Method and device for identifying molten steel at the end of a long nozzle

Abstract

The present invention discloses a method and device for identifying molten steel at the end of a shroud. These devices are used to identify molten steel and determine when to activate the automatic tundish level control program. During the production process, an image of the shroud above the tundish is captured and uploaded in real time. A deep learning object detection algorithm is used to detect the presence of the shroud in the image. Once the shroud is detected, image processing techniques are used to extract the molten steel region at the end of the shroud in the image and determine whether the molten steel is flowing normally. When normal flow is determined, the automatic tundish level control program is activated. This method not only automatically determines the activation of automatic tundish level control, but also reduces the workload of operators.

Description

Technical Field

[0001] The invention belongs to the field of steelmaking and continuous casting in the metallurgical industry, and particularly relates to a method and device for identifying molten steel at the end of a shroud. Background Art

[0002] The tundish acts as a buffer between the ladle and the crystallizer, regulating the flow of molten steel and maintaining a stable liquid level. Since multiple furnaces of molten steel are typically poured continuously during the production process, the ladle needs to be frequently replaced and restarted. During ladle pouring, the turret moves the ladle to the pouring position, and a robotic arm places the shroud onto the sliding nozzle of the ladle. The nozzle is then opened to allow the molten steel to flow into the tundish. If no molten steel flows out, the shroud is removed and oxygen is burned. When the ladle is almost finished pouring, manual or slag detection systems determine whether the molten steel has reached the slag layer. If slag is detected, the sliding nozzle is closed, and the shroud is removed to replace the ladle.

[0003] In continuous casting production, tundish liquid level automatic control technology is increasingly being used. Existing tundish liquid level automatic control systems lack access to external information and are unable to automatically adjust the operating state of the control system based on actual production conditions. During actual production, the tundish liquid level must first be manually checked to meet the set value range before automatic tundish level control can begin. This not only prevents automatic judgment and activation of the tundish liquid level automatic control, but also increases the workload on workers. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for identifying molten steel at the end of a long nozzle, and a device for identifying molten steel at the end of a long nozzle based on the above-mentioned molten steel identification method. This method can not only automatically determine and start automatic control of the liquid level in the intermediate tank, but also reduce the workload of operators.

[0005] The technical solution adopted in the present invention is:

[0006] A method for identifying molten steel at the end of a shroud is used to identify molten steel and thus determine the timing for activating an automatic tundish liquid level control program: during the production process, an image of the shroud position above the tundish is captured and uploaded in real time, and a deep learning target detection algorithm is used to detect whether the shroud exists in the image. Once the shroud is detected in the image, image processing technology is used to extract the molten steel area at the end of the shroud in the image and identify whether the molten steel is flowing out normally. When it is identified that the molten steel is flowing out normally, the automatic tundish liquid level control program is activated.

[0007] Furthermore, a set of programs for capturing images of the shroud position above the tundish, detecting the shroud, extracting the molten steel area, and identifying the molten steel is set as an identification program. After the automatic control program for the tundish liquid level is turned on, the identification program is stopped and the ladle pouring stage is entered. After that, the large ladle is lowered to a low position and the shroud is extended into the tundish until the ladle changing stage is entered, at which time the identification program is started again.

[0008] Furthermore, a method for identifying molten steel using image processing technology is: first extract the ROI area at the end of the long nozzle in the image for subsequent molten steel identification, and then perform grayscale processing, binarization processing, morphological opening operation, and connected domain processing on the ROI area in turn, so as to extract the molten steel area, and finally identify whether the molten steel flows out normally based on the shape and size of the molten steel area.

[0009] Furthermore, a model building method for a deep learning target detection algorithm for detecting whether there is a long shroud in an image is as follows: first, a number of original images are taken at the long shroud position above the intermediate tank under different working conditions, and then the images are pre-processed to produce a data set, and then a neural network is trained with the goal of detecting the long shroud, and then the optimal model is saved.

[0010] A device for identifying molten steel at the end of a shroud is used to identify molten steel and thus determine when to start an automatic control program for the liquid level in an intermediate tank. The device comprises an industrial camera and a controller; the industrial camera is used to photograph the position of the shroud above the intermediate tank and upload the photographs to the controller; the controller adopts the control strategy of the above-mentioned method for identifying molten steel at the end of the shroud, controls the operation of the industrial camera according to the control strategy, and is electrically connected to a control module of the automatic control program for the liquid level in the intermediate tank. The device uploads the detection results of whether the shroud exists and the identification results of whether the molten steel is flowing out normally to the control module. When the control module identifies that the molten steel is flowing out normally, the control module starts the automatic control program for the liquid level in the intermediate tank.

[0011] Furthermore, the industrial camera is installed at the base of the turntable and is equipped with high-temperature protection measures.

[0012] The beneficial effects of the present invention are:

[0013] This method uses a deep learning target detection algorithm to detect long nozzles, and can quickly and accurately detect long nozzles. Since the molten steel flows into the tundish through the long nozzle, that is, the molten steel is located below the long nozzle, this method first detects whether the long nozzle exists and then extracts the molten steel area at the end of the long nozzle in the image. This can quickly and accurately locate the range of the molten steel, thereby reducing the interference of other highlighted targets in the background and achieving accurate segmentation of the molten steel. This method can automatically obtain the status of the long nozzle and molten steel, and determine the timing of starting the automatic control program of the tundish liquid level by identifying the molten steel. No human intervention is required throughout the process. It can not only automatically determine the start of the automatic control of the tundish liquid level, but also reduce the workload of the operator. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 1 is a schematic diagram of the process and corresponding image of the method for identifying molten steel at the end of the long nozzle in an embodiment of the present invention.

[0015] Figure 2 It is a schematic diagram of the process of model establishment and use of the deep learning target detection algorithm in the present invention.

[0016] Figure 3 It is a schematic diagram of the process of identifying molten steel in the ROI area and the corresponding image in the present invention. DETAILED DESCRIPTION

[0017] The present invention will be further described below with reference to the accompanying drawings and examples.

[0018] Example 1

[0019] This embodiment provides a method for identifying molten steel at the end of a long nozzle, which is used to identify the molten steel and determine the start time of the automatic control program of the liquid level in the intermediate tank: during the production process, such as Figure 1 As shown, an image of the shroud above the tundish is captured and uploaded in real time. A deep learning object detection algorithm is used to detect the presence of the shroud in the image. Once the shroud is detected, image processing techniques are used to extract the molten steel region at the end of the shroud in the image and determine whether the molten steel is flowing normally. When normal flow is determined, the tundish liquid level automatic control program is activated. This method uses a deep learning object detection algorithm to detect the shroud, enabling rapid and accurate detection of the shroud. Since molten steel flows into the tundish through the shroud, meaning it is located below the shroud, this method first detects the presence of the shroud and then extracts the molten steel region at the end of the shroud in the image. This allows for rapid and accurate location of the molten steel, thereby reducing interference from other bright background objects and achieving precise segmentation of the molten steel. This method automatically determines the status of the shroud and molten steel, and determines when to activate the tundish liquid level automatic control program by identifying the molten steel. This process requires no human intervention, enabling automatic determination and activation of the tundish liquid level automatic control program while also reducing the workload of operators.

[0020] Suppose a set of programs for capturing images at the shroud position above the tundish, detecting the shroud, extracting the molten steel area, and identifying the molten steel is a recognition program. Regarding the start and stop timing of the recognition program, in this embodiment: after the tundish liquid level automatic control program is turned on, the recognition program is stopped and the ladle pouring stage begins. After that, the large ladle is lowered to a low position and the shroud is extended into the tundish until the ladle changing stage begins, at which time the recognition program is started again. The recognition program works during the production process. It is stopped after the tundish liquid level automatic control program is turned on and started during the ladle changing stage. It is only started to identify the molten steel at key nodes to avoid disordered consumption.

[0021] The model building method of the deep learning target detection algorithm for detecting whether there is a long nozzle in the image is as follows: Figure 2 As shown in the figure, several original images are first taken at the shroud position above the tundish under different working conditions, and then the images are preprocessed to create a data set. Then, the neural network is trained with the goal of detecting the shroud, and the optimal model is saved.

[0022] The method of using image processing technology to identify molten steel is: Figure 3 As shown in the figure, the ROI area is first extracted at the end of the long nozzle in the image for subsequent molten steel recognition. Then, the ROI area is processed with grayscale, binarized, morphologically opened, and connected domains in sequence to extract the molten steel area. Finally, whether the molten steel flows out normally is identified based on the shape and size of the molten steel area.

[0023] Example 2

[0024] This embodiment provides a device for identifying molten steel at the end of a long nozzle, which is used to identify molten steel and thus determine the timing for starting the automatic control program for the liquid level in the intermediate tank: it includes an industrial camera and a controller; the industrial camera is used to photograph the position of the long nozzle above the intermediate tank and upload it to the controller; the controller adopts the control strategy of the method for identifying molten steel at the end of the long nozzle in Example 1, and controls the operation of the industrial camera according to the control strategy, and is electrically connected to the control module of the automatic control program for the liquid level in the intermediate tank (generally a PLC controller), and uploads the detection results of whether the long nozzle exists and the identification results of whether the molten steel is flowing out normally to the control module. When the control module identifies that the molten steel is flowing out normally, it starts the automatic control program for the liquid level in the intermediate tank.

[0025] In order to prevent high temperature from harming the camera, the industrial camera is installed on the base of the turntable and equipped with high temperature protection measures.

[0026] The embodiments described above are part of the embodiments of the present application, rather than all of the embodiments. The detailed description of the embodiments of the present application is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

Claims

1. A method for identifying molten steel at the end of a long nozzle, characterized in that: Used to identify molten steel and thus determine when to start the automatic control program for the liquid level in the tundish: During the production process, an image of the shroud position above the tundish is captured and uploaded in real time. A deep learning target detection algorithm is used to detect whether there is a shroud in the image. Once the shroud is detected in the image, image processing technology is used to extract the molten steel area at the end of the shroud in the image and identify whether the molten steel is flowing out normally. When it is identified that the molten steel is flowing out normally, the automatic control program for the liquid level in the tundish is started.

2. The method for identifying molten steel at the end of a shroud according to claim 1, wherein: A set of procedures for capturing images of the shroud position above the tundish, detecting the shroud, extracting the molten steel area, and identifying the molten steel is defined as the recognition program. Once the automatic tundish liquid level control program is turned on, the recognition program is stopped and the ladle pouring phase begins. Afterwards, the ladle is lowered to a low position and the shroud is extended into the tundish until the ladle replacement phase begins, at which point the recognition program is started again.

3. The method for identifying molten steel at the end of a shroud according to claim 1, wherein: The method of using image processing technology to identify molten steel is: first extract the ROI area at the end of the long nozzle in the image for subsequent molten steel identification, then perform grayscale processing, binarization processing, morphological opening operation, and connected domain processing on the ROI area in turn to extract the molten steel area, and finally identify whether the molten steel is flowing out normally based on the shape and size of the molten steel area.

4. The method for identifying molten steel at the end of a shroud according to claim 1, wherein: The model establishment method of the deep learning target detection algorithm for detecting whether there is a long shroud in the image is as follows: first, several original images are taken at the long shroud position above the tundish under different working conditions, and then the images are preprocessed to produce a data set. Then, the neural network is trained with the detection of the long shroud as the goal, and then the optimal model is saved.

5. A molten steel identification device at the end of a long nozzle, characterized by: The device is used to identify molten steel and thus determine the timing for starting an automatic control program for the liquid level in an intermediate tank, comprising an industrial camera and a controller; the industrial camera is used to photograph the position of the shroud above the intermediate tank and upload the photograph to the controller; the controller adopts the control strategy of the method for identifying molten steel at the end of the shroud as described in any one of claims 1 to 4, controls the operation of the industrial camera according to the control strategy, and is electrically connected to a control module of the automatic control program for the liquid level in the intermediate tank, and uploads the detection results of whether the shroud exists and the identification results of whether the molten steel is flowing out normally to the control module, and the control module starts the automatic control program for the liquid level in the intermediate tank when it is identified that the molten steel is flowing out normally.

6. The molten steel identification device at the end of the shroud as claimed in claim 5, characterized in that: The industrial camera is installed at the base of the turntable and is equipped with high-temperature protection measures.

Images