Method, system, and medium for monitoring motion behavior of ferroalloy in rh refining process

Abstract

The application relates to the technical field of image processing, and discloses a method and system for monitoring the motion behavior of ferroalloy in an RH refining process and a medium. The method comprises the following steps: collecting real-time images of a ladle, a vacuum chamber and an immersion tube in an RH refining device, and obtaining an image sequence of ferroalloy particle motion; analyzing background change characteristics, separating a particle foreground area and extracting position coordinates; correlating time sequence data, establishing a single-particle continuous motion trajectory; fusing multi-view trajectory data, calculating three-dimensional space position changes and constructing a three-dimensional coordinate sequence; analyzing particle speed and direction changes, counting the residence time at the bottom of the ladle and the cycle number of the vacuum chamber, and outputting a speed distribution diagram and a residence time statistics table. The application improves the intelligent degree and spatial analysis accuracy of particle motion monitoring, and significantly enhances the automation level and visualization effect of the motion behavior analysis of the RH refining process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and in particular to a method and system for monitoring the motion behavior of ferroalloy in the RH refining process, and a medium. BACKGROUND

[0002] In the RH vacuum refining process of the steel industry, the existing technology mainly uses theoretical calculation, numerical simulation and simple physical observation methods to study the motion behavior of ferroalloy. Traditional monitoring methods include mathematical modeling based on fluid mechanics theory, computational fluid dynamics (CFD) software simulation, and indirect methods such as composition analysis and temperature measurement to infer the distribution state of the alloy. Some research institutions have developed physical water model experimental devices to simulate the RH refining process through a scaled-down model, mainly relying on manual observation or simple video recording to analyze particle motion trajectories. These existing technologies provide basic experimental means for understanding the RH refining mechanism to some extent.

[0003] However, the existing technology has significant technical defects and application limitations. Traditional manual observation methods are highly subjective and have low precision, cannot accurately capture the trajectories of fast-moving particles, and cannot simultaneously monitor the motion states of multiple targets; existing image recording systems lack intelligent image processing algorithms and can only provide qualitative visual information, cannot automatically extract quantitative parameters such as particle position, velocity, and acceleration; lack effective multi-view image fusion technology, cannot realize three-dimensional reconstruction and spatial trajectory analysis of particle motion; existing methods cannot establish a quantitative correlation between particle motion behavior and process parameters, lack process optimization guidance based on real-time monitoring data.

[0004] Based on the in-depth analysis of the above technical defects, the existing technology faces a series of progressive technical challenges: from the perspective of image data processing, how to achieve accurate collection and synchronous processing of multi-view images directly affects the data quality of subsequent analysis; from the perspective of target recognition, how to accurately separate and identify ferroalloy particles in a complex water flow background determines the basic performance of the monitoring system; from the perspective of trajectory tracking, how to establish a reliable multi-target correlation mechanism and continuous trajectory generation problem is related to the accuracy of motion behavior analysis; from the perspective of spatial reconstruction, how to fuse multi-view two-dimensional trajectory information into three-dimensional spatial coordinates is a prerequisite for in-depth analysis of motion characteristics; from the perspective of behavior analysis, how to extract meaningful motion parameters and statistical features from three-dimensional trajectory data directly determines the engineering application value of the monitoring results. These technical problems are interdependent and jointly constrain the development of intelligent monitoring technology for ferroalloy motion behavior in the RH refining process. SUMMARY

[0005] The application provides a method, system and medium for monitoring the motion behavior of ferroalloy in an RH refining process, to solve the problems that the existing technology cannot automatically identify and accurately track the multi-view motion trajectory of ferroalloy particles in the RH refining process, and lacks three-dimensional motion reconstruction and quantitative behavior analysis capabilities based on image processing. The intelligent degree and spatial analysis accuracy of particle motion monitoring are improved, and the automation level and visualization effect of RH refining process motion behavior analysis are significantly enhanced.

[0006] In a first aspect, the application provides a method for monitoring the motion behavior of ferroalloy in an RH refining process, which comprises:

[0007] Step S1, collecting real-time images of a ladle model, a vacuum chamber model and an immersion tube model in an RH refining simulation device to obtain image sequence data containing the motion state of ferroalloy particles;

[0008] Step S2, analyzing the background change characteristics in the image sequence data to separate the foreground area of the ferroalloy particles and obtain particle position coordinate information;

[0009] Step S3, correlating the change rule of the particle position coordinate information in time sequence to establish the continuous motion trajectory of a single particle and generate trajectory coordinate data;

[0010] Step S4, fusing multi-view trajectory coordinate data to calculate the position change of particles in three-dimensional space and construct a three-dimensional space coordinate sequence;

[0011] Step S5, analyzing the speed and motion direction change of particles in the three-dimensional space coordinate sequence, counting the residence time of particles at the bottom of the ladle and the circulation times in the vacuum chamber, and outputting a particle motion speed distribution map and a residence time statistics table.

[0012] In a second aspect, the application provides a system for monitoring the motion behavior of ferroalloy in an RH refining process, which comprises:

[0013] An extraction module is configured to collect real-time images of a ladle model, a vacuum chamber model and an immersion tube model in an RH refining simulation device to obtain image sequence data containing the motion state of ferroalloy particles;

[0014] A separation module is configured to analyze the background change characteristics in the image sequence data to separate the foreground area of the ferroalloy particles and obtain particle position coordinate information;

[0015] An analysis module is configured to correlate the change rule of the particle position coordinate information in time sequence to establish the continuous motion trajectory of a single particle and generate trajectory coordinate data;

[0016] a fusion module, which fuses the multi-view trajectory coordinate data, calculates the position change of the particles in the three-dimensional space, and constructs a three-dimensional space coordinate sequence;

[0017] a statistics module, which is configured to analyze the velocity and motion direction change of the particles in the three-dimensional space coordinate sequence, count the residence time of the particles on the bottom of the ladle and the circulation times of the particles in the vacuum chamber, and output a particle motion velocity distribution diagram and a residence time statistics table.

[0018] In a third aspect, a device for monitoring the motion behavior of ferroalloy in an RH refining process is provided, which comprises a memory and at least one processor, and the memory stores instructions; the at least one processor invokes the instructions in the memory to enable the device for monitoring the motion behavior of ferroalloy in an RH refining process to perform the method for monitoring the motion behavior of ferroalloy in an RH refining process described above.

[0019] In a fourth aspect, a computer readable storage medium is provided, which stores instructions, and when the instructions are run on a computer, the computer is enabled to perform the method for monitoring the motion behavior of ferroalloy in an RH refining process described above.

[0020] In the technical scheme provided in the present application, the multi-view image acquisition system of the complete RH refining simulation device including a ladle model, a vacuum chamber model and an immersion tube model is established, thereby solving the technical problem that the traditional single view cannot comprehensively monitor the motion trajectory of ferroalloy particles. The image segmentation processing based on the self-adaptive background modeling algorithm can effectively cope with the background change caused by water flow disturbance, accurately separate the foreground area of the ferroalloy particles, and significantly improve the accuracy and stability of particle recognition. The Kalman filter is used for multi-target tracking processing, a reliable particle trajectory connection mechanism is established through state prediction and Mahalanobis distance data association, and the defects of strong subjectivity and low precision of the traditional manual observation method are overcome. The application of the triangulation method realizes accurate conversion of the multi-view two-dimensional trajectory data to the three-dimensional space coordinates, and provides key technical support for in-depth analysis of the spatial motion behavior of particles.

[0021] In view of the special requirements of the analysis of the motion behavior of ferroalloy in the RH refining process, the region division algorithm and the time differential processing technology of the present application can accurately calculate the velocity vector, acceleration change and residence time distribution of the particles in each device region, and provide a quantitative data basis for process parameter optimization. By converting the complex three-dimensional motion trajectory into an intuitive velocity distribution diagram and a residence time statistics table, the automation degree and the visualization effect of the motion behavior monitoring of the RH refining process are enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0023] Figure 1 An embodiment of the method for monitoring the motion behavior of ferroalloy in the RH refining process in the embodiments of the present application is shown in the figure.

[0024] Figure 2 A flowchart of the particle motion trajectory tracking algorithm in the embodiments of the present application is shown in the figure.

[0025] Figure 3 An embodiment of the system for monitoring the motion behavior of ferroalloy in the RH refining process in the embodiments of the present application is shown in the figure.

[0026] Figure 4 An embodiment of the system for monitoring the motion behavior of ferroalloy in the RH refining process in the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0027] The embodiments of the present application provide a method, system and medium for monitoring the motion behavior of ferroalloy in the RH refining process. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily mean a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "comprising" or "having" and any variation thereof is intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0028] For the convenience of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 An embodiment of the method for monitoring the motion behavior of ferroalloy in the RH refining process in the embodiments of the present application comprises the following steps.

[0029] In step S1, real-time images of the ladle model, the vacuum chamber model and the immersion tube model in the RH refining simulation device are collected to obtain image sequence data containing the motion state of ferroalloy particles.

[0030] Specifically, real-time images of the ladle model, vacuum chamber model and immersion tube model in the RH refining simulation device are collected, and image data is obtained from different perspectives by multiple cameras. The first perspective image collection is performed on the transparent area of the side wall of the ladle model to obtain the movement state of the ferroalloy particles in the ladle. The second perspective image collection is performed on the front transparent wall of the vacuum chamber model to obtain the movement image of the particles in the vertical direction of the vacuum chamber. Meanwhile, the third perspective image collection is performed on the transparent area of the side of the immersion tube model to record the flow state of the ferroalloy particles in the immersion tube. All image data is processed synchronously by a high-frame-rate acquisition system to ensure the timeliness and accuracy of the data of each frame of image. The image data is taken from different angles, which can comprehensively capture the movement trajectories of the ferroalloy particles at different spatial positions, covering the flow characteristics of the particles in the ladle, vacuum chamber and immersion tube. After time synchronization processing, these image data form image sequence data containing the movement state of the ferroalloy particles, providing dynamic monitoring data support.

[0031] Step S2, analyze the background change characteristics in the image sequence data, separate the foreground area of the ferroalloy particles, and obtain the particle position coordinate information.

[0032] Specifically, the background change characteristics in the image sequence data are analyzed, the image is denoised by Gaussian filtering to remove noise interference caused by environmental changes, equipment vibration or water flow disturbance, and the clarity and accuracy of the image data are ensured. Based on the local image features, an adaptive threshold segmentation method is used to process the image, and the background and foreground are distinguished to accurately separate the foreground area of the ferroalloy particles. Through this process, the movement area of the particles in the image can be effectively extracted. The background difference algorithm is further used to detect and mark the foreground area, accurately positioning the position of the ferroalloy particles in the image, and obtaining the foreground area marking result of the particles in each frame of image. On this basis, the centroid coordinates of the particles are calculated, and the specific position coordinate information of the particles is obtained by combining the geometric feature extraction technology.

[0033] Step S3, correlate the change rule of the particle position coordinate information in the time sequence, establish the continuous movement trajectory of a single particle, and generate trajectory coordinate data.

[0034] Specifically, the change rule of the associated particle position coordinate information over time sequence is obtained by inputting the particle coordinate data obtained in the previous step into the trajectory generation algorithm in time sequence, combining a Kalman filter for state prediction processing, and predicting the possible position of the particle at each image update, so that the trajectory tracking process has robustness and continuity. After the predicted position is generated, the Mahalanobis distance calculation method is used to measure the similarity between the predicted point and the actually detected particle coordinate, so that accurate data association can be realized under complex water flow background and particle dense conditions, thereby avoiding the interruption of the trajectory caused by the intersection or shielding of the particle position. Based on the matching relationship matrix, the coordinates of the same particle in adjacent frames can be connected one by one to form a continuous time sequence position coordinate sequence, and further use the trajectory smoothing algorithm to remove incidental jumps or noise interference, so that the trajectory curve is more in line with the actual motion law. In this process, the system can not only assign a unique trajectory identifier to each particle, but also automatically update the trajectory set when the particle enters or exits the monitoring area, ensuring that all particles in the monitoring range are recorded completely. The trajectory data is stored in the form of time sequence, including the coordinate point set of each particle at each time, thereby providing basic data for subsequent three-dimensional space trajectory fusion and motion behavior analysis.

[0035] Step S4, fuse the multi-view trajectory coordinate data, calculate the position change of the particle in the three-dimensional space, and construct a three-dimensional space coordinate sequence.

[0036] Specifically, the multi-view trajectory coordinate data is fused, the two-dimensional coordinates obtained from each view are converted between the pixel coordinate system and the physical coordinate system according to the calibration parameters of the camera, the pixel points in the image are mapped to the real physical space position coordinates, so that the coordinate data under different views has a unified measurement basis. The multi-view data is fused by using the triangulation method, and the actual position coordinates of the particle in the three-dimensional space are calculated through the geometric relationship between different camera views, which avoids the problems of shielding or depth information loss caused by a single view. After obtaining the three-dimensional position coordinates fused by multiple views, the coordinates are input into a unified device coordinate system, and the alignment and unification of different camera data in the same space reference system are realized by means of a coordinate transformation matrix, so that the particle motion trajectory can completely cover the whole space of the ladle, the vacuum chamber and the immersion tube. The three-dimensional coordinate points are arranged and stored in time sequence order to form a continuous three-dimensional space coordinate sequence, and the abnormal fluctuations caused by noise are eliminated through a smoothing algorithm, so as to ensure the continuity and accuracy of the trajectory. The whole process realizes the reconstruction of the two-dimensional to three-dimensional motion trajectory, so that the position change of the particle in the RH refining simulation device can be completely recorded and presented.

[0037] Step S5, analyze the speed and direction change of the particles in the three-dimensional coordinate sequence, and count the residence time of the particles in the bottom of the ladle and the circulation number in the vacuum chamber, output the particle motion speed distribution map and the residence time statistics table.

[0038] Specifically, the speed and direction change of the particles in the three-dimensional coordinate sequence are analyzed, the three-dimensional velocity vector and acceleration vector of the particles in different time periods are calculated by time differentiation of the continuous coordinate points, and the motion direction of the particles is judged in combination with the change trend of the trajectory, so as to form the dynamic characterization of the particle motion state. Match the three-dimensional velocity vector and acceleration vector with the space area of the device, and judge whether the position of the particle belongs to the bottom of the ladle, the vacuum chamber or the immersion tube by means of area division algorithm, so as to accurately record the residence of the particles in each area. In the bottom area of the ladle, the residence time of the particles in the area is accumulated according to the time threshold, and the quantitative statistical data of the residence time is obtained; in the vacuum chamber area, the circulation number is counted according to the circulation path characteristics and the closed condition of the coordinate trajectory, and the number of cycles completed by the particles in the vacuum chamber is determined. All residence time data and circulation number data are converted into intuitive charts by visualization processing algorithm, wherein the particle speed distribution map represents the motion speed and direction change with different colors and vector directions, and the residence time statistics table intuitively displays the residence characteristics of the particles in each key area in a numerical form. Through the above process, not only the motion law of the ferroalloy particles in the RH refining simulation device can be quantitatively reflected, but also a visual reference basis for process optimization can be provided, further improving the practicability and intuitiveness of the monitoring method.

[0039] In a specific embodiment, the process of performing step S1 can specifically include the following steps:

[0040] Performing first-view image acquisition processing on the transparent area of the side wall of the ladle model to obtain first-view image data of the particle motion inside the ladle;

[0041] Performing second-view image acquisition processing on the front transparent wall of the vacuum chamber model to obtain second-view image data of the vertical motion of the particles in the vacuum chamber;

[0042] Performing third-view image acquisition processing on the transparent area of the side of the immersion tube model to obtain third-view image data of the flow state of the particles in the immersion tube;

[0043] Performing time synchronization processing on the first-view image data, the second-view image data and the third-view image data to obtain image sequence data containing the motion state of the ferroalloy particles.

[0044] Specifically, multi-view image acquisition is performed on the ladle model, vacuum chamber model and immersion tube model of the RH refining simulation device. A high-definition camera device is arranged on the transparent area of the side wall of the ladle model to record the first-view image of the movement process of the internal particles in the liquid environment, so that the transverse and bottom movement characteristics of the particles in the ladle can be completely captured. A second camera device is arranged on the front transparent wall of the vacuum chamber model to collect the vertical upward and downward movement of the particles in the vacuum chamber driven by air pressure. Through this view, the acceleration, deceleration and stagnation state of the particles in the circulating flow can be intuitively reflected. A third camera device is arranged on the transparent area of the side of the immersion tube model to collect the flow behavior of the particles in the pipeline during the process of entering or leaving the vacuum chamber, so that the movement mode and speed change of the particles in the narrow channel can be clearly recorded. The data of all first-view, second-view and third-view are time-stamped and synchronously processed by a unified acquisition system to ensure that each frame of image remains strictly consistent in time, thereby realizing the full-process dynamic tracking of the same particle at different spatial positions. The image data after multi-view acquisition is time-sequentially aligned and stored in a formatted manner to form image sequence data containing the complete movement state of ferroalloy particles in the ladle, vacuum chamber and immersion tube, which provides high-precision original data support for subsequent particle separation, trajectory reconstruction and movement behavior analysis, and can effectively avoid the monitoring blind area caused by single view, thereby ensuring the integrity and accuracy of the acquisition process.

[0045] For example, when performing image acquisition, a high-speed camera with a resolution of 1920x1080 can be installed on the side wall of the ladle model, with a sampling frame rate set to 200 fps, to capture the rapid movement of particles in the simulated liquid steel; another camera of the same type is installed on the front of the vacuum chamber model to capture the up-and-down flow state of the particles; a high-speed camera with a resolution of 1280x720 is installed on the side of the immersion tube model, with a frame rate set to 150 fps, to record the continuous flow of particles in the pipe. The three cameras are time-stamped by a time synchronization module to ensure spatial consistency of all data at the same time. In the image processing stage, after the original image is denoised by Gaussian filtering, the particle recognition accuracy is improved from 82% to more than 91%; in adaptive threshold segmentation, the threshold dynamic range is set to [0.15, 0.35], which can effectively eliminate false foreground caused by water flow fluctuations, reducing the false detection rate of foreground extraction by about 20%. Using Kalman filter to predict the next moment position of the particle, when the particle is blocked in 5 consecutive frames, the prediction deviation is still within 2 pixels; combined with the Mahalanobis distance matching algorithm, when the number of particles increases to 50, the trajectory correct association rate remains above 90%, effectively avoiding false association when the trajectories intersect. Using the triangulation method, the two-dimensional coordinates of the two cameras are converted into three-dimensional coordinates, with an error of ±1.5 mm; after the coordinate transformation matrix is unified, the particle trajectory across the region of the ladle and the vacuum chamber can be continuously spliced, avoiding the breakage of the trajectory caused by different coordinate systems. By time-differentiating the three-dimensional coordinate sequence, the average particle movement speed is calculated to be 0.35 m / s, with a maximum speed of 0.65 m / s, and the cycle number is 18 within 10 minutes; the average residence time at the bottom of the ladle is about 1.2 seconds, and the longest residence time is 3.5 seconds. The output speed distribution map shows that most particles are distributed in the interval of 0.3-0.4 m / s, and the particle residence time statistics table clearly reflects the particle residence law in different regions.

[0046] In a specific embodiment, the process of performing step S2 can specifically include the following steps:

[0047] Gaussian filtering and denoising are performed on the image sequence data to obtain filtered image sequence data;

[0048] Adaptive threshold segmentation is performed on the filtered image sequence data based on local image features to obtain a binary foreground target image;

[0049] The binary foreground target image is input into a background difference algorithm for foreground detection processing to obtain a foreground region marking result of the ferroalloy particles;

[0050] Centroid coordinate calculation and geometric feature extraction are performed on the foreground region marking result to obtain particle position coordinate information.

[0051] Specifically, the image sequence data is subjected to Gaussian filter denoising processing, and a weighted average value of neighboring pixels is assigned to a center pixel by using convolution operation, so as to eliminate random interference caused by water flow disturbance, uneven illumination or camera equipment noise, so that the image edge details remain clear, and the filtered image sequence is more smooth and stable. On the basis of the filtering result, the image is subjected to adaptive threshold segmentation processing based on local image features, the threshold is dynamically adjusted in different regions, the background and foreground are effectively distinguished, the foreground information of the ferroalloy particles can be accurately extracted under the environment of illumination change or water body fluctuation, and a binary foreground target image is generated. The binary image is input into a background difference algorithm for foreground detection, the pixel region which is changed is identified by frame-by-frame comparison with the modeling background, the foreground region marking result of the particles is obtained, and the multiple frame results are subjected to connected domain analysis to remove isolated noise points, so that the particle boundary detected is more complete. On the basis of the foreground region marking result, the centroid coordinate calculation is performed, the centroid position of the particle region is extracted by calculating the gray moment or geometric moment of the particle region, and the effectiveness of the particle is further confirmed in combination with the geometric features such as area, circularity and aspect ratio, so that the stable particle position coordinate information is obtained.

[0052] Taking the RH refining water model reduced in proportion as an example, clear water is injected into the ladle model and the liquid level is kept at about 650 mm, a negative pressure environment of 65 Pa is maintained through the air suction device at the top of the vacuum chamber, and gas is introduced into the side wall of the immersion tube to form a circulating flow field. In the experiment, ice blocks with a diameter of about 10 mm are used as substitutes for ferroalloy particles, and three high-speed cameras arranged on the side wall of the ladle, the front of the vacuum chamber and the side of the immersion tube are used to synchronously collect the motion state of the particles in different spatial regions. The image sequence collected is accompanied by water ripple and bubble noise interference, therefore, the image is subjected to Gaussian filter denoising processing first, so that the particle boundary is clearer, then the adaptive threshold segmentation is used to distinguish the foreground and background, the motion region of the ice block is extracted, the background difference method is further used to eliminate the false changes caused by liquid level fluctuation, the stable foreground region marking result of the particles is obtained, and finally the coordinate information of each ice block in different frames is obtained through centroid calculation and geometric feature extraction, thereby providing high-precision basic data for subsequent trajectory reconstruction and three-dimensional motion behavior analysis.

[0053] In a specific embodiment, the process of performing step S3 can specifically include the following steps:

[0054] The particle position coordinate information is input into a Kalman filter for state prediction processing, so as to obtain the predicted position coordinates of the particles in the next frame;

[0055] Based on the Mahalanobis distance calculation, the predicted position coordinates and the particle position in the current frame are subjected to data association processing, so as to obtain a particle matching relationship matrix;

[0056] According to the particle matching relationship matrix, the particle position coordinates between adjacent frames are connected to obtain a time sequence position coordinate sequence of a single particle.

[0057] The time sequence position coordinate sequence of a single particle is subjected to trajectory smoothing and coordinate storage processing to obtain trajectory coordinate data.

[0058] Specifically, the obtained particle position coordinate information is sequentially input into a Kalman filter for state prediction. The filter calculates the possible position of the particle in the next frame according to the velocity and position at the previous time, so that the continuity of the trajectory can be maintained even in the case of image noise interference or temporary occlusion of the particle. After the prediction result is generated, the Mahalanobis distance method is used to compare the predicted position with the actual position of the detected particle in the current frame. The matching degree is determined by calculating the distance between the two in a multi-dimensional feature space, thereby establishing the correspondence between the particles and avoiding false association when the number of particles is large or the motion trajectories intersect. According to the particle matching relationship matrix, the coordinates of the same particle in adjacent frames are connected one by one to form a complete time sequence position coordinate sequence, and the relationship between consecutive frames is used to ensure that the trajectory of each particle from entering the monitoring area to leaving the monitoring area will not be interrupted. During the trajectory establishment process, the time sequence coordinate sequence is subjected to trajectory smoothing processing. The weighted average or polynomial fitting method is used to remove the instantaneous jumps caused by environmental disturbances or detection errors, so that the trajectory curve is closer to the true motion law. The smoothed trajectory sequence is stored as trajectory coordinate data. Each particle corresponds to a unique trajectory number and a complete time sequence coordinate set, thereby realizing the conversion from single-frame position data to dynamic trajectory data and laying a data foundation for three-dimensional space fusion and motion behavior analysis. This processing process not only improves the robustness and accuracy of particle motion tracking, but also ensures that stable and reliable trajectory information can still be generated under complex water flow conditions, thereby ensuring a high correspondence between experimental data and actual production scenarios.

[0059] Taking the simulation experiment of ferroalloy in RH refining process as an example, a scaled water model system is built in the laboratory. The ladle model is filled with clean water and the liquid level is kept at 700 mm. The vacuum chamber model is connected to the air extraction device at the top to maintain a stable low pressure environment. The immersion tube model is connected to the air to form a circulating flow. In the experiment, ferroalloy particles with a diameter of about 10 mm are added. These particles are carried into the vacuum chamber by the water flow and continuously move up and down under the action of circulation. Three high-speed cameras are used to capture real-time images from the side wall of the ladle, the front of the vacuum chamber, and the side of the immersion tube. Continuous particle motion sequence images are obtained. After image acquisition, the images are denoised by Gaussian filtering to eliminate water waves and bubbles. Then, the adaptive threshold segmentation method is used to separate the particle foreground from the background. The background difference algorithm further eliminates false change areas, and the particle foreground marker results are extracted. Based on the marker area, the centroid coordinates are calculated and the geometric features are extracted to obtain the two-dimensional position data of the particles. These position data are input into the Kalman filter to predict the next time position. Even if the particles are temporarily obscured in the circulation path, the continuity of the trajectory can be maintained. The prediction results and the actual detection results are matched by Mahalanobis distance to establish the correspondence of the particles. The coordinates of adjacent frames are connected point by point to form the time sequence trajectory. The trajectory is smoothed to remove jitter errors, generating stable and reliable trajectory coordinate data, which provides a solid data support for the subsequent reconstruction of three-dimensional motion trajectory and analysis of alloy motion behavior. Figure 2 The figure shows the particle motion trajectory tracking algorithm flow.

[0060] In a specific embodiment, the process of performing step S4 can specifically include the following steps:

[0061] According to the camera calibration parameters, the two-dimensional coordinates in the trajectory coordinate data are converted from the pixel coordinate system to the physical coordinate system to obtain the trajectory coordinate data in the physical coordinate system;

[0062] Based on the triangulation method, the trajectory coordinate data in the physical coordinate system is subjected to multi-view coordinate fusion processing to obtain the three-dimensional position coordinates of the particles;

[0063] The three-dimensional position coordinates are input into the coordinate transformation matrix for device coordinate system processing to obtain the three-dimensional position coordinates in the device coordinate system;

[0064] The three-dimensional position coordinates in the device coordinate system are subjected to time sequence arrangement and coordinate sequence construction processing to obtain the three-dimensional space coordinate sequence.

[0065] Specifically, the trajectory coordinate data is converted according to the camera calibration parameters, the pixel coordinate points in the original image are mapped into the physical coordinate system, so that the data under different camera perspectives can be unified to the real spatial scale, and the comparability and physical meaning of each coordinate point are ensured. The physical coordinate data after conversion is fused by using the triangulation method, the three-dimensional position coordinates of the particles are obtained by calculating the intersection points of the rays under different perspectives in the three-dimensional space, so that the problem of unable to obtain depth information under single perspective is solved, and the coordinate deviation caused by partial occlusion is eliminated. After obtaining the three-dimensional position coordinates, the three-dimensional position coordinates are input into the pre-set coordinate transformation matrix, the coordinate systems corresponding to different cameras are unified into the coordinate system of the device as a whole, so that the regions of the ladle, the vacuum chamber and the immersion tube can be spliced and compared in the same reference frame. The three-dimensional coordinate data after the coordinate system unification is arranged in time sequence, so that the spatial position of each particle from entering the monitoring area to leaving the monitoring area can be continuously recorded, and the discrete points are connected into complete trajectories by constructing the time sequence coordinate sequence. The three-dimensional spatial coordinate sequence formed not only contains the change rule of the particles in the spatial position, but also reflects the continuity of the particles in the time dimension, and can provide accurate input data for subsequent speed calculation, direction analysis and residence time statistics.

[0066] For example, when calibrating the camera, a standard checkerboard calibration board is used, and the intrinsic matrix is calculated through 20 calibration images, and the mean square error of the distortion parameters is controlled within 0.35 pixels, so as to ensure that the subsequent coordinate conversion accuracy meets the requirements of three-dimensional reconstruction. In the multi-perspective triangulation fusion stage, the same particle point collected by the two cameras on the side wall of the ladle and the front of the vacuum chamber coincides in the three-dimensional space after calculation with an error of less than 2.1 mm, which indicates that the coordinate fusion accuracy can meet the demand of motion behavior analysis. The coordinate systems of the cameras are converted into a unified device coordinate system with the center of the bottom of the ladle as the origin, and after calculation by the coordinate transformation matrix, the spatial position difference of the same particle under different cameras is reduced to within 1.8 mm, thereby ensuring the continuity of the trajectory.

[0067] In a specific embodiment, the process of performing step S5 can specifically include the following steps:

[0068] Time-differentiating the three-dimensional spatial coordinate sequence to obtain three-dimensional velocity vectors and acceleration vectors of the particles;

[0069] Based on the region division algorithm, the three-dimensional velocity vectors and acceleration vectors are subjected to device region attribution judgment processing to obtain the position residence records of the particles in the bottom of the ladle, in the vacuum chamber and in the immersion tube;

[0070] According to the time threshold, the position residence records are subjected to residence time statistics and cycle number calculation processing to obtain residence time data and cycle number data of the particles in each region.

[0071] The residence time data and cycle count data are input into a chart generation algorithm for visualization data processing, resulting in a particle motion velocity distribution map and a residence time statistics table.

[0072] Specifically, the sequence of position coordinates of particles in three-dimensional space By performing time differentiation, its three-dimensional velocity vector is obtained: ,in These represent the particles in The velocity components in three directions are further differentiated in time to obtain the three-dimensional acceleration vector. ,in These represent the particles in The acceleration component in the direction. Based on a region partitioning algorithm, the three-dimensional space is divided into sub-regions such as the bottom of the ladle, the interior of the vacuum chamber, and the interior of the impregnation tube, and the particle velocity vector is determined. and acceleration vector By assigning it to different spatial regions, we obtain a record of its location within each region, that is, at any given time... The position coordinates of the particles The velocity and acceleration characteristics are matched to a specific region. Furthermore, a formula for calculating dwell time is introduced: ,in Indicates the particle in the region The cumulative time spent inside, and These represent the times of two consecutive entries into and exits from the area. This indicates the number of entries and exits within the area; it is also calculated using the following formula: The particles were obtained in the region The number of times the loop repeats within the loop. (Dwell time data) and loop count data The input chart generation algorithm is used for visualization processing, and the output charts include particle motion velocity distribution maps and regional residence time statistics tables. This visually presents the motion patterns and dynamic behaviors of particles in different regions such as the bottom of the ladle, the vacuum chamber, and the impregnation tube, thus providing quantifiable and visualized technical support for monitoring the motion behavior of ferroalloys in the RH refining process.

[0073] For example, a three-dimensional laser scanner is used to collect image data at a frequency of 50 frames per second in a ladle simulation device, obtaining the three-dimensional motion trajectory of iron alloy particles with a particle size of 10 mm in the water flow. The spatial coordinates of the particles at different time points are extracted by image processing algorithms, and the velocity and acceleration vectors of the particles in the molten steel are calculated by time differentiation method. By setting the boundaries of the ladle, vacuum chamber and immersion tube, the simulation device space is divided into three regions, and the region where the particle is located at each time is determined according to the relative relationship between the particle position and the region boundary. For example, the particle stays at the bottom of the ladle for 5 seconds, and the circulation number in the vacuum chamber is 3 times, and such data records are further analyzed to analyze the distribution and motion law of the particles. In the residence time statistics, the time threshold is set to 1 second, and the residence time of the particles in the ladle, vacuum chamber and immersion tube is counted according to the time interval. For example, in a 10-second observation time, the cumulative residence time of the particles at the bottom of the ladle is 4 seconds, the cumulative residence time in the vacuum chamber is 3 seconds, and the cumulative residence time in the immersion tube is 2 seconds. These data help to further optimize the RH refining process. The residence time and circulation number data are converted into particle motion velocity distribution diagram using chart generation algorithm, and the residence time of different regions is presented in the form of column chart, which directly shows the motion characteristics of the particles in each region. Through these visual charts, the operator can clearly see the distribution and dynamic change of the particles in different regions.

[0074] In a specific embodiment, the process of performing device region attribution judgment processing on the three-dimensional velocity vector and acceleration vector based on the region division algorithm to obtain the position residence record of the particles in the bottom of the ladle, in the vacuum chamber and in the immersion tube can specifically include the following steps:

[0075] According to the geometric boundary parameters of the ladle model, the vacuum chamber model and the immersion tube model, the device space is divided into three-dimensional regions, and the spatial boundary coordinates of the ladle region, the vacuum chamber region and the immersion tube region are obtained;

[0076] The position coordinates in the three-dimensional velocity vector and acceleration vector are matched with the spatial boundary coordinates to obtain the device region identifier to which the particle currently belongs;

[0077] The device region identifier is subjected to time sequence continuity analysis processing to obtain the transfer trajectory and region residence time record of the particles between regions;

[0078] According to the ladle bottom height threshold and the vacuum chamber circulation path characteristics, the region residence time record is classified and counted to obtain the position residence record of the particles in the bottom of the ladle, in the vacuum chamber and in the immersion tube.

[0079] Specifically, the device space is divided into a ladle region, a vacuum chamber region and an immersion tube region by three-dimensional region division processing of the geometric boundary parameters of the ladle model, the vacuum chamber model and the immersion tube model, and the space boundary coordinates of each region are obtained. The position coordinates in the three-dimensional velocity vector and the acceleration vector are matched with the space boundary coordinates, so as to determine the region attribution of the particles at different times, and the region identification of the particles is continuously updated in the process. Through the continuity analysis of the time sequence of the region identifications, the transfer trajectory of the particles between different regions can be reconstructed, and the residence time record of the particles in each region is further obtained. In combination with the height threshold condition of the ladle bottom and the circulation path characteristics in the vacuum chamber, the residence time data are classified and counted, so that the residence and distribution characteristics of the particles in the ladle bottom, the vacuum chamber and the immersion tube can be accurately obtained.

[0080] Taking the monitoring of the movement behavior of ferroalloy in the RH refining process as an example, in the construction module, the sensors are arranged on the side wall of the refining furnace and the lower part of the vacuum chamber, and the temperature and movement trajectory data of the ferroalloy particles at different depths are collected, for example, the sinking rate of the particles is about 0.15 m / s in 5 seconds, the local temperature fluctuation range is 1450-1550℃, and an initial coupling matrix is formed; the coupling weight between different depth layers is calculated by using the matrix, and the result shows that the correlation coefficient of the movement of the deep layer particles and the middle layer reaches 0.82, indicating that the energy and mass exchange between the two layers is strong; the change of the thermocline boundary is tracked in real time, and it is observed that the alloy particles are redistributed by the disturbance of the rising bubbles after 10 seconds, and the boundary moves up by about 0.3 m; the three-dimensional trajectory distribution of the particles in 20 seconds and the residence time of the particles at different depths are obtained by constraining the data reconstruction through the adaptive layered structure parameters; the trajectory, the temperature, the speed and the time sequence are fused to generate the holographic monitoring result of the movement behavior of the ferroalloy in the RH refining process, which can be used to judge the dissolution efficiency and uniformity of the alloy in the vacuum refining environment.

[0081] The above describes the method for monitoring the movement behavior of ferroalloy in the RH refining process in the embodiments of the application, and the following describes the system for monitoring the movement behavior of ferroalloy in the RH refining process in the embodiments of the application. Please refer to Figure 3 One embodiment of the system for monitoring the movement behavior of ferroalloy in the RH refining process in the embodiments of the application includes:

[0082] The extraction module is configured to collect real-time images of the ladle model, the vacuum chamber model and the immersion tube model in the RH refining simulation device, and obtain image sequence data containing the movement state of the ferroalloy particles.

[0083] The separation module is configured to analyze the background change characteristics in the image sequence data, separate the foreground region of the ferroalloy particles, and obtain the particle position coordinate information.

[0084] The analysis module is used for analyzing the change rule of the particle position coordinate information in the time sequence, establishing the continuous motion track of a single particle, and generating track coordinate data;

[0085] The fusion module is used for fusing the track coordinate data of multiple perspectives, calculating the position change of the particle in the three-dimensional space, and constructing a three-dimensional space coordinate sequence;

[0086] The statistical module is used for analyzing the speed and motion direction change of the particle in the three-dimensional space coordinate sequence, counting the residence time of the particle in the ladle bottom and the circulation number of the particle in the vacuum chamber, and outputting a particle motion speed distribution diagram and a residence time statistical table.

[0087] Through the cooperation of the above-mentioned components, the whole-process monitoring and intelligent analysis of the motion behavior of the ferroalloy particles in the RH refining process can be realized. The construction module provides the basic temperature, flow field and particle position information, the calculation module analyzes the key motion parameters, the monitoring module ensures that the state of the particles in different spatial positions is captured in real time, the fusion module uniformly models the multi-source data, and the output module finally forms intuitive charts and evaluation results. Each link is connected to each other, which not only ensures the comprehensiveness of data acquisition, but also improves the reliability of analysis conclusions, and provides a scientific basis and technical support for the optimization of the RH refining process.

[0088] The above Figure 3 The motion behavior monitoring system for the ferroalloy in the RH refining process in the embodiment of the present application is described in detail from the perspective of the modular functional entity, and the motion behavior monitoring device for the ferroalloy in the RH refining process in the embodiment of the present application is described in detail from the perspective of hardware processing.

[0089] Referring to Figure 4 In the embodiment of the present application, a motion behavior monitoring device for ferroalloy in the RH refining process is also provided, which can be a server, and its internal structure can be as follows Figure 4The processor of the computer is used to provide computing and control capabilities. The memory of the device for monitoring the motion behavior of ferrous alloy in RH refining process includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The database of the device for monitoring the motion behavior of ferrous alloy in RH refining process is used to store the corresponding data in the embodiment. The network interface of the device for monitoring the motion behavior of ferrous alloy in RH refining process is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to realize the above method.

[0090] Those skilled in the art can understand that, Figure 4 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the device for monitoring the motion behavior of ferrous alloy in RH refining process to which the scheme of the present application is applied.

[0091] The present application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium, and can also be a volatile computer readable storage medium. The computer readable storage medium stores instructions, and when the instructions are run on a computer, the computer executes the steps of the method for monitoring the motion behavior of ferrous alloy in RH refining process.

[0092] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-mentioned system, system and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0093] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a device (which can be a personal computer, a server, or a network device, etc.) for monitoring the motion behavior of ferroalloy in the RH refining process to execute all or part of the steps of the method described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0094] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring the motion behavior of ferroalloys in the RH refining process, characterized in that, The method includes: Step S1: Collect real-time images of the ladle model, vacuum chamber model, and impregnation tube model in the RH refining simulation device to obtain image sequence data containing the motion state of ferroalloy particles; Step S2: Analyze the background change features in the image sequence data, separate the foreground region of the iron alloy particles, and obtain the particle position coordinate information; Step S3: Associating the particle position coordinate information with the changing patterns over time to establish a continuous motion trajectory of a single particle and generate trajectory coordinate data includes: inputting the particle position coordinate information into a Kalman filter for state prediction processing to obtain the predicted position coordinates of the particle in the next frame; performing data association processing on the predicted position coordinates and the particle position in the current frame based on Mahalanobis distance calculation to obtain a particle matching relationship matrix; performing trajectory connection processing on the particle position coordinates between adjacent frames according to the particle matching relationship matrix to obtain a temporal position coordinate sequence of a single particle; and performing trajectory smoothing and coordinate storage processing on the temporal position coordinate sequence of the single particle to obtain the trajectory coordinate data. Step S4: Integrate the trajectory coordinate data from multiple perspectives, calculate the positional changes of the particles in three-dimensional space, and construct a three-dimensional spatial coordinate sequence; Step S5: Analyze the changes in the velocity and direction of motion of the particles in the three-dimensional spatial coordinate sequence, count the dwell time of the particles at the bottom of the ladle and the number of cycles in the vacuum chamber, and output a particle velocity distribution map and a dwell time statistics table. This includes: performing time differentiation processing on the three-dimensional spatial coordinate sequence to obtain the three-dimensional velocity vector and acceleration vector of the particles; performing device region classification processing on the three-dimensional velocity vector and acceleration vector based on a region partitioning algorithm to obtain the position dwell records of the particles at the bottom of the ladle, in the vacuum chamber, and in the impregnation tube; performing dwell time statistics and cycle count calculation processing on the position dwell records according to a time threshold to obtain the dwell time data and cycle count data of the particles in each region; inputting the dwell time data and cycle count data into a chart generation algorithm for visualization data processing to obtain the particle velocity distribution map and dwell time statistics table.

2. The method for monitoring the motion behavior of ferroalloys in the RH refining process according to claim 1, characterized in that, Step S1 includes: First-view image acquisition and processing were performed on the transparent area of ​​the side wall of the ladle model to obtain first-view image data of particle movement inside the ladle; Second-view image acquisition and processing were performed on the front transparent wall of the vacuum chamber model to obtain second-view image data of the vertical motion of particles inside the vacuum chamber. A third-view image was acquired and processed on the transparent side area of ​​the impregnated tube model to obtain third-view image data of the particle flow state inside the impregnated tube. The first-view image data, the second-view image data, and the third-view image data are time-synchronized to obtain the image sequence data containing the motion state of the iron alloy particles.

3. The method for monitoring the motion behavior of ferroalloys in the RH refining process according to claim 1, characterized in that, Step S2 includes: The image sequence data is subjected to Gaussian filtering for noise reduction to obtain filtered image sequence data; Based on local image features, adaptive threshold segmentation is performed on the filtered image sequence data to obtain a binarized foreground target image. The binarized foreground target image is input into the background subtraction algorithm for foreground detection processing to obtain the foreground region marking result of the iron alloy particles; The centroid coordinates of the foreground region marking results are calculated and geometric features are extracted to obtain the particle position coordinate information.

4. The method for monitoring the motion behavior of ferroalloys in the RH refining process according to claim 1, characterized in that, Step S4 includes: Based on the camera calibration parameters, the two-dimensional coordinates in the trajectory coordinate data are transformed from pixel coordinates to physical coordinates to obtain trajectory coordinate data in the physical coordinate system. Based on the triangulation method, the trajectory coordinate data in the physical coordinate system are processed by multi-view coordinate fusion to obtain the three-dimensional position coordinates of the particles. The three-dimensional position coordinates are input into the coordinate transformation matrix and processed in the device coordinate system to obtain the three-dimensional position coordinates in the device coordinate system. The three-dimensional position coordinates in the device coordinate system are processed by temporal arrangement and coordinate sequence construction to obtain the three-dimensional spatial coordinate sequence.

5. The method for monitoring the motion behavior of ferroalloys in the RH refining process according to claim 1, characterized in that, The region partitioning algorithm is used to determine the device region affiliation of the three-dimensional velocity vector and acceleration vector, resulting in records of particle position and residence at the bottom of the ladle, inside the vacuum chamber, and within the impregnation tube, including: The space of the device is divided into three-dimensional regions based on the geometric boundary parameters of the ladle model, vacuum chamber model, and impregnation tube model, so as to obtain the spatial boundary coordinates of the ladle region, vacuum chamber region, and impregnation tube region. The position coordinates in the three-dimensional velocity vector and acceleration vector are matched with the spatial boundary coordinates to obtain the device area identifier to which the particle currently belongs; The device area identifiers are subjected to time series continuity analysis to obtain the transfer trajectory of particles between each area and the record of the duration of residence in each area; Based on the ladle bottom height threshold and vacuum chamber circulation path characteristics, the residence time records in the area are classified and statistically processed to obtain the position and residence records of the particles at the bottom of the ladle, in the vacuum chamber, and in the impregnation tube.

6. A system for monitoring the motion behavior of ferroalloys in the RH refining process, characterized in that, For implementing the method for monitoring the motion behavior of ferroalloys in the RH refining process as described in any one of claims 1 to 5, the system for monitoring the motion behavior of ferroalloys in the RH refining process comprises: The extraction module is used to acquire real-time images of the ladle model, vacuum chamber model, and impregnation tube model in the RH refining simulation device, and obtain image sequence data containing the motion state of ferroalloy particles; The separation module is used to analyze the background change features in the image sequence data, separate the foreground region of the iron alloy particles, and obtain the particle position coordinate information; The analysis module correlates the changing patterns of the particle position coordinates over time to establish a continuous motion trajectory for a single particle and generate trajectory coordinate data. This includes: inputting the particle position coordinates into a Kalman filter for state prediction processing to obtain the predicted position coordinates of the particle in the next frame; performing data association processing on the predicted position coordinates and the particle position in the current frame based on Mahalanobis distance calculation to obtain a particle matching matrix; performing trajectory connection processing on the particle position coordinates between adjacent frames according to the particle matching matrix to obtain a temporal position coordinate sequence for a single particle; and performing trajectory smoothing and coordinate storage processing on the temporal position coordinate sequence for the single particle to obtain the trajectory coordinate data. The fusion module fuses the trajectory coordinate data from multiple perspectives, calculates the positional changes of particles in three-dimensional space, and constructs a three-dimensional spatial coordinate sequence. The statistics module is used to analyze the changes in the velocity and direction of motion of particles in the three-dimensional spatial coordinate sequence, count the residence time of particles at the bottom of the ladle and the number of cycles in the vacuum chamber, and output a particle velocity distribution map and a residence time statistics table. This includes: performing time differentiation processing on the three-dimensional spatial coordinate sequence to obtain the three-dimensional velocity vector and acceleration vector of the particles; performing device region classification processing on the three-dimensional velocity vector and acceleration vector based on a region partitioning algorithm to obtain the position residence records of particles at the bottom of the ladle, in the vacuum chamber, and in the impregnation tube; performing residence time statistics and cycle count calculation processing on the position residence records according to time thresholds to obtain the residence time data and cycle count data of particles in each region; and inputting the residence time data and cycle count data into a chart generation algorithm for visualization data processing to obtain the particle velocity distribution map and residence time statistics table.

7. A device for monitoring the motion behavior of ferroalloys during RH refining, characterized in that, The method includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the method for monitoring the motion behavior of ferroalloys in the RH refining process as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the method for monitoring the motion behavior of ferroalloys in the RH refining process as described in any one of claims 1 to 5.

Images