Method, device, electronic device and storage medium for determining the end point of converter smelting

Abstract

An embodiment of the present application provides a method, device, electronic device and storage medium for determining the end point of converter smelting, belonging to the field of artificial intelligence technology. To solve the problem that the accuracy of the current converter blowing end point is relatively low, the following solution is provided: obtaining the current furnace mouth flame feature set according to the initial flame image; determining the predicted mid- and late-stage conversion point of converter smelting according to the current furnace mouth flame feature set through a modified LSTM prediction model; matching the predicted flame features at the mid- and late-stage conversion point with the converter smelting conversion point flame feature set; if the predicted flame features at the mid- and late-stage conversion point match the converter smelting conversion point flame feature set, it is determined that the converter blowing enters the late stage; determining the converter blowing end point according to the carbon content at the mid- and late-stage conversion point of converter blowing through the converter blowing late-stage carbon model. In this way, based on the combination of flame image recognition of the mid- and late-stage conversion point of converter blowing and the converter blowing late-stage carbon model, the accuracy of determining the converter blowing end point is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to a method, device, electronic device and storage medium for determining the end point of converter smelting. Background Art

[0002] With the continuous development of the iron and steel industry, the production requirements for each production process in the iron and steel industry are also getting higher and higher. At present, the long-process steelmaking process requires the use of converters to melt molten steel. The carbon content of the molten steel is an important parameter for judging whether the molten steel is qualified. Different steel grades have strict carbon content requirements. When the judgment of the carbon content at the end point of smelting is inappropriate, secondary or multiple reblows are required, which affects production efficiency.

[0003] In the prior art, an artificial intelligence prediction model is used to judge the temperature of the molten steel in the converter based on the converter flame color. The existing artificial intelligence prediction model calibrates and trains each temperature flame feature through machine learning to form a data set, and judges the end point of converter blowing by judging the temperature of the molten steel in the converter in real time. In this solution, a large number of sample sets need to be trained to obtain the data set of the artificial intelligence prediction model. A large number of sample sets require a large number of personnel to measure the real-time indicators of the molten steel during the production process, consuming a large amount of manpower and financial resources and having low production efficiency. Summary of the Invention

[0004] To solve the above technical problems, embodiments of this application provide a method, device, electronic device and storage medium for determining the end point of converter smelting.

[0005] In a first aspect, embodiments of this application provide a method for determining the end point of converter smelting, the method including:

[0006] Obtain an initial flame image from a camera, and obtain a current converter mouth flame feature set according to the initial flame image;

[0007] Determine a predicted mid- to late-stage conversion point of converter smelting according to the current converter mouth flame feature set through a modified LSTM prediction model;

[0008] Obtain the mid- to late-stage conversion point predicted flame feature corresponding to the predicted mid- to late-stage conversion point;

[0009] Match the mid- to late-stage conversion point predicted flame feature with the converter smelting conversion point flame feature set;

[0010] If the mid- to late-stage conversion point predicted flame feature matches the converter smelting conversion point flame feature set, determine that the converter blowing enters the late stage;

[0011] Determine the end point of converter blowing according to the carbon content at the mid- to late-stage conversion point of converter blowing through a converter blowing late-stage carbon model.

[0012] In one embodiment, the step of obtaining the current furnace mouth flame feature set according to the initial flame image includes:

[0013] Extracting the current furnace mouth flame image from the initial flame image;

[0014] Extracting the current furnace mouth flame feature set from the current furnace mouth flame image.

[0015] In one embodiment, the step of extracting the current furnace mouth flame image from the initial flame image includes:

[0016] Extracting the observation port flame image from the initial flame image;

[0017] Performing grayscale processing on the observation port flame image to obtain the current furnace mouth flame image.

[0018] In one embodiment, the step of extracting the current furnace mouth flame feature set from the current furnace mouth flame image includes:

[0019] Respectively extracting the brightness value, rotation invariant local binary pattern feature value, and histogram of oriented gradients feature value from the current furnace mouth flame image;

[0020] Generating the current furnace mouth flame feature set according to the brightness value, the rotation invariant local binary pattern feature value, and the histogram of oriented gradients feature value.

[0021] In one embodiment, the step of obtaining the predicted flame feature of the mid-to-late conversion point corresponding to the predicted mid-to-late conversion point includes:

[0022] Obtaining the predicted flame image of the mid-to-late conversion point corresponding to the predicted mid-to-late conversion point;

[0023] Extracting the predicted flame feature of the mid-to-late conversion point from the predicted flame image of the mid-to-late conversion point.

[0024] In one embodiment, the step of obtaining the converter smelting conversion point flame feature set includes:

[0025] Determining the mid-to-late conversion point of converter smelting according to the converter flue gas data;

[0026] Obtaining the historical flame image of the mid-to-late conversion point corresponding to the mid-to-late conversion point from the historical converter flame images;

[0027] Obtaining the converter smelting conversion point flame feature set from the historical flame image of the mid-to-late conversion point.

[0028] In one embodiment, the step of determining the end point of converter blowing through the carbon model in the later stage of converter blowing according to the carbon content at the mid-to-late conversion point of converter blowing includes:

[0029] Calculate the blowing duration from the carbon content at the mid - to - late conversion point during the later stage of converter blowing to the historical data of the carbon content at the later stage of the converter on the same day through the carbon model for the later stage of converter blowing;

[0030] Determine the end point of converter blowing according to the blowing duration.

[0031] In a second aspect, an apparatus for determining the end point of converter smelting provided by an embodiment of the present application includes:

[0032] A first acquisition module, configured to acquire an initial flame image from a camera and obtain a current furnace mouth flame feature set according to the initial flame image;

[0033] A first determination module, configured to determine a predicted mid - to - late conversion point of converter smelting according to the current furnace mouth flame feature set through a modified LSTM prediction model;

[0034] A second acquisition module, configured to acquire the predicted flame features at the mid - to - late conversion point corresponding to the predicted mid - to - late conversion point;

[0035] A matching module, configured to match the predicted flame features at the mid - to - late conversion point with a converter smelting conversion point flame feature set;

[0036] A second determination module, configured to determine that the converter blowing enters the later stage if the predicted flame features at the mid - to - late conversion point match the converter smelting conversion point flame feature set;

[0037] A third determination module, configured to determine the end point of converter blowing according to the carbon content at the mid - to - late conversion point of converter blowing through a carbon model for the later stage of converter blowing.

[0038] In a third aspect, an electronic device provided by an embodiment of the present application includes a memory and a processor. The memory is used to store a computer program, and the computer program, when running on the processor, executes the method for determining the end point of converter smelting provided in the first aspect.

[0039] In a fourth aspect, a computer - readable storage medium provided by an embodiment of the present application stores a computer program, and the computer program, when running on a processor, executes the method for determining the end point of converter smelting provided in the first aspect.

[0040] The method, device, electronic device and storage medium for determining the end point of converter smelting provided by the present application obtain the current furnace mouth flame feature set according to the initial flame image; determine the predicted mid-late conversion point of converter smelting according to the current furnace mouth flame feature set through a modified LSTM prediction model; obtain the predicted flame features of the mid-late conversion point corresponding to the predicted mid-late conversion point; match the predicted flame features of the mid-late conversion point with the converter smelting conversion point flame feature set; if the predicted flame features of the mid-late conversion point match the converter smelting conversion point flame feature set, it is determined that the converter blowing enters the later stage; determine the end point of converter blowing according to the carbon content at the mid-late conversion point of converter blowing through the carbon model in the later stage of converter blowing. In this way, based on the combination of flame image recognition of the mid-late conversion point of converter blowing and the carbon model in the later stage of converter blowing, the accuracy of the end point of converter blowing is improved, the cost is reduced, and the production efficiency is increased. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the present application, the accompanying drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application and should not be regarded as limiting the protection scope of the present application. In each drawing, similar components are numbered similarly.

[0042] Figure 1 FIG. 1 shows a schematic flow chart of a method for determining the end point of converter smelting provided by an embodiment of the present application;

[0043] Figure 2 FIG. 2 shows a schematic structural diagram of a device for determining the end point of converter smelting provided by an embodiment of the present application.

[0044] Reference numerals: 200 - device for determining the end point of converter smelting; 201 - first acquisition module; 202 - first determination module; 203 - second acquisition module; 204 - matching module; 205 - second determination module; 206 - third determination module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments.

[0046] The components of the embodiments of the present application generally described and illustrated in the drawings herein may be arranged and designed in a variety of different configurations. Therefore, the detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the present application claimed, but merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts fall within the protection scope of the present application.

[0047] As used hereinafter, the terms "comprising", "having" and their cognates that may be used in various embodiments of the present application are only intended to indicate specific features, numbers, steps, operations, elements, components or combinations of the foregoing items, and should not be construed as precluding the existence or adding the possibility of one or more other features, numbers, steps, operations, elements, components or combinations of the foregoing items.

[0048] In addition, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.

[0049] Unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the various embodiments of the present application pertain. The terms (such as those defined in a commonly used dictionary) will be construed to have the same meaning as the contextual meaning in the relevant technical field and will not be construed to have an idealized meaning or an overly formal meaning unless clearly defined in the various embodiments of the present application.

[0050] There are several schemes for judging the end point of converter steelmaking in the prior art. The first scheme is as follows: The flame image at the converter mouth is uploaded to a computer through a camera, and then the image is processed to obtain flame characteristics, which are summarized as the current flame characteristic data set at the converter mouth; then, based on this flame characteristic data set, the flame characteristics at the converter mouth are observed in real time again through the camera, and the temperature of the molten steel at this time is matched according to the characteristics and output as data to the computer interface. However, the existing converter end point temperature prediction scheme has the following defects: In the process of training the artificial intelligence data set, a large amount of data is required as the calibrated sample data set, and the temperature curve of the molten steel in the converter needs to be detected a large number of times to be used as the characteristics required for training the artificial intelligence data set. The temperature curve in the converter is summarized after sampling and analysis at different time points, rather than the result of real-time monitoring. Therefore, there is a certain deviation, resulting in a relatively low accuracy of determining the end point of converter steelmaking by this scheme.

[0051] The second solution for determining the end point of converter steelmaking in the prior art is as follows: A scheme for on-line measurement of the carbon content of molten steel in a converter based on flame image features. The specific process is as follows: Take the average value of the number of Harris corner points in three adjacent frames of images as the basis for judging the three stages of steelmaking. Calculate the entropy of each pixel in the image, and extract the points with entropy greater than the threshold as the contour positions. Calculate the inclination angle of the right contour of the flame. After obtaining the contour of the flame image, calculate the angle between the line connecting two pixels at the edge of the window and the horizontal. Statistically analyze the angles at all contours, and take the average value as the inclination angle of the flame contour. In the off-line state, measure the carbon content of molten steel at multiple moments in the late stage of the flame, and at the same time obtain the corresponding inclination angle of the right contour of the flame. Save the multiple carbon content data of molten steel and the inclination angle data of the right contour of the flame as a binary relation table. Measure the carbon content of molten steel in real time by interpolation. However, when training the model in this scheme, a large number of images in the late stage of the flame and their corresponding molten steel composition and temperature are required. And if you want to obtain this aspect of data, you need to sample it manually. Each heat requires a large number of samples, which is instead more wasteful of manpower and financial resources.

[0052] In summary, the solutions for determining the end point of converter steelmaking in the prior art require training a large number of sample sets to obtain the data set of the artificial intelligence prediction model. A large number of sample sets require a large number of personnel to measure the real-time indicators of molten steel during the production process, consuming a large amount of manpower and financial resources, with low production efficiency and relatively low accuracy.

[0053] Embodiment 1

[0054] The present disclosure provides a method for determining the end point of converter smelting.

[0055] Specifically, referring to Figure 1 , the method for determining the end point of converter smelting includes:

[0056] Step S101, obtain an initial flame image from a camera, and obtain a current furnace mouth flame feature set according to the initial flame image.

[0057] In an embodiment, the step of obtaining a current furnace mouth flame feature set according to the initial flame image in step S101 includes:

[0058] Extract the current furnace mouth flame image according to the initial flame image;

[0059] Extract the current furnace mouth flame feature set from the current furnace mouth flame image.

[0060] It should be noted that the method for determining the end point of converter smelting in this embodiment can be applied to electronic devices, such as computer devices. Specifically, the electronic device is communicatively connected to the camera. The electronic device also has other modules. For example, the camera and the electronic device are connected through a network, a USB cable, or a serial port. The electronic device also includes devices such as a processor and a memory.

[0061] In this embodiment, the camera is installed at a preset distance in front of the converter fireproof hood door. Generally, it can be placed 20 m in front of the converter fireproof hood. The lens of the camera is facing the observation port position during the converter blowing process, and the lens field of view is within a preset angle range. For example, the preset angle range can be -30 to -30°. In this embodiment, the lens field of view is 0°, so that the camera can collect the furnace mouth flame image during the converter blowing process in real time, and the flame image collected by the camera is used as the initial flame image.

[0062] After the converter blowing starts, the camera enters the working state. The camera directly obtains the flame image at a preset speed. For example, the flame image is directly obtained at a speed of 2 times / s. When shooting, adjust the camera focus so that it can clearly capture the flame image as the initial flame image, and extract the flame image inside the furnace mouth observation port from each initial flame image to unify the image input size.

[0063] The camera is fixed by a special bracket for the heating and heat insulation protective cover. The heating and heat insulation protective cover is made of high-temperature resistant quartz glass. The height of the special bracket is the same as the height of the converter fireproof hood observation port, and it supports rotating the shooting angle. Fix the camera in the artificial fire-watching area and adjust the appropriate angle so that it can capture the furnace mouth flame of the converter.

[0064] In one embodiment, the step of extracting the current furnace mouth flame image according to the initial flame image includes:

[0065] Extract the observation port flame image from the initial flame image;

[0066] Perform gray-scale processing on the observation port flame image to obtain the current furnace mouth flame image.

[0067] Specifically, to extract the observation port flame image from the initial flame image, the background area in the initial flame image can be removed, and only the flame image corresponding to the observation port size is retained to improve the effectiveness of the image. Further, perform gray-scale preprocessing on the observation port flame image to obtain the current furnace mouth flame image, which can reduce the calculation amount and improve the calculation speed.

[0068] In one embodiment, the step of extracting the current furnace mouth flame feature set from the current furnace mouth flame image includes:

[0069] Extract the brightness value, rotation invariant local binary pattern feature value, and histogram of oriented gradients feature value from the current furnace mouth flame image respectively;

[0070] Generate the current furnace mouth flame feature set according to the brightness value, the rotation invariant local binary pattern feature value, and the histogram of oriented gradients feature value.

[0071] In this embodiment, the current furnace mouth flame feature set obtained in real time is processed respectively to obtain the current furnace mouth flame feature set, and the current furnace mouth flame features include but are not limited to the brightness value, the rotation invariant local binary pattern feature value, the histogram of oriented gradients feature value, the furnace mouth flame contour feature, and the gray value parameter.

[0072] For example, the HSI color space parameters can be used to establish a color space model for the current furnace mouth flame image to obtain the current furnace mouth flame feature set, which includes but is not limited to the contour feature and gray value parameter of the furnace mouth flame.

[0073] Step S102, determine the predicted mid - to - late conversion point of converter smelting according to the current furnace mouth flame feature set through the modified LSTM prediction model.

[0074] In this embodiment, the acquisition of the modified LSTM prediction model includes the following steps:

[0075] Extract multiple furnace mouth flame feature vectors, where the extracted furnace mouth flame feature vectors include but are not limited to the brightness value, rotation invariant LBP feature value, and HOG feature value; use the multiple furnace mouth flame feature vectors as the input variables of the initial LSTM prediction model;

[0076] Perform normalization processing on the input variables of the initial LSTM prediction model. Specifically, use the series tosupervised() function to convert the data set, and convert the data into multiple input variables and 1 output variable;

[0077] Randomly divide the data set into a training set and a test set. The training set is used for the training and learning of the initial LSTM prediction model, and the test set is used for the evaluation and testing of the LSTM prediction model to obtain the modified LSTM prediction model;

[0078] The modified LSTM prediction model can be used to predict the smelting temperature in the furnace corresponding to the feature data in the current furnace mouth flame feature data set. The ratio of the randomly generated training set to the test set is 9∶1.

[0079] Exemplarily, the following scheme can be adopted to extract the brightness value: use the PIL module in python to obtain the brightness value of the image;

[0080] The following scheme can be adopted to extract the rotation-invariant Local Binary Pattern (LBP) eigenvalue: LBP is used to extract image texture features. After determining the radius size and the number of sampling points, continuously rotate the positions of the sampling points within the prototype neighborhood, and select the minimum value from the obtained series of LBP eigenvalues as the LBP eigenvalue of the LBP central pixel point.

[0081] The following scheme can be adopted to extract the Histogram of Oriented Gradients (HOG) eigenvalue: HOG operates on local grid cells of an image and can maintain good invariance to both geometric and optical deformations of the image.

[0082] It should be noted that LSTM is a type of time-recurrent neural network that can perfectly solve the problem of multiple input variables. This embodiment provides an LSTM prediction model for developing multi-variable time series prediction in the Keras deep learning library. By preparing a dataset of flame image features at the mid-late transformation point of converter smelting for the LSTM prediction model, regarding the dataset as a supervised learning problem and normalizing the input variables, specifically using the series_to_supervised() function to transform the dataset of flame image features at the mid-late transformation point of converter smelting, converting the data into 3 input variables (brightness value, LBP, HOG) and 1 output variable (mid-late transformation point of converter smelting). Randomly divide the dataset of flame image features at the mid-late transformation point of converter smelting into a training set and a test set.

[0083] In this embodiment, the ratio of the randomly generated training set to the test set is 9:1. The training set is used for the training and learning of the LSTM prediction model, and the test set is used for the evaluation and testing of the LSTM prediction model. Reconstruct the input (X) into the 3D format expected by the LSTM prediction model, i.e., [samples, time steps, features] to obtain a modified LSTM prediction model.

[0084] Further supplementary explanation is that the number of hidden layers and the number of neurons in the hidden layer of the initial LSTM prediction model can both be adjusted according to actual needs. Use the Mean Absolute Error (MAE) loss function in the initial LSTM prediction model, and then track the training and testing failures during the training process by setting the validation_data parameter in the fit() function. At the end of the run, plot the training and testing losses. Combine the prediction set and the test set, and perform inverse scaling. It is also possible to inverse-scale the test dataset with the expected pollution numbers. Using the predicted values and the actual values, the error score of the model can be calculated, and the Root Mean Square Error (RMSE) of the error generated by units basically the same as the variables can also be calculated.

[0085] Step S103, obtain the mid-late transformation point prediction flame features corresponding to the predicted mid-late transformation point.

[0086] In one embodiment, step S103 includes the following steps:

[0087] Acquiring a corresponding mid-to-late transition point prediction flame image at the predicted mid-to-late transition point;

[0088] The mid-to-late transition point predicted flame feature is extracted from the mid-to-late transition point predicted flame image.

[0089] Step S104, matching the predicted flame characteristics of the mid-to-late transition point with the converter smelting transition point flame characteristic set.

[0090] In this embodiment, if the predicted flame characteristics of the mid-to-late transition point match the flame characteristic set of the converter smelting transition point, step S105 is executed; if the predicted flame characteristics of the mid-to-late transition point match the flame characteristic set of the converter smelting transition point, the process returns to step S101.

[0091] In one embodiment, the step of acquiring the converter smelting transition point flame feature set comprises:

[0092] Determine the mid- to late-stage transition point of converter smelting based on converter flue gas data;

[0093] Acquire a mid-to-late transition point historical flame image corresponding to the mid-to-late transition point from the historical converter flame image;

[0094] The converter smelting transition point flame feature set is obtained from the mid-to-late transition point historical flame images.

[0095] In this embodiment, the converter flue gas data may be the converter flue gas data of the past year, and the historical converter flame image may be the converter mouth flame image data of the past year. The historical converter flame image is calibrated by the converter flue gas data, and the flame features of the transition point in the middle and late stages of converter blowing are calibrated, and trained as the flame feature set of the converter smelting transition point. In the calibration process, when judging the transition point in the middle and late stages of converter blowing, it is necessary to eliminate obviously erroneous data.

[0096] In this embodiment, the converter flue gas data is used to analyze and locate the time of the mid-to-late transition point of blowing during the converter smelting process. The historical converter flame image can be the converter mouth flame image data of the past year. According to the time of the mid-to-late transition point, the historical flame image of the mid-to-late transition point is determined from the converter mouth flame image data of the past year, and finally the converter smelting transition point flame feature set of the mid-to-late converter is extracted from the historical flame image of the mid-to-late transition point. Specifically, the converter smelting transition point flame feature set may include the brightness value corresponding to the converter smelting transition point, the rotationally invariant local binary pattern (LBP) feature value, the directional gradient histogram (HOG) feature value, and may also include other types of feature values, which are not limited here.

[0097] Step S105: If the predicted flame characteristics at the mid-late conversion point match the flame characteristic set of the converter smelting conversion point, it is determined that the converter blowing enters the late stage.

[0098] In this embodiment, the predicted flame characteristics at the mid-late conversion point obtained in real time are compared with the flame characteristics in the flame characteristic set of the converter smelting conversion point. If they conform to the flame characteristics of the converter smelting conversion point set, a signal indicating that the mid-late conversion point of the converter smelting has been reached is generated, the conversion point of the converter smelting is predicted in real time, and it is determined that the converter blowing enters the late stage.

[0099] Step S106: Determine the end point of the converter blowing according to the carbon content at the mid-late conversion point of the converter blowing through the carbon model in the late stage of the converter blowing.

[0100] In this embodiment, the carbon model in the late stage of the converter blowing is used to represent the distribution law of the carbon content in the late stage of the converter blowing. The end point of the converter blowing is calculated based on the carbon content at the mid-late conversion point of the converter blowing through the carbon model in the late stage of the converter blowing.

[0101] This embodiment uses the converter flue gas to calibrate the mid-late conversion point of the converter smelting. The data is accurate, safe and reliable. Combined with artificial intelligence, it can realize the real-time prediction of the mid-late conversion point of the converter blowing. This embodiment can effectively improve the accuracy rate of the artificial intelligence recognition ability, and predict the end point of the converter blowing through the carbon model in the late stage of the converter blowing. Compared with only relying on manual experience to determine the end point of the converter blowing, the hit rate is increased by 40%. Compared with only relying on artificial intelligence to judge the end point temperature or end point carbon content, the secondary blowing ratio is decreased by 30%.

[0102] In one implementation manner, step S106 includes the following steps:

[0103] Calculate the blowing duration from the carbon content at the mid-late conversion point of the converter blowing to the historical data of the carbon content in the late stage of the converter on the same day through the carbon model in the late stage of the converter blowing;

[0104] Determine the end point of the converter blowing according to the blowing duration.

[0105] Specifically, the carbon model in the late stage of the converter blowing includes the following formula 1:

[0106] Formula 1:

[0107] Where k is a fixed value, which can be calculated from the data in the later stage of a certain historical steelmaking process in the converter. The value of k is the carbon content at the blowing end corresponding to the historical steelmaking process. %C represents the carbon content. For example, the value of k can be calculated from the data in the later stage of the first furnace of steelmaking on the same day. According to Formula 1, based on the carbon content at the turning point in the middle and later stages of converter blowing, calculate the blowing duration required for the carbon content to change from the carbon content at the turning point in the middle and later stages of converter blowing to the carbon content corresponding to the blowing end of the converter, so as to determine the blowing end of the converter from this blowing duration.

[0108] This embodiment uses converter flue gas data instead of a large number of manual samplings for detection, which not only saves the training cost but also reduces the risk of workers' work. Compared with the existing solution of continuously analyzing and training the converter smelting process, the artificial intelligence training adopted in this embodiment only targets the turning point in the middle and later stages of converter smelting. The flame characteristics change significantly in this stage, which can effectively improve the recognition accuracy of artificial intelligence. According to the decarburization law in the later stage of converter blowing, establish a decarburization curve model of the molten steel in the later stage of converter blowing, that is, establish a carbon model in the later stage of converter blowing, and predict the converter end point through the carbon content at the converter smelting end point in real time. In this way, the number of reblows and manual sampling detections can be effectively reduced, making the working environment of workers safer and improving production efficiency at the same time.

[0109] Compared with the prior art, the method for determining the converter smelting end point provided in this embodiment has the following beneficial effects: The data required for training the model in this embodiment is replaced by converter flue gas instead of manual sampling, which can save a large amount of manpower and financial resources and reduce the risk during workers' sampling analysis and detection; in addition, this embodiment only uses the flame image recognition method to predict the turning point in the middle and later stages of converter blowing. Compared with directly judging the carbon content or temperature in the molten steel based on the current flame characteristics, the flame characteristics change more significantly in the middle and later stages of converter blowing, and the result of determining the turning point in the middle and later stages based on the flame characteristics in the middle and later stages of converter blowing is more accurate, and the prediction of the turning point in the middle and later stages can be realized in real time. Based on the combination of flame image recognition of the turning point in the middle and later stages of converter blowing and the carbon model in the later stage of converter blowing, the accuracy of determining the converter blowing end point is improved. Compared with only relying on manual experience to judge the smelting end point, the accuracy is increased by about 40%; compared with only relying on artificial intelligence to judge the temperature or carbon content of the molten steel, the secondary blowing ratio is reduced by about 20%.

[0110] Embodiment 2

[0111] In addition, the embodiments of the present disclosure provide a device for determining the converter smelting end point.

[0112] Specifically, referring to Figure 2 , the device 200 for determining the converter smelting end point includes:

[0113] The first acquisition module 201 is configured to acquire an initial flame image from a camera and obtain a current furnace mouth flame feature set according to the initial flame image;

[0114] The first determination module 202 is configured to determine the predicted mid - to - late conversion point of converter smelting according to the current furnace mouth flame feature set through a corrected LSTM prediction model;

[0115] The second acquisition module 203 is configured to acquire the predicted flame features at the mid - to - late conversion point corresponding to the predicted mid - to - late conversion point;

[0116] The matching module 204 is configured to match the predicted flame features at the mid - to - late conversion point with the converter smelting conversion point flame feature set;

[0117] The second determination module 205 is configured to determine that the converter blowing enters the late stage if the predicted flame features at the mid - to - late conversion point match the converter smelting conversion point flame feature set;

[0118] The third determination module 206 is configured to determine the converter blowing end point according to the carbon content at the mid - to - late conversion point of converter blowing through a converter blowing late - stage carbon model.

[0119] In an embodiment, the first acquisition module 201 is further configured to extract the current furnace mouth flame image according to the initial flame image;

[0120] Extract the current furnace mouth flame feature set from the current furnace mouth flame image.

[0121] In an embodiment, the first acquisition module 201 is further configured to extract the observation port flame image from the initial flame image;

[0122] Perform grayscale processing on the observation port flame image to obtain the current furnace mouth flame image.

[0123] In an embodiment, the first acquisition module 201 is further configured to extract the brightness value, the rotation - invariant local binary pattern feature value, and the histogram of oriented gradients feature value from the current furnace mouth flame image respectively;

[0124] Generate the current furnace mouth flame feature set according to the brightness value, the rotation - invariant local binary pattern feature value, and the histogram of oriented gradients feature value.

[0125] In an embodiment, the second acquisition module 203 is further configured to acquire the predicted flame image at the mid - to - late conversion point corresponding to the predicted mid - to - late conversion point;

[0126] Extract the predicted flame features at the mid - to - late conversion point from the predicted flame image at the mid - to - late conversion point.

[0127] In an embodiment, the matching module 204 is further configured to determine the mid - to - late conversion point of converter smelting according to the converter flue gas data;

[0128] Obtain the historical flame image of the middle and late conversion point corresponding to the middle and late conversion point from the historical converter flame image;

[0129] Obtain the converter smelting conversion point flame feature set from the historical flame image of the middle and late conversion point.

[0130] In one embodiment, the third determination module is further configured to calculate the blowing duration from the carbon content at the middle and late conversion point of converter blowing to the historical data of the carbon content at the late stage of the converter on the same day through the carbon model in the late stage of converter blowing;

[0131] Determine the end point of converter blowing according to the blowing duration.

[0132] Compared with the prior art, the device for determining the end point of converter smelting provided in this embodiment has the following beneficial effects: The data required for training the model in this embodiment is replaced by converter flue gas instead of manual sampling, which can save a large amount of manpower and financial resources and reduce the risk during the sampling analysis and detection by workers; In addition, in this embodiment, only the flame image recognition method is used to predict the middle and late conversion points of converter blowing. Compared with directly judging the carbon content or temperature in the molten steel based on the current flame characteristics, the flame characteristics in the middle and late stages of converter blowing change more significantly, and the result of determining the middle and late conversion points based on the flame characteristics in the middle and late stages of converter blowing is more accurate. It can realize the real-time prediction of the middle and late conversion points. Based on the combination of flame image recognition of the middle and late conversion points of converter blowing and the carbon model in the late stage of converter blowing, the accuracy of determining the end point of converter blowing is improved. Compared with only relying on manual experience to determine the smelting end point, the accuracy is increased by about 40%; Compared with only relying on artificial intelligence to judge the temperature or carbon content of the molten steel, the secondary blowing ratio is reduced by about 20%.

[0133] Embodiment 3

[0134] In addition, the present disclosure embodiment provides an electronic device, including a memory and a processor, where the memory stores a computer program, and the computer program executes the method for determining the end point of converter smelting provided in Embodiment 1 when running on the processor.

[0135] The electronic device provided in this embodiment can implement the method for determining the end point of converter smelting provided in Embodiment 1. To avoid repetition, it will not be elaborated here.

[0136] Embodiment 4

[0137] This application also provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and the computer program executes the method for determining the end point of converter smelting provided in Embodiment 1 when running on a processor.

[0138] The computer-readable storage medium provided in this embodiment can implement the method for determining the end point of converter smelting provided in Embodiment 1. To avoid repetition, it will not be elaborated here.

[0139] It should be noted that in this text, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article or terminal including a series of elements not only includes those elements but also other elements not explicitly listed, or further includes elements inherent to such process, method, article or terminal. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or terminal including such element.

[0140] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on this understanding, the technical solution of the present application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc) and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present application.

[0141] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them fall within the protection scope of the present application.

Claims

1. A method for determining the end point of converter smelting, characterized in that, The method includes: Obtaining an initial flame image from a camera, and obtaining a current converter mouth flame feature set according to the initial flame image; Determining a predicted mid-to-late conversion point of converter smelting according to the current converter mouth flame feature set through a modified LSTM prediction model; Obtaining mid-to-late conversion point predicted flame features corresponding to the predicted mid-to-late conversion point; Matching the mid-to-late conversion point predicted flame features with a converter smelting conversion point flame feature set; If the mid-to-late conversion point predicted flame features match the converter smelting conversion point flame feature set, it is determined that the converter blowing enters the later stage; Determining the converter blowing end point according to the carbon content at the mid-to-late conversion point of converter blowing through a converter blowing late-stage carbon model; The step of determining the converter blowing end point according to the carbon content at the mid-to-late conversion point of converter blowing through the converter blowing late-stage carbon model includes: Calculating the blowing duration for updating the carbon content from the carbon content at the mid-to-late conversion point of converter blowing to the historical data of the converter late-stage carbon content on the same day through the converter blowing late-stage carbon model; Determining the converter blowing end point according to the blowing duration; The converter blowing late-stage carbon model includes the following formula 1: Formula 1: Where k is a fixed value, and the value of k can be calculated through the converter historical steelmaking late-stage data. k is the carbon content at the blowing end point corresponding to the historical steelmaking process. %C represents the carbon content. According to formula 1, based on the carbon content at the mid-to-late conversion point of converter blowing, calculate the blowing duration required for the carbon content to change from the carbon content at the mid-to-late conversion point of converter blowing to the carbon content corresponding to the converter blowing end point, so as to determine the converter blowing end point from this blowing duration.

2. The method according to claim 1, characterized in that, The step of obtaining the current converter mouth flame feature set according to the initial flame image includes: Extracting the current converter mouth flame image according to the initial flame image; Extracting the current converter mouth flame feature set from the current converter mouth flame image.

3. The method according to claim 2, characterized in that, The step of extracting the current converter mouth flame image according to the initial flame image includes: Extracting the observation port flame image from the initial flame image; Performing grayscale processing on the observation port flame image to obtain the current converter mouth flame image.

4. The method according to claim 2, wherein The step of extracting the current converter mouth flame feature set from the current converter mouth flame image includes: Respectively extracting the brightness value, rotation invariant local binary pattern feature value, and histogram of oriented gradients feature value from the current converter mouth flame image; Generating the current converter mouth flame feature set according to the brightness value, the rotation invariant local binary pattern feature value, and the histogram of oriented gradients feature value.

5. The method according to claim 1, characterized in that, The step of obtaining the mid-to-late conversion point predicted flame features corresponding to the predicted mid-to-late conversion point includes: Obtaining the corresponding mid-to-late conversion point predicted flame image at the predicted mid-to-late conversion point; Extracting the mid-to-late conversion point predicted flame features from the mid-to-late conversion point predicted flame image.

6. The method according to claim 1, characterized in that, The step of obtaining the converter smelting conversion point flame feature set includes: Determining the mid-to-late conversion point of converter smelting according to converter flue gas data; Obtaining the mid-to-late conversion point corresponding mid-to-late conversion point historical flame image from historical converter flame images; Obtaining the converter smelting conversion point flame feature set from the mid-to-late conversion point historical flame image.

7. An apparatus for determining the end point of converter smelting, characterized in that, The device includes: A first acquisition module, configured to acquire an initial flame image from a camera and obtain a current furnace mouth flame feature set according to the initial flame image; A first determination module, configured to determine a predicted mid-late conversion point of converter smelting according to the current furnace mouth flame feature set through a modified LSTM prediction model; A second acquisition module, configured to acquire mid-late conversion point predicted flame features corresponding to the predicted mid-late conversion point; A matching module, configured to match the mid-late conversion point predicted flame features with a converter smelting conversion point flame feature set; A second determination module, configured to determine that the converter blowing enters the later stage if the mid-late conversion point predicted flame features match the converter smelting conversion point flame feature set; A third determination module, configured to determine the converter blowing end point according to the carbon content at the mid-late conversion point of converter blowing through a converter blowing later-stage carbon model; The step of determining the converter blowing end point according to the carbon content at the mid-late conversion point of converter blowing through the converter blowing later-stage carbon model includes: Calculating, through the converter blowing later-stage carbon model, the blowing duration for updating the carbon content at the mid-late conversion point of converter blowing to the historical data of the carbon content in the later stage of the converter on the same day; Determining the converter blowing end point according to the blowing duration; The converter blowing later-stage carbon model includes the following formula 1: Formula 1: Where k is a fixed value, and the value of k can be calculated through the historical converter steelmaking later-stage data. k is the carbon content at the blowing end point corresponding to the historical steelmaking process. %C represents the carbon content. According to formula 1, based on the carbon content at the mid-late conversion point of converter blowing, calculate the blowing duration required for the carbon content to change from the carbon content at the mid-late conversion point of converter blowing to the carbon content corresponding to the converter blowing end point, so as to determine the converter blowing end point from this blowing duration.

8. An electronic device, characterized in that, It includes a memory and a processor. The memory stores a computer program, and the computer program executes the method for determining the converter smelting end point according to any one of claims 1 to 6 when running on the processor.

9. A computer-readable storage medium, characterized in that, It stores a computer program, and the computer program executes the method for determining the converter smelting end point according to any one of claims 1 to 6 when running on the processor.

Images