Method for automatically identifying abnormity of hot rolling plate shape

Abstract

The invention discloses a hot rolling plate shape abnormity automatic identification method, and belongs to the field of AI computer vision, and the method comprises the steps: video collection and transmission: arranging a camera between hot rolling and finish rolling field racks, and collecting a strip steel plate shape video stream in real time, the video stream is transmitted to the AI intelligent identification system through the streaming media server or the switch or is directly transmitted to the AI intelligent identification system through the switch; the AI intelligent recognition system carries out frame extraction on the video stream, and point position coordinates of four corner points of the strip steel plate shape are recognized through a key point detection model; voice early warning is automatically recognized through the AI computer vision technology, the fluctuation value of a specific certain point of the plate shape is pointed out in time, accurate plate shape position control key points are provided, and rolling tends to be normal; and meanwhile, a plate shape abnormal position overrun signal can be triggered to send an L1 control signal, field equipment is triggered to act, field rolling is stopped, and serious equipment safety accidents caused by waste steel are avoided.

Description

Technical Field

[0001] The present invention relates to the field of AI computer vision technology, and in particular to a method for automatically identifying abnormal hot-rolled plate shapes. Background Art

[0002] In the conventional hot rolling process, when the strip is rolled in the finishing mill, due to defects in the incoming strip, equipment failure, process changes, etc., it may cause abnormal plate shapes such as the strip head warping, process breakage, large rolling, edge wave shape, and middle wave shape. Manual real-time monitoring of the camera video is required.

[0003] During the rolling process of the finishing mill, the strip shape of the finishing mill needs to be confirmed manually in real time. If the strip shape monitoring four-point position (such as Figure 2 There is a significant fluctuation among the four points ABCD (point A fluctuates). After manual monitoring is discovered, manual intervention is made to the roll gap between the stands to make the rolled plate shape between the stands return to normal, avoiding further deterioration and causing scrap steel accidents.

[0004] like Figure 1 : The plate shape between the racks is normal, and there is no obvious fluctuation at the four points.

[0005] Therefore, a method is needed to use computer vision technology to perform AI intelligent analysis of the four-point position of the strip shape, identify abnormal fluctuations in a timely manner, and push voice messages to on-site operators to achieve AI intelligent recognition of abnormal warnings; when the fluctuation exceeds the limit, the computer vision platform will simultaneously forward the AI ​​recognition abnormality signal to the L1 control system, triggering the finishing rolling stop signal, and avoiding the occurrence of scrap steel-induced equipment safety accidents. Automatic identification method of hot rolling plate shape abnormalities. Summary of the Invention

[0006] In view of the deficiencies in the prior art, the present invention provides a method for automatically identifying abnormal hot-rolled plate shapes, which solves the problems raised in the above-mentioned background technology.

[0007] Technical solution: To solve the above technical problems, according to one aspect of the present invention, more specifically, a method for automatically identifying abnormal hot-rolled plate shape, comprising the following steps:

[0008] S1. Video acquisition and transmission: A camera is installed between the stands of the hot rolling and finishing rolling site to capture the strip shape video stream in real time, and transmit the video stream to the AI ​​intelligent recognition system through a streaming media server or switch (or directly to the AI ​​intelligent recognition system through a switch);

[0009] S2. Flatness anomaly detection: The AI ​​intelligent recognition system extracts frames from the video stream and uses the key point detection model to identify the coordinates of the four corner points of the strip flatness. It monitors the up and down fluctuations of the four-point coordinates in real time. When the up and down fluctuations of the coordinates of any point exceed the set limit, the point is determined to be abnormal and the abnormal location information is obtained;

[0010] S3. Abnormal warning and linkage control: The AI ​​intelligent recognition system broadcasts the abnormal location information to the on-site operator in the form of voice. At the same time, when the fluctuation value exceeds the limit value, it sends an abnormal signal to the L1 control system. The L1 control system triggers the on-site equipment to stop rolling according to the abnormal signal.

[0011] Furthermore, the key point detection model adopts a high-resolution feature map maintenance network, connects multiple high-resolution to low-resolution sub-networks in parallel, learns by placing a Gaussian distribution at each corner point position, and uses the point with the largest response value on each layer of heatmap as the detected corner point coordinate.

[0012] Furthermore, the set limit is 20 mm, which is adjusted according to actual production process requirements.

[0013] Furthermore, the voice warning includes broadcasting the specific location of the abnormal corner point and the fluctuation value, providing the operator with clear intervention guidance.

[0014] Furthermore, after receiving the abnormal signal, the L1 control system triggers the emergency stop button of the finishing mill to achieve an emergency stop of the on-site rolling.

[0015] Furthermore, it also includes a historical data storage step. The AI ​​intelligent recognition system will store the detected corner point coordinates, fluctuation values, abnormal records and other information for subsequent production data analysis and model optimization.

[0016] Furthermore, the key point detection model performs the following steps: first, it receives the signal and determines whether to perform a frame extraction operation. If the frame extraction conditions are met, frame extraction is performed, and the image after frame extraction is input into the detection model. After processing by the post-processing module, it is determined whether the four-point positions of the plate shape are abnormal; if the frame extraction conditions are not met, the historical record is stored.

[0017] The beneficial effects of the method for automatically identifying abnormal hot-rolled plate shape of the present invention are:

[0018] The present invention uses AI computer vision technology to automatically identify voice warnings, promptly point out the fluctuation value of a specific point in the plate shape, and provide accurate plate position control points to make rolling return to normal; at the same time, it can trigger an abnormal plate position over-limit signal to the L1 control signal, triggering the on-site equipment to stop on-site rolling, and avoid serious equipment safety accidents caused by scrap steel. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0020] Figure 1 Schematic diagram of the four-point position of the finishing mill shape monitoring system (points A, B, C, and D) of the present invention;

[0021] Figure 2 Schematic diagram comparing the normal and abnormal plate shapes of the present invention;

[0022] Figure 3 This is a diagram of the architecture of the automatic abnormality identification system of the present invention;

[0023] Figure 4 This is a flowchart of the AI ​​intelligent recognition system algorithm implementation of the present invention;

[0024] Figure 5 Schematic diagram of the network structure of the key point detection model of the present invention;

[0025] Figure 6 The overall system architecture and signal transmission flow chart of the present invention;

[0026] Figure 7 Schematic diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0028] In order to make the technical solution of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] Reference Figure 1-Figure 7 ,A method for automatically identifying abnormal hot-rolled plate shape, video acquisition and transmission: a camera is set up between the stands at the hot-rolled finishing rolling site to capture the strip shape scene in real time, and the video stream is transmitted to the AI ​​intelligent recognition system through a streaming server or switch.

[0030] Flatness Anomaly Identification: The AI-powered recognition system extracts frames from the video stream using a key point detection model to identify the coordinates of the four corner points (A, B, C, and D) of the strip shape, and monitors the fluctuations of these coordinates in real time. If the fluctuation of any coordinate point exceeds a set limit (e.g., 20mm), the system automatically identifies the abnormal location.

[0031] The key point detection model adopts a high-resolution feature map maintenance network and connects multiple high-resolution to low-resolution sub-networks in parallel. It learns by placing a Gaussian distribution at each corner point and uses the point with the largest response value on each layer of heatmap as the detected key point to improve spatial detection accuracy.

[0032] Abnormal warning and handling:

[0033] Voice warning: The identified abnormal position information will be broadcast in voice form and pushed to the on-site operators to guide them to intervene in time to adjust the stand roll gap to make the rolled plate shape return to normal.

[0034] Linkage control: When the four-point fluctuation value of the plate shape exceeds the limit value (scrap steel risk), the AI ​​intelligent recognition system sends an abnormal signal to the L1 control system. The L1 control system triggers the on-site equipment to stop rolling and avoid equipment safety accidents caused by scrap steel.

[0035] Example

[0036] like Figure 7 As shown, the present invention is currently implemented in the 2250 production line of Zhanjiang Hot Rolling Mill. The abnormal information system promptly notifies the site through voice, reducing the labor intensity of manual real-time recognition of the finished rolled plate shape and improving the recognition accuracy.

[0037] The specific implementation method is as follows:

[0038] 1. Abnormalities at the four points of the plate shape are identified in real time and reported to the operator in the form of voice to guide corrections, so that the rolled plate shape between the stands tends to be normal, avoiding further deterioration and causing scrap steel accidents.

[0039] 2. If the scrap steel exceeds the limit, the L1 stop signal can be triggered in real time to avoid safety accidents of on-site equipment.

[0040] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A method for automatically identifying abnormal hot-rolled plate shape, characterized in that: The following steps are involved: S1. Video acquisition and transmission: A camera is installed between the stands of the hot rolling and finishing rolling site to capture the strip shape video stream in real time, and transmit the video stream to the AI ​​intelligent recognition system through a streaming media server or switch; S2. Flatness anomaly detection: The AI ​​intelligent recognition system extracts frames from the video stream and uses the key point detection model to identify the coordinates of the four corner points of the strip flatness. It monitors the up and down fluctuations of the four-point coordinates in real time. When the up and down fluctuations of the coordinates of any point exceed the set limit, the point is determined to be abnormal and the abnormal location information is obtained; S3. Abnormal warning and linkage control: The AI ​​intelligent recognition system broadcasts the abnormal location information to the on-site operator in the form of voice. At the same time, when the fluctuation value exceeds the limit value, it sends an abnormal signal to the L1 control system. The L1 control system triggers the on-site equipment to stop rolling according to the abnormal signal.

2. The method for automatically identifying abnormal hot-rolled plate shape according to claim 1, characterized in that: The key point detection model adopts a high-resolution feature map maintenance network, connects multiple high-resolution to low-resolution sub-networks in parallel, and learns by placing a Gaussian distribution at each corner point position. The point with the largest response value on each layer of heatmap is used as the detected corner point coordinate.

3. The method for automatically identifying abnormal hot-rolled plate shape according to claim 1, characterized in that: The set limit is 20mm, which is adjusted according to actual production process requirements.

4. The method for automatically identifying abnormal hot-rolled plate shape according to claim 1, characterized in that: The voice warning includes broadcasting the specific location of the abnormal corner point and the fluctuation value, providing clear intervention guidance for the operator.

5. The method for automatically identifying abnormal hot-rolled plate shape according to claim 1, characterized in that: After receiving the abnormal signal, the L1 control system triggers the emergency stop button of the finishing mill to achieve an emergency stop of the on-site rolling.

6. The method for automatically identifying abnormal hot-rolled plate shape according to claim 1, characterized in that: It also includes a historical data storage step. The AI ​​intelligent recognition system will store the detected corner point coordinates, fluctuation values ​​and abnormal record information for subsequent production data analysis and model optimization.

7. The method for automatically identifying abnormal hot-rolled plate shape according to claim 1, characterized in that: The key point detection model performs the following steps: first, receiving a signal, determining whether to perform a frame extraction operation, and if the frame extraction conditions are met, performing frame extraction. The frame-extracted image is input into the detection model, and after processing by the post-processing module, determining whether the positions of the four points of the plate shape are abnormal; if the frame extraction conditions are not met, storing the historical records.

Images