Automatic converter lifting and shaking method and system for converter tapping based on machine vision and medium

Abstract

The invention provides an automatic lifting and rocking method and system for converter tapping based on machine vision and a medium, and the method comprises the following steps: step 1, when judging that a slag blocking condition is met, triggering an infrared industrial camera to start to continuously collect an image sequence in a steelmaking process in real time; 2, an image preprocessing unit carries out filtering and noise reduction preprocessing on the obtained image sequence, the steel flow and the background are separated after preprocessing, and interference of a background area is eliminated; 3, an image feature extraction unit extracts feature parameters of the steel flow and calculates the proportion of the steel slag in the steel flow; whether slag stopping succeeds or not is judged through feature fusion of the slag stopping judgment unit; and 4, the prediction model in the online prediction unit receives the tapping parameters to predict the furnace lifting opportunity, a prediction result is transmitted to the converter control unit, and furnace shaking in advance when slag stopping fails or furnace lifting operation when slag stopping succeeds is automatically triggered. Real-time monitoring and automatic judgment in the slag stopping process are achieved, and the accuracy and reliability of judgment are improved. Meanwhile, the furnace lifting opportunity is dynamically optimized according to the slag stopping result, the steel making efficiency is improved while the molten steel quality is guaranteed, and unnecessary energy consumption and material waste are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of steelmaking, and in particular, relates to a method, system and medium for automatically lifting and shaking a converter for steel tapping based on machine vision. Background Art

[0002] Steelmaking is an important process in the steel industry. In the steelmaking process, slag blocking and furnace lifting are two key links, which play a decisive role in the quality of molten steel and production efficiency. Slag blocking refers to the use of slag cones or other devices to block the steel flow during the process of pouring molten steel from the converter into the ladle to prevent slag from entering the ladle, thereby ensuring the purity of the molten steel. The success or failure of slag blocking directly affects the quality of molten steel. Furnace lifting refers to lifting the converter mouth after steelmaking is completed to prevent slag from flowing into the ladle with the molten steel. The timing of furnace lifting is crucial. If the furnace is lifted too late, as the molten steel level drops, some slag will be drawn into the molten steel. These mixed slags will seriously affect the purity of the molten steel, introduce a large number of non-metallic inclusions, and reduce the mechanical properties, toughness, and fatigue resistance of the finished steel. If the furnace is lifted too early, although the mixing of slag can be avoided, it also means that a considerable part of the qualified molten steel is not fully utilized, and these molten steels usually have to be returned to the converter for re-oxidation, which not only prolongs the steelmaking cycle, but also consumes a lot of energy and raw materials. The ideal time to lift the furnace should be at the last moment before the slag is mixed in, which can maximize the utilization of the molten steel in the converter and ensure the cleanliness of the molten steel in the ladle.

[0003] Traditional slag blocking and furnace lifting control mainly rely on the experience and judgment of operators. Operators observe phenomena such as steel flow and decide the timing of furnace lifting based on experience. This method has the following problems: it is highly subjective and limited by the experience and skills of operators. The judgment results vary from person to person and lack objective standards. Manual observation is difficult to capture the instantaneous changes of steel flow, and the judgment is delayed, affecting the timeliness of process control. The harsh working environment poses a threat to the health of operators and also affects their judgment.

[0004] The Chinese patent with the patent publication number CN112458231A discloses a method and system for detecting slag in a converter, including: a laser distance measuring device measures the distance from the ladle car to the center of the converter in real time; a slag analysis device calculates the location of the steel flow according to the rotation angle, and controls the ladle car to move to the location of the steel flow according to the location of the steel flow and the distance from the ladle car to the center of the converter; an infrared thermal imager device collects steel flow images in real time; a slag analysis device processes the steel flow images and calculates the slag content; when the slag content exceeds the abnormal threshold, the slag blocking device is controlled to close the steel outlet, and the converter is controlled to rotate back to the initial angle. This method mainly uses automation equipment and image processing technology to monitor the slag content in the flow of molten steel, and controls the opening of the steel outlet and the rotation of the converter through these data. However, this technology relies on the pixel ratio of slag in the steel flow image as a basis for judgment. This method is potentially affected by image acquisition conditions, such as lighting changes, the accuracy of the laser distance measuring device, the performance of the infrared thermal imager and the stability of its interval with the on-site environment. The pixel ratio of slag may cause errors due to changes in camera angle, vibration of converter body, dust accumulation of laser rangefinder or infrared thermal imager during steelmaking process, and other external factors. The accuracy of image processing algorithm is limited by the set threshold. Once the threshold is improperly selected, it will directly affect the result of slag detection. At the same time, it is difficult to adapt to the dynamic changes of slag content caused by changes in steelmaking process if the threshold is fixed.

[0005] The Chinese patent with the patent publication number CN110438284A discloses an intelligent steel-tapping device and control method for a converter, including a secondary control system for converter steelmaking, a converter slag detection system, a steel-tapping slag blocking system, a furnace mouth monitoring module, a steel-tapping monitoring module, a communication control module, and a deep learning host. However, first of all, in terms of image acquisition and preprocessing technology, this method cannot accurately extract image sequences and perform accurate analysis during the steelmaking process, and a large amount of slag will enter the iron ladle during the actual production process. Secondly, in terms of accuracy and reliability, this method uses a conventional learning computer host with integrated multi-channel GPUs to improve the accuracy of target state recognition in the image through a neural network model, but the accuracy is not high and the judgment is not accurate enough. Finally, in terms of prediction models and control strategies, the scheme does not have a prediction model and automatic control logic based on steel-tapping parameters for prediction and judgment, and the prediction and judgment process is too primitive. Summary of the invention

[0006] In view of the defects in the prior art, the purpose of the present invention is to provide a method, system and medium for automatically lifting and shaking a converter for steel tapping based on machine vision.

[0007] According to the present invention, a method for automatically lifting and shaking a converter for steel tapping based on machine vision comprises:

[0008] Step 1: The position sensor L1 is used to obtain the converter angle data in real time. When the slag blocking condition is met, the infrared industrial camera is triggered to start real-time continuous acquisition of the image sequence of the steelmaking process;

[0009] Step 2: The image preprocessing unit performs filtering and noise reduction preprocessing on the acquired image sequence. After preprocessing, the steel flow is separated from the background by a method based on a combination of region growing and threshold segmentation to eliminate interference from the background area;

[0010] Step 3: The image feature extraction unit extracts the characteristic parameters of the steel flow and calculates the proportion of slag in the steel flow; the slag blocking determination unit determines whether the slag blocking is successful through feature fusion;

[0011] Step 4: The prediction model in the online prediction unit accepts the steel-tapping parameters to predict the timing of lifting the furnace, and transmits the prediction results to the converter control unit, which automatically triggers the operation of shaking the furnace in advance if the slag blocking fails or lifting the furnace if the slag blocking succeeds.

[0012] Preferably, step 1 comprises:

[0013] Step 1.1: The position sensor L1 collects the converter angle data θ(t) in real time and transmits the data to the control system;

[0014] Step 1.2: The control system determines whether the slag blocking condition is met according to the preset steel tapping starting angle threshold θ0;

[0015] The judgment condition is: θ(t)≥θ0, which means the slag blocking process starts;

[0016] Step 1.3: After the slag blocking conditions are met, the control system sends a trigger signal to the infrared industrial camera to start continuously collecting the image sequence of the converter steel tapping process.

[0017] Preferably, step 2 comprises:

[0018] Step 2.1: Use Gaussian filter to smooth the image and reduce noise interference; the kernel function of Gaussian filter is:

[0019]

[0020] Among them, σ is the standard deviation of the Gaussian kernel, which controls the degree of smoothing;

[0021] Filtered image I g (t i ) is: I g (t i )=I(t i )*G(x,y);

[0022] Among them, * represents the convolution operation;

[0023] Step 2.2: After preprocessing the image I g (t i ), selecting one or more pixel points in the steel flow area as seed points;

[0024] Starting from the seed point, the adjacent pixel points are included in the steel flow area; let the current pixel point be (x, y), and its gray value is I p (x, y), and compare it with the gray value I of the seed point p (x0, y0) for comparison, if it satisfies:

[0025] |I p (x, y)-I p (x0,y0)|≤T g

[0026] Then (x, y) is included in the steel flow area, and this process is repeated until no new pixel points are added;

[0027] Among them, T g is the growth threshold, which controls the grayscale similarity of the steel flow area;

[0028] Step 2.3: Perform threshold segmentation on the grown steel flow area, extract the precise boundary of the steel flow, and define the threshold T s The formula is:

[0029]

[0030] Among them, the average gray value of the steel flow area is μ s , the average gray value of the background area is μ b ;

[0031] The gray value is greater than T s The pixel points are marked as steel flow, and the rest are marked as background, and the binary steel flow mask M is obtained. s (t i ):

[0032]

[0033] Step 2.4: Steel Flow Mask M s (t i ) to perform morphological processing, fill holes, and smooth boundaries. The formula is:

[0034]

[0035] in, and denote expansion and erosion operations respectively, and S is the structural element;

[0036] Step 2.5: Apply the morphologically processed steel flow mask M s ′(ti ) is applied to the original image I(t i ), and obtain the separated steel flow image I s (t i ):

[0037]

[0038] Preferably, step 3 comprises:

[0039] Step 3.1: Extract the characteristic parameter velocity v(t i ), flow q(t i ) and shape characteristics f(t i );

[0040] Step 3.2: Use weighted average fusion to extract multiple feature parameters to determine whether slag blocking is successful;

[0041] Step 3.3: According to the grayscale difference between molten steel and slag in the infrared image, set the grayscale threshold g0;

[0042] According to the grayscale threshold g0, the steel flow area image I after the preprocessing image s (t i ) is binarized to obtain the regions of molten steel and slag;

[0043] Traverse each pixel point (x, y) of the image, extract its gray value g(x, y), and compare the gray value g(x, y) with the threshold g0: if g(x, y)>g0, then mark the pixel point as a molten steel pixel, otherwise mark it as a slag pixel. After marking all the pixels, the molten steel area R is obtained. s and slag area R g A binary image of

[0044] Count the number of molten steel pixels n respectively s (t i ) and the number of slag pixels n g (t i ), according to the number of molten steel pixels n s (t i ) and the number of slag pixels n g (t i ), calculate the slag ratio r(t i ):

[0045] Among them, r(t i ) indicates that at the i-th time t i The slag ratio ranges from [0,1].

[0046] Preferably, step 3.1 comprises:

[0047] Step 3.1.1: Calculate the flow velocity v(t i ), for two adjacent frames of image I(t i ) and I(t i+1 ) to grayscale the image and calculate the time derivative I of the grayscale image t and spatial derivative I x ,I y , according to the optical flow constraint equation I x u+I y v+I t = 0, combined with global or local optimization methods, solve the velocity vector (u, v), use median filtering on the velocity vector to obtain a smooth velocity field, extract the velocity vector of the steel flow area, and calculate the average velocity:

[0048]

[0049] Among them, n is the number of pixels in the steel flow area;

[0050] Step 3.1.2: Analyze the pixel area of ​​the steel flow area to estimate the flow rate q(t i );

[0051] For image I(t i ) is binarized to extract the binary mask B(t i ), perform morphological processing on the binary mask to remove holes and burrs, and count the number of pixels n with a pixel value of 1 in the binary mask s (t i ), estimate the cross-sectional area of ​​the steel flow based on the number of pixels and pixel size:

[0052] S(t i )=n s (t i )*s p

[0053] Among them, s p is the physical size of a single pixel, the volume flow rate q(t i )=v(t i )*S(t i ), * indicates a multiplication operation;

[0054] Step 3.1.3: Extract the shape features of the steel flow f(t i );

[0055] For the binary mask B(t i ) to perform connected domain analysis and extract connected blocks in the steel flow area; for each connected block, calculate the circularity C = 4πA / P by its shape parameters area A and perimeter P 2; Calculate the aspect ratio R = L / W, where L and W are the length and width of the minimum circumscribed rectangle of the connected block respectively; Quantify the shape feature into a scalar value:

[0056] f(t i )=A*C / R.

[0057] Preferably, step 3.2 comprises:

[0058] Normalize each characteristic parameter to make its value range consistent;

[0059] Then, according to the importance and credibility of each feature parameter, the weight coefficient w is set v , w q , w f ;

[0060] Calculate the comprehensive eigenvalue F(t i )=w v *v(t i )+w q *q(t i )+w f *f(t i ); Set the threshold F0, when F(t i )≥F0, it is judged that the slag blocking is successful; when F(t i )<F0, it is judged as slag blocking failure.

[0061] Preferably, step 4 comprises:

[0062] Step 4.1: Select the prediction features for the timing of lifting the furnace: slag blocking judgment result D, slag proportion r(t i ), molten steel weight m, steel tapping history time t h , molten steel flow rate v(t i ), converter parameters p(t i ), environmental factors e(t i ) features, combined into a feature vector x(t i )=[D,r(t i ), m, t h , v(t i ), p(t i ), e(t i )], as the input of the prediction model;

[0063] Step 4.2: Deploy the trained prediction model in the online prediction unit and receive the feature vector x(t i ), calculate the predicted value of the furnace lifting opportunity

[0064] Step 4.3: Prediction results It is transmitted to the converter control unit for automatic operation of lifting and shaking the converter.

[0065] Preferably, the image sequence in step 1 is represented as I(t i )|i=1,2,...,n;

[0066] Among them, t i Indicates the acquisition time of the i-th frame image, and monitors the changes in the steel flow state in real time.

[0067] A converter steel-tapping automatic furnace lifting and shaking system based on machine vision, comprising:

[0068] Module 1: used to obtain the converter angle data in real time through the position sensor L1. When the slag blocking conditions are met, the infrared industrial camera is triggered to start real-time continuous acquisition of the image sequence of the steelmaking process;

[0069] Module 2: The image preprocessing unit performs filtering and noise reduction preprocessing on the acquired image sequence. After preprocessing, the steel flow is separated from the background by a method based on a combination of region growing and threshold segmentation to eliminate interference from the background area;

[0070] Module 3: The image feature extraction unit is used to extract the characteristic parameters of the steel flow and calculate the proportion of slag in the steel flow; the slag blocking judgment unit is used to determine whether the slag blocking is successful through feature fusion;

[0071] Module 4: The prediction model used in the online prediction unit accepts the steel-making parameters to predict the timing of lifting the furnace, and transmits the prediction results to the converter control unit, automatically triggering the operation of shaking the furnace in advance if the slag blocking fails or lifting the furnace if the slag blocking succeeds.

[0072] A computer-readable storage medium storing a computer program, characterized in that when the computer program is executed by a processor, the steps of a method for automatically lifting and shaking a converter for steel tapping based on machine vision are implemented.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] 1. The slag blocking situation during the converter steel-making process is monitored in real time by infrared industrial cameras, and the state changes of the steel flow are analyzed by specific image processing algorithms, which can accurately determine whether the slag blocking is successful with high accuracy;

[0075] 2. By adopting a method based on infrared visual monitoring and machine vision intelligent analysis, the problem of traditional slag blocking judgment relying on operator experience, subjectivity and uncertainty is solved, and real-time monitoring and automatic judgment of the slag blocking process are realized, which improves the accuracy and reliability of judgment;

[0076] 3. Based on the slag blocking results, the system can automatically determine the best time to lift or shake the furnace, effectively avoiding the problem of molten steel quality degradation and waste, and improving steelmaking efficiency;

[0077] 4. Through the automation and intelligence of slag blocking judgment and furnace lifting timing selection, the operational complexity and risk are reduced, and the safety and reliability of the steelmaking process are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0079] Figure 1 This is a flow chart of the automatic furnace lifting method for converter steel tapping based on machine vision;

[0080] Figure 2 This is a schematic diagram of the structure of the automatic lifting system for converter steel-tapping based on machine vision;

[0081] Figure 3 This is the effect diagram of steel flow area segmentation after successful slag blocking;

[0082] Figure 4 This is the effect diagram of steel flow area segmentation when slag blocking fails. DETAILED DESCRIPTION

[0083] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0084] The invention provides a method, system and medium for automatically lifting and shaking a converter for steel tapping based on machine vision.

[0085] like Figure 1 As shown, a method for automatically lifting and shaking a converter for steel tapping based on machine vision comprises:

[0086] Step 1: Obtain the converter angle data in real time through the position sensor L1. When the converter rotates to the preset angle, it indicates that the steel tapping process has begun and the slag blocking conditions are met, triggering the infrared industrial camera to start capturing the image sequence I(t i )|i=1,2,...,n,monitor the state of the steel flow. Among them, t i Indicates the acquisition time of the i-th frame image, and monitors the changes in the steel flow state in real time.

[0087] Step 1.1: The position sensor L1 collects the converter angle data θ(t) in real time and transmits the data to the control system.

[0088] Step 1.2: The control system determines whether the slag blocking condition is met according to the preset steel tapping starting angle threshold θ0. The judgment condition is expressed as: θ(t) ≥ θ0. When the converter angle θ(t) is greater than or equal to the preset threshold θ0, it is determined that the slag blocking process has started.

[0089] Step 1.3: After the slag blocking conditions are met, the control system sends a trigger signal to the infrared industrial camera to start continuously collecting images of the converter steel tapping process.

[0090] Step 2: The image preprocessing unit performs filtering and noise reduction preprocessing on the acquired image sequence. After preprocessing, a method based on combining region growing and threshold segmentation is used to separate the steel flow from the background, eliminate interference from the background area, and effectively improve the recognition accuracy.

[0091] Step 2.1: Infrared images usually contain high-frequency noise. Use a Gaussian filter to smooth the image and reduce noise interference. The kernel function of the Gaussian filter is:

[0092] Among them, σ is the standard deviation of the Gaussian kernel, which controls the degree of smoothing.

[0093] Filtered image I g (t i ) is: I g (t i )=I(t i )*G(x,y).

[0094] Among them, * represents the convolution operation.

[0095] Step 2.2: After preprocessing the image I g (t i ), select one or more pixels in the steel flow area as seed points. The selection of seed points can be determined based on prior knowledge (such as the location of the steel outlet) or automatic detection (the point with the maximum brightness). Starting from the seed point, the adjacent pixels are included in the steel flow area. Let the current pixel be (x, y), and its gray value is I p (x, y), and compare it with the gray value I of the seed point p (x0, y0) is compared, and if the following conditions are met, (x, y) is included in the steel flow area: p (x, y)-I p (x0,y0)|≤T g . Repeat this process until no new pixels are added.

[0096] Among them, T g is the growth threshold, which controls the grayscale similarity of the steel flow area.

[0097] Step 2.3: Perform threshold segmentation on the steel flow area after growth to extract the precise boundary of the steel flow. Assume that the average gray value of the steel flow area is μ s , the average gray value of the background area is μ b , define the threshold T s :

[0098]

[0099] The gray value is greater than T s The pixel points are marked as steel flow, and the rest are marked as background, and the binary steel flow mask M is obtained. s (t i ):

[0100]

[0101] Step 2.4: Use morphological processing to mask the steel flow M s (t i ) performs morphological processing to fill holes and smooth boundaries. in, and They represent dilation and erosion operations respectively, and S is the structural element.

[0102] Step 2.5: Apply the morphologically processed steel flow mask M s ′(t i ) is applied to the original image I(t i ), and obtain the separated steel flow image I s (t i ):

[0103]

[0104] Step 3: The image feature extraction unit extracts the characteristic parameters of the steel flow and calculates the proportion of slag in the steel flow, and then determines whether the slag blocking is successful through feature fusion of the slag blocking judgment unit.

[0105] Step 3.1: Extract the characteristic parameter velocity v(t i ), flow q(t i ) and shape characteristics f(t i ).

[0106] Step 3.1.1: Apply the optical flow method to calculate the velocity v(t i ). For two adjacent frames of image I(t i ) and I(t i+1 ) to grayscale the image and calculate the time derivative I of the grayscale image t and spatial derivative I x , I y , according to the optical flow constraint equation Ix u+I y v+I t = 0, combined with global or local optimization methods, solve the velocity vector (u, v), use median filtering on the velocity vector to obtain a smooth velocity field, extract the velocity vector of the steel flow area, and calculate the average velocity size Where n is the number of pixels in the steel flow area.

[0107] Step 3.1.2: Analyze the pixel area of ​​the steel flow area to estimate the flow rate q(t i ). For image I(t i ) is binarized to extract the binary mask B(t i ), perform morphological processing on the binary mask to remove holes and burrs, and count the number of pixels n with a pixel value of 1 in the binary mask s (t i ), and estimate the steel flow cross-sectional area S(t i )=n s (t i )*s p , where s p is the physical size of a single pixel, the volume flow rate q(t i )=v(t i )*S(t i );

[0108] Among them, * represents the multiplication operation.

[0109] Step 3.1.3: Extract the shape features of the steel flow f(t i ). Before and after the slag blocking is successful, the molten steel will present different shapes when flowing. If the slag blocking is unsuccessful, the steel flow is more regular and continuous, while if the slag blocking is successful, the steel flow is more scattered and intermittent. The difference can be distinguished by analyzing the shape parameters of the steel flow area. First, the binary mask B(t i ) to perform connected domain analysis and extract connected blocks in the steel flow area. For each connected block, calculate the circularity C = 4πA / P by its shape parameters area A and perimeter P. 2 ; Calculate the aspect ratio R = L / W, where L and W are the length and width of the minimum circumscribed rectangle of the connected block respectively; Quantify the shape feature into a scalar value f(t i )=A*C / R.

[0110] Step 3.2: Use weighted average fusion to extract multiple feature parameters to determine whether slag blocking is successful.

[0111] Normalize each characteristic parameter to make its value range consistent;

[0112] Then, according to the importance and credibility of each feature parameter, the weight coefficient w is setv , w q , w f ;

[0113] Calculate the comprehensive eigenvalue F(t i )=w v *v(t i )+w q *q(t i )+w f *f(t i ); Set the threshold F0, when F(t i )≥F0, it is judged that the slag blocking is successful; when F(t i )<F0, it is judged as slag blocking failure.

[0114] Step 3.3: According to the grayscale difference between molten steel and slag in the infrared image, set the grayscale threshold g0. Use the grayscale threshold g0 to calculate the steel flow area image I after the preprocessing image. s (t i ) is binarized to obtain the areas of molten steel and slag. Traverse each pixel point (x, y) of the image, extract its grayscale value g(x, y), and compare the grayscale value g(x, y) with the threshold g0: if g(x, y)>g0, then mark the pixel point as a molten steel pixel, otherwise mark it as a slag pixel. After marking all pixels, the molten steel area R is obtained. s and slag area R g Binary image. Count the number of molten steel pixels n s (t i ) and the number of slag pixels n g (t i ), according to the number of molten steel pixels n s (t i ) and the number of slag pixels n g (t i ), calculate the slag ratio r(t i ): Among them, r(t i ) indicates that at the i-th time t i The slag ratio ranges from [0, 1].

[0115] Step 4: The prediction model in the online prediction unit accepts the steel-tapping parameters to predict the timing of lifting the furnace, and transmits the prediction results to the converter control unit, which automatically triggers the operation of shaking the furnace in advance if the slag blocking fails or lifting the furnace if the slag blocking succeeds. Figure 3 and Figure 4 This is the effect diagram of steel flow area segmentation for successful and failed slag blocking.

[0116] Step 4.1: Based on historical data and expert experience, select the features that have an important impact on the prediction of the timing of lifting the furnace: slag blocking judgment result D, steel slag proportion r(ti ), molten steel weight m, steel tapping history time t h , molten steel flow rate v(t i ), converter parameters p(t i ), environmental factors e(t i ) and other features are combined into a feature vector x(t i )=[D,r(t i ), m, t h , v(t i ), p(t i ), e(t i )] as the input of the prediction model.

[0117] Step 4.2: The trained prediction model is deployed in the online prediction unit, which receives the feature vector x(t i ), calculate the predicted value of the furnace lifting opportunity

[0118] Step 4.3: Prediction results The data are transmitted to the converter control unit for automatic operation of lifting and shaking the converter.

[0119] The present invention also provides an automatic furnace lifting and shaking system for converter steel tapping based on machine vision. The automatic furnace lifting and shaking system for converter steel tapping based on machine vision can be realized by executing the process steps of the automatic furnace lifting and shaking method for converter steel tapping based on machine vision, that is, those skilled in the art can understand the automatic furnace lifting and shaking method for converter steel tapping based on machine vision as a preferred implementation of the automatic furnace lifting and shaking system for converter steel tapping based on machine vision.

[0120] To facilitate understanding of the system structure, such as Figure 2 The figure shows the structure diagram of the automatic lifting system for converter steel-tapping based on machine vision, which shows the layout relationship of the main components such as converter, ladle, infrared camera, etc. in the form of a side view.

[0121] The automatic furnace lifting and shaking system for converter steel tapping based on machine vision includes:

[0122] Module 1: used to obtain the converter angle data in real time through the position sensor L1. When the slag blocking conditions are met, the infrared industrial camera is triggered to start real-time continuous acquisition of the image sequence of the steelmaking process;

[0123] Module 2: The image preprocessing unit performs filtering and noise reduction preprocessing on the acquired image sequence. After preprocessing, the steel flow is separated from the background by a method based on a combination of region growing and threshold segmentation to eliminate interference from the background area;

[0124] Module 3: The image feature extraction unit is used to extract the characteristic parameters of the steel flow and calculate the proportion of slag in the steel flow; the slag blocking judgment unit is used to determine whether the slag blocking is successful through feature fusion;

[0125] Module 4: The prediction model used in the online prediction unit accepts the steel-making parameters to predict the timing of lifting the furnace, and transmits the prediction results to the converter control unit, automatically triggering the operation of shaking the furnace in advance if the slag blocking fails or lifting the furnace if the slag blocking succeeds.

[0126] A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the steps of a method for automatically lifting and shaking a converter for steel tapping based on machine vision are implemented.

[0127] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0128] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for automatically lifting and shaking a converter for steel tapping based on machine vision, characterized in that: include: Step 1: The position sensor L1 is used to obtain the converter angle data in real time. When the slag blocking condition is met, the infrared industrial camera is triggered to start real-time continuous acquisition of the image sequence of the steelmaking process; Step 2: The image preprocessing unit performs filtering and noise reduction preprocessing on the acquired image sequence. After preprocessing, the steel flow is separated from the background by a method based on a combination of region growing and threshold segmentation to eliminate interference from the background area; Step 3: The image feature extraction unit extracts the characteristic parameters of the steel flow and calculates the proportion of slag in the steel flow; the slag blocking determination unit determines whether the slag blocking is successful through feature fusion; Step 4: The prediction model in the online prediction unit accepts the steel-tapping parameters to predict the timing of lifting the furnace, and transmits the prediction results to the converter control unit, which automatically triggers the operation of shaking the furnace in advance if the slag blocking fails or lifting the furnace if the slag blocking succeeds.

2. The automatic furnace lifting and shaking method for converter steel tapping based on machine vision according to claim 1 is characterized in that: The step 1 comprises: Step 1.1: The position sensor L1 collects the converter angle data θ(t) in real time and transmits the data to the control system; Step 1.2: The control system determines whether the slag blocking condition is met according to the preset steel tapping starting angle threshold θ0; The judgment condition is: θ(t)≥θ0, which means the slag blocking process starts; Step 1.3: After the slag blocking conditions are met, the control system sends a trigger signal to the infrared industrial camera to start continuously collecting the image sequence of the converter steel tapping process.

3. The automatic furnace lifting and shaking method for converter steel tapping based on machine vision according to claim 1 is characterized in that: The step 2 comprises: Step 2.1: Use Gaussian filter to smooth the image and reduce noise interference; the kernel function of Gaussian filter is: Among them, σ is the standard deviation of the Gaussian kernel, which controls the degree of smoothing; Filtered image I g (t i ) is: I g (t i )=I(t i )*G(x,y); Among them, * represents the convolution operation; Step 2.2: After preprocessing the image I g (t i ), selecting one or more pixel points in the steel flow area as seed points; Starting from the seed point, the adjacent pixels are included in the steel flow area; let the current pixel be (x, y), and its gray value is I p (x, y), and compare it with the gray value I of the seed point p (x0, y0) for comparison, if it satisfies: |I p (x,y)-I p (x0,y0)|≤T g Then (x, y) is included in the steel flow area, and this process is repeated until no new pixel points are added; Among them, T g is the growth threshold, which controls the grayscale similarity of the steel flow area; Step 2.3: Perform threshold segmentation on the grown steel flow area, extract the precise boundary of the steel flow, and define the threshold T s The formula is: Among them, the average gray value of the steel flow area is μ s , the average gray value of the background area is μ b ; The gray value is greater than T s The pixel points are marked as steel flow, and the rest are marked as background, and the binary steel flow mask M is obtained. s (t i ): Step 2.4: Steel Flow Mask M s (t i ) to perform morphological processing, fill holes, and smooth boundaries. The formula is: in, and denote expansion and erosion operations respectively, and S is the structural element; Step 2.5: Apply the morphologically processed steel flow mask M s ′(t i ) is applied to the original image I(t i ), and obtain the separated steel flow image I s (t i ):

4. The automatic furnace lifting and shaking method for converter steel tapping based on machine vision according to claim 1 is characterized in that: The step 3 comprises: Step 3.1: Extract the characteristic parameter velocity v(t o ), flow q(t i ) and shape characteristics f(t o ); Step 3.2: Use weighted average fusion to extract multiple feature parameters to determine whether slag blocking is successful; Step 3.3: According to the grayscale difference between molten steel and slag in the infrared image, set the grayscale threshold g0; According to the grayscale threshold g0, the steel flow area image I after preprocessing the image s (t i ) is binarized to obtain the regions of molten steel and slag; Traverse each pixel point (x, y) of the image, extract its gray value g(x, y), and compare the gray value g(x, y) with the threshold g0: if g(x, y)>g0, then mark the pixel point as a molten steel pixel, otherwise mark it as a slag pixel. After marking all the pixels, the molten steel area R is obtained. s and slag area R g A binary image of Count the number of molten steel pixels n respectively s (t i ) and the number of slag pixels n g (t i ), according to the number of molten steel pixels n s (t i ) and the number of slag pixels n g (t i ), calculate the slag ratio r(t i ): Among them, r(t i ) indicates that at the i-th time t i The slag ratio ranges from [0,1].

5. The automatic furnace lifting and shaking method for converter steel tapping based on machine vision according to claim 1 is characterized in that: Step 3.1 shown includes: Step 3.1.1: Calculate the flow velocity v(t i ), for two adjacent frames of image I(t i ) and I(t i+1 ) to grayscale the image and calculate the time derivative I of the grayscale image t and spatial derivative I x ,I y , according to the optical flow constraint equation I x u+I y v+I t = 0, combined with global or local optimization methods, solve the velocity vector (u, v), use median filtering on the velocity vector to obtain a smooth velocity field, extract the velocity vector of the steel flow area, and calculate the average velocity: Among them, n is the number of pixels in the steel flow area; Step 3.1.2: Analyze the pixel area of ​​the steel flow area to estimate the flow rate q(t i ); For image I(t i ) is binarized to extract the binary mask B(t i ), perform morphological processing on the binary mask to remove holes and burrs, and count the number of pixels n with a pixel value of 1 in the binary mask s (t i ), estimate the cross-sectional area of ​​the steel flow based on the number of pixels and pixel size: S(t i )=n s (t i )*s p Among them, s p is the physical size of a single pixel, the volume flow rate q(t i )=v(t i )*S(t i ), * indicates a multiplication operation; Step 3.1.3: Extract the shape features of the steel flow f(t i ); For the binary mask B(t i ) to perform connected domain analysis and extract connected blocks in the steel flow area; for each connected block, calculate the circularity X = 4πA / P by its shape parameters area A and perimeter P 2 ; Calculate the aspect ratio R = L / W, where L and W are the length and width of the minimum circumscribed rectangle of the connected block respectively; Quantify the shape feature into a scalar value: f(t i )=A*C / R。 6. The automatic furnace lifting and shaking method for converter steel tapping based on machine vision according to claim 1 is characterized in that: The step 3.2 comprises: Normalize each characteristic parameter to make its value range consistent; Then, according to the importance and credibility of each feature parameter, the weight coefficient w is set v ,w q ,w f ; Calculate the comprehensive eigenvalue F(t i ) = w v * v(t i ) + w q * q(t i ) + w f * f(t i ) ; Set the threshold F0. When F(t i ) ≥ F0, it is judged that the slag blocking is successful; when F(t i ) < F0, it is judged that the slag blocking fails.

7. The automatic furnace lifting and shaking method for converter steel tapping based on machine vision according to claim 1 is characterized in that: The step 4 comprises: Step 4.1: Select the prediction features for the timing of lifting the furnace: slag blocking judgment result D, slag proportion r(t i ), molten steel weight m, steel tapping history time t h , molten steel flow rate v(t i ), converter parameters p(t i ), environmental factors e(t i ) features, combined into a feature vector x(t i )=[D,r(t i ),m,t h ,v(t i ),p(t i ),e(t i )], as the input of the prediction model; Step 4.2: Deploy the trained prediction model in the online prediction unit and receive the feature vector x(t i ), calculate the predicted value of the furnace lifting opportunity Step 4.3: Prediction results It is transmitted to the converter control unit for automatic operation of lifting and shaking the converter.

8. The automatic furnace lifting and shaking method for converter steel tapping based on machine vision according to claim 1 is characterized in that: The image sequence in step 1 is represented as I(t i )|i=1,2,…,n; Among them, t i Indicates the acquisition time of the i-th frame image, and monitors the changes in the steel flow state in real time.

9. A converter steel tapping automatic furnace lifting and shaking system based on machine vision, characterized in that: include: Module 1: used to obtain the converter angle data in real time through the position sensor L1. When the slag blocking conditions are met, the infrared industrial camera is triggered to start real-time continuous acquisition of the image sequence of the steelmaking process; Module 2: The image preprocessing unit performs filtering and noise reduction preprocessing on the acquired image sequence. After preprocessing, the steel flow is separated from the background by a method based on a combination of region growing and threshold segmentation to eliminate interference from the background area; Module 3: The image feature extraction unit is used to extract the characteristic parameters of the steel flow and calculate the proportion of slag in the steel flow; the slag blocking judgment unit is used to determine whether the slag blocking is successful through feature fusion; Module 4: The prediction model used in the online prediction unit accepts the steel-making parameters to predict the timing of lifting the furnace, and transmits the prediction results to the converter control unit, automatically triggering the operation of shaking the furnace in advance if the slag blocking fails or lifting the furnace if the slag blocking succeeds.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for automatically lifting and shaking a converter for steel tapping based on machine vision according to any one of claims 1 to 8 are implemented.

Images