Submerged chain conveyor chain breakage detection system and detection method based on deep learning

Abstract

The invention discloses a submerged chain conveyor chain breakage detection system and detection method based on deep learning. The system comprises an image acquisition device, an image processing unit, a judgment unit and an alarm unit, the method comprises the following steps: capturing a scraper image in the operation process of a submerged chain conveyor by using an image acquisition device, establishing a data set, training a YOLOV5 neural network on the established data set, detecting the scraper by using a trained YOLOV5 neural network model to obtain position data of the scraper, and judging whether the scraper moves or not according to the positions of the scraper among different frames. Whether the submerged chain conveyor works normally or not is judged, so that whether the chain is broken or not is judged. The deep learning neural network is used for detecting the scraper blade in the operation process of the submerged chain conveyor to judge whether the scraper blade operates normally or not, and the problems that manual inspection efficiency is low, and the false alarm rate of a position sensor method is high are solved.

Description

Technical Field

[0001] The present invention relates to the fields of deep learning algorithms, computer vision technology, image processing and industrial machinery monitoring, and specifically proposes a slag machine chain break detection system and detection method based on deep learning. Background Art

[0002] Slag picker is an important part of industrial processing equipment, mainly used in solid waste treatment in power generation, coal, steel, chemical and other industries. During long-term use, the chain of the slag picker may wear out, break and other problems, affecting the normal operation of the equipment. However, there are two existing methods for detecting broken chains of slag pickers, relying on manual inspection or using position sensors. The method relying on manual inspection is inefficient, and the broken chain may not be discovered in time due to human negligence, thus affecting the production process; while the method relying on position sensors will hinder the judgment accuracy of the position sensors due to the harsh and complex production environment on site, resulting in frequent false alarms. Therefore, it is of great significance to develop a broken chain detection system for slag pickers based on computer vision. Summary of the invention

[0003] In view of the problems existing in the above-mentioned prior art, the present invention provides a slag recovery machine chain break detection system based on deep learning, which can automatically, accurately and in real time detect the chain breakage of the slag recovery machine.

[0004] Another object of the present invention is to provide a slag scraper chain break detection method based on deep learning.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A slag scraper chain break detection system based on deep learning, comprising an image acquisition device, an image processing unit, a judgment unit and an alarm unit; the image acquisition device is used to capture real-time images of the slag scraper in operation; the image processing unit uses a target detection algorithm to process and identify scraper images; the judgment unit relies on the scraper positions between different frames to determine whether the slag scraper is stopped, and then determines whether the chain is broken; when the judgment unit confirms the chain break, the alarm unit triggers an alarm mechanism to notify the operator to handle it in time.

[0006] The image acquisition device splits the acquired real-time images of the slag removal machine into frames, and retains one frame for every 30 frames of images to reduce the amount of calculation in the operation process.

[0007] The image processing unit adds random Gaussian blur to the image to remove noise, and changes the HSV parameters of the image to increase contrast and brightness.

[0008] A method for detecting chain breakage of a slag extractor based on deep learning comprises the following steps: Step 1: collecting and preprocessing the operation images of the slag scoop machine. The image acquisition device collects the operation images of the slag scoop machine in real time. Step 2, picture preprocessing, the image processing unit uses the target detection algorithm to process and identify the scraper image; Step 3, chain break judgment, judging whether the scraper moves normally according to the detected scraper center position coordinates; Step 4, issuing an alarm. When the chain break is confirmed in step 3, the alarm mechanism is triggered to notify the operator to handle it in time.

[0009] In the step 1, the image acquisition device is placed in a position where it can clearly and accurately capture the real-time image of the scraper operation process of the slag machine, record the video in real time and split it frame by frame, and retain one image every 30 frames.

[0010] In the step 2, the scraper position of the slag scraper is detected based on the target detection algorithm of YOLOV5, which specifically includes: Step 2.1, labeling: Use an image acquisition device to collect images of the slag scoop machine during operation to construct a data set. After obtaining the scraper images of the slag scoop machine, use the scraper images of the slag scoop machine as a training set to manually label them, and label the scraper position coordinates in each image; Step 2.2, data set expansion: Since the collected images have high similarity, the trained algorithm is prone to overfitting. Therefore, the Mosaic data enhancement method is used to enhance the image data set to improve the algorithm detection accuracy. Step 2.3, training: Use the expanded dataset to train the YOLOV5 deep neural network; Step 2.4, use: Apply the trained YOLOV5 neural network to detect the scraper image of the slag machine, directly identify the scraper position in the frame to be detected and output the pixel coordinates of its center position.

[0011] In the step 3, judging whether the scraper moves normally specifically includes: Step 3.1, calculate the change of the scraper center position coordinates, and divide the slag machine operation status into three categories according to the change of the center position pixels: normal, fault (Q state), and shutdown (S state). When the scraper center position pixel change value of two adjacent pictures is greater than the upper threshold, it is considered to be in a normal operation state; when the scraper center position pixel change value of two adjacent pictures is between the upper and lower thresholds, it is considered to be in a fault state; when the scraper center position pixel change value of two adjacent pictures is less than the lower threshold, it is considered to be in a shutdown operation state; Step 3.2, according to the obtained operating status, in order to avoid false alarms, if the S state appears three times in a row, it is considered that the slag machine has stopped operating, that is, the chain is broken; if the Q state appears six times in a row, it is considered that the slag machine has stopped, that is, the chain is broken.

[0012] In the step 3.1, the camera parameters and shooting position are converted into the image to obtain the pixel value per second. According to the above data combined with the characteristic of taking 1 frame every 30 frames for detection, a threshold of 60 is selected to judge whether the scraper is moving, that is, if the pixel coordinate change of the center position of the scraper in two images with a difference of 30 frames is greater than 60, it is considered to be in normal motion. If the pixel change value of the center position of the scraper in two adjacent images is between 60 and 30, it is considered to have a problem in its operation (Q state). If the pixel change value of the center position of the scraper in two adjacent images is less than 30, it is considered to be in a stopped state (S state).

[0013] In the step 3.1, the scraper moving speed is 4m / s in normal operation. According to the camera parameters and shooting position, IMX219 camera, pixel: 3264x2464, frame rate: 24fps, converted to image is 62 pixels / s.

[0014] In the step 3.2, the occurrence of the Q state for 6 times in a row is considered as the slag extractor stopping, i.e., the chain breaking state, and the occurrence of the S state is also counted.

[0015] In step 4, the GPIO interface of the system hardware is used to send a switch signal to the PLC to implement the disconnection alarm function.

[0016] The advantages and effects of the present invention are as follows: The chain break detection algorithm of the present invention is trained and implemented using the YOLO v5 target detection algorithm based on a deep learning model. Through training with a large number of scraper images, the model can accurately identify the position of the scraper in the picture, and judge whether it is operating normally according to the position change of the scraper in different frames of the video, so as to serve as a detection algorithm for whether the chain is broken. In addition, the alarm unit outputs a switch signal (DO) to the PLC through the output channel on the plate to communicate signals. After being received by the PLC, the operator can be notified by sound, light, or other means on the system screen display, so as to process at the fastest speed.

[0017] The advantages of the present invention also lie in that computer vision technology is used to achieve automatic, real-time and accurate detection of chain breakage in the slag recovery machine, reducing the workload of manual inspections and improving the maintenance efficiency of the equipment; at the same time, the setting of the alarm mechanism enables chain breakage to be discovered and handled in a timely manner, thereby ensuring the stable operation of the slag recovery machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a structural block diagram of a slag machine chain break detection system based on deep learning in the present invention. DETAILED DESCRIPTION

[0019] In order to deepen the understanding of the present invention, the present invention will be further described in detail below in conjunction with examples. The examples are only used to explain the present invention and do not constitute a limitation on the protection scope of the present invention. Example

[0020] A slag scraper chain break detection system based on deep learning, comprising an image acquisition device, an image processing unit, a judgment unit and an alarm unit; the image acquisition device is used to capture real-time images of the slag scraper in operation; the image processing unit uses a target detection algorithm to process and identify scraper images; the judgment unit relies on the scraper positions between different frames to determine whether the slag scraper is stopped, and then determines whether the chain is broken; when the judgment unit confirms the chain break, the alarm unit triggers an alarm mechanism to notify the operator to handle it in time.

[0021] The image acquisition device splits the acquired real-time images of the slag removal machine into frames, and retains one frame for every 30 frames of images to reduce the amount of calculation in the operation process.

[0022] The image processing unit adds random Gaussian blur to the image for denoising, and changes the HSV parameters of the image to achieve the effect of increasing contrast and brightness, so as to improve the scraper detection effect.

[0023] A method for detecting chain breakage of a slag extractor based on deep learning comprises the following steps: Step 1: Capture and preprocess the operation screen of the slag scraper. The image acquisition device captures the operation image of the slag scraper in real time. The image acquisition device is placed in a position where it can clearly and accurately capture the real-time image of the slag scraper operation process. The video is recorded in real time and split frame by frame, retaining one image every 30 frames.

[0024] Step 2: Image preprocessing: the image processing unit uses the YOLOV5 target detection algorithm to process and identify the scraper image; specifically, Step 2.1, labeling: Use an image acquisition device to collect images of the slag scoop machine during operation to construct a data set. After obtaining the scraper images of the slag scoop machine, use the scraper images of the slag scoop machine as a training set to manually label them, and label the scraper position coordinates in each image; Step 2.2, data set expansion: Since the collected images have high similarity, the trained algorithm is prone to overfitting. Therefore, the Mosaic data enhancement method is used to enhance the image data set to improve the algorithm detection accuracy. Step 2.3, training: Use the expanded dataset to train the YOLOV5 deep neural network; Step 2.4, use: Apply the trained YOLOV5 neural network to detect the scraper image of the slag machine, directly identify the scraper position in the frame to be detected and output the pixel coordinates of its center position.

[0025] Step 3, chain break judgment, judging whether the scraper moves normally according to the detected scraper center position coordinates; specifically including: Step 3.1, calculate the change of the scraper center position coordinates, and divide the slag machine operation status into three categories according to the change of the center position pixels: normal, fault (Q state), and shutdown (S state). When the scraper center position pixel change value of two adjacent pictures is greater than the upper threshold, it is considered to be in a normal operation state; when the scraper center position pixel change value of two adjacent pictures is between the upper and lower thresholds, it is considered to be in a fault state; when the scraper center position pixel change value of two adjacent pictures is less than the lower threshold, it is considered to be in a shutdown operation state; Step 3.2, according to the obtained operating status, in order to avoid false alarms, if the S state appears three times in a row, it is considered that the slag machine has stopped operating, that is, the chain is broken; if the Q state appears six times in a row, it is considered that the slag machine has stopped, that is, the chain is broken.

[0026] In the step 3.1, the camera parameters and shooting position are converted into the image to obtain the pixel value per second. According to the above data combined with the characteristic of taking 1 frame every 30 frames for detection, a threshold of 60 is selected to judge whether the scraper is moving, that is, if the pixel coordinate change of the center position of the scraper in two images with a difference of 30 frames is greater than 60, it is considered to be in normal motion. If the pixel change value of the center position of the scraper in two adjacent images is between 60 and 30, it is considered to have a problem in its operation (Q state). If the pixel change value of the center position of the scraper in two adjacent images is less than 30, it is considered to be in a stopped state (S state).

[0027] In the step 3.1, the scraper moving speed is 4m / s in normal operation. According to the camera parameters and shooting position, IMX219 camera, pixel: 3264x2464, frame rate: 24fps, converted to image is 62 pixels / s.

[0028] In the step 3.2, the occurrence of the Q state for 6 times in a row is considered as the slag extractor stopping, i.e., the chain breaking state, and the occurrence of the S state is also counted.

[0029] Step 4, issuing an alarm. When the chain break is confirmed in step 3, the alarm mechanism is triggered to notify the operator to handle it in time. In step 4, the GPIO interface of the system hardware is used to send a switch signal to the PLC to realize the chain break alarm function.

[0030] The above is an example of a patent application for a slag machine chain break detection system based on computer vision of the present invention, but it does not limit the specific implementation mode of the present invention. Any changes and modifications based on the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. A slag machine chain break detection system based on deep learning, characterized in that It includes an image acquisition device, an image processing unit, a judgment unit and an alarm unit; the image acquisition device is used to capture real-time images of the slag scoop machine in operation; the image processing unit uses a target detection algorithm to process and identify the scraper image; the judgment unit relies on the scraper position between different frames to judge whether the slag scoop machine is stopped, and then judges whether the chain is broken; when the judgment unit confirms that the chain is broken, the alarm mechanism is triggered to notify the operator to handle it in time.

2. According to the deep learning-based slag machine chain break detection system of claim 1, it is characterized in that The image acquisition device splits the acquired real-time images of the slag removal machine into frames, and retains one frame for every 30 frames of images to reduce the amount of calculation in the operation process.

3. The deep learning-based chain break detection system for slag extractor according to claim 1 is characterized in that The image processing unit adds random Gaussian blur to the image to remove noise, and changes the HSV parameters of the image to increase contrast and brightness.

4. The detection method of the slag machine chain break detection system based on deep learning according to claim 1 is characterized in that The steps include: Step 1: collecting and preprocessing the operation images of the slag scoop machine. The image acquisition device collects the operation images of the slag scoop machine in real time. Step 2, picture preprocessing, the image processing unit uses the target detection algorithm to process and identify the scraper image; Step 3, chain break judgment, judging whether the scraper moves normally according to the detected scraper center position coordinates; Step 4, issuing an alarm. When the chain break is confirmed in step 3, the alarm mechanism is triggered to notify the operator to handle it in time.

5. The method for detecting chain breakage of a slag extractor based on deep learning according to claim 1 is characterized in that In the step 1, the image acquisition device is placed in a position where it can clearly and accurately capture the real-time image of the scraper operation process of the slag machine, record the video in real time and split it frame by frame, and retain one image every 30 frames.

6. The method for detecting chain breakage of a slag extractor based on deep learning according to claim 5 is characterized in that In the step 2, the scraper position of the slag scraper is detected based on the target detection algorithm of YOLOV5, which specifically includes: Step 2.1, labeling: Use an image acquisition device to collect images of the slag scoop machine during operation to construct a data set. After obtaining the scraper images of the slag scoop machine, use the scraper images of the slag scoop machine as a training set to manually label them, and label the scraper position coordinates in each image; Step 2.2, data set expansion: Since the collected images have high similarity, the trained algorithm is prone to overfitting. Therefore, the Mosaic data enhancement method is used to enhance the image data set to improve the algorithm detection accuracy. Step 2.3, training: Use the expanded dataset to train the YOLOV5 deep neural network; Step 2.4, use: Apply the trained YOLOV5 neural network to detect the scraper image of the slag machine, directly identify the scraper position in the frame to be detected and output the pixel coordinates of its center position.

7. The method for detecting chain breakage of a slag extractor based on deep learning according to claim 5 is characterized in that In the step 3, judging whether the scraper moves normally specifically includes: Step 3.1, calculate the change of the scraper center position coordinates, and divide the slag machine operation status into three categories according to the change of the center position pixels: normal, fault (Q state), and shutdown (S state). When the scraper center position pixel change value of two adjacent pictures is greater than the upper threshold, it is considered to be in a normal operation state; when the scraper center position pixel change value of two adjacent pictures is between the upper and lower thresholds, it is considered to be in a fault state; when the scraper center position pixel change value of two adjacent pictures is less than the lower threshold, it is considered to be in a shutdown operation state; Step 3.2, according to the obtained operating status, in order to avoid false alarms, if the S state appears three times in a row, it is considered that the slag machine has stopped operating, that is, the chain is broken; if the Q state appears six times in a row, it is considered that the slag machine has stopped, that is, the chain is broken.

8. The method for detecting chain breakage of a slag extractor based on deep learning according to claim 7 is characterized in that In the step 3.1, the camera parameters and shooting position are converted into the image to obtain the pixel value per second. According to the above data combined with the characteristic of taking 1 frame every 30 frames for detection, a threshold of 60 is selected to judge whether the scraper is moving, that is, if the pixel coordinate change of the center position of the scraper in two images with a difference of 30 frames is greater than 60, it is considered to be in normal motion. If the pixel change value of the center position of the scraper in two adjacent images is between 60 and 30, it is considered to have a problem in its operation (Q state). If the pixel change value of the center position of the scraper in two adjacent images is less than 30, it is considered to be in a stopped state (S state).

9. A method for detecting chain breakage of a slag extractor based on deep learning according to claim 8, characterized in that In the step 3.1, the scraper moving speed is 4m / s in normal operation. According to the camera parameters and shooting position, the IMX219 camera has 3264x2464 pixels and a frame rate of 24fps, which is converted to 62 pixels / s in the image. In the step 3.2, the occurrence of the Q state for 6 times in a row is considered as the slag extractor stopping, i.e., the chain breaking state, and the occurrence of the S state is also counted.

10. The method for detecting chain breakage of a slag extractor based on deep learning according to claim 5, characterized in that In the step 4, the GPIO interface of the system hardware is used to send a switch signal to the PLC to implement the disconnection alarm function.

Images