Intelligent identification system for slag samples in steelmaking refining process

Abstract

The application discloses a kind of slag sample intelligent identification system in steelmaking refining process, including image acquisition unit, core algorithm processing unit, industrial data interaction unit, QMS system docking unit;Image acquisition unit receives the image capture instruction issued by QMS system docking unit, collects slag sample image and is transmitted to core algorithm processing unit;Core algorithm processing unit completes feature extraction, quantification calculation and standardization grading to received slag sample image, and outputs the slag sample grading result of three dimensions of color, pore, thickness;Industrial data interaction unit binds the slag sample grading result with the furnace number, steel grade, refining process stage basic information corresponding to the furnace, after completing data encapsulation, transmission to QMS system docking unit;QMS system docking unit is used for the two-way real-time communication of system and steel plant QMS platform.The advantage of the present application is to realize the non-contact, fast, accurate, objective identification and quality grading of refining slag sample, improve the efficiency of slag sample detection and determination.

Description

Technical Field

[0001] This invention belongs to the field of metallurgical steelmaking technology, and particularly relates to a smart identification system for slag samples in the steelmaking refining process. Background Technology

[0002] In the entire steelmaking process, the refining process, as the core intermediate link connecting the electric arc furnace roughing of molten steel and the continuous casting of steel products, undertakes the crucial functions of precise control of molten steel composition, removal of inclusions, and temperature homogenization. Its process control level directly determines the purity and mechanical properties of the final steel. Among these processes, the slag state at different refining stages is the core carrier reflecting the metallurgical reaction process of molten steel. The composition and macroscopic morphology of the slag sample are directly related to the desulfurization efficiency, alloy element recovery rate, and oxygen content control level of the corresponding heat. The extraction and testing results of the slag sample are the core basis for determining whether the molten steel quality meets the requirements for the next process, playing an irreplaceable supporting role in the overall steel quality control.

[0003] Currently, the quality assessment of refining slag samples still relies on a combination of manual on-site visual inspection and subsequent laboratory component analysis by the Science and Technology Quality Department. This traditional operating model has gradually revealed numerous shortcomings in practical application: First, the manual visual identification process in front of the furnace relies heavily on the long-term experience of senior operators. There are significant individual differences in the technical level, judgment standards, and working conditions of different operators, which leads to strong subjectivity in the slag sample evaluation results and poor consistency in cross-shift evaluation. The misjudgment rate of manual identification can reach more than 5%, and a large number of misjudgment risks directly flow into subsequent processes.

[0004] Secondly, the supporting laboratory component testing process requires multiple steps, including slag sample packaging, cross-regional testing, sample preparation, instrumental analysis, and manual report entry. The overall testing cycle is long, and the feedback of results is significantly delayed compared to the real-time production rhythm of the refining process. It is impossible to respond quickly to unqualified slag samples that occur during the refining process, and it is difficult to initiate the corresponding process correction process in a timely manner.

[0005] The defects of the aforementioned traditional testing methods can easily lead to a series of chain quality problems in actual production: at best, it can cause excessive oxygen content in molten steel, decreased recovery rate of alloying elements, increased number of non-metallic inclusions, and desulfurization efficiency that does not meet the process design requirements; at worst, it can cause batch process accidents such as unqualified composition of the entire furnace of molten steel and interruption of casting in the continuous casting process, resulting in additional quality cost losses and production efficiency losses. Summary of the Invention

[0006] The purpose of this invention is to address the problems of high reliance on manual labor, low detection efficiency, delayed result feedback, strong subjectivity in judgment, and high misjudgment rate in existing refining slag sample testing. It provides a smart slag sample identification system for the steelmaking refining process, enabling non-contact, rapid, accurate, and objective identification and quality grading of slag samples throughout the refining process. This improves the efficiency of slag sample detection and judgment, reduces labor costs and quality risks, and provides standardized and traceable data support for optimizing steelmaking refining processes.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A smart slag sample identification system for steelmaking refining processes comprises an image acquisition unit, a core algorithm processing unit, an industrial data interaction unit, and a QMS system interface unit, which are sequentially connected via communication. These four units achieve real-time data transmission across the entire process via industrial Ethernet, replacing the traditional manual visual identification method to automate and digitize slag sample identification throughout the refining process. The specific structure and connection relationships of each unit are as follows: The image acquisition unit is a high-resolution industrial vision image acquisition device, equipped with an adaptive supplementary lighting component. The signal input end of the unit is connected to the command output end of the QMS system docking unit, and receives the image capture command issued by the QMS system docking unit in real time, accurately triggering the slag sample image acquisition action of the corresponding process node.

[0008] The acquisition sequence of this unit is perfectly matched with the key metallurgical nodes of the refining process, covering all core process stages from 5 minutes, 10 minutes, 20 minutes, 60 minutes of refining, final slag, to vacuum entry. After acquisition, the original image of the uncompressed slag sample is transmitted in real time to the image input port of the core algorithm processing unit through the industrial network. Non-contact image acquisition can be completed without human intervention, avoiding the safety hazards to operators caused by the high temperature environment in front of the furnace.

[0009] The core algorithm processing unit is the core computing module of the system. Its hardware carrier is an edge computing host equipped with an industrial-grade computing chip. It incorporates a slag sample feature analysis model built on Python and OpenCV technology stacks. The unit's image receiving port is communicatively connected to the output of the image acquisition unit, used for feature extraction, quantization calculation, and standardization grading of the received slag sample images. The unit integrates four independently operating functional sub-modules: Color Grading Module: The module has a built-in HSV color space conversion program. It automatically filters image background interference through K-means clustering algorithm, accurately extracts the main color tone features of the slag sample, and linearly divides the full visual range of slag sample color from deep black to pure white into ten quantization levels from 0 to 9. This completely eliminates the subjective bias of manual color judgment and realizes an objective digital description of slag sample color.

[0010] Porosity grading module: The module has a built-in image segmentation and contour analysis program that can accurately identify all the pore contours on the surface of the slag sample under complex furnace background, automatically calculate the proportion of the pore area to the total area of ​​the slag sample fragments, and divide the proportion into ten quantitative levels from 0 to 9, so as to realize the standardized quantitative statistics of the pore characteristics of the slag sample.

[0011] Thickness Grading Module: The module has a built-in focal length stack and edge detection fusion program. Through the fusion calculation of multiple frames of images with different focal lengths, it realizes non-contact visual measurement of slag sample thickness. Using pixels as a unified unit of measurement, the slag sample thickness is divided into ten quantification levels from 0 to 9. The thickness parameters can be accurately collected without contacting the slag sample.

[0012] Encoding Generation Module: The module has a built-in encoding fusion program that merges the grading results of the three dimensions of color, porosity, and thickness into a unique three-digit code in the order of "color C-porosity P-thickness T". This code serves as a unique "digital fingerprint" for the quality of the slag sample, realizing a standardized and traceable digital description of the quality characteristics of complex slag samples, and providing a unique identifier for subsequent full-process data traceability.

[0013] The input end of the industrial data interaction unit is communicatively connected to the output end of the core algorithm processing unit. The unit is equipped with an industrial data matching database to bind slag sample identification data with production information. The slag sample digital code, grading results of each dimension, and preliminary qualification judgment information generated by the core algorithm processing unit are matched and bound one by one with the basic information such as the furnace number, steel grade, and refining process stage of the corresponding furnace. After the data is structurally encapsulated, the complete data packet is transmitted to the data input port of the QMS system docking unit to avoid mismatch and confusion of slag sample data from different furnaces.

[0014] The QMS system interface unit is equipped with a standardized industrial communication protocol interface, enabling bidirectional real-time data communication between the system and the steel plant's existing QMS platform. On one hand, the unit synchronously obtains the corresponding refining furnace number and metallurgical process information from the QMS platform in real time, and actively sends the corresponding stage image capture signals to the image acquisition unit according to the preset process nodes, achieving complete synchronization between the acquisition action and the production rhythm. On the other hand, the unit synchronizes the received slag sample full-dimensional identification data and qualification judgment results to the corresponding page of the QMS platform for real-time display, and synchronously outputs the slag sample qualification / abnormal judgment results based on the preset slag sample qualification range for different steel grades, as well as the corresponding slag condition abnormality solutions and process adjustment guidelines, guiding furnace operators to complete process corrections as soon as possible.

[0015] In summary, the overall closed-loop workflow of the system of this invention is as follows: the QMS system docking unit obtains the current furnace and metallurgical process information from the steel plant's QMS platform in real time → sends image capture instructions to the image acquisition unit according to preset nodes → the image acquisition unit acquires slag sample images of the corresponding stage and transmits them back → the core algorithm processing unit completes multi-dimensional feature extraction, quantitative grading and digital encoding of slag samples → the industrial data interaction unit completes the matching and binding of slag sample identification data with the corresponding furnace production information → the QMS system docking unit completes result transmission, real-time on-site display and judgment output.

[0016] Compared with the prior art, the advantages of the technical solution of the present invention are as follows: (1) This invention replaces the traditional slag sample detection mode, shortens the slag sample detection cycle from several hours to minutes, and improves the detection efficiency by nearly 10 times. It can realize real-time identification and early warning of slag sample abnormalities during the refining process, provide timely data support for on-site process adjustment, and avoid steel quality defects caused by continuous slag condition abnormalities from the source. (2) This invention transforms the three subjective evaluation indicators of color, porosity and thickness, which traditionally rely on human experience, into objective digital grading results of 0 to 9 levels, eliminating the subjective bias of human evaluation, reducing the misjudgment rate of slag sample identification from more than 5% to less than 1%, improving detection accuracy and stability, and effectively avoiding process rework and batch quality loss caused by human misjudgment. (3) This invention can replace the manual inspection process, directly saving hundreds of thousands of yuan in labor costs per year; by reducing quality misjudgment and reducing process abnormality losses, it can directly save more than one million yuan in labor and quality costs per year, with significant economic benefits; (4) The system-generated full-process standardized slag sample quality data stream of the present invention can provide accurate data support for the iterative optimization of smelting process. Based on the three-in-one digital coding system of "color-porosity-thickness", a unique "digital ID card" is established for each slag sample, realizing the standardized, traceable and fully digital management of the quality characteristics of refined slag samples. Detailed Implementation Example

[0017] To make the present invention clearer, the following description further illustrates a smart slag sample identification system in the steelmaking refining process of the present invention. The specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0018] This embodiment uses the refining process of a steel plant as an application scenario. The slag sample intelligent identification system of the present invention can be implemented according to the following steps: (a) Implementation environment: This system is deployed at the refining station and is equipped with an industrial-grade visual image acquisition camera, an edge computing industrial control computer, and the steel plant's existing QMS quality management platform. The system is developed based on Python and OpenCV technologies and runs on the Windows industrial operating system.

[0019] (II) System Hardware and Deployment: The image acquisition unit uses a 5-megapixel industrial area array camera, paired with a ring-shaped supplementary light source, and is installed at a fixed position in front of the refining furnace slag sample sampling station to ensure the consistency of illumination and the stability of angle in slag sample image acquisition. The camera communicates with the edge computing industrial control computer via gigabit Ethernet.

[0020] The core algorithm processing unit and the industrial data interaction unit are both deployed in the edge computing industrial control computer. The industrial control computer is configured with an industrial-grade processor, more than 8GB of RAM, and more than 256GB of solid-state drive to meet the requirements of real-time algorithm operation and local data caching.

[0021] The QMS system interface unit communicates with the existing QMS quality management platform via the steel plant's industrial Ethernet to ensure the real-time performance and security of data transmission.

[0022] (III) System Implementation and Operation Steps: (1) After the system is initialized and started, the QMS system docking unit obtains the furnace number, steel grade and refining process data of the corresponding refining station of the electric furnace production line from the steel plant's QMS platform in real time. When the refining process reaches the preset key nodes, namely 5min, 10min, 20min, 60min, final slag, and vacuum entry, it automatically sends an image capture command to the image acquisition unit.

[0023] (2) After receiving the instruction, the image acquisition unit completes the image acquisition of the corresponding slag sample and transmits the high-definition original image of the acquired slag sample to the core algorithm processing unit of the edge computing industrial control computer.

[0024] (3) After preprocessing the slag sample image, the core algorithm processing unit initiates the quantization and grading calculation of the three major features: Color grading: Convert the image to the HSV color space, remove background interference through K-means clustering algorithm, extract the main color tone of the sample, and output the color grading results of 0 to 9 levels according to the preset color level range; Porosity grading: Through threshold segmentation and contour extraction algorithms, the porosity region in the slag sample is identified, the ratio of the total area of ​​pores to the total area of ​​the effective region of the slag sample is calculated, and the porosity grading results of 0 to 9 are output according to the preset ratio range. Thickness grading: The image sequence of slag sample under different focal lengths is obtained by focal length stacking technology, and the depth information of the slag sample contour is extracted by edge detection algorithm. The thickness value of the slag sample is calculated, and the thickness grading results of 0 to 9 levels are output according to the preset thickness range. Code generation: Generate a three-digit code in the order of "color grade - porosity grade - thickness grade". For example, if the color grade is 3, the porosity grade is 4, and the thickness grade is 7, then the code "347" will be generated as the unique quality number identifier for the slag sample.

[0025] (4) The industrial data interaction unit matches and binds the above-mentioned classification results and digital codes with the corresponding furnace number, steel grade, refining stage, collection time and other information to complete the data standardization and encapsulation.

[0026] (5) The QMS system docking unit transmits the encapsulated slag sample data in all dimensions to the steel plant's QMS platform and displays it in real time on the page corresponding to the furnace number; at the same time, it calls the preset slag sample qualification judgment range and outputs the judgment result of "qualified / abnormal" for the slag sample. If it is judged as abnormal, it outputs the abnormal cause of the corresponding slag condition, the impact on the molten steel and the standardized solution measures. For example, if the slag color is black and the corresponding code is 0XX, it outputs the process guidance "insufficient deoxidation, FeO>2%, it is recommended to alloy precipitation deoxidation, and add deoxidizer and reduce oxygen level".

[0027] Other abnormal conditions in slag conditions and their solutions include: If the slag is yellowish-green, the output will be "Insufficient deoxidation, FeO is between 1% and 2%, it is recommended to continue deoxidation and maintain a reducing atmosphere"; If the slag is viscous and crusty, the output will be "The alkalinity is too high or the temperature is too low, or it has been stored for a long time. It is recommended to heat up the slag to dissolve it. The situation will improve once the aluminum oxide products are generated. Alternatively, you can adjust the amount of refining slag or fluorite." If the slag is too thin, the output will be "Alkalinity too low, it is recommended to add lime before the mid-game". If it is calcium carbide slag, the output will be "Carbon powder or silicon carbide not cleared or excessive and not fully stirred. It is recommended to supply low power, increase argon gas, stop adding deoxidizer until it is broken down".

[0028] (6) All slag sample data are stored locally and synchronously on the QMS platform to achieve full-process traceability and form a slag sample quality database to provide data support for subsequent refining process optimization.

[0029] After the system of this invention was implemented on the production line, it achieved unmanned automatic identification of refining slag samples throughout the entire process. The detection time for a single sample was ≤30s, and the detection efficiency was more than 10 times higher than that of the traditional mode. The misjudgment rate of slag sample identification was reduced from the original 5.2% to less than 0.8%. At the same time, it provided standardized full-process data support for the optimization of refining process parameters, achieving the design expectations.

[0030] In addition to the embodiments described above, the present invention may have other implementations. All technical solutions formed by equivalent substitution or equivalent transformation fall within the protection scope claimed by the present invention.

Claims

1. A smart identification system for slag samples in steelmaking refining processes, characterized in that: It includes an image acquisition unit, a core algorithm processing unit, an industrial data interaction unit, and a QMS system docking unit that are connected in sequence. The signal input terminal of the image acquisition unit is communicatively connected to the instruction output terminal of the QMS system docking unit. It is used to receive the image capture instruction issued by the QMS system docking unit, acquire slag sample images of key nodes in the refining process, and transmit the original slag sample image to the core algorithm processing unit in real time. The core algorithm processing unit has a built-in slag sample feature analysis model, which is used to extract features, perform quantitative calculations and standardize grading on the received slag sample images, and output slag sample grading results in three dimensions: color, porosity and thickness. The industrial data interaction unit is used to match and bind the slag sample classification results output by the core algorithm processing unit with the basic information of the corresponding furnace number, steel grade, and refining process stage, and transmit the data to the QMS system docking unit after data encapsulation. The QMS system docking unit is used for bidirectional real-time communication between the system and the steel plant's QMS platform. On the one hand, it obtains the on-site refining furnace number and metallurgical process information from the QMS platform and sends the corresponding stage image capture signal to the image acquisition unit. On the other hand, it synchronizes the received slag sample identification data to the QMS platform for real-time display and outputs the judgment result of whether the slag sample is qualified or abnormal.

2. The intelligent slag sample identification system in the steelmaking refining process according to claim 1, characterized in that: The intelligent slag sample identification system operates as follows: The QMS system docking unit obtains the current furnace batch and metallurgical process information from the steel plant's QMS platform in real time, and then sends image capture instructions to the image acquisition unit according to preset nodes. Next, the image acquisition unit acquires slag sample images of the corresponding stage and transmits them back. Then, the core algorithm processing unit completes the multi-dimensional feature extraction, quantitative grading and digital encoding of the slag sample. Finally, the industrial data interaction unit completes the matching and binding of slag sample identification data with the corresponding furnace batch production information. The QMS system docking unit completes the result transmission, real-time on-site display and judgment output.

3. The intelligent slag sample identification system in the steelmaking refining process according to claim 1 or 2, characterized in that: The image acquisition unit is a high-resolution industrial vision image acquisition device equipped with an adaptive supplementary lighting component. Its acquisition nodes cover all core process stages of refining at 5 minutes, 10 minutes, 20 minutes, 60 minutes, final slag, and vacuum entry. The acquisition process is a non-contact, non-human intervention mode.

4. The intelligent slag sample identification system in the steelmaking refining process according to claim 1 or 2, characterized in that: The core algorithm processing unit's slag sample feature analysis model is built on Python and OpenCV technology stacks, and integrates four independent functional sub-modules, including a color grading module, a porosity grading module, a thickness grading module, and an encoding generation module.

5. The intelligent slag sample identification system in the steelmaking refining process according to claim 4, characterized in that: The color grading module extracts the main color tone of the slag sample using the K-means clustering algorithm in the HSV color space, and linearly divides the slag sample color from dark black to pure white into ten quantization levels from 0 to 9.

6. The intelligent slag sample identification system in the steelmaking refining process according to claim 5, characterized in that: The pore grading module identifies pores in the slag sample through image segmentation and contour analysis, calculates the proportion of the pore area to the total area of ​​the slag fragments, and divides this proportion into ten quantitative levels from 0 to 9.

7. The intelligent slag sample identification system in the steelmaking refining process according to claim 6, characterized in that: The thickness grading module uses a focal length stacking and edge detection fusion method to perform non-contact measurement of slag sample thickness, and divides the slag sample thickness into ten quantification levels from 0 to 9 using pixels as the unit of measurement.

8. The intelligent slag sample identification system in the steelmaking refining process according to claim 7, characterized in that: The encoding generation module merges the grading results of the three dimensions of color, pores, and thickness in the order of "color C-pores P-thickness T" to generate a unique three-digit code as the "digital fingerprint" of the sample.