A blast furnace molten iron temperature on-line detection method and system

Abstract

The application discloses a kind of blast furnace molten iron temperature on-line detection method and system, blast furnace molten iron temperature on-line detection method, including following process: the RGB image of the molten iron collected is carried out gray scale processing according to temperature interval, obtains first gray scale image;The first gray scale image is carried out space transformation and is carried out interpolation to first gray scale image using interpolation algorithm and obtains second gray scale;Second gray scale is carried out low-pass filtering;The gray scale of second gray scale after low-pass filtering is carried out image enhancement, and third gray scale is obtained;The molten iron temperature of the RGB image of the molten iron collected time is obtained by using the convolution neural network model that has been trained to third gray scale processing.The application realizes blast furnace molten iron temperature on-line detection by the way of molten iron RGB image processing, avoids approaching molten iron by artificial to detect molten iron temperature, and the detection result is relatively accurate.

Description

Technical Field

[0001] This invention relates to the field of online blast furnace molten iron temperature detection technology, and in particular to an online blast furnace molten iron temperature detection method and system. Background Technology

[0002] In the steelmaking process, the quality of molten iron, the energy consumption level and stability of the furnace are reflected in the molten iron temperature information. Online monitoring of molten iron temperature is an important means to ensure product specifications, improve product quality, and increase the yield of high-quality products. However, the blast furnace tapping area environment is complex and variable, and the raw material quality fluctuates frequently. There is a lack of accurate online detection methods for molten iron temperature information both domestically and internationally. Currently, the main methods used in production are: 1. Manual handheld thermocouple detection; 2. Blackbody cavity temperature measurement. Both of these methods require manual insertion of equipment into the molten iron or sampling near the tapping spout, which is fraught with safety hazards. The ambient temperature, reaching 40-43℃, undoubtedly increases the labor intensity of workers and also increases labor costs, making it difficult to meet the needs of high-quality and safe development in the steel industry. Therefore, how to achieve unmanned online detection of molten iron temperature is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0003] To address the problems existing in the prior art, the present invention aims to provide an online detection method and system for blast furnace molten iron temperature. The present invention achieves online detection of blast furnace molten iron temperature by processing RGB images of molten iron, avoiding the need for manual approach to the molten iron to detect the temperature, while also providing more accurate detection results.

[0004] The technical solution adopted in this invention is as follows:

[0005] A method for online detection of molten iron temperature in a blast furnace includes the following steps:

[0006] The RGB image of the collected molten iron is processed into grayscale to obtain a first grayscale image. The formula for grayscale processing is as follows;

[0007]

[0008] In the formula, g(m,n) is the color pixel value of the image after grayscale processing, R(m,n) is the red pixel value, G(m,n) is the green pixel value, B(m,n) is the blue pixel value, and T is the temperature of the molten iron at the time of acquiring the RGB image of the molten iron.

[0009] A second grayscale image is obtained by performing a spatial transformation on the first grayscale image and then interpolating the first grayscale image using an interpolation algorithm.

[0010] Apply a low-pass filter to the second grayscale image;

[0011] Image enhancement is performed on the grayscale image obtained after low-pass filtering of the second grayscale image to obtain the third grayscale image;

[0012] The third grayscale image is processed using a pre-trained convolutional neural network model to obtain the temperature of the molten iron at the moment the RGB image of the molten iron is captured.

[0013] Preferably, when the size of the acquired RGB image is large, the acquired RGB image of molten iron needs to be cropped first, and the cropped RGB image is then processed into grayscale.

[0014] Preferably, the grayscale image obtained by low-pass filtering of the second grayscale image is enhanced using the following formula to obtain grayscale image O:

[0015]

[0016] In the formula, low(m,n) represents the pixel value of the image after low-pass filtering, and R0 represents the passband radius.

[0017] Preferably, the training process of the convolutional neural network model includes:

[0018] The convolutional neural network model was trained using an image-temperature database to obtain a trained convolutional neural network model.

[0019] The image-temperature database includes the third grayscale images with different temperature labels. The third grayscale images with the same temperature label include several third grayscale images corresponding to RGB images of molten iron acquired at different angles.

[0020] Preferred method: After processing the third grayscale image using a pre-trained convolutional neural network model to obtain the molten iron temperature at the moment the RGB image of the molten iron is acquired, temperature compensation is then performed on the molten iron temperature.

[0021] Preferably, the compensation temperature for molten iron is calculated using the following formula:

[0022]

[0023] In the formula, T c To compensate for the temperature value, γ is the iron temperature loss coefficient, which is between 0.75 and 0.9, and V is the effective volume of the blast furnace.

[0024] The present invention also provides an online blast furnace hot metal temperature detection system, comprising:

[0025] First grayscale processing module: used to perform grayscale processing on the acquired RGB image of molten iron to obtain a first grayscale image. The formula for grayscale processing is as follows;

[0026]

[0027] In the formula, g(m,n) is the color pixel value of the image after grayscale processing, R(m,n) is the red pixel value, G(m,n) is the green pixel value, B(m,n) is the blue pixel value, and T is the temperature of the molten iron at the time of acquiring the RGB image of the molten iron.

[0028] The second grayscale processing module is used to perform spatial transformation on the first grayscale image and interpolate the first grayscale image using an interpolation algorithm to obtain a second grayscale image.

[0029] Filtering module: Used to perform low-pass filtering on the second grayscale image;

[0030] The third grayscale processing module is used to enhance the grayscale image after the second grayscale image has been low-pass filtered to obtain the third grayscale image.

[0031] Calculation module: Used to process the third grayscale image using a pre-trained convolutional neural network model to obtain the temperature of the molten iron at the moment the RGB image of the molten iron is captured.

[0032] Preferably, the online blast furnace hot metal temperature detection system of the present invention further includes:

[0033] Data acquisition module: used to acquire RGB images of molten iron at the molten iron trough of the blast furnace.

[0034] Preferably, the calculation module includes a first calculation module and a second calculation module:

[0035] First calculation module: used to process the third grayscale image to obtain the temperature of the molten iron at the moment the RGB image of the molten iron was captured;

[0036] The second calculation module is used to perform temperature compensation on the molten iron temperature at the moment of acquiring the RGB image of the molten iron.

[0037] Preferably, the second calculation module calculates the compensation temperature for the molten iron using the following formula:

[0038]

[0039] In the formula, T c To compensate for the temperature value, γ is the iron temperature loss coefficient, which is between 0.75 and 0.9, and V is the effective volume of the blast furnace.

[0040] The present invention has the following beneficial effects:

[0041] In this invention, the online blast furnace molten iron temperature detection method directly uses RGB images instead of the infrared images used in conventional measurement methods. This reduces equipment requirements, as a common optical camera suffices. When processing the acquired RGB images of molten iron for grayscale, the image is processed according to temperature ranges. This method improves the accuracy of the molten iron temperature detection results. Spatial transformation is applied to the first grayscale image to correct system errors and random instrument position errors introduced by the image acquisition system during image acquisition. High frequencies represent a wide color range; low-pass filtering filters out high-frequency signals. Since the RGB images of molten iron have smooth color transitions and a large number of low-frequency signals are the image features, low-pass filtering is applied to the second grayscale image to remove high-frequency signals. This invention achieves online blast furnace molten iron temperature detection through RGB image processing, avoiding manual proximity to the molten iron for temperature measurement. The detection results are more accurate, and the equipment requirements are lower compared to infrared detection. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute a limitation of the present invention.

[0043] Figure 1 This is a schematic diagram of the overall structure of the online blast furnace hot metal temperature detection system in an embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of the installation of the online blast furnace molten iron temperature detection system in an embodiment of the present invention.

[0045] In the diagram: 1. High-temperature resistant glass cover; 2. Optical camera; 3. Fireproof cotton aluminum silicate insulation layer; 4. Chassis; 5. Raspberry Pi; 6. Cooling fan; 7. Battery module; 8. Computer terminal; 9. Iron trough; 10. Molten iron; 11. Swinging spout; 12. Molten iron ladle; 13. Molten iron ladle number; 14. Temperature measurement system; 15. Rotating mechanism; 16. Long shaft of swinging spout; 17. Rail. Detailed Implementation

[0046] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0047] Example 1

[0048] This embodiment of the blast furnace molten iron temperature online detection system is a blast furnace molten iron temperature online detection system based on Raspberry Pi and convolutional neural network. The system for online molten iron temperature detection includes a high temperature protection layer, a Raspberry Pi control terminal and a computer terminal. The Raspberry Pi control terminal and the computer terminal communicate wirelessly.

[0049] The high-temperature protective layer includes a high-temperature resistant glass cover 1, a fireproof cotton aluminum silicate insulation layer 3, and a chassis 4. The fireproof cotton aluminum silicate insulation layer 3 covers the outside of the chassis 4, and the high-temperature glass cover 1 is installed at the front of the chassis 4 (e.g., Figure 1 (shown on the left end);

[0050] The Raspberry Pi control unit includes a Raspberry Pi 5, an optical camera 2 connected to the Raspberry Pi 5, a battery module 7, a rotating mechanism 15, and a cooling fan 6; the optical camera 2 can be a standard optical camera. In this embodiment, the optical camera 2 serves as an image acquisition module; the Raspberry Pi 5 is used for grayscale processing and filtering, and is an integration of the first, second, and third grayscale processing modules and the filtering module; the Raspberry Pi control unit is installed at a point on the long axis of the oscillating spout away from the spout's molten iron trough.

[0051] The computer terminal, deployed in a control room far from the site, is used to receive images and temperature information of molten iron transmitted from the site, display the location information of the molten iron ladle, and interact with the Raspberry Pi control terminal. In this embodiment, the computer terminal acts as a computing module, processing the obtained grayscale images and storing the established image-temperature database.

[0052] See Figures 1-2 The online detection method for blast furnace molten iron temperature in this embodiment includes the following steps:

[0053] 1) Establishment of image-temperature database:

[0054] ① Take molten iron samples from 9 locations in the molten iron ditch, place them in containers, take photos of the molten iron, and measure the temperature of the molten iron at that moment, with a measurement accuracy of 0.1℃;

[0055] ② Use OpenCV on Raspberry Pi 5 to crop the molten iron image to a size of m*n pixels; the following formula is used to crop the acquired RGB image:

[0056]

[0057] The image is processed to obtain a grayscale image G (i.e., the first grayscale image mentioned above in this invention), where g(m,n) are the color pixel values ​​of the image after grayscale processing, R(m,n) are the red pixel values, G(m,n) are the green pixel values, B(m,n) are the blue pixel values, and T is the temperature of the molten iron at the time of acquiring the RGB image of the molten iron. Experiments show that the prediction effect is significantly improved after processing the RGB image according to the temperature range.

[0058] ③ Perform spatial transformation on the grayscale image G to correct the systematic error caused by the image acquisition system and the random error of the instrument position. Then, use an interpolation algorithm to interpolate the grayscale image to obtain the grayscale image H (i.e., the second grayscale image mentioned above in this invention).

[0059] ④ Perform low-pass filtering on the grayscale image H, and use the formula

[0060]

[0061] Image enhancement is performed to obtain grayscale image O (i.e., the third grayscale image mentioned above in this invention), where low(m,n) is the pixel value of the image after low-pass filtering, and R0 represents the passband radius. Since high frequency represents a large color range, low-pass filtering filters high frequency signals. This invention uses molten iron images, where the image color transition is smooth, and a large number of low-frequency signals are the image features, so high frequency signals are filtered out.

[0062] ⑤ Input the temperature of the molten iron at the moment the RGB image of the molten iron is acquired in the computer terminal, and create this temperature folder to store the corresponding grayscale image O of the molten iron at that moment. Acquire 10 grayscale images O at different angles at each temperature to extract different features; acquire 10 grayscale images O at different angles to improve prediction accuracy.

[0063] ⑥ Allow the molten iron sample to cool naturally in the container to 1480.0℃, and collect the molten iron temperature and grayscale image O during the cooling process. Then heat it to 1520.0℃ and collect the molten iron temperature and grayscale image O during the process. A total of 400 sets of molten iron temperature image information were collected, with 10 photos in each set (collected at the same temperature), for a total of 4000 molten iron grayscale images O, and labeled with the corresponding temperature label.

[0064] 2) Control the Raspberry Pi rotation mechanism through the computer terminal, adjust the camera angle to capture photos of molten iron, and process the image through steps ②, ③ and ④ in step 1) to obtain the grayscale image O of the molten iron at this moment.

[0065] 3) The steps for building a convolutional neural network model to predict molten iron temperature include:

[0066] a. Establish a sequential module;

[0067] b. After establishing a convolutional layer, max pooling is used, and finally Dropout is used to discard some neurons. The model has a total of 11 convolutional layers and 10 fully connected layers.

[0068] c. Input the 4000 images of molten iron obtained in step 1)⑥ into the convolutional neural network, iterate 100 times to train the model, and obtain the trained convolutional neural network.

[0069] d. Input the grayscale image O of the molten iron obtained in step 2) into the trained convolutional neural network, and calculate the temperature information of the molten iron at this moment through the trained convolutional neural network;

[0070] e. The convolutional neural network program presets a molten iron temperature compensation model to compensate for the temperature loss of molten iron from the tapping spout to the swinging chute, and finally obtains more accurate molten iron temperature information in the furnace.

[0071] The hot metal temperature compensation model includes the following steps:

[0072] 1) Measure the distance d from the blast furnace taphole to the swing chute;

[0073] 2) The temperature loss coefficient γ of molten iron with distance is obtained by fitting the algorithm;

[0074] 3) Through the formula The compensated temperature is obtained;

[0075] f. Display the current molten iron temperature information on the display module.

Claims

1. A method for on-line detection of the temperature of molten iron in a blast furnace, characterized by, The method comprises the following steps: carrying out gray processing on the collected molten iron RGB image to obtain a first gray image, and the formula of the gray processing is as follows: wherein g(m, n) is the color pixel value of the image after the gray processing, R(m, n) is the red pixel value, G(m, n) is the green pixel value, B(m, n) is the blue pixel value, and T is the temperature of the molten iron at the time of collecting the RGB image of the molten iron; carrying out space transformation on the first gray image and carrying out interpolation on the first gray image by using an interpolation algorithm to obtain a second gray image; carrying out low-pass filtering on the second gray image; carrying out image enhancement on the gray image after the low-pass filtering of the second gray image to obtain a third gray image; and the image enhancement on the gray image after the low-pass filtering of the second gray image is carried out by using the following formula to obtain a gray image O: In the formula, low(m, n) is the pixel value of the image after low-pass filtering, denotes the passband radius; processing the third gray image by using a trained convolutional neural network model to obtain the temperature of the molten iron at the time of collecting the RGB image of the molten iron, and then carrying out temperature compensation on the temperature of the molten iron, and the compensation temperature of the molten iron is calculated by the following formula: In the formula, To compensate for temperature values; The molten iron temperature loss coefficient is 0.75-0.9; The effective volume of the blast furnace.

2. The method for on-line detection of the temperature of molten iron in a blast furnace according to claim 1, characterized in that, carrying out clipping on the collected RGB image of the molten iron first, and then carrying out gray processing on the clipped RGB image.

3. The method for on-line detection of the temperature of molten iron in a blast furnace according to claim 1, characterized in that, The training process of the convolutional neural network model comprises: training the convolutional neural network model by using an image-temperature database to obtain the trained convolutional neural network model; the image-temperature database comprises the third gray images with different temperature labels, and the third gray images with the same temperature label comprise a plurality of third gray images corresponding to the RGB images of the molten iron collected at different angles.

4. A high-temperature molten iron on-line detection system for a blast furnace, characterized by, comprise: a first gray processing module: configured to carry out gray processing on the collected RGB image of the molten iron to obtain a first gray image, and the formula of the gray processing is as follows: wherein g(m, n) is the color pixel value of the image after the gray processing, R(m, n) is the red pixel value, G(m, n) is the green pixel value, B(m, n) is the blue pixel value, and T is the temperature of the molten iron at the time of collecting the RGB image of the molten iron; a second gray processing module: configured to carry out space transformation on the first gray image and carry out interpolation on the first gray image by using an interpolation algorithm to obtain a second gray image; a filtering module: configured to carry out low-pass filtering on the second gray image; a third gray processing module: configured to carry out image enhancement on the gray image after the low-pass filtering of the second gray image to obtain a third gray image; and the image enhancement on the gray image after the low-pass filtering of the second gray image is carried out by using the following formula to obtain a gray image O: In the formula, low(m, n) is the pixel value of the image after low-pass filtering, denotes the passband radius; a calculation module: configured to process the third gray image by using a trained convolutional neural network model to obtain the temperature of the molten iron at the time of collecting the RGB image of the molten iron, and then carry out temperature compensation on the temperature of the molten iron, and the compensation temperature of the molten iron is calculated by the following formula: In the formula, is the compensation temperature value; is the molten iron temperature loss coefficient, taking a value between 0.75 and 0.9; is the effective volume of the blast furnace.

5. The on-line detection system for the temperature of molten iron in a blast furnace according to claim 4, characterized in that, further comprise: a data acquisition module: configured to collect the RGB image of the molten iron at the molten iron channel of the blast furnace.

6. The on-line detection system for the temperature of molten iron in a blast furnace according to claim 4, characterized in that, the calculation module comprises a first calculation module and a second calculation module: the first calculation module: configured to process the third gray image to obtain the temperature of the molten iron at the time of collecting the RGB image of the molten iron; the second calculation module: configured to carry out temperature compensation on the temperature of the molten iron at the time of collecting the RGB image of the molten iron.

Images