Method for discriminating cold and hot cells of aluminum electrolysis cell based on image recognition

Abstract

The invention discloses an aluminum electrolysis cell cold and hot cell distinguishing method based on image recognition, which comprises the following steps: carrying out image acquisition on a fire eye area for crust breaking and blanking of an aluminum electrolysis cell in production, preprocessing the acquired image, and carrying out feature and image recognition on the preprocessed image; and establishing a flame characteristic performance evaluation model according to the extracted and identified characteristics, and dividing the state of the electrolytic cell into a normal operation cell, a hot cell and a cold cell through the characteristic performance evaluation model. The cold and hot states of the electrolytic cell are judged by using an image recognition technology, so that the judgment efficiency and accuracy are greatly improved; the judgment result is objective and standardized, and the influence of human factors is reduced; the online detection of the working condition state of the aluminum electrolysis cell is realized, and the production efficiency and the product quality consistency are improved; the production process can be timely adjusted, and the current efficiency and operation efficiency of the aluminum electrolysis cell are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of heat dissipation discrimination of anode covering materials of aluminum electrolytic cells, and relates to a method for discriminating hot and cold cells of aluminum electrolytic cells based on image recognition. Background Art

[0002] In the aluminum electrolysis production process, each electrolytic cell is equipped with a certain number of point feeders, and the normal feeding interval is in the range of 1.5~3min. The point feeder system is combined with a computer control system, and the feeder quantity can be adjusted by automatically adjusting the feeder interval through the control system, thus forming a variety of quasi-continuous "feeder on demand" technologies to meet the requirements of modern aluminum electrolysis process for alumina concentration control.

[0003] The electrolytic cell has intact Al 2 O 3 The crust can cover the periphery and sides of the anode to reduce the oxidation and burning of the anode. At the same time, it is a poor conductor of heat and can act as an insulator to reduce heat loss in the tank.

[0004] When the electrolytic cell is unloaded, the shell is first broken to expose the melt area of ​​the electrolytic cell. 2 O 3 In order to better dissolve the electrolyte. Such operation will cause flames to be generated at the shelling and unloading point of the electrolytic cell and accompanied by the generation of flames.

[0005] The flame condition of the aluminum electrolysis cell is an important parameter reflecting the state of the aluminum electrolysis process. It can reflect the thermal state of the electrolysis cell (hot and cold cell) and is directly related to the current efficiency of the electrolysis cell. Figure 3 The flame condition shown. Traditional judgment methods mostly rely on the experience of workers and lack objectivity and accuracy. In order to overcome this shortcoming, the present invention proposes a method for distinguishing hot and cold tanks from the flame eye condition of aluminum electrolytic cells based on visual image recognition technology to achieve fast and accurate tank condition judgment. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a method for distinguishing hot and cold tanks in an aluminum electrolytic cell based on image recognition, so as to solve the problems existing in the prior art.

[0007] The scheme implemented by the present invention is: A method for distinguishing hot and cold aluminum electrolytic cells based on image recognition, which collects images of the flame eye area of ​​the aluminum electrolytic cell during shelling and unloading in production and preprocesses the collected images, performs feature and image recognition on the preprocessed images, establishes a characteristic performance evaluation model of the flame according to the extracted and recognized features, and distinguishes the status of the electrolytic cell into a normal operating cell, a hot cell and a cold cell through the characteristic performance evaluation model.

[0008] A method for distinguishing hot and cold tanks in an aluminum electrolytic cell based on image recognition specifically comprises the following steps: Step 1: Image acquisition: image acquisition is performed on the fire eye area of ​​the aluminum electrolytic cell for shelling and unloading, and a flame image of the fire eye area is obtained; Step 2: Image processing, pre-processing the collected images, and distinguishing the flame area, anode covering material area, and anode steel claw area according to the images; Step 3: Feature extraction: extract edge, corner and ridge image features from the preprocessed flame area flame image, and analyze the flame core, inner flame and outer flame area according to the edge, corner and ridge image features to obtain the width, height, size, color and brightness of the flame; Step 4: According to the width, height, size, color and brightness of the flame core, inner flame and outer flame, the characteristic performance evaluation model is used to determine whether the electrolytic cell is a normal operating cell, a hot cell or a cold cell.

[0009] Furthermore, the characteristic performance evaluation model in the above step 4 uses a machine learning model to distinguish between hot and cold aluminum electrolytic cells.

[0010] Furthermore, the preprocessing in the above step 2 includes grayscale conversion, filtering and denoising, and edge enhancement.

[0011] Furthermore, the images collected above are divided into hot slots through machine learning if the flames are yellow, there is no rising flame and there is local smoke, or there is rising white smoke and dazzling red light; the images collected are divided into cold slots through machine learning if the flames are weak and blue-white in color; and the images collected are divided into normal slots through machine learning if the flames are light blue and strong.

[0012] A method for distinguishing hot and cold tanks of aluminum electrolytic cells based on image recognition also includes optimizing and adjusting the normal operation tank, hot tank or cold tank according to the judgment, wherein the adjustment method of the cold tank is to find out the cause, eliminate the factors leading to the cold tank, and restore the thermal balance to normal, specifically: 1) Insufficient energy input caused by low cell voltage setting or control value. Find out the cause of this problem and restore the working voltage to the normal range, or adjust the working voltage to a level slightly higher than the normal range to speed up the recovery of thermal balance. When the thermal balance is restored to the set level, adjust the cell voltage to the normal range. 2) If the amount of aluminum fluoride added is too large, the amount of aluminum output is too small, or the anode covering material is insufficient (the height of the anode covering material is less than 100mm), resulting in excessive energy expenditure, then find out the causes of these problems and restore the amount of aluminum fluoride added, the amount of aluminum output or the thickness of the anode covering material to the normal range; 3) If the manual operation of pole changing is not standardized (such as the pole changing time is too long, exceeding 30 minutes; or the additional material feeding is too much, exceeding 5% of the normal average value), causing excessive heat dissipation, the working voltage should be temporarily increased appropriately, the anode insulation should be strengthened, or the reference material feeding interval should be extended to restore the thermal balance to normal; The adjustment method of the cold tank is: for the initial hot tank, find out and eliminate its cause to restore the electrolytic cell to normal, specifically: 1) If the slot voltage is high (i.e. the slot voltage is higher than 2.5% of the normal average value) or there is room for the slot voltage to be reduced, the slot voltage setting value should be appropriately reduced; 2) If the electrolyte molecular ratio is too high (i.e. the electrolyte molecular ratio is 5% higher than the normal average set value), then increase the amount of aluminum fluoride added appropriately; 3) If the aluminum liquid height in the electrolytic tank is too low (the aluminum liquid height in the electrolytic tank is lower than 18 cm, or the aluminum liquid height is 10% lower than the average aluminum liquid height in the electrolytic tank in the work area), the aluminum output should be appropriately reduced; 4) If the material feeding control is abnormal (i.e. the daily material feeding amount of the electrolytic cell is 5% higher than the normal average), it should be restored to normal as soon as possible; The hot-slot symptoms can be gradually eliminated by restoring the deviated technical conditions to normal.

[0013] Effects of the invention: The invention uses image recognition technology to judge the hot and cold states of the electrolytic cell, greatly improving the efficiency and accuracy of the judgment; the judgment results are objective and standardized, reducing the influence of human factors; it realizes the online detection of the working state of the aluminum electrolytic cell, improving the production efficiency and product quality consistency; it is conducive to timely adjustment of the production process and improving the current efficiency and operation efficiency of the aluminum electrolytic cell. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the anode cover and fire eye area of ​​the aluminum electrolysis cell; Figure 1 In the figure, 1: electrolytic cell; 2: anode cover; 3: anode; 4: flame; Figure 2 is a schematic diagram of the side structure of the image acquisition device; Figure 3 It is a structural schematic diagram of the image acquisition device. DETAILED DESCRIPTION

[0015] The present invention is further described below in conjunction with specific embodiments.

[0016] Example 1: Figure 1-3As shown, a method for distinguishing hot and cold aluminum electrolytic cells based on image recognition is provided, which collects images of the fire eye area of ​​the aluminum electrolytic cell during shelling and unloading in production, preprocesses the collected images, partitions the fire eye area of ​​the shelling and unloading, distinguishes the covering material layer and the flame, performs feature and image recognition on the preprocessed images, establishes a feature performance evaluation model of the flame 4 according to the extracted and recognized features, and distinguishes the status of the electrolytic cell 1 into a normal operating cell, a hot cell and a cold cell through the feature performance evaluation model.

[0017] A method for distinguishing hot and cold tanks in an aluminum electrolytic cell based on image recognition specifically comprises the following steps: Step 1: Image acquisition: image acquisition is performed on the fire eye area of ​​the aluminum electrolytic cell for shelling and unloading, and a flame image of the fire eye area is obtained; The image acquisition device used for image acquisition includes an industrial-grade high-definition camera 101, which is mounted on a mounting plate 103 through a flange 102. The mounting plate 103 is fixedly connected to the lower end of an electric push rod 104. The housing of the electric push rod 104 is fixedly connected to a horizontal sliding plate 105. The horizontal sliding plate 105 is connected to a horizontal beam 107 through two slider guide pairs 106. The horizontal beam 107 is mounted on one side of a crane beam 108. The horizontal sliding plate 105 is driven to move by a connected screw nut pair 109. The nut of the screw nut pair 109 is fixedly connected to a nut seat at the bottom of the horizontal sliding plate 105. The screw of the screw nut pair 108 The two ends are fixedly connected to the horizontal beam 107 through the bearing seat, and one end is connected to the drive motor 110 after extending out of the bearing seat. The drive motor 110 is installed on the horizontal beam 107 through the motor frame. The two ends of the overhead crane beam 108 move through the walking mechanism 111. Through the walking mechanism, the industrial-grade high-definition camera 101 can move vertically, thereby realizing the acquisition of images of different electrolytic cells. The horizontal sliding plate driven by the driving motor moves horizontally (walking along the length direction of the electrolytic cell), which can realize the image acquisition of different positions of a single electrolytic cell. The telescopic mechanism of the electric push rod can realize the acquisition of the best distance control, which can avoid damage to the image acquisition equipment and realize the best image quality acquisition. When in use, open the electrolytic cell cover, control the position of the camera horizontally and vertically, and realize the image acquisition of different electrolytic cells and a single electrolytic cell.

[0018] Limit baffles 113 are provided at both ends of the horizontal beam 107, and travel limit switches 112 are installed on the limit baffles 113. When the slider of the slider guide pair 106 touches the travel limit switch 112, the driving motor stops moving, which can realize the control of the safe travel and play a protective role.

[0019] In order to achieve accurate acquisition, a plurality of vibration-damping grooves 114 are arranged on the horizontal beam 107. Steel balls 115 of different diameters are installed in each vibration-damping groove 114, which can reduce the vibration generated by the movement of the entire overhead crane or the operation of its motor, thereby improving the stability and accuracy of image acquisition. In addition, a shock-absorbing rubber pad 116 is arranged on the connecting surface between the horizontal beam 107 and the overhead crane beam 108 to further achieve a vibration isolation effect.

[0020] In order to save costs, the housing of the electric push rod 104 can also be fixedly connected to the walking mechanism of the shell breaking device on the overhead crane to achieve walking, thereby achieving positioning at different positions.

[0021] Step 2: Image processing, pre-processing the collected images, and distinguishing the flame area, anode covering material area, and anode steel claw area according to the images; Step 3: Feature extraction: extract edge, corner and ridge image features from the preprocessed flame area flame image, and analyze the flame core, inner flame and outer flame area according to the edge, corner and ridge image features to obtain the width, height, size, color and brightness of the flame; Step 4: According to the width, height, size, color and brightness of the flame core, inner flame and outer flame, the characteristic performance evaluation model is used to determine whether the electrolytic cell is a normal operating cell, a hot cell or a cold cell.

[0022] Furthermore, the characteristic performance evaluation model in the above step 4 uses a machine learning model to distinguish between hot and cold aluminum electrolytic cells.

[0023] Furthermore, the preprocessing in the above step 2 includes grayscale conversion, filtering and denoising, and edge enhancement.

[0024] Furthermore, the images collected above are divided into hot slots through machine learning if the flames are yellow, there is no rising flame and there is local smoke, or there is rising white smoke and dazzling red light; the images collected are divided into cold slots through machine learning if the flames are weak and blue-white in color; and the images collected are divided into normal slots through machine learning if the flames are light blue and strong.

[0025] like Figure 1 , image recognition of flame 4, judging the color and brightness of flame 4 to determine the cold or hot condition of electrolytic cell 1; when the flame of flame 4 is yellow and weak, the electrolytic cell is a hot cell; when the flame of flame 4 is weak and blue-white, it is a cold cell. The specific judgment is as follows: 1) Hot tank: yellow and weak flame: early stage of hot tank; no flame rising, local smoke: middle stage of hot tank; white smoke rising, bright red light: late stage of hot tank; the result of the electrolytic tank state of hot tank: the side legs are thin and short, or even no side legs, the aluminum liquid and electrolyte liquid are spread out, directly contacting the side lining. This kind of tank has a large amount of aluminum loss; second, it is easy to have side leakage, which greatly reduces the current efficiency; third, it is easy to burn through the side, causing side leakage.

[0026] 2) Cold tank: The flame is weak and the color is blue-white, which is a cold tank. The result of the electrolytic cell state of a cold tank is: the edges are enlarged and long, extending to below the anode, the furnace bottom is cold and prone to precipitation, the electrolyte temperature is too low and sticky, and the alumina solubility is poor. Over time, the furnace bottom will grow into a crust, making the electrolytic cell difficult to manage. In order to maintain production, the tank operating voltage has to be increased.

[0027] 3) Normal cell: The side legs of a normal cell are evenly distributed on the edge of the positive projection of the anode, the aluminum liquid is squeezed in the center of the cell, and the current passes vertically from the anode to the cathode. The electrolytic cell with this type of tank has stable technical conditions, is easy to manage, and has high current efficiency.

[0028] A method for distinguishing hot and cold tanks of aluminum electrolytic cells based on image recognition also includes optimizing and adjusting the normal operation tank, hot tank or cold tank according to the judgment, wherein the adjustment method of the cold tank is to find out the cause, eliminate the factors leading to the cold tank, and restore the thermal balance to normal, specifically: 1) Insufficient energy input caused by low cell voltage setting or control value. Find out the cause of this problem and restore the working voltage to the normal range, or adjust the working voltage to a level slightly higher than the normal range to speed up the recovery of thermal balance. When the thermal balance is restored to the set level, adjust the cell voltage to the normal range. 2) If the amount of aluminum fluoride added is too large, the amount of aluminum output is too small, or the anode covering material is insufficient (the height of the anode covering material is less than 100mm), resulting in excessive energy expenditure, then find out the causes of these problems and restore the amount of aluminum fluoride added, the amount of aluminum output or the thickness of the anode covering material to the normal range; 3) If the manual operation of pole changing is not standardized (such as the pole changing time is too long, exceeding 30 minutes; or the additional material feeding is too much, exceeding 5% of the normal average value), causing excessive heat dissipation, the working voltage should be temporarily increased appropriately, the anode insulation should be strengthened, or the reference material feeding interval should be extended to restore the thermal balance to normal; The adjustment method of the cold tank is: for the initial hot tank, find out and eliminate its cause to restore the electrolytic cell to normal, specifically: 1) If the slot voltage is high (i.e. the slot voltage is higher than 2.5% of the normal average value) or there is room for the slot voltage to be reduced, the slot voltage setting value should be appropriately reduced; 2) If the electrolyte molecular ratio is too high (i.e. the electrolyte molecular ratio is 5% higher than the normal average set value), then increase the amount of aluminum fluoride added appropriately; 3) If the aluminum liquid height in the electrolytic tank is too low (the aluminum liquid height in the electrolytic tank is lower than 18 cm, or the aluminum liquid height is 10% lower than the average aluminum liquid height in the electrolytic tank in the work area), the aluminum output should be appropriately reduced; 4) If the material feeding control is abnormal (i.e. the daily material feeding amount of the electrolytic cell is 5% higher than the normal average), the feeder should be repaired as soon as possible to restore it to normal; The hot-slot symptoms can be gradually eliminated by restoring the deviated technical conditions to normal.

[0029] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for distinguishing hot and cold tanks in aluminum electrolytic cells based on image recognition, characterized in that: By collecting images of the flame eye area of ​​the aluminum electrolytic cell during production and preprocessing the collected images, the preprocessed images are subjected to feature and image recognition, and a characteristic performance evaluation model of the flame (4) is established based on the extracted and recognized features. The characteristic performance evaluation model is used to distinguish the status of the electrolytic cell (1) into a normal operating cell, a hot cell, and a cold cell.

2. The method for distinguishing hot and cold tanks of an aluminum electrolytic cell based on image recognition according to claim 1, characterized in that: The specific method includes the following steps: Step 1: Image acquisition: image acquisition is performed on the fire eye area of ​​the aluminum electrolytic cell for shelling and unloading, and a flame image of the fire eye area is obtained; Step 2: Image processing, pre-processing the collected images, and distinguishing the flame area, anode covering material area, and anode steel claw area according to the images; Step 3: Feature extraction: extract edge, corner and ridge image features from the preprocessed flame area flame image, and analyze the flame core, inner flame and outer flame area according to the edge, corner and ridge image features to obtain the width, height, size, color and brightness of the flame; Step 4: According to the width, height, size, color and brightness of the flame core, inner flame and outer flame, the characteristic performance evaluation model is used to determine whether the electrolytic cell is a normal operating cell, a hot cell or a cold cell.

3. The method for distinguishing hot and cold aluminum electrolytic cells based on image recognition according to claim 2, characterized in that: In step 4, the feature performance evaluation model uses a machine learning model to distinguish between hot and cold aluminum electrolytic cells.

4. The method for distinguishing hot and cold tanks of aluminum electrolytic cells based on image recognition according to claim 1 is characterized in that: The preprocessing in step 2 includes grayscale, filtering denoising and edge enhancement.

5. The method for distinguishing hot and cold tanks of aluminum electrolytic cells based on image recognition according to claim 2 is characterized in that: Through machine learning, the collected images are divided into hot slots if the flames are yellow, there is no rising flame and there is local smoke, or there is rising white smoke and dazzling red light; the collected images are divided into cold slots if the flames are weak and blue-white in color through machine learning; the collected images are divided into normal slots if the flames are light blue and strong through machine learning.

6. The method for distinguishing hot and cold aluminum electrolytic cells based on image recognition according to claim 2, characterized in that: It also includes optimizing adjustments based on the normal operating tank, hot tank or cold tank. Among them, the adjustment method of the cold slot is to find out the cause, eliminate the factors that cause the cold slot, and restore the thermal balance to normal, specifically: 1) Insufficient energy input caused by low cell voltage setting or control value. Find out the cause of this problem and restore the working voltage to the normal range, or adjust the working voltage to a level slightly higher than the normal range to speed up the recovery of thermal balance. When the thermal balance is restored to the set level, adjust the cell voltage to the normal range. 2) If the amount of aluminum fluoride added is too large, the amount of aluminum output is too small, or the anode covering material is insufficient, resulting in excessive energy expenditure, then find out the causes of these problems and restore the amount of aluminum fluoride added, the amount of aluminum output or the thickness of the anode covering material to the normal range; 3) If the excessive heat dissipation is caused by improper manual operation of pole replacement, the working voltage should be temporarily increased appropriately, the anode insulation should be strengthened, or the reference material unloading interval should be extended to restore the thermal balance to normal; The adjustment method of the cold tank is: for the initial hot tank, find out and eliminate its cause to restore the electrolytic cell to normal, specifically: 1) If the slot voltage is high or there is room for reduction, reduce the slot voltage setting value; 2) If the electrolyte molecular ratio is too high, increase the amount of aluminum fluoride added; 3) If the aluminum liquid height in the electrolytic cell is too low, the aluminum output will be reduced; 4) If the material feeding control is abnormal, it will be restored to normal; The hot-slot symptoms can be gradually eliminated by restoring the deviated technical conditions to normal.

Images