Blast furnace tuyere abnormal condition detection method and system and storage medium

Abstract

The invention discloses a blast-furnace tuyere abnormal condition detection method, which belongs to the technical field of blast-furnace smelting, and comprises the following steps: firstly, continuously acquiring video data in a blast-furnace tuyere, and then splitting the video data into image data according to frames, acquiring image data through a tuyere anomaly detection model, and extracting the color, shape, edge and other features of the image through convolution calculation; and finally, comparing the image features with a standard data set of the tuyere anomaly detection model, and outputting an anomaly condition corresponding to the standard data set. By utilizing the method, the abnormal condition of the air outlet can be accurately detected in real time, so that a guarantee is provided for the generalization performance of the model, the operation stability of the blast furnace is improved, meanwhile, the maintenance cost is reduced, and the diversity and accuracy of data are increased.

Description

Technical Field

[0001] The present invention relates to the technical field of blast furnace tuyere detection, and in particular to a blast furnace tuyere abnormality detection algorithm, system and storage medium. Background Art

[0002] The blast furnace tuyere is a key equipment in the blast furnace air supply system and an important part to ensure the smooth progress of chemical reactions inside the blast furnace. Therefore, the state of the tuyere directly affects the blast furnace's production efficiency, energy consumption and the quality of the final product.

[0003] In actual production, various abnormal conditions may occur during the operation of the blast furnace tuyere, such as slag hanging on the tuyere, water leakage, burn-through, coal shortage, etc. If these abnormal conditions are not discovered and handled in time, it will lead to unstable blast furnace operation and may even cause production accidents, resulting in huge economic losses.

[0004] At present, the common detection means mainly rely on observation by experienced technical workers. However, this method has problems such as long inspection cycle, poor operating environment, and inability to provide real-time feedback. In view of this, in order to improve the blast furnace smelting quality and reduce labor costs, the present invention is proposed. Summary of the invention

[0005] To this end, the present invention provides a method, system and storage medium for detecting abnormal conditions of blast furnace tuyere, so as to solve the problems in the prior art of long inspection cycle and untimely feedback due to manual inspection of blast furnace tuyere.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] According to the first aspect of the present invention;

[0008] The invention discloses a method for detecting abnormal conditions of a blast furnace tuyere, comprising the following steps:

[0009] Continuously collect video data in the blast furnace tuyere and split the video data into image data by frame;

[0010] The image data is acquired through the tuyere anomaly detection model, and the image features of the image data are extracted, wherein the extracted image features are mainly the color, shape, edge and other features of the image;

[0011] Furthermore, the image features are compared with the standard data set of the tuyere anomaly detection model to output the abnormal conditions corresponding to the standard data set. Specifically, the abnormal conditions include abnormal categories and corresponding image data. The abnormal categories include slag hanging, water leakage, burn-through and coal breakage.

[0012] In a possible implementation, the deep learning model is trained before detection to obtain an air outlet anomaly detection model, which specifically includes:

[0013] The video data of the blast furnace tuyere is collected in real time, and then some normal video data and all abnormal video data in the video data are selected, and on this basis, the normal video data and the abnormal video data are split into corresponding normal images and abnormal images;

[0014] Finally, normal images and abnormal graphics are classified and labeled according to the defined standard data set to obtain labeled data, and the deep model is trained based on the labeled data to obtain the vent anomaly detection model.

[0015] In a possible implementation, normal images and abnormal graphics are classified and labeled according to a defined standard data set to obtain labeled data, and then data enhancement is performed on the labeled data.

[0016] Further, data augmentation can be performed by color space conversion, blurring, random cropping, flipping, increasing contrast, and occlusion.

[0017] In a possible implementation, after real-time acquisition of video data of the blast furnace tuyere, it is necessary to manually classify and annotate normal images and abnormal graphics according to a defined standard data set to obtain annotated data.

[0018] In one possible implementation, the deep learning model includes a convolutional neural network model and a residual neural network structure.

[0019] In one possible implementation, it further includes:

[0020] Accumulating the abnormal conditions to obtain accumulated abnormal data;

[0021] The parameters of the tuyere anomaly detection model are adjusted according to the accumulated abnormal data to obtain an optimized tuyere anomaly detection model.

[0022] Furthermore, the dross refers to the fact that the edge of the tuyere is not smooth and there is a dark area with a large gray scale at the dross location for a long time;

[0023] The water leakage refers to the appearance of dark areas or water marks on the edge of the tuyere, and the inner area of ​​the tuyere becomes dark for a long time;

[0024] Burn-through refers to abnormal brightness in the dark area of ​​the coal gun blowpipe, the bright area becomes significantly larger, and the brightness disappears after the coal is stopped;

[0025] The so-called coal-breaking means that the edge of the tuyere is smooth, the color inside the tuyere is bright, and there is no dark area of ​​coal injection.

[0026] The present invention has the following advantages:

[0027] In the present invention, the blast furnace tuyere abnormal situation detection method is utilized to detect the abnormal situation of the tuyere in real time and accurately, and for the blast furnace tuyere with complex environment, it is helpful to obtain high-quality and representative data. Combined with the preprocessing of the data, it is possible to achieve the goal of building a deep learning model that can adapt to different blast furnace conditions and has good generalization ability, which is conducive to the subsequent improvement of blast furnace operation stability, reduction of maintenance costs, and increase of data diversity, providing a guarantee for the generalization performance of the model.

[0028] According to a second aspect of the present invention,

[0029] Disclosed is a blast furnace tuyere abnormality detection system, which applies the blast furnace tuyere abnormality detection method proposed above, wherein the blast furnace tuyere abnormality detection system comprises:

[0030] A video shooting device, suitable for continuously collecting video data in the blast furnace tuyere from multiple angles;

[0031] a data processing module, which is communicatively connected to the video shooting device, and is adapted to acquire the video data and split the video data into image data by frame, acquire the image data through an air outlet anomaly detection model and extract image features of the image data, compare the image features with a standard data set of the air outlet anomaly detection model, and output an abnormality corresponding to the standard data set;

[0032] The present invention has the following advantages:

[0033] In the furnace tuyere abnormal situation detection system, the abnormal situation of the tuyere video data can be detected in real time, and an intelligent detection platform can be provided for ironmaking workers. Through this system, not only can alarm information be provided for abnormal situations, but also abnormal data can be saved. Therefore, by accumulating a large amount of abnormal data, the model can be reversely adjusted to further enhance the generalization performance of the model and realize the continuous optimization of the detection accuracy of the algorithm.

[0034] According to a third aspect of the present invention,

[0035] Disclosed is a storage medium storing a computer program, wherein when the computer program is executed by a processor, the method for detecting abnormal conditions of blast furnace tuyere as proposed above is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0037] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0038] Figure 1 A flow chart of a method for detecting abnormal conditions of blast furnace tuyere provided by the present invention;

[0039] Figure 2 The present invention provides a preprocessing flow chart. DETAILED DESCRIPTION

[0040] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Please refer to Figure 1 and Figure 2 The present invention discloses a method for detecting abnormal conditions of blast furnace tuyere, which is mainly used to replace the method of manually checking the blast furnace tuyere and then judging the smelting condition of the blast furnace in the prior art, thereby improving the monitoring accuracy of blast furnace smelting and the feedback speed of blast furnace conditions. The technical solution of the present invention will be explained in the form of specific embodiments below.

[0042] In a specific embodiment disclosed in the present invention, the method for detecting abnormal conditions in a blast furnace tuyere specifically includes the following steps. First, it is necessary to continuously collect video data in the blast furnace tuyere, and then split the video data into image data by frame, wherein the image data is an RGB color picture. Next, the image data is acquired through the tuyere abnormality detection model, and the image features of the image data are extracted, wherein the image features refer to the color, shape, edge and other features of the image extracted by convolution calculation. Finally, the image features are compared with the standard data set of the tuyere abnormality detection model, and the abnormal conditions corresponding to the standard data set are output. In this embodiment, the abnormal conditions include abnormal categories and corresponding image data, and the abnormal categories refer to a variety of abnormal conditions including slag hanging, water leakage, burn-through and coal failure.

[0043] In this embodiment, the abnormal categories are explained in detail, among which slag refers to the uneven edge of the tuyere, and there is a dark area with a large grayscale at the slag location for a long time; water leakage refers to the appearance of a dark area or water mark at the edge of the tuyere, and the internal area of ​​the tuyere becomes dark for a long time; burn-through refers to the abnormal brightness of the dark area of ​​the coal gun blowpipe, and the bright area becomes significantly larger, and the light disappears after the coal is stopped; and coal outage refers to a smooth edge of the tuyere, a bright color inside the tuyere, and no dark area with coal injection.

[0044] In some embodiments, it is necessary to call the tuyere anomaly detection model to judge the image features of the obtained blast furnace tuyere, and before that, it is necessary to train the deep learning model to obtain stable model parameters, and then obtain the tuyere anomaly detection model. On this basis, it is necessary to train the tuyere anomaly detection model, and the training step includes real-time acquisition of video data of the blast furnace tuyere, and then, in the obtained video data, select part of the normal video data and all the abnormal video data in the video data, and split all the normal and abnormal video data into corresponding normal images and abnormal images. Finally, the normal images and abnormal graphics are classified and labeled according to the defined standard data set, and the labeled data is obtained, and the deep model is trained based on the labeled data to obtain the tuyere anomaly detection model.

[0045] In this embodiment, the vent anomaly detection model is mainly composed of a convolutional neural network model and a residual neural network structure. In the specific training process, the vent anomaly detection model outputs a feature vector after inputting the corresponding normal image and abnormal image, as well as the corresponding abnormal category. After the feature vector is extracted, the result is predicted by the idea of ​​classification, and the difference between the predicted result and the true result is obtained by subtracting the predicted result from the true result to obtain the difference between the true value and the predicted value. The formula for calculating the difference is the loss function, and the calculation formula is:

[0046]

[0047] Among them, Loss represents the loss value; i represents the sample; M represents the number of categories; yic Sign function, if the true category of sample i is equal to c, the value is 1, otherwise the value is 0; p ic Indicates the probability that the observed sample i belongs to category c. In addition, specifically, data enhancement is used during the training process to prevent overfitting, and an optimization algorithm is used to adjust the model parameters to improve the generalization ability of the model. When the loss value drops to a certain level, the model training is completed.

[0048] In this embodiment, a standard data set is defined in the trained vent anomaly detection model, wherein the standard data set is obtained through correct classification and is mainly used as a reference for the data set. On this basis, normal images and abnormal graphics are classified and labeled according to the defined standard data set to obtain labeled data. Furthermore, after obtaining the labeled data, the labeled data needs to be enhanced. Specifically, the data enhancement methods include color space conversion, blurring, random cropping, flipping, contrast enhancement and occlusion.

[0049] Specifically, by enhancing the image data, the distinguishing features between normal images and abnormal graphics can be significantly highlighted. In addition, after being input into the air outlet anomaly detection model, the enhanced image data can also significantly improve the processing speed of the air outlet anomaly detection model, thereby classifying and labeling abnormal graphics at a faster speed.

[0050] In a possible embodiment, normal images and abnormal graphics are classified and labeled manually according to a defined standard data set to obtain labeled data, thereby selecting part of the normal video data and all the abnormal video data in the video data.

[0051] In a specific embodiment disclosed in the present invention, continuous image data can be obtained during the monitoring of blast furnace tuyere, and abnormal conditions can be accumulated in the continuous image data to obtain accumulated abnormal data. Furthermore, according to the accumulated abnormal data, the parameters of the tuyere abnormality detection model can be reversely adjusted, thereby obtaining a further optimized tuyere abnormality detection model, thereby further enhancing the generalization performance of the model and optimizing the detection accuracy of the algorithm.

[0052] Based on the same inventive concept, the present invention also discloses a blast furnace tuyere abnormality detection system, which applies the blast furnace tuyere abnormality detection method proposed above, and includes a video shooting device, a data processing module and a display module, wherein the video shooting device is suitable for continuously collecting video data in the blast furnace tuyere from multiple angles. The data processing module is connected to the video shooting device in communication, and the data processing module is suitable for obtaining video data and splitting the video data into image data by frame, and obtaining image data and extracting image features of the image data through a tuyere abnormality detection model, comparing the image features with the standard data set of the tuyere abnormality detection model, outputting abnormal conditions corresponding to the standard data set, and finally displaying the abnormal conditions through a display module, wherein the display module is connected to the data processing module in communication, and the abnormal categories and corresponding image data can be fully displayed and published through the display module.

[0053] Based on the same inventive concept, the present invention also discloses a storage medium storing a computer program, which implements the above-mentioned blast furnace tuyere abnormal condition detection method when executed by a processor.

[0054] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for detecting abnormal conditions of blast furnace tuyere, characterized in that: The following steps are involved: Continuously collecting video data in the blast furnace tuyere; Splitting the video data into image data by frame; Acquiring the image data through an air outlet anomaly detection model and extracting image features of the image data; The image features are compared with a standard data set of the tuyere anomaly detection model, and an abnormality corresponding to the standard data set is output, wherein the abnormality includes an abnormality category and the corresponding image data, and the abnormality category includes slag hanging, water leakage, burn-through and coal breakage.

2. The method for detecting abnormal conditions of blast furnace tuyere according to claim 1, characterized in that: The method further includes training a deep learning model before detection to obtain the air outlet anomaly detection model, which specifically includes: Collecting the video data of the blast furnace tuyere in real time; Selecting a portion of normal video data and all abnormal video data from the video data; Splitting the normal video data and the abnormal video data into corresponding normal images and abnormal images; Classify and annotate the normal image and the abnormal image according to the defined standard data set to obtain annotated data; The deep model is trained based on the labeled data to obtain the air outlet anomaly detection model.

3. The method for detecting abnormal conditions of blast furnace tuyere according to claim 2, characterized in that: The normal image and the abnormal image are classified and labeled according to the defined standard data set to obtain labeled data, and then the method further includes: performing data enhancement on the labeled data.

4. The method for detecting abnormal conditions of blast furnace tuyere according to claim 3, characterized in that: The data enhancement methods include color space conversion, blurring, random cropping, flipping, increasing contrast and occlusion.

5. The method for detecting abnormal conditions of blast furnace tuyere according to any one of claims 2 to 4, characterized in that: The normal images and the abnormal graphics are classified and labeled manually according to the defined standard data set to obtain the labeled data.

6. The method for detecting abnormal conditions of blast furnace tuyere according to any one of claims 1 to 4, characterized in that: The deep learning model includes a convolutional neural network model and a residual neural network structure.

7. The method for detecting abnormal conditions of blast furnace tuyere according to any one of claims 1 to 4, characterized in that: Also includes: Accumulating the abnormal conditions to obtain accumulated abnormal data; The parameters of the tuyere anomaly detection model are adjusted according to the accumulated abnormal data to obtain an optimized tuyere anomaly detection model.

8. The method for detecting abnormal conditions of blast furnace tuyere according to any one of claims 1 to 4, characterized in that: The dross refers to the fact that the edge of the tuyere is not smooth and there is a dark area with a large gray scale at the dross location for a long time; The water leakage refers to the appearance of dark areas or water marks on the edge of the tuyere, and the inner area of ​​the tuyere becomes dark for a long time; Burn-through refers to abnormal brightness in the dark area of ​​the coal gun blowpipe, the bright area becomes significantly larger, and the brightness disappears after the coal is stopped; The so-called coal-breaking means that the edge of the tuyere is smooth, the color inside the tuyere is bright, and there is no dark area of ​​coal injection.

9. A blast furnace tuyere abnormality detection system, characterized in that: Using the blast furnace tuyere abnormality detection method according to any one of claims 1 to 8, the blast furnace tuyere abnormality detection system comprises: A video shooting device, suitable for continuously collecting video data in the blast furnace tuyere from multiple angles; a data processing module, which is communicatively connected to the video shooting device, and is adapted to acquire the video data and split the video data into image data by frame, acquire the image data through an air outlet anomaly detection model and extract image features of the image data, compare the image features with a standard data set of the air outlet anomaly detection model, and output an abnormality corresponding to the standard data set; A display module is connected to the data processing module for displaying the abnormal situation.

10. A storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a processor, the method for detecting abnormal conditions of a blast furnace tuyere as described in any one of claims 1 to 8 is implemented.

Images