic control of slag splashing for furnace protection.

[0054] In one embodiment of this application, the method for obtaining the slag-splashing lance position parameter configuration table includes: determining the slag-splashing furnace protection position and slag-splashing target according to the current slag-splashing mode; dividing the slag-splashing process into a static slag adjustment stage, a dynamic slag-splashing stage, and an endpoint determination stage according to a preset stage division rule; determining the lance position height and lance position holding time for the static slag adjustment stage, the lance lifting rules and lance lowering rules for the dynamic slag-splashing stage, and the endpoint standard level and endpoint lance position for the endpoint determination stage, based on the slag-splashing furnace protection position and slag-splashing target, thereby obtaining the slag-splashing lance position parameter configuration table, wherein the lance lifting rules include the lance lifting standard level and the corresponding lance lifting height and lance lifting holding time, and the lance lowering rules include the lance lowering standard level and the corresponding lance lowering height and lance lowering holding time.

[0055] In some embodiments, during actual production operations, after the converter tapping is completed, the current erosion status of the furnace lining and the lining locations requiring priority maintenance for this slag splashing are determined based on the furnace lining erosion detection results or the current converter image. Slag splashing locations refer to different areas of the converter furnace lining, including the furnace bottom area and the furnace body area. A corresponding slag splashing mode is selected based on the different slag splashing locations; for example, the furnace bottom splashing mode is selected when the furnace bottom erosion is severe and requires repair; the furnace body splashing mode is selected when the furnace body erosion is severe and requires repair. The slag splashing location and target (i.e., the desired furnace protection effect, such as restoring the furnace bottom thickness to the target value or achieving the required slag layer coverage on the furnace body) are determined based on the selected current slag splashing mode.

[0056] Based on the inherent technological principles of slag splashing for furnace protection, the system divides the slag splashing process into three sequentially executed control stages according to preset stage division rules: static slag adjustment stage, dynamic slag splashing stage, and endpoint determination stage. Specifically, in the static slag adjustment stage, an open-loop slag adjustment operation is performed according to a preset fixed lance position and fixed duration, the purpose of which is to adjust the physical state of the slag to improve the subsequent dynamic slag splashing effect; in the dynamic slag splashing stage, the lance position is adjusted in real time based on visual feedback information to achieve adaptive slag splashing coating; in the endpoint determination stage, the system automatically determines whether the slag splashing endpoint has been reached according to preset conditions.

[0057] Based on the location and target of slag splashing for furnace protection, the system determines the specific parameters for each stage, forming a complete slag splashing lance position parameter configuration table. Specifically, this includes: Lance position safety boundaries: Configure the highest and lowest lance position parameters as the upper and lower safety limits for oxygen lance operation, ensuring that the oxygen lance will not exceed these limits during dynamic lance adjustment. The safety boundaries differ for different slag splashing modes and can be set according to actual needs.

[0058] Static slag adjustment stage parameters: Configure the gun position height and gun position holding time during the static slag adjustment stage. This stage is performed with a fixed gun position, and its purpose is to adjust the slag and improve the slag splashing effect in the subsequent dynamic stage.

[0059] Dynamic slag splash stage parameters: Configure the raising and lowering rules for the dynamic slag splash stage. Raising rules include the standard raising level, corresponding raising height, and raising hold duration; lowering rules include the standard lowering level, corresponding lowering height, and lowering hold duration. Both the raising and lowering standard levels are threshold values ​​used to compare with the real-time identified target slag splash level to determine whether the corresponding gun position action needs to be performed.

[0060] Endpoint determination stage parameters: Configure the endpoint standard level and endpoint gun position for the endpoint determination stage, which are used to determine the slag splash endpoint.

[0061] The slag splashing lance position parameter configuration table supports setting multiple tables with different parameter configurations to switch between different slag splashing modes, including at least the furnace bottom splashing mode and the furnace body splashing mode. Different slag splashing modes correspond to different furnace protection positions and slag splashing targets, and their various parameters (including lance position safety boundaries, static slag adjustment stage parameters, dynamic slag splashing stage lance raising / lowering rules, endpoint determination parameters, etc.) can be configured independently according to process requirements. Users can quickly load the corresponding parameter configuration table by selecting different slag splashing modes, achieving flexible switching between different process requirements.

[0062] Once the above parameters are configured, a complete slag splashing lance position parameter configuration table is obtained, which can be directly used to control the oxygen lance position action.

[0063] In one embodiment of this application, the slag splashing lance position parameter configuration table includes the lance position height and lance position holding time during the static slag adjustment stage, the lance lifting standard level and corresponding lifting height and lifting holding time during the dynamic slag splashing stage, the lance lowering standard level and corresponding lowering height and lowering holding time, and the endpoint standard level and endpoint lance position during the endpoint determination stage. Based on the slag splashing lance position parameter configuration table corresponding to the target slag splashing level and the current slag splashing mode, a lance position control command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the lance position control command, including: generating a lance position holding command based on the lance position height and lance position holding time, and controlling the oxygen lance to maintain a fixed lance position during the static slag adjustment stage based on the lance position holding command until static slag adjustment. The static slag adjustment stage ends; after the static slag adjustment stage ends, the dynamic slag splashing stage begins. Based on the comparison between the target slag splashing level and the standard levels for raising and lowering the lance, a lance position action command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the lance position action command. After the lance position action command is executed, the target slag splashing level is compared with the standard levels for raising and lowering the lance again. In the dynamic slag splashing stage, the target slag splashing level and the current actual lance position are monitored in real time. If the target slag splashing level is less than or equal to the endpoint standard level, and the current actual lance position is less than or equal to the endpoint lance position, a termination command is generated to end the dynamic control of slag splashing and furnace protection. The lance position control command includes a lance position holding command, a lance position action command, and a termination command.

[0064] In one embodiment of this application, a lance position action command is generated based on the comparison result between the target slag splash level and the standard lance raising and lowering levels. The lance is then controlled to perform the corresponding lance position action according to the lance position action command. This includes: if the target slag splash level is greater than the standard lance raising level, a lance raising action command is generated, and the lance is raised to the corresponding lance raising height and held for the specified duration according to the lance raising action command; if the target slag splash level is less than or equal to the standard lance lowering level, a lance lowering action command is generated, and the lance is lowered to the corresponding lance lowering height and held for the specified duration according to the lance lowering action command.

[0065] In this embodiment, after the slag splashing program is started, it first enters the static slag conditioning stage. Based on the lance height and lance holding duration in the slag splashing lance position parameter configuration table, a lance holding command is generated. Based on this command, the oxygen lance is controlled to maintain a fixed lance position during the static slag conditioning stage until the stage ends. The slag splashing effect in the subsequent dynamic slag splashing stage is improved by adjusting the physical state of the slag (such as viscosity and fluidity).

[0066] After the static slag adjustment stage, the dynamic slag splashing stage automatically begins. In this stage, based on the comparison between the target slag splashing level and the standard levels for raising and lowering the lance, a lance position action command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the command. Specifically, the standard levels for raising and lowering the lance are obtained from the slag splashing lance position parameter configuration table, and the target slag splashing level is compared with both the standard levels for raising and lowering the lance. Lifting the lance: If the target slag splash level is greater than the standard lifting level, it means that the current slag splash is too violent. A lifting action command is generated, and the oxygen lance is controlled to be raised to the corresponding lifting height and held for the specified duration according to the lifting action command.

[0067] Lowering the lance: If the target slag splash level is less than or equal to the lowering standard level, it means that the current slag splash level is insufficient. A lowering action command is generated, and the oxygen lance is controlled to be lowered to the corresponding lowering height and maintained for the specified duration according to the lowering action command.

[0068] After each raising or lowering of the oxygen lance, the target slag splash level is compared with the standard raising and lowering levels to determine whether the operation needs to be performed again. This process is repeated to form a closed-loop feedback adjustment mechanism of sensing, comparison, execution, and re-sensing, enabling the oxygen lance position to dynamically and adaptively adjust according to the actual slag splash state at any given moment.

[0069] Understandably, in the slag splashing lance position parameter configuration table, the lance raising standard level and the lance lowering standard level can be configured to the same or different values ​​according to process requirements. When the lance raising standard level and the lance lowering standard level are set to the same value, the dynamic slag splashing stage uses this single value as a comparison benchmark: when the target slag splashing level is greater than the benchmark value, the lance raising operation is performed; when the target slag splashing level is less than or equal to the benchmark value, the lance lowering operation is performed. When the lance raising standard level and the lance lowering standard level are set to different values, the typical configuration is that the lance raising standard level is higher than the lance lowering standard level. In this case, a transition zone is formed between the lance raising standard level and the lance lowering standard level. The existence of this transition zone can avoid frequent raising and lowering of the oxygen lance due to fluctuations in the target slag splashing level near the threshold, which is beneficial to improving the stability of lance position control and reducing the wear of the actuator. When the target slag splash level is between the standard lance lowering level and the standard lance raising level (i.e., the target slag splash level is greater than the standard lance lowering level and less than or equal to the standard lance raising level), it indicates that the current slag splash intensity is moderate. No lance raising or lowering command is generated; the current lance position is maintained, and slag splashing continues. The assessment is repeated in the next comparison cycle. This control logic allows the oxygen lance position to adaptively adjust according to real-time changes in the slag splash state, while maintaining lance position stability when the slag splash intensity is moderate, effectively balancing control response speed and stability.

[0070] It should be noted that during the dynamic slag splashing stage, the oxygen lance position is always constrained by the safe range of the highest and lowest lance positions in the slag splashing lance position parameter configuration table during the execution of the above-mentioned lance raising and lowering operations, ensuring that the safe boundary is not exceeded during dynamic lance adjustment.

[0071] During the dynamic slag splashing phase, the target slag splashing level and the current actual gun position are detected in real time to determine the endpoint. The endpoint determination is performed synchronously with the closed-loop adjustment steps described above, meaning that the endpoint conditions are checked simultaneously in each comparison cycle.

[0072] The specific conditions for determining the endpoint are as follows: when the target slag splashing level is less than or equal to the endpoint standard level, and the current actual lance position is less than or equal to the endpoint lance position, the slag splashing endpoint is determined to have been reached. A termination command is generated to end the dynamic control of slag splashing furnace protection only when both of the above conditions are met simultaneously.

[0073] The aforementioned lance position control commands include lance position holding commands generated during the static slag conditioning stage, lance position action commands generated during the dynamic slag splashing stage, and termination commands generated during the endpoint determination stage. The entire process, from the start of the static slag conditioning stage to the automatic termination of the endpoint determination stage, requires no manual intervention, achieving fully automated and intelligent control of the slag splashing furnace protection process.

[0074] In an exemplary embodiment, to better understand the solution of this application, the following embodiments are provided: First, a high-temperature resistant industrial vision sensor deployed at the target converter furnace mouth is used to collect a real-time video stream of the furnace mouth during the slag splashing protection process. The slag splashing image at the furnace mouth at the current moment is obtained from the video stream, thus obtaining the target slag splashing image. The current slag splashing mode is also acquired; in this embodiment, the furnace bottom splashing mode is taken as an example.

[0075] Then, the slag level recognition model is pre-built and trained in the following way: First, establish a standard for classifying slag splash levels. Based on the density and intensity of slag particle splashing, slag splash images are divided into nine levels, referencing... Figure 3 , Figure 3 This is a schematic diagram illustrating different levels of slag splashing images in an exemplary embodiment of this application; the higher the level, the denser the slag particles and the more intense the splashing; the lower the level, the slag particles are sparser and the splashing is less intense. To accommodate different lighting conditions in various scenes and the shift in the slag splashing flame area caused by converter adjustments and changes in furnace mouth thickness, 1000 images with different furnace mouth brightness and slag splashing flame positions were collected for each level, resulting in a total dataset of 9000 images. Based on the aforementioned level classification criteria, all sample images were labeled with their respective levels, generating image samples with slag splashing level labels.

[0076] Then, the initial network model is trained based on image samples labeled with slag splash levels. This embodiment uses ResNet50 as the core feature extraction network, and instead of using pre-trained weights on general datasets like ImageNet, it employs a strategy of training from scratch. The model input is a continuous sequence of images, with a uniform image size of 200 pixels × 500 pixels. During training, data augmentation operations such as random horizontal flipping and brightness fine-tuning are applied. The model receives five consecutive frames as input sequences each time, feeding these five frames sequentially into the same ResNet50 network to extract high-dimensional feature sequences from each frame. The temporal feature fusion module integrates these five feature sequences, capturing the temporal evolution of slag splashing and outputting a comprehensive feature representation that fuses the spatiotemporal context. This comprehensive feature representation undergoes dimensionality reduction via a global average pooling layer, is then fed into a Dropout layer with a dropout rate of 0.2 to suppress overfitting, and finally mapped to the original logical values ​​of the nine slag splash levels through a fully connected layer. End-to-end training was performed using the cross-entropy loss function and the Adam optimizer, with a training period of 100 epochs. The dataset was divided into training, validation, and test sets in chronological order. The trained model achieved a classification accuracy of 98%. An epoch can be understood as the process of a complete dataset passing through the neural network once and returning once; it can also be understood as training all training samples once.

[0077] After model training converges, INT8 quantization technology is used to perform post-training quantization compression on the model weights and activation values, converting the data precision from 32-bit floating-point numbers to 8-bit integers. After model quantization compression, hardware requirements are extremely low. It does not rely on expensive high-performance graphics cards; only a standard-configuration industrial computer is needed to achieve high-speed, real-time inference operations. No additional high-performance computing resources are required, significantly reducing the hardware threshold and implementation cost for on-site deployment, demonstrating good economic efficiency and scalability. For example, the quantized model file size is 17MB, and the quantized model can achieve a real-time inference speed of 20fps (Frames Per Second) on a 4-core, 4GB memory industrial computer without a GPU.

[0078] In this embodiment, the slag splashing gun position parameter configuration table is pre-designed in the following way: The furnace bottom splashing mode is selected based on the furnace lining erosion condition. Based on this mode, the furnace bottom area is determined as the location for slag splashing protection, and the target is to repair the furnace bottom. According to the inherent process rules of slag splashing protection, the process is divided into a static slag adjustment stage, a dynamic slag splashing stage, and an endpoint determination stage.

[0079] Based on the location of the slag splashing protection and the slag splashing target, the specific parameters for each stage are determined, resulting in the slag splashing lance position parameter configuration table: In the splashed furnace bottom mode, the highest lance position is set to 2000mm and the lowest lance position is set to 250mm, which serve as the safety boundary for the oxygen lance position operation. In the static slag adjustment stage, the lance position height is set to 2000mm and the lance position holding time is set to 20s. During this stage, the oxygen lance maintains a fixed lance position to adjust the slag state. In the dynamic splashed slag stage, the lance lifting standard level is set to 5, the corresponding lance lifting height is 100mm and the lance lifting holding time is 5s, and the lance lowering standard level is set to 5, the corresponding lance lowering height is 200mm and the lance lowering holding time is 5s. In the endpoint determination stage, the endpoint standard level is set to 1 and the endpoint lance position is set to 600mm.

[0080] In the actual slag splashing control process, after the target slag splashing image is input into the pre-trained slag splashing level recognition model, the model performs the following online inference steps: First, flame tracking is performed on the target slag splash image. The core area of ​​the slag splash is located based on real-time image processing algorithms, and the area of ​​interest (ROI) is automatically adjusted according to the location results, effectively overcoming the interference caused by furnace flame jumping, external obstruction and camera vibration.

[0081] Then, a predetermined number of consecutive frames (e.g., 5 frames) are selected from the target slag splash image, and the analysis region of each frame is determined based on the region of interest. After size normalization (200 pixels × 500 pixels), each analysis region is input into the ResNet50 feature extraction network of the slag splash level recognition model to extract the high-dimensional feature sequence of each frame. The temporal feature fusion module of the slag splash level recognition model integrates the 5 high-dimensional feature sequences to capture the temporal evolution of slag splashing and generate a comprehensive feature representation that integrates the spatiotemporal context. This comprehensive feature representation is then sequentially processed through a global average pooling layer for dimensionality reduction, a Dropout layer (with a dropout rate of 0.2) to suppress overfitting, and a fully connected layer for mapping, outputting the original slag splash level.

[0082] After obtaining the original slag splash level, the intelligent rectification module integrated within the model performs post-processing on the original slag splash level. Specifically, this includes: maintaining a configurable time window (e.g., 1 second); applying a moving average filter to the original slag splash level within the time window to obtain the current slag splash level for the current time window; calculating the level change amplitude based on the current slag splash level and the previous slag splash level in the previous time window, with a preset maximum allowable level jump value (i.e., preset change amplitude) of 2 levels. If the change amplitude exceeds 2 levels, a smooth preset transition value is used to replace the current slag splash level, resulting in a constrained slag splash level after jump constraints; and normalizing the constrained slag splash level at a preset frequency (e.g., once per second) to obtain the rectified and optimized target slag splash level signal. This target slag splash level signal is stable, smooth, and reliable, and can be directly used for downstream gun position closed-loop control.

[0083] After the slag splashing program is started, it first enters the static slag conditioning stage. Based on the lance position height (2000mm) and lance position holding time (20s) in the slag splashing lance position parameter configuration table, a lance position holding command is generated to control the oxygen lance to maintain a fixed lance position of 2000mm during the static slag conditioning stage until the 20s end.

[0084] After the static slag adjustment stage, the system automatically enters the dynamic slag splashing stage. In this stage, based on a comparison of the target slag splashing level with the standard levels for raising and lowering the oxygen lance, a lance position action command is generated, and the oxygen lance is controlled to perform the corresponding lance position action. Specifically: If the target slag splash level is greater than the standard level for raising the lance (level 5), it indicates that the current slag splashing is too intense. A lance raising command is generated, controlling the oxygen lance to rise by 100mm (raising height) and hold for 5 seconds (raising hold duration). If the target slag splash level is less than or equal to the standard level for lowering the lance (level 5), it indicates that the current slag splashing is insufficient. A lance lowering command is generated, controlling the oxygen lance to lower by 200mm (lowering height) and hold for 5 seconds (lower hold duration). After each lance raising or lowering action is completed, the target slag splash level is compared with the standard levels for raising and lowering the lance. This process is repeated continuously, forming a closed-loop feedback adjustment mechanism of perception, comparison, execution, and re-perception.

[0085] During the dynamic slag splashing stage, the target slag splashing level and the current actual lance position are detected synchronously in real time. When the target slag splashing level is less than or equal to the endpoint standard level (Level 1), and the current actual lance position is less than or equal to the endpoint lance position (600mm), it is determined that the slag splashing endpoint has been reached, a termination command is generated, and the dynamic control of slag splashing and furnace protection is automatically terminated.

[0086] The entire process, from start-up to automatic termination, requires no manual intervention, achieving fully automated and intelligent control of the converter slag splashing protection process.

[0087] This application achieves closed-loop automatic control of the slag splashing furnace protection process by moving from image perception to intelligent decision-making and then to automatic execution. This frees operators from intensive visual observation and manual operation, and represents an improvement from experience-based slag splashing to intelligent slag splashing. The slag splashing furnace protection process of this application does not require manual intervention during the process. Based on real-time visual feedback, it dynamically adapts to the slag splashing condition of each furnace, avoiding the impact of human fatigue and inconsistent operator skill levels. This results in more stable slag splashing effects, more precise control of furnace lining and bottom thickness, effectively extending the converter's service life and improving the uniformity and stability of the furnace protection effect. This application solidifies the experience of excellent operators into the slag splashing level recognition model and operating rules. The slag splashing effect can be precisely and stably adjusted through the slag splashing gun position parameter configuration table, providing flexible control and forming a digital asset that can be inherited and reused by the enterprise, achieving the accumulation and standardization of core process knowledge. The model provided by this application has strong generalization ability. Through training with a large dataset, it can adapt to different furnace conditions, accurately determine its slag splashing state, and also has strong transferability, enabling rapid deployment in different sites.

[0088] Figure 4 This is a block diagram illustrating a dynamic control device for slag splashing furnace protection, as shown in an exemplary embodiment of this application. This device can be applied to... Figure 1 The implementation environment shown is specifically configured in computer device 102. This device can also be applied to other exemplary implementation environments and specifically configured in other devices. This embodiment does not limit the implementation environment to which this device is applicable.

[0089] like Figure 4 As shown, the exemplary dynamic control device for slag splashing protection includes: a data acquisition module 410, a grade identification module 420, and a protection control module 430.

[0090] The data acquisition module 410 is used to acquire the target slag splashing image at the target converter mouth and obtain the current slag splashing mode of the target converter; the level recognition module 420 is used to input the target slag splashing image into a preset slag splashing level recognition model to obtain the target slag splashing level used to characterize the intensity of the current slag splashing, and the slag splashing level recognition model is trained based on image samples with slag splashing level labels; the furnace protection control module 430 is used to generate lance position control commands according to the slag splashing lance position parameter configuration table corresponding to the target slag splashing level and the current slag splashing mode, and control the oxygen lance to perform the corresponding lance position actions according to the lance position control commands to complete the dynamic control of slag splashing furnace protection.

[0091] It should be noted that the dynamic control device for slag splashing protection provided in the above embodiments and the dynamic control method for slag splashing protection provided in the above embodiments belong to the same concept. The specific operation methods of each module and unit have been described in detail in the method embodiments and will not be repeated here. In practical applications, the dynamic control device for slag splashing protection provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0092] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the dynamic control method for slag splashing furnace protection as described in any of the above embodiments.

[0093] Figure 5 A schematic diagram of a computer system suitable for implementing the embodiments of this application is shown. It should be noted that... Figure 5 The computer system 500 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of the embodiments of this application.

[0094] like Figure 5 As shown, the computer system 500 includes a Central Processing Unit (CPU) 501, which can perform various appropriate actions and processes, such as executing the methods provided in the various embodiments above, based on a program stored in Read-Only Memory (ROM) 502 or a program loaded from Storage Unit 508 into Random Access Memory (RAM) 503. The RAM 503 also stores various programs and data required for system operation. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An Input / Output (I / O) interface 505 is also connected to the bus 504.

[0095] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. Removable media 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 510 as needed so that computer programs read from them can be installed into storage section 508 as needed.

[0096] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit (CPU) 501, it performs various functions defined in the system of this application.

[0097] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0098] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0099] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0100] Another aspect of this application provides a computer-readable storage medium storing a computer program that is executed by a processor to implement the dynamic control method for slag splashing furnace protection as described in any of the above embodiments. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not incorporated into the electronic device.

[0101] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0102] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the dynamic control method for slag splashing furnace protection provided in the various embodiments described above.

[0103] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this application.

[0104] Other embodiments of this application will readily conceive of by those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0105] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A dynamic control method for slag splashing and furnace protection, characterized in that, include: Acquire the target slag splashing image at the target converter mouth, and obtain the current slag splashing mode of the target converter; The target slag splashing image is input into a preset slag splashing level recognition model to obtain a target slag splashing level that characterizes the intensity of the current slag splashing. The slag splashing level recognition model is trained based on image samples with slag splashing level labels. Based on the target slag splashing level and the slag splashing lance position parameter configuration table corresponding to the current slag splashing mode, a lance position control command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the lance position control command to complete the dynamic control of slag splashing for furnace protection.

2. The dynamic control method for slag splashing and furnace protection according to claim 1, characterized in that, The methods for obtaining the splash level identification model include: Acquire slag splash level labels and sample slag splash images, wherein the slag splash level labels are determined based on the density of slag particle splashing; The slag-splashing images of the samples are labeled according to the slag-splashing level labels to generate image samples with slag-splashing level labels; The slag level recognition model is trained based on the image samples with slag level labels to obtain the slag level recognition model.

3. The dynamic control method for slag splashing and furnace protection according to claim 1, characterized in that, After inputting the target slag splash image into a preset slag splash level recognition model, the method further includes: Flame tracking is performed on the target slag splash image to locate the core area of ​​the slag splash, and the area of ​​interest is adjusted based on the location results; A predetermined number of consecutive frames are selected from the multiple frames of the target slag splash images, and the analysis area of ​​each frame is determined according to the region of interest. The analysis regions of each frame image are input into the feature extraction network of the slag splashing level recognition model to extract the high-dimensional feature sequence of each frame image; The temporal feature fusion module of the splashing slag level identification model integrates the high-dimensional feature sequences of the preset number of frame images to generate a comprehensive feature representation that integrates the spatiotemporal context. The comprehensive feature representation is sequentially processed through the global average pooling layer for dimensionality reduction, the random deactivation layer for overfitting suppression, and the fully connected layer for mapping in the slag splashing level identification model to obtain the original slag splashing level.

4. The dynamic control method for slag splashing and furnace protection according to claim 3, characterized in that, After obtaining the original slag grade, the method further includes: Statistical analysis is performed on the original slag splash levels within the preset time window to obtain the current slag splash level for the current time window; The change in slag level is calculated based on the current slag splashing level and the previous slag splashing level in the previous time window. If the change range of the level is greater than the preset change range, then the current slag splashing level is replaced by a preset transition value to obtain the constrained slag splashing level after the jump constraint; The constrained slag splashing level is normalized and output according to a preset frequency to obtain the target slag splashing level after rectification and optimization.

5. The dynamic control method for slag splashing and furnace protection according to any one of claims 1 to 4, characterized in that, The methods for obtaining the slag-splashing gun position parameter configuration table include: Determine the slag splashing location and slag splashing target based on the current slag splashing mode; According to the preset stage division rules, the slag splashing process is divided into a static slag adjustment stage, a dynamic slag splashing stage, and an endpoint determination stage. Based on the slag splashing furnace protection position and the slag splashing target, the lance position height and lance position holding time in the static slag adjustment stage, the lance raising rules and lance lowering rules in the dynamic slag splashing stage, and the endpoint standard level and endpoint lance position in the endpoint determination stage are determined respectively, thereby obtaining the slag splashing lance position parameter configuration table. The lance raising rules include the lance raising standard level and the corresponding lance raising height and lance raising holding time, and the lance lowering rules include the lance lowering standard level and the corresponding lance lowering height and lance lowering holding time.

6. The dynamic control method for slag splashing and furnace protection according to any one of claims 1 to 4, characterized in that, The slag splashing gun position parameter configuration table includes the gun position height and gun position holding time during the static slag adjustment stage, the gun lifting standard level and corresponding gun lifting height and holding time during the dynamic slag splashing stage, the gun lowering standard level and corresponding gun lowering height and holding time, and the endpoint standard level and endpoint gun position during the endpoint determination stage. Based on the target slag splashing level and the slag splashing mode corresponding to the slag splashing lance position parameter configuration table, a lance position control command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the lance position control command, including: Based on the lance height and the lance holding time, a lance holding command is generated, and the oxygen lance is controlled to maintain a fixed lance position during the static slag adjustment stage until the static slag adjustment stage ends. After the static slag adjustment stage is completed, the dynamic slag splashing stage begins. Based on the comparison results between the target slag splashing level and the standard level for raising and lowering the lance, a lance position action command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the lance position action command. After the lance position action command is completed, the target slag splashing level is compared with the standard level for raising and lowering the lance again. During the dynamic slag splashing stage, the target slag splashing level and the current actual lance position are detected in real time. If the target slag splashing level is less than or equal to the endpoint standard level and the current actual lance position is less than or equal to the endpoint lance position, a termination command is generated to end the dynamic control of slag splashing and furnace protection. The lance position control command includes the lance position holding command, the lance position action command, and the termination command.

7. The dynamic control method for slag splashing and furnace protection according to claim 6, characterized in that, Based on the comparison results between the target slag splash level and the standard levels for raising and lowering the oxygen lance, a lance position action command is generated, and the oxygen lance is controlled to perform the corresponding lance position action according to the lance position action command, including: If the target slag splash level is greater than the standard level for raising the lance, a lance raising action command is generated, and the oxygen lance is raised to the corresponding lance raising height and maintained for the specified duration according to the lance raising action command. If the target slag splash level is less than or equal to the lance lowering standard level, a lance lowering action command is generated, and the oxygen lance is controlled to be lowered to the corresponding lance lowering height and maintained for the specified duration according to the lance lowering action command.

8. A dynamic control device for slag splashing and furnace protection, characterized in that, include: The data acquisition module is used to acquire the target slag splashing image at the furnace mouth of the target converter and to acquire the current slag splashing mode of the target converter; The level recognition module is used to input the target slag splashing image into a preset slag splashing level recognition model to obtain the target slag splashing level used to characterize the intensity of the current slag splashing. The slag splashing level recognition model is trained based on image samples with slag splashing level labels. The furnace protection control module is used to generate lance position control commands based on the target slag splashing level and the slag splashing mode corresponding to the slag splashing lance position parameter configuration table, and to control the oxygen lance to perform the corresponding lance position actions according to the lance position control commands, so as to complete the dynamic control of slag splashing furnace protection.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the dynamic control method for slag splashing furnace protection as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the dynamic control method for slag splashing furnace protection as described in any one of claims 1 to 7.

Images