A method for collaborative inspection and maintenance work of a robot in a coal gas tank

Abstract

This invention relates to the field of gas holder technology, specifically to a method for collaborative inspection and maintenance of gas holders using robots. The method includes: a track-mounted robot running on a rail on a second-floor truss circular walkway; a tracked robot moving on a first-floor piston-shaped walkway; processing data collected by the track-mounted robot to obtain high-altitude defect features and coordinates, and processing data collected by the tracked robot to obtain piston area defect features and coordinates; matching the high-altitude defect features and coordinates with the piston area defect features and coordinates to obtain leakage features and coordinates; generating a three-dimensional hazard map based on the leakage features and coordinates, the map including hazard type and location; the track-mounted and tracked robots performing maintenance operations based on the hazard type and location; and transmitting inspection data and maintenance operation records to the main control system to optimize the inspection cycle and maintenance strategy. This invention achieves comprehensive inspection of the gas holder structure without blind spots.

Description

Technical Field

[0001] This invention relates to the field of gas holder technology, and specifically to a method for collaborative inspection and maintenance of gas holders using robots. Background Technology

[0002] As a critical facility for storing flammable and explosive gases, gas holders have complex internal structures and operate in harsh environments (such as high temperature, high humidity, and high dust levels), requiring extremely high integrity and sealing. Traditional operation and maintenance methods rely on manual inspections and periodic maintenance, but the limited space inside the holder (e.g., the distance between the piston and the side plate is less than 1 meter) and the presence of corrosive media (such as H2S and CO) result in high risks and low efficiency for manual entry operations. Furthermore, key parameters such as piston lifting and sealing device wear during the dynamic operation of the gas holder require real-time monitoring, but manual methods are insufficient for continuous data acquisition and rapid response to anomalies, posing a potential safety hazard.

[0003] In existing technologies, the operation and maintenance of gas holders mainly adopts a combination of fixed sensors (such as laser rangefinders and gas detectors) and localized mechanized tools (such as ash removal equipment). Fixed monitoring equipment only covers key areas (such as piston stroke limit points) and cannot achieve full-area detection within the holder; while mechanized tools have limited functions (such as limited to ash removal or localized oil replenishment), requiring manual assistance and unable to perform collaborative operations. For example, a certain type of dry gas holder uses a fixed camera on the top of the holder to monitor piston tilt, but there are blind spots (occupying more than 40% of the space inside the holder), and emergency handling (such as oil replenishment in the oil trench) cannot be performed simultaneously. In addition, existing technologies lack multi-device linkage mechanisms, resulting in fragmented data acquisition, analysis, and execution, making it difficult to meet the real-time closed-loop management requirements under complex operating conditions. These problems significantly restrict the safety, continuity, and intelligence level of gas holder operation and maintenance. Summary of the Invention

[0004] To address the aforementioned technical problems of incomplete monitoring coverage, lack of coordination between inspection and maintenance, and data lag, this invention provides a method for collaborative inspection and maintenance of gas holders using robots. This invention primarily utilizes tracked and tracked robots deployed in different spatial layers to work collaboratively. It generates a three-dimensional hazard map through the fusion of multi-source heterogeneous data, and uses this map for intelligent decision-making and task scheduling, thereby achieving unmanned, precise inspection and real-time maintenance of the entire gas holder area, 24 / 7.

[0005] The technical means employed in this invention are as follows:

[0006] A method for collaborative inspection and maintenance of a gas holder using robots includes the following steps: The track-mounted robot runs on the rails of the two-story truss circular walkway, collecting visual images, infrared thermal imaging data and acoustic signals of the high-altitude structure. The tracked robot moves on a piston-ring-shaped platform to collect visual data from ground equipment and environmental gas data. Data collected by the tracked robot is processed to obtain high-altitude defect features and coordinates; data collected by the tracked robot is processed to obtain piston area defect features and coordinates. Based on the high-altitude defect features and coordinates and the piston region defect features and coordinates, leakage features and coordinates are obtained by matching them. A three-dimensional hazard map is generated based on the leakage characteristics and coordinates, and the three-dimensional hazard map includes the hazard type and location; The tracked and tracked robots perform maintenance operations based on the type and location of the hazard, and generate inspection data and maintenance operation records; The inspection data and maintenance operation records are transmitted to the main control system to generate an inspection report containing defect information. Based on the inspection report, the inspection cycle and maintenance strategy are optimized.

[0007] Furthermore, the processing of the data collected by the orbital robot to obtain high-altitude defect features and coordinates includes: The visual image is processed using the YOLOv8 target detection model to identify the surface defect categories such as steel structure corrosion, weld cracks, and loose bolts, and to identify the coordinates of the defect targets. The infrared thermal imaging data is classified using the DenseNet classification model to identify infrared anomaly categories and extract temperature anomaly coordinates. A Gaussian mixture model was used to perform cluster analysis on the acoustic signals to distinguish and identify abnormal acoustic anomaly patterns caused by sealing leaks, loose bolts, or structural friction, and to locate the coordinates of the abnormal sound sources. Calculate the Euclidean distance in three-dimensional space between the target coordinates of the defect, the temperature anomaly coordinates, and the abnormal sound source coordinates. If the Euclidean distance is less than the fusion threshold, it is determined to be the same physical defect. The target coordinates of the defect, the temperature anomaly coordinates, and the abnormal sound source coordinates are weighted and averaged to obtain the coordinates of the high-altitude defect. Surface defect categories, infrared anomaly categories, and acoustic anomaly patterns, along with their corresponding confidence scores, are constructed as high-altitude defect features.

[0008] Furthermore, the processing of the data collected by the tracked robot to obtain the defect features and coordinates of the piston region includes: The visual data was processed using the ResNet-50 image classification model to identify the surface condition categories of aging and cracking of the sealing rubber curtain, drying and missing sealing grease, and excessive dust accumulation on the reflector, and the platform defect coordinates were calculated. The gas concentration data is analyzed using a time-series prediction model based on a long short-term memory network to identify abnormal concentration fluctuation patterns. Combined with the implementation pose of the tracked robot, the coordinates of the concentration anomaly points are determined. Based on the surface state category and the concentration anomaly fluctuation pattern, piston region defect features are constructed, and piston region defect coordinates are constructed based on the platform defect coordinates and the coordinates of the concentration anomaly points.

[0009] Further, the step of matching the high-altitude defect features and coordinates with the piston region defect features and coordinates to obtain leakage features and coordinates includes: The coordinates of the high-altitude defect are vertically projected onto the piston plane to obtain the projected coordinates; Calculate the planar distance between the projected coordinates and the defect coordinates of each piston region; If there are high-altitude defects and piston area defects with a planar distance less than the vertical correlation threshold, and the high-altitude defects contain acoustic or infrared anomalies related to leakage, while the piston area defects contain gas concentration anomalies, then a confirmed leak is determined to have occurred. Based on the coordinates of the high-altitude defect and the coordinates of the piston region defect, the fused leak coordinates of the leak event are calculated and confirmed.

[0010] Furthermore, the formula for calculating the fused leak coordinates of the leak event was confirmed as follows:

[0011] in, x l To fuse the x-coordinate of the leak coordinates, y l To merge the y-coordinate of the leak coordinates, z l To fuse the z-coordinate of the leak coordinates, x f The x-coordinate of the high-altitude defect is given. x g The x-coordinate of the piston region defect coordinates. y f The y-coordinate of the high-altitude defect is given. y g Let y be the coordinate of the defect in the piston region. z f To fuse the z-coordinate of the leak coordinates, λ This is the planar blending correction coefficient.

[0012] Furthermore, the track-mounted robot is equipped with a high-definition camera, an infrared thermal imager, and acoustic sensors for inspecting trusses, welds, guide rollers, cabinet walls, and high-altitude equipment. The tracked robot is equipped with a vision camera, a gas detector, and a retractable robotic arm. The robotic arm integrates a cleaning brush, a sampling probe, and an oil injection gun, and is used to inspect and maintain seals, reflectors, and instruments.

[0013] Furthermore, the types of potential hazards include low-risk defects, medium-risk defects, and high-risk defects. For low-risk defects, the robotic arm of the tracked robot is controlled to perform automatic maintenance operations. For medium-risk defects, the tracked robot or the tracked robot is controlled to perform maintenance operations. For high-risk defects, the tracked robot and the tracked robot locate the high-risk defects and simultaneously trigger emergency procedures.

[0014] Furthermore, the low-risk defect includes dust accumulation on the reflector, and the automatic maintenance operation is a dust removal operation; the medium-risk defect includes loose bolts, and the targeted maintenance operation is a tightening operation; the high-risk defect includes piston tilt warning, and the emergency procedure includes cutting off the gas supply or activating the backup sealing mechanism.

[0015] Compared with the prior art, the present invention has the following advantages: 1. This invention constructs an air-ground collaborative operation system by deploying tracked and tracked robots, and generates a three-dimensional hidden danger map based on multi-sensor data fusion and intelligent analysis, realizing high-precision, blind-angle inspection of cabinet structure, sealing system and operating equipment, making the automatic defect identification rate reach more than 98% and reducing the risk of personnel exposure by 100%.

[0016] 2. By establishing a collaborative judgment mechanism for leaks based on multi-source data vertical projection matching and evidence verification, this invention enables rapid identification and accurate location of penetrating high-risk defects such as gas leaks, shortens the emergency response time to less than 30 seconds, and significantly improves the intelligent monitoring, early warning and real-time control capabilities of major hazard sources.

[0017] 3. This invention, through a data closed-loop management system and by training a prediction model based on historical data to dynamically optimize inspection strategies and maintenance cycles, has achieved a shift in the operation and maintenance mode from post-maintenance to predictive maintenance. This reduces the labor cost of single-cabinet operation and maintenance by 75% and saves more than one million yuan in annual operation and maintenance costs.

[0018] Based on the above reasons, this invention can be widely promoted in fields such as gas holders. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the structure of the explosion-proof robot for zoned inspection and maintenance inside the gas holder of the present invention.

[0021] In the diagram: 1. Tracked robot; 2. Truss circular walkway; 21. Truss; 22. Guide roller; 23. Maintenance and charging station; 3. Tracked robot; 4. Piston circular walkway; 5. Detection instrument; 6. Piston plate inside the cabinet; 7. Piston plate weld; 8. Gas inside the piston; 9. Counterweight mechanism; 10. Leveling wire rope; 11. Reflector; 12. Sealing mechanism; 13. Cabinet wall. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] This invention provides a method for collaborative inspection and maintenance of gas holders using robots, comprising the following steps: S1. The track-mounted robot runs on the rails of the two-story truss circular walkway, collecting visual images, infrared thermal imaging data and acoustic signals of the high-altitude structure.

[0025] S2. The tracked robot moves on a piston-ring-shaped platform to collect visual data from ground equipment and environmental gas data.

[0026] Among them, the track-mounted robot is equipped with a high-definition camera, an infrared thermal imager, and acoustic sensors to inspect trusses, welds, guide rollers, cabinet walls, and high-altitude equipment.

[0027] The tracked robot is equipped with a vision camera, a gas detector, and a retractable robotic arm. The robotic arm integrates a cleaning brush, a sampling probe, and an oil injection gun for the inspection and maintenance of seals, reflectors, and instruments.

[0028] Figure 1 This is a structural diagram of the internal structure of the gas holder used in this invention. The gas holder includes a cabinet, a piston disposed inside the cabinet, and a tracked robot and a tracked robot deployed at different heights inside the cabinet.

[0029] The cabinet includes cylindrical cabinet walls, with trusses on the outer side of the cabinet walls for support and reinforcement. A truss-shaped walkway surrounds the cabinet on the truss. Guide rollers for guiding a track-mounted robot are installed between the truss-shaped walkway and the cabinet walls.

[0030] The piston includes a horizontal piston plate inside the cabinet, a sealing mechanism is provided between the edge of the piston plate inside the cabinet and the cabinet wall, and a piston ring-shaped walkway is provided above the piston plate inside the cabinet.

[0031] The track-mounted robot is deployed on a truss-shaped circular walkway and moves along a track set on the walkway. A truss is located beneath the truss-shaped walkway. The tracked robot is deployed on a piston-ring-shaped walkway and moves on the walkway.

[0032] The truss circular walkway is also equipped with a maintenance and charging station for providing energy to the tracked robot.

[0033] The piston ring walkway or the piston plate inside the cabinet is equipped with a detection instrument for monitoring the status inside the cabinet.

[0034] The piston plate inside the cabinet is also equipped with a reflector for laser ranging or position calibration.

[0035] The piston plate inside the cabinet is welded from multiple steel plates, forming piston plate welds, which are the inspection targets of the tracked robot.

[0036] The space below the piston plate inside the cabinet stores the gas inside the piston.

[0037] The cabinet is equipped with a counterweight mechanism and leveling steel wire ropes connecting the counterweight mechanism to multiple points on the edge of the piston plate inside the cabinet. By adjusting the tension of each leveling steel wire rope, the levelness of the piston plate inside the cabinet can be dynamically corrected.

[0038] S3. Process the data collected by the tracked robot to obtain the high-altitude defect features and coordinates, and process the data collected by the tracked robot to obtain the piston area defect features and coordinates.

[0039] Specifically, the steps for processing the data collected by the track-mounted robot to obtain the features and coordinates of high-altitude defects are as follows: S31. The YOLOv8 target detection model is used to process the visual images, identify the surface defect categories such as steel structure corrosion, weld cracks, and loose bolts, and identify the coordinates of the defect targets.

[0040] S32. The DenseNet classification model is used to classify infrared thermal imaging data, identify infrared anomaly categories, and extract temperature anomaly coordinates.

[0041] S33. A Gaussian mixture model is used to perform cluster analysis on the acoustic signals to distinguish and identify abnormal acoustic anomaly patterns caused by sealing leaks, loose bolts or structural friction, and to locate the coordinates of the abnormal sound sources.

[0042] S34. Calculate the Euclidean distance between the target coordinates, temperature anomaly coordinates, and abnormal sound source coordinates in three-dimensional space. If the Euclidean distance is less than the fusion threshold, it is determined to be the same physical defect. The target coordinates, temperature anomaly coordinates, and abnormal sound source coordinates are weighted and averaged to obtain the high-altitude defect coordinates.

[0043] S35. Surface defect categories, infrared anomaly categories, and acoustic anomaly patterns, along with their corresponding confidence scores, are used to construct high-altitude defect features. The formula for calculating high-altitude defect coordinates is as follows:

[0044] in, P f The coordinates of the high-altitude defect. w v This is a weighting coefficient for visual positioning accuracy. C v The confidence score for the visual inspection results. P v The coordinates of the defect target. w t This is a weighting coefficient for infrared positioning accuracy. C t This represents the confidence score of the infrared detection results. P t For temperature anomaly coordinates, w a This is a weighting coefficient for temperature positioning accuracy. C a This represents the confidence score of the acoustic test results. P a The coordinates of the abnormal sound source.

[0045] The data collected by the tracked robot is processed to obtain the defect features and coordinates of the piston area, specifically including: S36. The ResNet-50 image classification model is used to process the visual data to identify the surface condition categories of aging and cracking of the sealing rubber curtain, drying and missing sealing grease, and excessive dust accumulation on the reflector, and to calculate the platform defect coordinates.

[0046] S37. A time-series prediction model based on a long short-term memory network is used to analyze the gas concentration data, identify abnormal concentration fluctuation patterns, and determine the coordinates of abnormal concentration points by combining the implementation pose of the tracked robot.

[0047] S38. Based on the surface state category and concentration anomaly fluctuation pattern, construct the piston region defect features, and based on the platform defect coordinates and the coordinates of the concentration anomaly points, construct the piston region defect coordinates.

[0048] S4. Based on the high-altitude defect features and coordinates and the piston area defect features and coordinates, the leakage features and coordinates are matched to obtain the leakage features and coordinates.

[0049] Step S4 includes: S41. Project the coordinates of the high-altitude defect vertically onto the piston plane to obtain the projected coordinates.

[0050] S42. Calculate the planar distance between the projected coordinates and the defect coordinates of each piston region.

[0051] S43. If there are high-altitude defects and piston area defects with a planar distance less than the vertical correlation threshold, and the high-altitude defects contain acoustic or infrared anomalies related to leakage, while the piston area defects contain gas concentration anomalies, then it is determined that a confirmed leak has occurred.

[0052] S44. Based on the coordinates of the high-altitude defect and the piston region defect, calculate the merged leak coordinates to confirm the leak event. The formula for calculating the merged leak coordinates to confirm the leak event is:

[0053] in, x l To fuse the x-coordinate of the leak coordinates, y l To merge the y-coordinate of the leak coordinates, z l To fuse the z-coordinate of the leak coordinates, x f The x-coordinate of the high-altitude defect is given. x g The x-coordinate of the piston region defect coordinates. y f The y-coordinate of the high-altitude defect is given. y g Let y be the coordinate of the defect in the piston region. zf To fuse the z-coordinate of the leak coordinates, λ This is the planar blending correction coefficient.

[0054] S5. Generate a 3D hazard map based on leakage characteristics and coordinates. The 3D hazard map includes hazard type and location.

[0055] The types of potential hazards include low-risk defects, medium-risk defects, and high-risk defects. For low-risk defects, the robotic arm of the tracked robot is controlled to perform automatic maintenance operations. For medium-risk defects, the tracked robot or the tracked robot is controlled to perform maintenance operations. For high-risk defects, the tracked robot and the tracked robot locate the high-risk defects and trigger emergency procedures simultaneously.

[0056] Low-risk defects include dust accumulation on the reflector, which can be addressed by automatic dust removal during maintenance; medium-risk defects include loose bolts, which can be addressed by tightening during maintenance; high-risk defects include piston tilt warnings, which can be addressed by cutting off the gas supply or activating the backup sealing mechanism during emergency procedures.

[0057] S6. Tracked and tracked robots perform maintenance operations based on the type and location of the hazard, and generate inspection data and maintenance operation records.

[0058] S7. Transmit inspection data and maintenance operation records to the main control system, generate an inspection report containing defect information, and optimize the inspection cycle and maintenance strategy based on the inspection report.

[0059] The operational scenarios used in this invention are as follows: Routine inspections: Tracked robots perform weekly cyclical inspections of trusses, piston sealing mechanisms, buffer limit devices, etc., and complete tasks such as lubricating guide rollers; Tracked robots cover piston rings, cabinet position monitoring instruments, measuring reflectors, piston welds, sealing oil tanks, counterweight systems, etc., and complete tasks such as cleaning reflectors once a week. All inspection data is synchronized to the main control system in real time.

[0060] Special protection inspection: When the operating pressure of the gas holder is greater than 80% of the rated value, the system automatically triggers a high-frequency inspection, focusing on monitoring the condition of the seals and welds.

[0061] Emergency Response: If the concentration of combustible gas detected is greater than 20% LEL and a temperature change occurs in the infrared thermal imaging, the robot will automatically lock the coordinates of the leak point, the main control system will remotely start the nitrogen purging device, and the maintenance personnel will be notified at the same time.

[0062] The orbital and crawler robots operate autonomously 24 / 7, with a daily patrol frequency of up to 12 times (only once a week in the traditional manual mode), achieving "automation with reduced personnel"; the operation and maintenance personnel remotely monitor through the main control system and only intervene in complex decision-making. The number of on-site operators for a single cabinet is ≤1, meeting the "reduced personnel" job setting standard. It is configured with explosion protection certification (Ex d IIB T4Gb), IP64 protection level, and fault self-diagnosis function, meeting the requirements of "intrinsic safety transformation of high-risk equipment"; data transmission uses 5G private network encryption (AES-256), conforming to the industrial Internet security specification, and supporting the implementation of the "intelligent unmanned" scenario.

[0063] Embodiment 1 This embodiment provides a collaborative inspection system for robots inside a gas cabinet, which consists of four core modules: an orbital robot system, a crawler robot system, a main control system, and production equipment and valve groups supporting the gas cabinet. Each module exchanges data and transmits instructions through a hybrid network of industrial Ethernet and wireless communication, achieving all-round and intelligent inspection and maintenance of the cabinet environment.

[0064] Orbital robot system: Deployed along the hanging rail of the annular walking platform guardrail on the second floor truss of the gas cabinet, meeting the explosion protection level of Ex d IIB T4Gb and the IP64 protection level, ensuring safe operation under complex working conditions. Adopting the gear-rack drive method, the speed can be accurately adjusted within the range of 0.5 - 1 m / s. Equipped with a high-energy density battery, the single-charge endurance time is more than 8 hours, and it supports returning to the preset work position for automatic charging during the inspection interval to ensure uninterrupted operation. Equipped with a 12-million-pixel explosion-proof camera (integrated with adaptive LED light supplement) to ensure clear images in the dim environment inside the cabinet; equipped with an infrared thermal imager (temperature measurement accuracy of ±2°C) to accurately monitor temperature anomalies at key parts such as the cabinet wall welds and sealing membranes; integrated with a microphone array (noise localization accuracy of ±0.5m) to spatially locate the abnormal sound source and assist in judging mechanical failures. Interacts with the main control system in real time through the dedicated wireless Wi-Fi network inside the cabinet to synchronize high-definition video, thermal imaging data, and gas concentration information.

[0065] Crawler robot system: Adopting a wide-track crawler chassis design, it has excellent adhesion and obstacle-crossing ability, and can stably adapt to the inclined environment with a piston surface inclination of 5° to 35°. It integrates lidar and visual navigation modules to build a high-precision environmental map, enabling autonomous positioning and path planning in complex terrain inside the cabinet. It is equipped with a 6-degree-of-freedom explosion-proof manipulator with flexible movement ability. A tool quick-change device is provided at the end, which can quickly switch the laser methane telemeter, high-definition detection probe, ash cleaning brush, sampler or oil injector according to task requirements to achieve various operations such as detection, cleaning, and maintenance. In addition to basic sensors, the core equipment includes a combustible gas detector (range 0~100%LEL, accuracy 1%), which can sensitively detect trace leaks; and an ultrasonic thickness gauge (resolution 0.1mm) for non-contact measurement of corrosion and thinning of cabinet walls, pipelines, etc. An integrated positioning module is built-in. After the rail robot makes a preliminary positioning of anomalies (such as finding high-temperature points through thermal imaging), the crawler robot can be dispatched to the target area, and the manipulator is used to approach for precise gas concentration measurement, wall thickness measurement or maintenance actions.

[0066] Main control system: As the system center, it is responsible for unified task scheduling, data aggregation analysis and equipment management. Based on algorithms such as image recognition, infrared temperature measurement analysis, gas concentration trend analysis, and sound spectrum analysis, it can automatically identify faults such as equipment leakage, temperature anomalies, and abnormal noises, and give early warnings. The system can also analyze the height data of each point of the piston, calculate and warn the tilt state of the piston in real time. It provides a Web end and a mobile App, which can be used to watch the inspection pictures in real time, receive alarms, remotely control the robot, generate inspection reports, etc.

[0067] Production equipment and valve groups supporting the gas holder: Through protocols such as industrial Ethernet and OPC UA, access the device data of existing PLC control systems, fixed gas alarms, piston height radar level gauges, etc. in the gas holder. Deploy an edge computing gateway in the cabinet area to conduct preliminary local analysis on the video stream and sensor data transmitted back by the robot, and only upload key events and metadata to the cloud or data center to reduce bandwidth pressure and improve response speed.

[0068] Embodiment 2 This embodiment provides a working method for a robot to work inside a gas holder, and the steps are as follows: 1. Manipulator ash cleaning operation: Check the status of the manipulator ash cleaning tool: The operator remotely checks the wear condition of the high-pressure purge nozzle and the air compressor pressure gauge in the control room. Confirm the safety of the environment inside the gas holder, and the camera confirms that there are no obstacles in the operation area. Based on the BIM model, set the specific area of the cabinet wall as the ash cleaning path, and turn on the power of the robot chassis, the manipulator control module, and the ash cleaning tool power supply in sequence.

[0069] Performing the dust removal operation: The robotic arm automatically moves to the designated position on the cabinet wall, opens the high-pressure air valve, and observes the dust removal effect through a high-definition camera. If stubborn dust accumulation is observed in certain areas, the operator increases the pressure, reduces the speed, and repeats the cleaning several times to achieve the desired result.

[0070] Post-dust cleaning process: Confirm dust removal is complete, close the high-pressure air valve, and reset the pressure gauge to zero. Retract the robotic arm to the robot's storage position and lock the joints. The system automatically records the operation time, dust removal area, air consumption, and any abnormalities. Move the robot to the recovery point and activate the built-in vacuum cleaner.

[0071] 2. Robotic arm lubrication operation: Preparation: Confirm the grease pump's oil level and check that the grease nozzle is clean and unobstructed. Based on the maintenance manual, this operation requires lubricating the guide rail slider on the cabinet wall; its position coordinates have been imported into the system. Plan the lubrication path from bottom to top. Start the dedicated power supply for the lubrication system; the oil circuit pressure self-check is normal.

[0072] Performing the lubrication action: The robotic arm positions itself at the lubrication starting point, starts the lubrication tool to perform the operation, monitors the lubrication effect in real time, and adjusts the lubrication parameters.

[0073] Post-lubrication processing: Turn off the lubrication tool, return the robotic arm to the standby position, record lubrication data, and clean up any residual oil.

[0074] 3. Robotic arm sampling operation: Preparation: Check the status of the sampling tool, confirm the location of the sampling point, set the sampling path plan, and start the power supply of the robotic arm.

[0075] Perform the sampling action: The robotic arm positions itself at the sampling starting point, starts the sampling tool to perform the operation, monitors the sampling effect in real time, and adjusts the sampling parameters.

[0076] Post-sampling processing: The sampling tool is turned off, the robotic arm returns to the standby position, the sampling data is recorded, and the sample is packaged and labeled.

[0077] Example 3 This embodiment provides a multi-sensor data fusion processing flow, the steps of which are as follows: A. Track-mounted robot: High-altitude structural defect detection process 1. Image acquisition and preprocessing: A high-resolution camera acquires images of the tank surface at a fixed frame rate. Image denoising (non-local mean filtering), illumination compensation (Retinex algorithm), and geometric distortion correction are performed. The U-Net network is then used to segment and extract key components such as steel structures, welds, and bolts to generate regions of interest (ROIs).

[0078] 2. Feature extraction and matching: Within the Region of Interest (ROI), the Canny operator combined with Hough transform is used for edge and contour enhancement, and the LBP algorithm is used for texture analysis. Target tracking between consecutive frames is achieved through ORB feature point matching, and combined with robot odometry and IMU data, coarse 3D reconstruction and pose estimation of defective targets are performed, providing a foundation for coordinate localization.

[0079] 3. Defect identification and classification: Visual signal processing: A pre-trained YOLOv8 model is used to detect high-definition images of high-altitude areas, identify defect categories (steel structure corrosion, weld cracks, loose bolts), corresponding defect target coordinates, and confidence scores.

[0080] Infrared signal processing: The DenseNet classification model was used to classify thermal image patches. Infrared anomaly categories (localized low-temperature anomalies, gas leak temperature spectrum patterns), temperature anomaly coordinates, and confidence scores were identified.

[0081] Acoustic signal processing: Audio signals from a microphone array are acquired, and features such as MFCC are extracted. A Gaussian mixture model (GMM) is used to cluster and identify normal and abnormal sounds (hissing sounds, bolt vibration sounds, and structural friction sounds). The categories of acoustic anomalies, the coordinates of abnormal sound sources (obtained through an acoustic array localization algorithm), and confidence scores are also identified.

[0082] 4. Multimodal data association and coordinate fusion: The coordinates of visual defects, infrared anomalies, and sound sources detected within the same time period are transformed to a unified global three-dimensional coordinate system. The Euclidean distance between each pair of coordinates is calculated. If a set of coordinates exists where the distances between all pairs of coordinates are less than a preset fusion threshold, they are determined to originate from the same physical defect. A confidence-weighted average is then applied to the coordinates of this defect to obtain the fused high-altitude defect coordinates.

[0083] B. Tracked Robot: Piston Sealing Area Status Monitoring Process 1. Image acquisition and preprocessing: The camera acquires images of the piston sealing system (rubber curtain, grease distribution, reflector). Noise reduction, enhancement, and perspective correction are performed to adapt to the platform's planar viewpoint.

[0084] 2. Feature extraction and matching: It primarily focuses on macroscopic analysis of surface texture and color features to aid in classification.

[0085] 3. Defect identification and classification: Piston surface visual inspection: The ResNet-50 model was used to classify high-resolution images of the piston sealing area (rubber curtain, grease, and reflector). The surface condition categories (aging and cracking of the sealing rubber curtain, drying and missing sealing grease, and excessive dust accumulation on the reflector) and the corresponding platform defect coordinates were obtained.

[0086] Gas concentration analysis: Concentration data is analyzed using an LSTM time-series prediction model. The model learns normal background concentration fluctuations and triggers an alarm when it detects abnormal concentration spikes or continuous upward trends that cannot be explained by environmental factors. The model acquires the concentration anomaly fluctuation patterns (instantaneous spikes), "gradient ascents," and the robot's pose coordinates at the time of the anomaly, i.e., the coordinates of the concentration anomaly point.

[0087] 4. Defect generation in the piston region: For piston region defect coordinates, coordinates of points with abnormal gas concentrations are usually preferred because they directly indicate the accumulation points of leaked substances. When there are no gas anomalies but severe surface defects exist, coordinates of platform defects are used.

[0088] By combining surface state categories and concentration anomaly fluctuation patterns, defect features of the piston region are constructed.

[0089] C. Cross-robot data fusion and precise leak location 1. Data synchronization and alignment: All data includes high-precision, synchronized timestamps. A unified global coordinate system ensures spatial comparability between aerial and ground coordinates.

[0090] 2. Vertical spatial correlation matching: The coordinates of each high-altitude defect are projected vertically along the direction of gravity onto the horizontal plane where the piston platform is located, yielding the projected coordinates. The Euclidean distance between these projected coordinates and the coordinates of each defect in the piston region on the XY plane is then calculated.

[0091] If the aerial coordinates and ground coordinates meet the following three conditions, it is determined to be a confirmed leak event: The Euclidean distance is less than the vertical correlation threshold; the high-altitude defect features corresponding to the high-altitude coordinates include leakage-related acoustic patterns or infrared features; the piston area defect features corresponding to the ground coordinates include abnormal gas concentration features.

[0092] 3. Calculation of leakage coordinates: For a confirmed leak event, the final fused leak coordinates are calculated, which combine the location information of the leak source (high altitude) and the leak manifestation (ground).

[0093] 4. Decision and Early Warning Generation: Using DS evidence theory or expert system reasoning, uncertainty modeling and comprehensive decision-making are performed on successfully correlated multi-source evidence to generate a comprehensive defect assessment report. Based on the severity of the leak (combined with concentration values, leak size estimates, etc.), tiered real-time early warnings (prompt, warning, severe) are initiated.

[0094] D. Inspection route coordination and system optimization 1. Dynamic task allocation: Based on the tank structure diagram, a high-altitude fan-shaped inspection area is designated for the tracked robot, and a bow-shaped path covering the piston platform is planned for the tracked robot. Once a suspected or confirmed defect is detected, the central scheduler will dynamically adjust the task priority.

[0095] 2. Resource Scheduling and Communication: The raw video stream is processed locally, transmitting only small data such as features, coordinates, and alarm signals. However, upon confirmation of a leak, a snapshot of multi-sensor data for the relevant time period is automatically transmitted back for manual verification. The system monitors the robot's battery level, plans the optimal path back to the charging and maintenance station when the battery is low, and dispatches a backup robot to take over the task. When the main communication link is interrupted, it automatically switches to a backup Mesh network or 5G link to ensure data transmission and command issuance.

[0096] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0097] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for collaborative inspection and maintenance of a gas holder using robots, characterized in that, Includes the following steps: The track-mounted robot runs on the rails of the two-story truss circular walkway, collecting visual images, infrared thermal imaging data and acoustic signals of the high-altitude structure. The tracked robot moves on a piston-ring-shaped platform to collect visual data from ground equipment and environmental gas data. Data collected by the tracked robot is processed to obtain high-altitude defect features and coordinates; data collected by the tracked robot is processed to obtain piston area defect features and coordinates. Based on the high-altitude defect features and coordinates and the piston region defect features and coordinates, leakage features and coordinates are obtained by matching them. A three-dimensional hazard map is generated based on the leakage characteristics and coordinates, and the three-dimensional hazard map includes the hazard type and location; The tracked and tracked robots perform maintenance operations based on the type and location of the hazard, and generate inspection data and maintenance operation records; The inspection data and maintenance operation records are transmitted to the main control system to generate an inspection report containing defect information. Based on the inspection report, the inspection cycle and maintenance strategy are optimized.

2. The method for collaborative inspection and maintenance of a gas holder using robots according to claim 1, characterized in that, The process of processing the data collected by the orbital robot to obtain high-altitude defect features and coordinates includes: The visual image is processed using the YOLOv8 target detection model to identify the surface defect categories such as steel structure corrosion, weld cracks, and loose bolts, and to identify the coordinates of the defect targets. The infrared thermal imaging data is classified using the DenseNet classification model to identify infrared anomaly categories and extract temperature anomaly coordinates. A Gaussian mixture model was used to perform cluster analysis on the acoustic signals to distinguish and identify abnormal acoustic anomaly patterns caused by sealing leaks, loose bolts, or structural friction, and to locate the coordinates of the abnormal sound sources. Calculate the Euclidean distance in three-dimensional space between the target coordinates of the defect, the temperature anomaly coordinates, and the abnormal sound source coordinates. If the Euclidean distance is less than the fusion threshold, it is determined to be the same physical defect. The target coordinates of the defect, the temperature anomaly coordinates, and the abnormal sound source coordinates are weighted and averaged to obtain the coordinates of the high-altitude defect. Surface defect categories, infrared anomaly categories, and acoustic anomaly patterns, along with their corresponding confidence scores, are constructed as high-altitude defect features.

3. The method for collaborative inspection and maintenance of a gas holder using robots according to claim 1, characterized in that, The process of processing the data collected by the tracked robot to obtain the defect features and coordinates of the piston region includes: The visual data was processed using the ResNet-50 image classification model to identify the surface condition categories of aging and cracking of the sealing rubber curtain, drying and missing sealing grease, and excessive dust accumulation on the reflector, and the platform defect coordinates were calculated. The gas concentration data is analyzed using a time-series prediction model based on a long short-term memory network to identify abnormal concentration fluctuation patterns. Combined with the implementation pose of the tracked robot, the coordinates of the concentration anomaly points are determined. Based on the surface state category and the concentration anomaly fluctuation pattern, piston region defect features are constructed, and piston region defect coordinates are constructed based on the platform defect coordinates and the coordinates of the concentration anomaly points.

4. The method for collaborative inspection and maintenance of a gas holder using robots according to claim 1, characterized in that, The process of matching the high-altitude defect features and coordinates with the piston region defect features and coordinates to obtain leakage features and coordinates includes: The coordinates of the high-altitude defect are vertically projected onto the piston plane to obtain the projected coordinates; Calculate the planar distance between the projected coordinates and the defect coordinates of each piston region; If there are high-altitude defects and piston area defects with a planar distance less than the vertical correlation threshold, and the high-altitude defects contain acoustic or infrared anomalies related to leakage, while the piston area defects contain gas concentration anomalies, then a confirmed leak is determined to have occurred. Based on the coordinates of the high-altitude defect and the coordinates of the piston region defect, the fused leak coordinates of the leak event are calculated and confirmed.

5. The method for collaborative inspection and maintenance of a gas holder using robots according to claim 4, characterized in that, The formula for calculating the fused leak coordinates of the confirmed leak event is as follows: in, x l To fuse the x-coordinate of the leak coordinates, y l To merge the y-coordinate of the leak coordinates, z l To fuse the z-coordinate of the leak coordinates, x f The x-coordinate of the high-altitude defect is given. x g The x-coordinate of the piston region defect coordinates. y f The y-coordinate of the high-altitude defect is given. y g Let y be the coordinate of the defect in the piston region. z f To fuse the z-coordinate of the leak coordinates, λ This is the planar blending correction coefficient.

6. The method for collaborative inspection and maintenance of a gas holder using robots according to claim 1, characterized in that, The track-mounted robot is equipped with a high-definition camera, an infrared thermal imager, and acoustic sensors for inspecting trusses, welds, guide rollers, cabinet walls, and high-altitude equipment. The tracked robot is equipped with a vision camera, a gas detector, and a retractable robotic arm. The robotic arm integrates a cleaning brush, a sampling probe, and an oil injection gun, and is used to inspect and maintain seals, reflectors, and instruments.

7. The method for collaborative inspection and maintenance of a gas holder using robots according to claim 1, characterized in that, The types of potential hazards include low-risk defects, medium-risk defects, and high-risk defects. For low-risk defects, the robotic arm of the tracked robot is controlled to perform automatic maintenance operations. For medium-risk defects, the tracked robot or the tracked robot is controlled to perform maintenance operations. For high-risk defects, the tracked robot and the tracked robot locate the high-risk defects and simultaneously trigger emergency procedures.

8. The method for collaborative inspection and maintenance of a gas holder using robots according to claim 7, characterized in that, The low-risk defects include dust accumulation on the reflector, and the automatic maintenance operation is a dust removal operation; the medium-risk defects include loose bolts, and the targeted maintenance operation is a tightening operation; the high-risk defects include piston tilt warning, and the emergency procedure includes cutting off the gas supply or activating the backup sealing mechanism.

Images