Method for quickly and intelligently identifying gas leakage of underground gas extraction pipeline

Abstract

A method for rapidly and intelligently recognizing gas leakage of an underground gas extraction pipeline comprises the following steps that S1, a plurality of gas flow meters are installed in the gas extraction pipeline; s2, arranging a gas leakage detection visualization device at a pair of flange plates between two adjacent sections of gas extraction pipes; s3, arranging a crossing guide device outside each pair of flange plates; and S4, when the numerical value of a certain gas flow meter changes and exceeds the set threshold value, the terminal gives an alarm and automatically starts the pipeline inspection robot to inspect the abnormal gas extraction pipe so as to determine a gas leakage source. The air leakage pipeline is determined through the numerical value change of the flow meter, and the air leakage source is positioned through the pipeline inspection robot based on the computer vision software model. The gas extraction pipeline can be analyzed and detected in real time, so that the abnormal states of the pipeline, such as visual traces of bubbles and gas leakage, can be identified. And the detection rate and precision of the gas leakage phenomenon can be obviously improved in a complex underground environment, so that potential safety accidents are reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of mining engineering, and in particular relates to a method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines. Background Art

[0002] In modern mining production, the safety of underground gas extraction pipelines is directly related to the life safety of miners and the overall environmental safety of the mine. With the development of smart mining and automation technology, higher requirements are placed on the detection and identification of gas leaks. Traditional gas leak detection often relies on manual inspections, which is inefficient and easily affected by subjective factors, making it difficult to achieve timely discovery and rapid response to potential dangers. With the continuous advancement of computer vision and artificial intelligence technology, how to quickly and intelligently identify gas leaks in underground gas extraction pipelines has gradually become a hot topic for scientific researchers. Summary of the invention

[0003] In order to solve the above technical problems existing in the prior art, the present invention provides a method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines which is safe, reliable, efficient and highly accurate.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solution: a method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines, comprising the following steps:

[0005] S1. Install several gas flow meters in the gas extraction pipelines. The real-time monitoring signals of the gas flow meters are transmitted to the pipeline inspection robot and the operator terminal through the signal transmission module.

[0006] S2. A gas leakage detection visualization device is installed at a pair of flanges between two adjacent gas extraction pipes;

[0007] S3. A spanning guide device is arranged outside each pair of flanges;

[0008] S4. When a gas flow count value changes and exceeds the set threshold, the terminal alarms and automatically starts the pipeline inspection robot to inspect the abnormal gas extraction pipe to determine the source of the leak.

[0009] The pipeline inspection robot includes an annular cylinder shell, and at least three crawler walking devices are evenly arranged along the inner circle of the annular cylinder shell in a circular manner. Each crawler walking device is connected to the inner circle of the annular cylinder shell through at least two springs. The inside of the annular cylinder shell is provided with a plurality of high-resolution cameras, a plurality of LED lighting beads, a control chip, a signal transmission module, a power battery and a backup power battery. A computer vision leakage detection model is integrated in the control chip for real-time detection of whether a pipeline leakage occurs.

[0010] The annular cylindrical shell comprises two semi-cylindrical shells, and one end and the other end of the two semi-cylindrical shells are respectively connected by a hinge and a buckle structure.

[0011] The visualization device for gas leakage detection includes an annular water bag wrapped with a pair of flanges. The annular water bag and two gas extraction pipes are sealed with waterproof glue. A water injection hole is provided in the annular water bag. After the annular water bag is filled with water through the water injection hole, the water injection hole is connected to a plastic hose. The plastic hose is connected to a water level gauge arranged along the length direction of the gas extraction pipe. The water level gauge is bonded to the outer circle of the gas extraction pipe. There is a small ball with a clear mark in the water level gauge.

[0012] The crossing guide device includes a front cone cylinder, a middle cylinder and a rear cone cylinder which are coaxial with the gas extraction pipe. The front cone cylinder is smaller in front and larger in rear, while the rear cone cylinder is smaller in rear and larger in front. The front end of the front cone cylinder contacts the outer circle of the gas extraction pipe on the front side of the flange, the front and rear ends of the middle cylinder are respectively connected with the rear end of the front cone cylinder and the front end of the rear cone cylinder, the rear end of the rear cone cylinder contacts the outer circle of the gas extraction pipe on the rear side of the flange, the outer circle of the gas extraction pipe on the front side of the flange is provided with a front bracket connected with the inner circle of the front cone cylinder, and the outer circle of the gas extraction pipe on the rear side of the flange is provided with a rear bracket connected with the inner circle of the rear cone cylinder; a small hole for passing a water level gauge is provided on the front side of the front cone cylinder or the rear side of the rear cone cylinder.

[0013] The computer vision gas leakage detection model uses the YOLOv8 detection algorithm, which is mainly responsible for quickly and intelligently identifying the images transmitted by the high-resolution camera to determine whether there is a gas extraction pipeline leak. The signal transmission module is integrated at the other end of the high-resolution camera to process the warning signal output by the control chip.

[0014] The training process of using the YOLOv8 detection algorithm to identify gas leakage in underground gas extraction pipelines includes the following steps:

[0015] 1) Data preparation: First, we need to collect image data of gas extraction pipelines with gas leakage and normal conditions. The image data set should be as diverse as possible, including images under various lighting, angles and environmental conditions. The location of the gas leakage should be marked in each image, and the bounding box annotation tool LabelImg should be used for annotation. After the annotation is completed, the data is divided into training set, validation set and test set, with a general ratio of 70%, 20% and 10%;

[0016] 2) Environment configuration: Before training, you need to configure the running environment of YOLOv8; using a deep learning framework such as PyTorch, first install the dependent libraries required by YOLOv8, and install them using the following command:

[0017] pip install torch torchvisionpip install-Ur requirements.txt

[0018] Next, clone the YOLOv8 repository:

[0019] git clone https: / / github.com / ultralytics / yolov8cd yolov8

[0020] If a remote server in the cloud is used, take AutoDL computing cloud as an example, and use the pre-set container instance to simplify the environment configuration steps;

[0021] 3) Dataset format conversion: YOLOv8 requires the dataset to be in a specific format; usually, the annotation file needs to be converted to the YOLO format, that is, each image corresponds to a text file, the file name is the same as the image, and the content format is in the form of "class index x_center y_center width height", and all values ​​are normalized relative coordinates;

[0022] 4) Configure training parameters: In YOLOv8, set hyperparameters such as learning rate, batch size, and epoch number by modifying the configuration file; refer to the default configuration of YOLOv8 and adjust these parameters through experiments to achieve better results;

[0023] 5) Train the model: Use the following command to start training, assuming that the dataset has been prepared and stored in the specified path:

[0024] python train.py --data your_dataset.yaml --cfg yolov8n.yaml --weights" --name gas_leak_detection --img640 --epochs100

[0025] In this command, your_dataset.yaml contains the configuration information of the training and validation dataset paths, yolov8n.yaml is the network configuration of YOLOv8, and --epochs specifies the training cycle;

[0026] 6) Verification and adjustment: During the training process, the model performance can be evaluated by observing the training and verification losses and the mean average accuracy (mAP) indicator. If the model is found to be overfitting or underfitting, the learning rate can be adjusted, the data enhancement strategy can be added, or a more complex model architecture can be used.

[0027] 7) Model evaluation and reasoning: After training is completed, use the validation set to evaluate model performance and perform reasoning. The following commands are used in the model reasoning phase:

[0028] python detect.py--weights runs / train / gas_leak_detection / weights / best.pt--img640--source your_image_or_video

[0029] Check the leak identification effect in a visual way and make further adjustments based on actual usage feedback;

[0030] 8) Deployment and application: After the model has been verified and achieved the expected results, it will be deployed and applied to actual underground gas extraction pipeline monitoring to ensure safety and efficiency;

[0031] The above eight steps are the basic process of using the YOLOv8n detection algorithm to train the gas extraction pipeline leakage recognition to obtain the computer vision gas leakage detection model. Through continuous experiments and adjustments, the recognition accuracy and stability of the model are continuously improved;

[0032] The specific steps to embed the trained computer vision air leak detection model into the control chip are as follows:

[0033] (1) Model export: After training is completed, the model is exported. YOLOv8 saves the best weights in the file runs / train / your training name / weights / best.pt. This file is the model for subsequent deployment.

[0034] (2) Environmental preparation: Ensure that the pipeline inspection robot has installed the necessary deep learning framework. The computing platform uses Raspberry Pi or NVIDIA Jetson. The deep learning framework is usually PyTorch or TensorFlow, and supports the YOLOv8 model. At this stage, the corresponding dependent libraries need to be installed, such as:

[0035] pip install torch torchvision

[0036] pip install opencv–python;

[0037] (3) Robot control system: Integration of the control system of the inspection robot. The control system of the inspection robot is designed using ROS or a custom framework. The YOLOv8 model is integrated into it in the following ways:

[0038] ROS: In Ros, you can create a node for image processing and use cv_bridge to convert image messages to and from OpenCV format.

[0039] Custom framework design: Create a Python script to capture camera images and call YOLOv8 for processing;

[0040] (4) Image capture: The pipeline inspection robot uses a high-resolution camera to obtain real-time images. The sample code for capturing images using OpenCV is as follows:

[0041]

[0042]

[0043] (5) Perform object detection: Apply the YOLOv8 model to the captured image for object detection using the following code:

[0044] Python import

[0045] #Inference results = model (frame)

[0046] #Analysis results detections = results.pandas().xyxy[0] #Get the detection results, including category, confidence, and box coordinates;

[0047] (6) Behavior decision-making: The robot makes behavioral decisions based on the output of the model; if a leak is detected, the robot performs specific actions such as stopping movement, sounding an alarm, or sending data to the monitoring center; logical judgment is made based on the detection results:

[0048]

[0049] (7) Deployment and testing: Run the entire program on the robot platform and conduct real-time testing to ensure that the model’s detection capabilities are tested under different lighting and angle conditions. If problems are found, the model can be returned for tuning.

[0050] (8) Overall system optimization: In the actual application of the robot, data is continuously collected, model performance is evaluated, and retraining is performed to improve detection accuracy. At the same time, computing efficiency is optimized, such as using model compression technology, to ensure that the model runs smoothly on devices with limited resources.

[0051] (9) Monitoring and feedback: Add monitoring functions to the robot system, record the detection results and corresponding treatment measures for subsequent analysis and optimization; in this way, the model and robot behavior are continuously improved based on actual usage feedback.

[0052] The specific process of step S2 is: put the annular water bag on the outside of a pair of flanges, use the glue of resin material to bond the front annular mouth of the annular water bag to the outer circle of the gas extraction pipe on the front side of the flange, and then use the glue of resin material to bond the rear annular mouth of the annular water bag to the outer circle of the gas extraction pipe on the rear side of the flange, and then inject water into the annular water bag through the water injection hole on the upper part of the annular water bag, the water will completely submerge the flange, and then use a plastic hose to connect to the water injection hole through a quick-plug connector, and then bond the plastic hose and the water level gauge to the gas extraction pipe along the length direction of the gas extraction pipe.

[0053] The specific process of step S3 is as follows: the front cone cylinder, the middle cylinder and the rear cone cylinder all adopt a two-half structure, firstly, the inner end of the front bracket is bonded to the outer circle of the front gas extraction pipe, then the two halves of the inner circle of the front cone cylinder are bonded to the outer end of the front bracket, and at the same time, the front port of the front cone cylinder is bonded to the outer circle of the front gas extraction pipe, and the connecting seam between the two halves of the front cone cylinder is also bonded, and then the inner end of the rear bracket is bonded to the outer circle of the rear gas extraction pipe, then the two halves of the inner circle of the rear cone cylinder are bonded to the outer end of the rear bracket, and the rear port of the rear cone cylinder is bonded to the outer circle of the rear gas extraction pipe, the water level meter passes through the small hole on the rear side of the rear cone cylinder, and the connecting seam between the two halves of the rear cone cylinder is also bonded, and finally, the front and rear ends of the middle cylinder are respectively bonded to the rear end of the front cone cylinder and the front end of the rear cone cylinder, and at the same time, the connecting seam between the two halves of the middle cylinder is also bonded, and the annular water bag is located between the front bracket, the rear bracket and the middle cylinder.

[0054] The specific process of step S4 is as follows: when the gas flow meter value on the gas extraction main pipeline is abnormal, the gas flow meters installed on each segmented gas extraction pipe are observed to find the gas extraction pipe with abnormal flow, and then the annular shell of the pipeline inspection robot is opened and put on the gas extraction pipe, the buckle structure is closed, and the crawler walking device of the inner circle of the pipeline inspection robot contacts the outer circle of the segmented gas extraction pipe, the power of the pipeline inspection robot is started, and the pipeline inspection robot moves from front to back along the gas extraction pipe to check the outer circle of the gas extraction pipe. Visual leakage detection is carried out on the surface. If a gas leakage is found somewhere on the gas extraction pipe, an alarm is sounded and the signal is transmitted to the terminal or monitoring center. When passing through the flange, the crawler walking device first moves backward along the outer circle of the front cone cylinder, and the crawler walking device compresses the connected spring. After passing through the middle cylinder and then the rear cone cylinder, the spring stretches. A high-resolution camera observes the small ball in the water level gauge. If the small ball fluctuates in the water level gauge, it means that the flange here is leaking, and an alarm is sounded at the same time, and the signal is transmitted to the terminal or monitoring center.

[0055] The working process of the pipeline inspection robot in step S4 is as follows:

[0056] ① System initialization: Turn on the power of the pipeline inspection robot, enter the operation mode, and then perform system self-check;

[0057] ②Environmental perception: Activate high-resolution cameras and start capturing real-time images;

[0058] ③. Data processing: When the crawler device moves along the pipeline, the high-resolution camera captures the images collected on the surface of the gas extraction pipe in real time and transmits the images to the control chip. The control chip inputs the captured images into the YOLOv8 model, identifies whether there is a gas leakage area in the image through the deep learning algorithm, and saves the detection results, including the coordinates and confidence of the gas leakage;

[0059] ④. Behavioral decision: Determine whether there is leakage in the output of the YOLOv8 model. If leakage is detected, perform the following operations:

[0060] Alarm: Trigger visual or audible alarm to alert the operator;

[0061] Record events: record the time and location of the leak detection;

[0062] Automatically adjust the path: According to the location of the gas leak, the pipeline inspection robot selectively bypasses the dangerous area to maintain safety;

[0063] ⑤. Task execution: The robot moves according to the predetermined path and patrol tasks, and continuously monitors the surrounding environment; when safety is confirmed, data is collected regularly to monitor the pipeline status.

[0064] ⑥. Data communication: The signal transmission module of the pipeline inspection robot uploads real-time data and detection records to the monitoring center through a wireless network or a wired network for subsequent analysis; the signal transmission module receives instructions or updates task information from the monitoring center;

[0065] ⑦ Emergency response: When the high-resolution camera detects cracks on the outer surface of the gas extraction pipe or the ball in the water level gauge shakes, it indicates a gas leak. The pipeline inspection robot can sound an alarm and start a preset emergency response program, such as sound and light alarm or contact the operator;

[0066] ⑧Model and system updates, continuous learning: During the operation, collect leak detection data, regularly retrain the YOLOv8 model, and optimize the detection capabilities; if the detection model is found to be deficient, update the training data and improve the model;

[0067] ⑨Termination and maintenance

[0068] Mission completed: After completing the scheduled inspection mission, the pipeline inspection robot returns;

[0069] Data integration and analysis: Integrate and analyze the collected data and generate a test report for subsequent reference;

[0070] System maintenance: Clean and inspect the equipment regularly to ensure that all components are working properly and improve the overall reliability of the robot.

[0071] By adopting the above technical solution, compared with the prior art, the present invention has the following technical effects:

[0072] (1) Real-time detection and high-speed response to gas extraction pipeline leaks.

[0073] By using the YOLOv8 model, the pipeline inspection robot can quickly analyze real-time video streams and quickly identify gas leaks, thereby reducing response time and taking prompt action.

[0074] Continuous monitoring: Pipeline inspection robots are able to continuously capture and analyze images during operation, while traditional methods usually rely on periodic inspections or manual patrols, which may lead to delays in leak detection.

[0075] (2) High degree of automation

[0076] Unmanned operation: Pipeline inspection robots can perform inspection tasks completely independently, reducing human intervention, which reduces dependence on human resources and reduces the possibility of human operational errors.

[0077] Intelligent decision-making: The pipeline inspection robot can automatically respond to the detection results (such as issuing alarms, adjusting paths, etc.), improving the efficiency of responding to emergencies.

[0078] (3) Strong flexibility and adaptability

[0079] Adaptability to complex environments: The pipeline inspection robot is designed to operate in complex and changeable underground environments, and can effectively detect terrain changes and obstacles. Traditional methods often require manual entry into complex terrain for inspection.

[0080] Can be integrated with multiple sensors: The system can combine with other types of sensors (such as gas concentration sensors) to make comprehensive judgments, avoid relying on a single detection method, and thus improve the accuracy and reliability of leak detection.

[0081] (4) Improve detection accuracy

[0082] Advantages of deep learning: The YOLOv8 model uses advanced deep learning technology to effectively extract features and identify leaking areas, thereby significantly reducing the missed detection rate and false alarm rate of leaks.

[0083] Continuous learning: Pipeline inspection robots can be retrained by continuously collecting new inspection data, optimizing algorithms over time to further improve the accuracy of leak detection.

[0084] (5) Data recording and analysis

[0085] Data transparency: The system automatically records detected leak data, including location, time and other important information, to facilitate subsequent analysis, review and data backtracking.

[0086] Real-time monitoring and feedback: Detection information is transmitted to the monitoring center in real time, allowing operators to instantly view on-site conditions and respond quickly, enhancing information transparency and operational efficiency.

[0087] (6) Improved security

[0088] Quick response: Once a gas leak is detected, the pipeline inspection robot will immediately trigger an alarm and initiate emergency response measures, thereby dealing with potential dangers in a timely manner and minimizing the occurrence of accidents.

[0089] Reduce human exposure: By allowing pipeline inspection robots to replace manual inspections of dangerous areas, the possibility of workers being exposed to high-risk environments can be effectively reduced, ensuring employee safety.

[0090] (7) Cost-effectiveness

[0091] Low long-term operating costs: Although the initial equipment investment may be high, in the long run, it can significantly reduce the cost of manual supervision and avoid potential economic losses caused by leaks.

[0092] Reduce labor burden: Automated inspection systems can reduce the need for professional technicians, thereby further reducing human resource costs and optimizing staffing.

[0093] The present invention realizes intelligent and rapid identification of gas drainage pipeline leakage by combining a flow meter, a gas leakage detection visualization device and a deep learning algorithm. The leaking pipeline is determined by the numerical change of the flow meter, and the source of the gas leakage is located by a pipeline inspection robot based on a computer vision software model. The gas drainage pipeline can be analyzed and detected in real time, thereby identifying abnormal conditions of the pipeline, such as bubbles, visual traces of gas leakage, etc. In the complex environment underground, this fast and intelligent identification method can significantly improve the detection rate and accuracy of gas leakage, thereby reducing the occurrence of potential safety accidents.

[0094] The importance of this invention lies not only in ensuring the safety of workers, but also in reducing economic losses and environmental pollution caused by gas leakage. Rapid identification and response to gas leakage risks can help take timely measures to prevent further expansion of accidents. By realizing an intelligent and automated detection system, the intelligence level of underground safety management can be enhanced and the safety operation efficiency of mines can be improved.

[0095] At the same time, the application of the present invention is also promoting the transformation of the mining industry towards digitalization and intelligence, promoting safe production in mines and efficient use of resources. In summary, the method of quickly and intelligently identifying gas leakage in underground gas extraction pipelines is an important means to ensure mine safety and improve production efficiency, and has very broad application prospects and social value.

[0096] In summary, the present invention has shown significant advantages in both scientific principles and innovative designs. It realizes efficient and accurate detection of gas leakage in gas extraction pipelines. It provides reliable guarantee for the safety management of coal mines. In terms of construction costs, it significantly reduces the human and material investment required for traditional detection methods. The pipeline inspection robot can autonomously complete gas leakage detection, reducing the dependence on manual inspections, thereby reducing the labor intensity and risks of operators. This efficient operation method effectively reduces labor costs and effectively improves economic benefits while rationally developing and utilizing coalbed methane resources. It greatly improves the safety of coal mines. The present invention not only brings more economic benefits to coal mines, but also provides new ideas for the intelligent mining of coal mines. By integrating advanced technologies and scientific principles, it promotes the intelligent transformation of the coal mining industry and improves the mining efficiency and safety management level of coalbed methane resources. All of this has opened up new paths for the future development of the coal mining industry and demonstrated broader application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a schematic diagram of the air leakage detection principle of the present invention;

[0098] Figure 2 This is a schematic diagram of the installation of the flow meter on the gas extraction pipeline;

[0099] Figure 3 It is a structural schematic diagram of a gas leakage detection visualization device and a spanning guide device arranged outside a pair of flanges between two adjacent gas extraction pipes;

[0100] Figure 4 yes Figure 3 Schematic diagram of the side structure of the pipeline inspection robot. DETAILED DESCRIPTION

[0101] The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings and examples.

[0102] like Figure 1-Figure 4 As shown, a method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines of the present invention comprises the following steps:

[0103] S1. Install several gas flow meters 3 in the gas extraction pipeline (including the gas extraction trunk pipeline 1 and several sections of gas extraction pipes 2); the real-time monitoring signal of the gas flow meter 3 is transmitted to the pipeline inspection robot 4 and the operator terminal through the signal transmission module;

[0104] S2. A gas leakage detection visualization device is provided at a pair of flanges 5 between two adjacent gas extraction pipes 2;

[0105] S3, a spanning guide device is arranged outside each pair of flanges 5;

[0106] S4. When the value of a gas flow meter 3 changes and exceeds the set threshold, the terminal alarms and automatically starts the pipeline inspection robot 4 to inspect the abnormal gas extraction pipe 2 to determine the source of the gas leakage.

[0107] The pipeline inspection robot 4 includes an annular shell, and at least three crawler walking devices 6 are evenly arranged along the inner circle of the annular shell. Each crawler walking device 6 is connected to the inner circle of the annular shell through at least two springs 7. The inside of the annular shell is provided with a plurality of high-resolution cameras 8, a plurality of LED lighting beads 9, a control chip 10, a signal transmission module 11, a power battery 12 and a backup power battery 25. A computer vision leakage detection model is integrated in the control chip 10 for real-time detection of whether a pipeline leakage occurs.

[0108] The annular cylindrical shell includes two semi-cylindrical shells 13 , and one end and the other end of the two semi-cylindrical shells 13 are connected by a hinge 14 and a buckle structure 15 respectively.

[0109] The visualization device for gas leakage detection includes an annular water bag 16 that wraps a pair of flanges 5. The annular water bag 16 and the two gas extraction pipes 2 are sealed by waterproof glue. A water injection hole is provided in the annular water bag 16. After the annular water bag 16 is filled with water through the water injection hole, the water injection hole is connected to a plastic hose 17. The plastic hose 17 is connected to a water level gauge 18 arranged along the length direction of the gas extraction pipe 2. The water level gauge 18 is bonded to the outer circle of the gas extraction pipe 2. There is a small ball 19 with a clear mark in the water level gauge 18.

[0110] The crossing guide device includes a front cone cylinder 20, a middle cylinder 21 and a rear cone cylinder 22 which are coaxial with the gas extraction pipe 2. The front cone cylinder 20 is smaller at the front and larger at the rear, while the rear cone cylinder 22 is smaller at the rear and larger at the front. The front end of the front cone cylinder 20 contacts the outer circle of the gas extraction pipe 2 on the front side of the flange 5, the front and rear ends of the middle cylinder 21 are respectively connected to the rear end of the front cone cylinder 20 and the front end of the rear cone cylinder 22, and the rear end of the rear cone cylinder 22 contacts the outer circle of the gas extraction pipe 2 on the rear side of the flange 5. The outer circle of the gas extraction pipe 2 on the front side of the flange 5 is provided with a front bracket 23 connected to the inner circle of the front cone cylinder 20, and the outer circle of the gas extraction pipe 2 on the rear side of the flange 5 is provided with a rear bracket 24 connected to the inner circle of the rear cone cylinder 22; a small hole for passing the water level gauge 18 is opened on the front side of the front cone cylinder 20 or the rear side of the rear cone cylinder 22.

[0111] The computer vision gas leakage detection model adopts the YOLOv8 detection algorithm, which is mainly responsible for quickly and intelligently identifying the images transmitted by the high-resolution camera 8 to determine whether there is a leakage in the gas extraction pipe 2; the signal transmission module 11 is integrated at the other end of the high-resolution camera 8 to process the warning signal output by the control chip 10;

[0112] The training process of using the YOLOv8 detection algorithm to identify two-way gas leakage in underground gas extraction pipes includes the following steps:

[0113] 1) Data preparation: First, we need to collect image data of two gas extraction pipes, one leaking and the other in good condition. The image data set should be as diverse as possible, including images under various lighting, angles, and environmental conditions. The location of the leak should be marked on each image, using the bounding box annotation tool LabelImg. After the annotation is completed, the data is divided into a training set, a validation set, and a test set, with a general ratio of 70%, 20%, and 10%;

[0114] 2) Environment configuration: Before training, you need to configure the running environment of YOLOv8; using a deep learning framework such as PyTorch, first install the dependent libraries required by YOLOv8, and install them using the following command:

[0115] pip install torch torchvisionpip install-Ur requirements.txt

[0116] Next, clone the YOLOv8 repository:

[0117] git clone https: / / github.com / ultralytics / yolov8cd yolov8

[0118] If a remote server in the cloud is used, take AutoDL computing cloud as an example, and use the pre-set container instance to simplify the environment configuration steps;

[0119] 3) Dataset format conversion: YOLOv8 requires the dataset to be in a specific format; usually, the annotation file needs to be converted to the YOLO format, that is, each image corresponds to a text file, the file name is the same as the image, and the content format is in the form of "class index x_center y_center width height", and all values ​​are normalized relative coordinates;

[0120] 4) Configure training parameters: In YOLOv8, set hyperparameters such as learning rate, batch size, and epoch number by modifying the configuration file; refer to the default configuration of YOLOv8 and adjust these parameters through experiments to achieve better results;

[0121] 5) Train the model: Use the following command to start training, assuming that the dataset has been prepared and stored in the specified path:

[0122] python train.py --data your_dataset.yaml --cfg yolov8n.yaml --weights" --name gas_leak_detection --img640 --epochs100

[0123] In this command, your_dataset.yaml contains the configuration information of the training and validation dataset paths, yolov8n.yaml is the network configuration of YOLOv8, and --epochs specifies the training cycle;

[0124] 6) Verification and adjustment: During the training process, the model performance can be evaluated by observing the training and verification losses and the mean average accuracy (mAP) indicator. If the model is found to be overfitting or underfitting, the learning rate can be adjusted, the data enhancement strategy can be added, or a more complex model architecture can be used.

[0125] 7) Model evaluation and reasoning: After training is completed, use the validation set to evaluate model performance and perform reasoning. The following commands are used in the model reasoning phase:

[0126] python detect.py--weights runs / train / gas_leak_detection / weights / best.pt--img640--source your_image_or_video

[0127] Check the leak identification effect in a visual way and make further adjustments based on actual usage feedback;

[0128] 8) Deployment and application: After the model has been verified and achieved the expected results, it will be deployed and applied to the actual underground gas extraction pipe 2-way monitoring to ensure safety and efficiency;

[0129] The above eight steps are the basic process of using the YOLOv8n detection algorithm to train the two-way gas leakage recognition of the gas extraction pipe to obtain the computer vision gas leakage detection model. Through continuous experiments and adjustments, the recognition accuracy and stability of the model are continuously improved;

[0130] The specific steps of embedding the trained computer vision air leakage detection model into the control chip 10 are as follows:

[0131] (1) Model export: After training is completed, the model is exported. YOLOv8 saves the best weights in the file runs / train / your training name / weights / best.pt. This file is the model for subsequent deployment.

[0132] (2) Environment preparation: Ensure that the necessary deep learning framework has been installed on the pipeline inspection robot 4. The computing platform uses Raspberry Pi or NVIDIA Jetson. The deep learning framework is usually PyTorch or TensorFlow, and supports the YOLOv8 model. At this stage, the corresponding dependent libraries need to be installed, such as:

[0133] pip install torch torchvision

[0134] pip install opencv–python;

[0135] (3) Robot control system: Integration of the control system of the inspection robot. The control system of the inspection robot is designed using ROS or a custom framework. The YOLOv8 model is integrated into it in the following ways:

[0136] ROS: In Ros, you can create a node for image processing and use cv_bridge to convert image messages to and from OpenCV format.

[0137] Custom framework design: Create a Python script to capture camera images and call YOLOv8 for processing;

[0138] (4) Image capture: Real-time images are acquired through the high-resolution camera 8 of the pipeline inspection robot 4. The sample code for capturing images using OpenCV is as follows:

[0139]

[0140] (5) Perform object detection: Apply the YOLOv8 model to the captured image for object detection using the following code:

[0141] Python import

[0142] #Inference results = model (frame)

[0143] #Analysis results detections = results.pandas().xyxy[0] #Get the detection results, including category, confidence, and box coordinates;

[0144] (6) Behavior decision-making: The robot makes behavioral decisions based on the output of the model; if a leak is detected, the robot performs specific actions such as stopping movement, sounding an alarm, or sending data to the monitoring center; logical judgment is made based on the detection results:

[0145]

[0146] (7) Deployment and testing: Run the entire program on the robot platform and conduct real-time testing to ensure that the model’s detection capabilities are tested under different lighting and angle conditions. If problems are found, the model can be returned for tuning.

[0147] (8) Overall system optimization: In the actual application of the robot, data is continuously collected, model performance is evaluated, and retraining is performed to improve detection accuracy. At the same time, computing efficiency is optimized, such as using model compression technology, to ensure that the model runs smoothly on devices with limited resources.

[0148] (9) Monitoring and feedback: Add monitoring functions to the robot system, record the detection results and corresponding treatment measures for subsequent analysis and optimization; in this way, the model and robot behavior are continuously improved based on actual usage feedback.

[0149] The specific process of step S2 is: put the annular water bag 16 on the outside of a pair of flanges 5, use the glue of resin material to bond the front annular mouth of the annular water bag 16 to the outer circle of the gas extraction pipe 2 on the front side of the flange 5, and then use the glue of resin material to bond the rear annular mouth of the annular water bag 16 to the outer circle of the gas extraction pipe 2 on the rear side of the flange 5, and then inject water into the annular water bag 16 through the water injection hole on the upper part of the annular water bag 16, the water completely submerges the flange 5, and then use the plastic hose 17 to connect with the water injection hole through the quick plug connector, and then bond the plastic hose 17 and the water level meter 18 to the gas extraction pipe 2 along the length direction of the gas extraction pipe 2.

[0150] The specific process of step S3 is as follows: the front cone 20, the middle cone 21 and the rear cone 22 all adopt a two-half structure, firstly the inner end of the front bracket 23 is bonded to the outer circle of the front gas extraction pipe 2, then the two inner circles of the front cone 20 are bonded to the outer end of the front bracket 23, and at the same time the front port of the front cone 20 is bonded to the outer circle of the front gas extraction pipe 2, the connecting seam between the two halves of the front cone 20 is also bonded, then the inner end of the rear bracket 24 is bonded to the outer circle of the rear gas extraction pipe 2, and then the rear cone 24 is bonded to the outer circle of the rear gas extraction pipe 2. The two inner circles of the two halves of 22 are bonded to the outer end of the rear bracket 24, the rear port of the rear cone cylinder 22 is bonded to the outer circle of the rear gas extraction pipe 2, the water level gauge 18 passes through the small hole on the rear side of the rear cone cylinder 22, and the connecting seam between the two halves of the rear cone cylinder 22 is also bonded. Finally, the front and rear ends of the middle cylinder 21 are respectively bonded to the rear end of the front cone cylinder 20 and the front end of the rear cone cylinder 22, and the connecting seam between the two halves of the middle cylinder 21 is also bonded. The annular water bag 16 is located between the front bracket 23, the rear bracket 24 and the middle cylinder 21.

[0151] The specific process of step S4 is as follows: when the value of the gas flow meter 3 on the gas extraction main pipeline 1 is abnormal, the gas flow meters 3 installed on each segmented gas extraction pipe 2 are observed to find the gas extraction pipe 2 with abnormal flow, and then the annular shell of the pipeline inspection robot 4 is opened and put on the gas extraction pipe 2, and the buckle structure 15 is closed. The crawler walking device 6 of the inner circle of the pipeline inspection robot 4 contacts the outer circle of the segmented gas extraction pipe 2, and the power of the pipeline inspection robot 4 is started. The pipeline inspection robot 4 moves from front to back along the gas extraction pipe 2 to inspect the outer surface of the gas extraction pipe 2. Visual gas leakage detection is performed. If a gas leakage is found somewhere on the gas extraction pipe 2, an alarm is sounded and the signal is transmitted to the terminal or monitoring center. When passing through the flange 5, the crawler walking device 6 first moves backward along the outer circle of the front cone 20, and the crawler walking device 6 compresses the connected spring 7. After passing through the middle cylinder 21, and then passing through the rear cone 22, the spring 7 stretches. The high-resolution camera 8 observes the small ball 19 in the water level gauge 18. If the small ball 19 fluctuates in the water level gauge 18, it means that the flange 5 here is leaking, and an alarm is sounded at the same time, and the signal is transmitted to the terminal or monitoring center.

[0152] The working process of the pipeline inspection robot 4 in step S4 is as follows:

[0153] ① System initialization: Turn on the power of the pipeline inspection robot 4, enter the operation mode, and then perform system self-check;

[0154] ② Environmental perception: Activate the high-resolution camera 8 and start capturing real-time images;

[0155] ③. Data processing: When the crawler walking device 6 moves along the pipeline, the high-resolution camera 8 captures the collected images on the surface of the gas extraction pipe 2 in real time, and transmits the images to the control chip 10. The control chip 10 inputs the captured images into the YOLOv8 model, identifies whether there is a gas leakage area in the image through the deep learning algorithm, and saves the detection results, including the coordinates and confidence of the gas leakage;

[0156] ④. Behavioral decision: Determine whether there is leakage in the output of the YOLOv8 model. If leakage is detected, perform the following operations:

[0157] Alarm: Trigger visual or audible alarm to alert the operator;

[0158] Record events: record the time and location of the leak detection;

[0159] Automatically adjust the path: According to the location of the gas leak, the pipeline inspection robot 4 selectively bypasses the dangerous area to maintain safety;

[0160] ⑤. Task execution: The robot moves according to the predetermined path and patrol tasks, and continuously monitors the surrounding environment; when safety is confirmed, data is collected regularly to monitor the pipeline status.

[0161] ⑥. Data communication: The signal transmission module 11 of the pipeline inspection robot 4 uploads the real-time data and detection records to the monitoring center through a wireless network or a wired network for subsequent analysis; the signal transmission module 11 receives instructions or updates task information from the monitoring center;

[0162] ⑦ Emergency response: When the high-resolution camera 8 detects cracks on the outer surface of the gas extraction pipe 2 or the ball 19 in the water level gauge 18 shakes, it indicates that there is a gas leak. The pipeline inspection robot 4 can sound an alarm and start a preset emergency response program, such as sound and light alarm or contact the operator;

[0163] ⑧Model and system updates, continuous learning: During the operation, collect leak detection data, regularly retrain the YOLOv8 model, and optimize the detection capabilities; if the detection model is found to be deficient, update the training data and improve the model;

[0164] ⑨Termination and maintenance

[0165] Mission completion: After completing the scheduled inspection mission, the pipeline inspection robot 4 returns;

[0166] Data integration and analysis: Integrate and analyze the collected data and generate a test report for subsequent reference;

[0167] System maintenance: Clean and inspect the equipment regularly to ensure that all components are working properly and improve the overall reliability of the robot.

[0168] The pipeline inspection robot 4 of the present invention also has the following functions:

[0169] User interface and monitoring, real-time monitoring interface: Provides the operator with a graphical interface that displays the current detection status, visual feedback and alarm information, allowing the operator to intuitively understand the robot's working status.

[0170] Remote control: allows the operator to remotely control the pipeline inspection robot 4 when necessary and instruct to adjust the task direction.

[0171] The pipeline inspection robot 4 can effectively perform the gas leakage identification task of the gas extraction pipe 2 in the underground environment to ensure safety and work efficiency. At the same time, the system is scalable and can integrate more sensors and functions to meet the ever-changing operational needs. The pipeline inspection robot 4 can also be used for gas leakage detection of other gas pipelines in the industrial field.

[0172] The above embodiments illustrate the basic principles and features of the present invention, but the above only illustrates the preferred embodiments of the present invention and is not limited to the embodiments. Under the inspiration of this patent, a person skilled in the art can make many forms of deformation and improvement without departing from the scope of protection of the present invention and the claims, which are all within the protection scope of the present invention. Therefore, the patent and protection scope of the present invention shall be subject to the attached claims.

Claims

1. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines, characterized in that: The following steps are involved: S1. Install several gas flow meters in the gas extraction pipelines. The real-time monitoring signals of the gas flow meters are transmitted to the pipeline inspection robot and the operator terminal through the signal transmission module. S2. A gas leakage detection visualization device is installed at a pair of flanges between two adjacent gas extraction pipes; S3. A spanning guide device is arranged outside each pair of flanges; S4. When a gas flow count value changes and exceeds the set threshold, the terminal alarms and automatically starts the pipeline inspection robot to inspect the abnormal gas extraction pipe to determine the source of the leak.

2. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 1, characterized in that: The pipeline inspection robot includes an annular cylinder shell, and at least three crawler walking devices are evenly arranged along the inner circle of the annular cylinder shell in a circular manner. Each crawler walking device is connected to the inner circle of the annular cylinder shell through at least two springs. The inside of the annular cylinder shell is provided with a plurality of high-resolution cameras, a plurality of LED lighting beads, a control chip, a signal transmission module, a power battery and a backup power battery. A computer vision leakage detection model is integrated in the control chip for real-time detection of whether a pipeline leakage occurs.

3. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 2, characterized in that: The annular cylindrical shell comprises two semi-cylindrical shells, and one end and the other end of the two semi-cylindrical shells are respectively connected by a hinge and a buckle structure.

4. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 3, characterized in that: The visualization device for gas leakage detection includes an annular water bag wrapped with a pair of flanges. The annular water bag and two gas extraction pipes are sealed with waterproof glue. A water injection hole is provided in the annular water bag. After the annular water bag is filled with water through the water injection hole, the water injection hole is connected to a plastic hose. The plastic hose is connected to a water level gauge arranged along the length direction of the gas extraction pipe. The water level gauge is bonded to the outer circle of the gas extraction pipe. There is a small ball with a clear mark in the water level gauge.

5. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 4, characterized in that: The crossing guide device includes a front cone cylinder, a middle cylinder and a rear cone cylinder which are coaxial with the gas extraction pipe. The front cone cylinder is smaller in front and larger in rear, while the rear cone cylinder is smaller in rear and larger in front. The front end of the front cone cylinder contacts the outer circle of the gas extraction pipe on the front side of the flange, the front and rear ends of the middle cylinder are respectively connected with the rear end of the front cone cylinder and the front end of the rear cone cylinder, the rear end of the rear cone cylinder contacts the outer circle of the gas extraction pipe on the rear side of the flange, the outer circle of the gas extraction pipe on the front side of the flange is provided with a front bracket connected with the inner circle of the front cone cylinder, and the outer circle of the gas extraction pipe on the rear side of the flange is provided with a rear bracket connected with the inner circle of the rear cone cylinder; a small hole for passing a water level gauge is provided on the front side of the front cone cylinder or the rear side of the rear cone cylinder.

6. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 5, characterized in that: The computer vision gas leakage detection model uses the YOLOv8 detection algorithm, which is mainly responsible for quickly and intelligently identifying the images transmitted by the high-resolution camera to determine whether there is a gas extraction pipeline leak. The signal transmission module is integrated at the other end of the high-resolution camera to process the warning signal output by the control chip. The training process of using the YOLOv8 detection algorithm to identify gas leakage in underground gas extraction pipelines includes the following steps: 1) Data preparation: First, we need to collect image data of gas extraction pipelines with gas leakage and normal conditions. The image data set should be as diverse as possible, including images under various lighting, angles and environmental conditions. The location of the gas leakage should be marked in each image, and the bounding box annotation tool LabelImg should be used for annotation. After the annotation is completed, the data is divided into training set, validation set and test set, with a general ratio of 70%, 20% and 10%; 2) Environment configuration: Before training, you need to configure the running environment of YOLOv8; using a deep learning framework such as PyTorch, first install the dependent libraries required by YOLOv8, and install them using the following command: pip install torch torchvisionpip install-Ur requirements.txt Next, clone the YOLOv8 repository: git clone https: / / github.com / ultralytics / yolov8cd yolov8 If a remote server in the cloud is used, take AutoDL computing cloud as an example, and use the pre-set container instance to simplify the environment configuration steps; 3) Dataset format conversion: YOLOv8 requires the dataset to be in a specific format; usually, the annotation file needs to be converted to the YOLO format, that is, each image corresponds to a text file, the file name is the same as the image, and the content format is in the form of "class index x_center y_center width height", and all values ​​are normalized relative coordinates; 4) Configure training parameters: In YOLOv8, set hyperparameters such as learning rate, batch size, and epoch number by modifying the configuration file; refer to the default configuration of YOLOv8 and adjust these parameters through experiments to achieve better results; 5) Train the model: Use the following command to start training, assuming that the dataset has been prepared and stored in the specified path: python train.py --data your_dataset.yaml --cfg yolov8n.yaml --weights" --namegas_leak_detection --img640 --epochs100 In this command, your_dataset.yaml contains the configuration information of the training and validation dataset paths, yolov8n.yaml is the network configuration of YOLOv8, and --epochs specifies the training cycle; 6) Verification and adjustment: During the training process, the model performance can be evaluated by observing the training and verification losses and the mean average accuracy (mAP) indicator. If the model is found to be overfitting or underfitting, the learning rate can be adjusted, the data enhancement strategy can be added, or a more complex model architecture can be used. 7) Model evaluation and reasoning: After training is completed, use the validation set to evaluate model performance and perform reasoning. The following commands are used in the model reasoning phase: python detect.py--weights runs / train / gas_leak_detection / weights / best.pt--img640--source your_image_or_video Check the leak identification effect in a visual way and make further adjustments based on actual usage feedback; 8) Deployment and application: After the model has been verified and achieved the expected results, it will be deployed and applied to actual underground gas extraction pipeline monitoring to ensure safety and efficiency; The above eight steps are the basic process of using the YOLOv8n detection algorithm to train the gas extraction pipeline leakage recognition to obtain the computer vision gas leakage detection model. Through continuous experiments and adjustments, the recognition accuracy and stability of the model are continuously improved; The specific steps to embed the trained computer vision air leak detection model into the control chip are as follows: (1) Model export: After training is completed, the model is exported. YOLOv8 saves the best weights in the file runs / train / your training name / weights / best.pt. This file is the model for subsequent deployment. (2) Environmental preparation: Ensure that the pipeline inspection robot has installed the necessary deep learning framework. The computing platform uses Raspberry Pi or NVIDIA Jetson. The deep learning framework is usually PyTorch or TensorFlow, and supports the YOLOv8 model. At this stage, the corresponding dependent libraries need to be installed, such as: pip install torch torchvision pip install opencv–python; (3) Robot control system: Integration of the control system of the inspection robot. The control system of the inspection robot is designed using ROS or a custom framework. The YOLOv8 model is integrated into it in the following ways: ROS: In Ros, you can create a node for image processing and use cv_bridge to convert image messages to and from OpenCV format. Custom framework design: Create a Python script to capture camera images and call YOLOv8 for processing; (4) Image capture: The pipeline inspection robot uses a high-resolution camera to obtain real-time images. The sample code for capturing images using OpenCV is as follows: (5) Perform object detection: Apply the YOLOv8 model to the captured image for object detection using the following code: Python import #Inference results = model (frame) #Analysis results detections = results.pandas().xyxy[0] #Get the detection results, including category, confidence, and box coordinates; (6) Behavior decision-making: The robot makes behavioral decisions based on the output of the model; if a leak is detected, the robot performs specific actions such as stopping movement, sounding an alarm, or sending data to the monitoring center; logical judgment is made based on the detection results: Python for index,row in detections.iterrows(): if row['class'] == 'gas_leak' and row['confidence']>threshold: #threshold is the confidence threshold print("Leak detected!") #Perform corresponding actions, such as stopping moving forward or backward; (7) Deployment and testing: Run the entire program on the robot platform and conduct real-time testing to ensure the detection capability of the model under different lighting and angle environmental conditions; If you find problems, you can go back and tune the model. (8) Overall system optimization: In the actual application of the robot, data is continuously collected, model performance is evaluated, and retraining is performed to improve detection accuracy. At the same time, computing efficiency is optimized, such as using model compression technology, to ensure that the model runs smoothly on devices with limited resources. (9) Monitoring and feedback: Add monitoring functions to the robot system, record the detection results and corresponding treatment measures for subsequent analysis and optimization; in this way, the model and robot behavior are continuously improved based on actual usage feedback.

7. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 6, characterized in that: The specific process of step S2 is: put the annular water bag on the outside of a pair of flanges, use the glue of resin material to bond the front annular mouth of the annular water bag to the outer circle of the gas extraction pipe on the front side of the flange, and then use the glue of resin material to bond the rear annular mouth of the annular water bag to the outer circle of the gas extraction pipe on the rear side of the flange, and then inject water into the annular water bag through the water injection hole on the upper part of the annular water bag, the water will completely submerge the flange, and then use a plastic hose to connect to the water injection hole through a quick-plug connector, and then bond the plastic hose and the water level gauge to the gas extraction pipe along the length direction of the gas extraction pipe.

8. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 8, characterized in that: The specific process of step S3 is as follows: the front cone cylinder, the middle cylinder and the rear cone cylinder all adopt a two-half structure, firstly, the inner end of the front bracket is bonded to the outer circle of the front gas extraction pipe, then the two halves of the inner circle of the front cone cylinder are bonded to the outer end of the front bracket, and at the same time, the front port of the front cone cylinder is bonded to the outer circle of the front gas extraction pipe, and the connecting seam between the two halves of the front cone cylinder is also bonded, and then the inner end of the rear bracket is bonded to the outer circle of the rear gas extraction pipe, then the two halves of the inner circle of the rear cone cylinder are bonded to the outer end of the rear bracket, and the rear port of the rear cone cylinder is bonded to the outer circle of the rear gas extraction pipe, the water level meter passes through the small hole on the rear side of the rear cone cylinder, and the connecting seam between the two halves of the rear cone cylinder is also bonded, and finally, the front and rear ends of the middle cylinder are respectively bonded to the rear end of the front cone cylinder and the front end of the rear cone cylinder, and at the same time, the connecting seam between the two halves of the middle cylinder is also bonded, and the annular water bag is located between the front bracket, the rear bracket and the middle cylinder.

9. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 8, characterized in that: The specific process of step S4 is as follows: when the gas flow meter value on the gas extraction main pipeline is abnormal, the gas flow meters installed on each segmented gas extraction pipe are observed to find the gas extraction pipe with abnormal flow, and then the annular shell of the pipeline inspection robot is opened and put on the gas extraction pipe, the buckle structure is closed, and the crawler walking device of the inner circle of the pipeline inspection robot contacts the outer circle of the segmented gas extraction pipe, the power of the pipeline inspection robot is started, and the pipeline inspection robot moves from front to back along the gas extraction pipe to check the outer circle of the gas extraction pipe. Visual leakage detection is carried out on the surface. If a gas leakage is found somewhere on the gas extraction pipe, an alarm is sounded and the signal is transmitted to the terminal or monitoring center. When passing through the flange, the crawler walking device first moves backward along the outer circle of the front cone cylinder, and the crawler walking device compresses the connected spring. After passing through the middle cylinder and then the rear cone cylinder, the spring stretches. A high-resolution camera observes the small ball in the water level gauge. If the small ball fluctuates in the water level gauge, it means that the flange here is leaking, and an alarm is sounded at the same time, and the signal is transmitted to the terminal or monitoring center.

10. A method for quickly and intelligently identifying gas leakage in underground gas extraction pipelines according to claim 9, characterized in that: The working process of the pipeline inspection robot in step S4 is as follows: ① System initialization: Turn on the power of the pipeline inspection robot, enter the operation mode, and then perform system self-check; ②Environmental perception: Activate high-resolution cameras and start capturing real-time images; ③. Data processing: When the crawler device moves along the pipeline, the high-resolution camera captures the images collected on the surface of the gas extraction pipe in real time and transmits the images to the control chip. The control chip inputs the captured images into the YOLOv8 model, identifies whether there is a gas leakage area in the image through the deep learning algorithm, and saves the detection results, including the coordinates and confidence of the gas leakage; ④. Behavioral decision: Determine whether there is leakage in the output of the YOLOv8 model. If leakage is detected, perform the following operations: Alarm: Trigger visual or audible alarm to alert the operator; Record events: record the time and location of the leak detection; Automatically adjust the path: According to the location of the gas leak, the pipeline inspection robot selectively bypasses the dangerous area to maintain safety; ⑤. Task execution: The robot moves according to the predetermined path and patrol tasks, and continuously monitors the surrounding environment; when safety is confirmed, data is collected regularly to monitor the pipeline status. ⑥. Data communication: The signal transmission module of the pipeline inspection robot uploads real-time data and detection records to the monitoring center through a wireless network or a wired network for subsequent analysis; the signal transmission module receives instructions or updates task information from the monitoring center; ⑦ Emergency response: When a high-resolution camera detects cracks on the outer surface of the gas extraction pipe or the ball in the water level gauge shakes, indicating a gas leak, the pipeline inspection robot can sound an alarm and initiate a preset emergency response procedure, such as sound and light alarm or contacting the operator; ⑧Model and system updates, continuous learning: During the operation, collect leak detection data, regularly retrain the YOLOv8 model, and optimize the detection capabilities; if the detection model is found to be deficient, update the training data and improve the model; ⑨Termination and maintenance Mission completed: After completing the scheduled inspection mission, the pipeline inspection robot returns; Data integration and analysis: Integrate and analyze the collected data and generate a test report for subsequent reference; System maintenance: Clean and inspect the equipment regularly to ensure that all components are working properly and improve the overall reliability of the robot.

Images