An intelligent inspection robot management and control platform, method, system and device

Abstract

The application discloses an intelligent inspection robot management and control platform, method, system and equipment, and belongs to the technical field of industrial automation and Internet of Things. The platform adopts a cloud-edge collaborative architecture, and comprises an edge side arranged on a site and a cloud platform layer adopting a front-end and back-end separation design. A back-end service layer of the cloud platform layer builds a micro-service cluster through a Go language, and a front-end application layer realizes cross-platform adaptation based on a Qt framework. The method comprises the steps of task arrangement and cloud-side issuing, edge-side execution and data acquisition, data uploading and edge-side preprocessing, cloud-side intelligent analysis and fusion diagnosis, real-time early warning and visual presentation, data closed loop and iterative optimization. Through multi-source data fusion analysis, AI intelligent diagnosis and multi-robot collaborative scheduling, the application solves the problems of rigid system architecture, data islands, poor real-time performance and collaboration of existing systems, realizes the intelligentization, collaboration and high efficiency of inspection operation, and is suitable for device inspection scenes in the energy industry such as electric power, coal and chemical industry.

Description

Technical Field

[0001] This invention relates to the fields of industrial automation, robot control and Internet of Things, specifically to an intelligent inspection robot management platform, method, system and equipment based on cloud-edge collaborative architecture, which can realize multi-robot collaborative scheduling, intelligent fusion analysis of multi-source data and real-time visual monitoring, and is applicable to equipment inspection scenarios in energy industries such as power, coal and chemical industries. Background Technology

[0002] In the energy industry, including power, coal, and chemical industries, routine inspections of equipment and facilities are crucial for ensuring safe and stable production. Traditional manual inspection methods rely on on-site inspections and recording by staff, which suffer from numerous problems, including high labor intensity, low inspection efficiency, significant susceptibility to environmental factors (such as high temperature, high pressure, and toxic environments) and human factors (such as fatigue and negligence), incomplete data recording, and difficulty in achieving digital management. These methods cannot meet the high-efficiency operation and maintenance needs of modern industrial production.

[0003] In recent years, inspection robot technology has been gradually applied to industrial inspection scenarios, but the supporting control platform still has significant shortcomings, which restricts the intelligent upgrading of inspection operations: 1. Rigid system architecture: Most existing platforms adopt a monolithic or tightly coupled architecture, with the front-end interface and back-end business logic deeply bound together. This makes system upgrades and modifications difficult, and they cannot flexibly adapt to various terminal devices such as large PC screens, tablets, and embedded screens for robots. They also have poor scalability and are unable to cope with the diverse needs of complex industrial environments.

[0004] 2. Insufficient Data Value Mining: The video, infrared, and sensor data collected by the inspection robot are isolated from the on-site fixed monitoring data and equipment management system data, forming "data silos." The platform lacks effective multi-source data fusion and in-depth analysis capabilities, and its early warning of equipment failures relies heavily on simple threshold judgments, resulting in low accuracy and an inability to achieve precise perception of equipment status and predictive maintenance.

[0005] 3. Poor real-time performance and collaboration: High data interaction latency between robots and the management platform, as well as between various functional modules within the platform, makes it difficult to support rapid response to emergencies such as equipment failures. Weak task coordination and scheduling among multiple robots, and weak linkage capabilities between robots and fixed monitoring equipment, result in unreasonable allocation of inspection resources and limited inspection efficiency.

[0006] 4. High integration and deployment costs: The lack of unified standards for existing system interfaces makes integration with existing industrial systems such as MES (Manufacturing Execution System) and EAM (Enterprise Asset Management System) difficult. Furthermore, centralized data processing places extremely high demands on network bandwidth and central server performance, resulting in persistently high system deployment and maintenance costs.

[0007] Therefore, there is an urgent need for an integrated management and control solution for inspection robots that is flexible in architecture, intelligent and efficient, highly collaborative, and easily scalable, in order to address the many shortcomings of existing technologies. Summary of the Invention

[0008] To address the aforementioned issues, this invention aims to propose an intelligent inspection robot management platform, method, system, and equipment. By introducing a cloud-edge collaborative and front-end / back-end separation architecture, it enhances the system's real-time response capabilities, scalability, and cross-platform adaptability, resolving the rigidity and poor adaptability issues of traditional architectures. Through deep integration of multi-source data and the introduction of AI intelligent analysis models, it achieves accurate perception, intelligent diagnosis, and predictive maintenance of equipment status, improving fault warning accuracy and reducing operation and maintenance costs. Furthermore, through standardized interface design and distributed processing solutions, it reduces the integration difficulty with existing industrial systems, optimizing deployment and operation and maintenance costs.

[0009] To achieve the above objectives, the technical solution of the present invention is implemented as follows: An intelligent inspection robot management platform, adopting a cloud-edge collaborative architecture, includes: Deployed on the edge side in the field, it includes at least one inspection robot for performing inspection tasks and collecting data, and edge computing nodes for data preprocessing, caching and network outage survival. The cloud platform layer adopts a front-end and back-end separation architecture, including: The backend service layer uses Go to build a microservice cluster, including a unified access and communication module, a task scheduling and collaboration module, a data intelligence analysis engine, a data resource management and service module, and a unified permission and security management module. The front-end application layer is built using the Qt framework to provide a cross-platform visual monitoring and operation interface; The edge side and the cloud platform layer interact with each other via a standardized RESTful API based on the HTTP / HTTPS protocol. The unified access and communication module integrates WebSocket service to establish a real-time communication link with the front-end application layer, enabling low-latency push of robot dynamics and alarm information.

[0010] Furthermore, the data intelligence analysis engine includes: The image recognition unit is used to identify instrument readings, determine status indicator lights, and detect appearance defects in equipment images collected by the robot using a pre-trained deep learning model. The sensor data analysis unit is used to perform real-time calculations, trend analysis, and abnormal fluctuation detection on the time-series data of acoustic, gas, and temperature sensors uploaded by the robot. The multi-source data fusion and cross-validation unit is used to combine the recognition results of the image recognition unit, the analysis results of the sensor data analysis unit, and the historical operating data of the equipment to make a comprehensive judgment to confirm the abnormal state of the equipment and reduce the false alarm rate.

[0011] Furthermore, the data resource management and service module adopts a layered hybrid storage architecture, including: a relational database for storing structured data such as device information, inspection tasks, and alarm records; a distributed object storage system for storing unstructured data such as images, videos, and infrared thermal images; a memory database for caching real-time robot status and frequently accessed data; and a distributed file system or time-series database for storing massive amounts of historical sensor data and supporting big data analysis.

[0012] Furthermore, the task scheduling and collaboration module has a built-in intelligent scheduling engine that can dynamically allocate and adjust inspection tasks for multiple robots based on the priority of inspection points, the robot's real-time battery level, health status, and location information, and perform collaborative path planning to avoid task conflicts and resource waste.

[0013] Furthermore, the front-end application layer is built on Qt 5.15 and above, and is compatible with Windows and Linux desktop operating systems and embedded Linux touch devices. It provides a unified visual monitoring interface that integrates electronic maps, data charts, video monitoring windows and alarm information lists, including panoramic visual monitoring, multi-dimensional data cockpit, task and asset configuration, alarm center and emergency response function modules.

[0014] Furthermore, the unified permission and security management module adopts the RBAC model to achieve fine-grained permission control, meets the Level 3 security requirements of the Information Security Protection System, and provides JWTAPI authentication and authorization, operation log auditing, data transmission encryption and access control functions.

[0015] To achieve the above objectives, the present invention also provides a method for controlling an intelligent inspection robot based on the above platform, comprising the following steps: S1. Task orchestration and cloud distribution: Managers configure inspection plans through the front-end application layer, and the task scheduling and collaboration module of the cloud platform layer decomposes the plans into specific tasks and distributes them asynchronously to the target inspection robot through a message queue. S2. Edge-side execution and data acquisition: After receiving the task, the inspection robot performs autonomous navigation and dynamic obstacle avoidance based on the environmental map constructed by SLAM technology, triggering the onboard high-definition visible light camera, infrared thermal imager, acoustic sensor and other equipment to collect raw data. S3. Data Uplink and Edge Preprocessing: Edge computing nodes encode, compress, preliminarily filter, and cache the collected raw data. Key data and alarm information are uploaded to the cloud platform layer through a standardized RESTful API. S4. Cloud-based intelligent analysis and fusion diagnosis: The data intelligent analysis engine at the cloud platform layer processes the uploaded multi-source data in parallel. Image data is used to identify defects through a deep learning model, sensor data is used for real-time calculation and trend analysis, and then the abnormal status of the equipment is confirmed through multi-source data fusion and cross-validation mechanism. S5. Real-time early warning and visualization: When an equipment malfunction is confirmed, alarm information is pushed to all online front-end application layers in real time via WebSocket service, accurately located and highlighted on the electronic map, and supports linkage to retrieve on-site video or dispatch nearby robots for verification. S6. Data closed loop and iterative optimization: All data in the inspection process is stored in the corresponding database. Equipment health reports are generated based on historical data, the inspection cycle and route are optimized, and the marked abnormal samples are fed back to the deep learning model to realize the model iterative upgrade.

[0016] Furthermore, the specific process of the multi-source data fusion and cross-validation mechanism described in step S4 is as follows: S41. Receive image data, sensor data, and corresponding device ID; S42. Parallel execution of image AI recognition and sensor data trend analysis; S43. If the image AI recognition does not find any abnormalities, the process ends; if a suspected abnormality is found, retrieve the sensor data of the device during the same time period. S44. If the sensor data shows an abnormal trend, confirm the defect alarm and execute step S5; if the sensor data is not abnormal, mark it as a low confidence alarm and store it in the database for manual review.

[0017] To achieve the above objectives, the present invention also provides an intelligent inspection robot control system, comprising: the aforementioned intelligent inspection robot control platform; at least one inspection robot equipped with a high-definition visible light camera, an infrared thermal imager, an acoustic sensor, a gas sensor, a lidar, and a navigation module, for performing inspection tasks and data acquisition; an edge computing node, implemented using an industrial gateway or server, for data preprocessing, caching, and basic functional assurance in offline states; and a third-party system interface for integration with existing industrial systems such as MES and EAM to achieve data interoperability.

[0018] To achieve the above objectives, the present invention also provides a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the intelligent inspection robot control method described above.

[0019] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: 1. Advanced architecture and elastic scaling: Adopting a front-end and back-end separation architecture of "Go back-end microservices + Qt cross-platform front-end", combined with cloud-edge collaboration mode, each module of the system can be deployed, expanded and maintained independently, which greatly improves the system's flexibility, maintainability and adaptability to complex industrial environments, and can flexibly adapt to a variety of terminal devices.

[0020] 2. Intelligent perception and precise decision-making: By introducing AI visual analysis models, sensor data trend analysis, and multi-source data fusion and cross-validation mechanisms, a leap from "passive alarm" to "proactive early warning and intelligent diagnosis" has been achieved, significantly improving the accuracy and reliability of equipment status perception and providing strong support for predictive maintenance.

[0021] 3. Collaborative and efficient, with real-time response: It realizes unified scheduling and collaborative operation of multiple robots, as well as intelligent linkage between robots and fixed monitoring facilities; the real-time communication mechanism based on WebSocket ensures extremely low latency push of monitoring information and alarms, improving the emergency response speed to emergencies.

[0022] 4. Open integration and cost optimization: The standardized API design facilitates integration with third-party industrial systems such as MES and EAM; edge-side data preprocessing reduces cloud computing load and network bandwidth requirements; the tiered hybrid storage solution balances storage performance and cost, thereby reducing the overall total cost of ownership (TCO) of the system. Attached Figure Description

[0023] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the overall system architecture of the intelligent inspection robot control platform according to an embodiment of the present invention; Figure 2 This is an exemplary layout diagram of the Qt framework visual monitoring interface of the front-end application layer in the intelligent inspection robot management platform described in this embodiment of the invention; Figure 3 This is a flowchart illustrating the logic of multi-source data cross-validation performed by the data intelligent analysis engine described in this embodiment of the invention. Figure 4This is a flowchart of the intelligent inspection robot control method described in an embodiment of the present invention. Detailed Implementation

[0024] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0025] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] Example 1 See Figure 1 A smart inspection robot management platform, adopting a cloud-edge collaborative architecture, includes: Deployed on the edge side in the field, it includes at least one inspection robot for performing inspection tasks and collecting data, and edge computing nodes for data preprocessing, caching and network outage survival. The cloud platform layer adopts a front-end and back-end separation architecture, including: The backend service layer uses Go to build a microservice cluster, including a unified access and communication module, a task scheduling and collaboration module, a data intelligence analysis engine, a data resource management and service module, and a unified permission and security management module. The front-end application layer is built using the Qt framework (QWidgets / QtQuick) to provide a cross-platform visual monitoring and operation interface; The edge side and the cloud platform layer interact with each other via a standardized RESTful API based on the HTTP / HTTPS protocol. The unified access and communication module integrates WebSocket service to establish a real-time communication link with the front-end application layer, enabling low-latency push of robot dynamics and alarm information.

[0027] It should be noted that the unified access and communication module uses the Gin framework to provide RESTful APIs (such as POST / api / v1 / robot / data for data upload and GET / api / v1 / task / assign for task retrieval) and uses the GorillaWebSocket library to maintain a long-lived connection with the front end.

[0028] Task scheduling and coordination module: High-concurrency task processing is achieved using Go goroutines and channels. The scheduling algorithm comprehensively considers robot battery level, distance, and current load to achieve load balancing.

[0029] Data intelligence analysis engine: This engine uses cgo or gRPC to call TensorFlow / PyTorch models deployed in a Python environment to perform intelligent analysis on uploaded device images. Simultaneously, a stream processing module written in Go performs windowing calculations and threshold determinations on the sensor data stream.

[0030] Data Resource Management and Service Module: Uses the GORM framework to operate the PostgreSQL database 135, storing device files and task logs. Images and videos are stored in MinIO object storage via the MinIOGoSDK. The Go-Redis client is used to store the robot's real-time status (such as location and battery level) in a Redis cache. Historical detailed data is periodically archived to HDFS.

[0031] Unified permissions and security management module: Integrates JWT for API authentication and authorization.

[0032] In a specific example, the data intelligence analysis engine includes: The image recognition unit is used to identify instrument readings, determine status indicator lights, and detect appearance defects (rust, looseness, leakage) in equipment images collected by the robot using pre-trained deep learning models (YOLOv5 / v8, SegNet). The sensor data analysis unit is used to perform real-time calculations, trend analysis, and abnormal fluctuation detection on the time-series data of acoustic, gas, and temperature sensors uploaded by the robot. The multi-source data fusion and cross-validation unit is used to combine the recognition results of the image recognition unit, the analysis results of the sensor data analysis unit, and the historical operating data of the equipment to make a comprehensive judgment to confirm the abnormal state of the equipment and reduce the false alarm rate.

[0033] It should be noted that by using pre-trained deep learning models (such as YOLOv5 / v8, SegNet), the instrument readings, status indicator status, and appearance defect features in the device image are quantified, and quantifiable recognition results (such as reading values, defect confidence, and indicator light on / off status) are output.

[0034] Instrument readings (general)

[0035] Note: Vmin / Vmax is the instrument range, θ act The AI ​​identifies the pointer angle and directly maps the reading.

[0036] Severity of appearance defects

[0037] Note: ConfD represents the AI ​​defect identification confidence level (0-1).

[0038] Multi-source data fusion and cross-validation unit: Conf total =0.4×S img +0.4×S sensor +0.2×S hist Note: S img Image recognition results (defect severity S) defect / 3, 0-1); S sensor Sensor abnormal results (abnormal = 1, normal = 0); S hist Historical anomaly probability (frequency of anomalies in the last 30 days / number of inspections, 0-1); Judgment rules: Conf total ≥0.7 (confirmed as abnormal), otherwise pending review.

[0039] In a specific example, the data resource management and service module adopts a layered hybrid storage architecture, including: a relational database (PostgreSQL cluster) for storing structured data such as device information, inspection tasks, and alarm records; a distributed object storage system (MinIO) for storing unstructured data such as images, videos, and infrared thermograms; a memory database (Redis cluster) for caching real-time robot status and frequently accessed data; and a distributed file system (HDFS) or time-series database for storing massive amounts of historical sensor data and supporting big data analysis.

[0040] In the specific implementation, the data storage scheme of the substation platform is as follows: Relational database (PostgreSQL cluster): storing structured data, including equipment information (such as transformer model, installation location, rated parameters), inspection tasks (task ID, executing robot, inspection time), alarm records (alarm ID, equipment ID, anomaly type, processing result), etc., storing a total of about 100,000 structured data records.

[0041] The distributed object storage system (MinIO) stores unstructured data, including equipment images collected by robots (approximately 5,000 images per day), infrared thermal images (approximately 2,000 images per day), and inspection video clips (10-30 seconds per clip, 100 clips per day), with a total storage capacity of approximately 5TB.

[0042] In-memory database (Redis cluster): caches the robot's real-time status (such as current location, remaining battery power, task progress) and frequently accessed basic information of devices (such as the rated parameters of key devices). The cached data is valid for 1 hour, ensuring that the front-end query response speed is ≤100ms.

[0043] InfluxDB (Time Series Database): Stores massive amounts of historical sensor data, including temperature, humidity, gas concentration, and acoustic data uploaded by each robot every 10 seconds, accumulating to approximately 180 million time series data entries over one year. It supports quick querying of trend curves by device and time range.

[0044] In a specific example, the task scheduling and collaboration module has a built-in intelligent scheduling engine that can dynamically allocate and adjust inspection tasks for multiple robots based on the priority of inspection points, the real-time battery level of the robots, their health status and location information, and perform collaborative path planning to avoid task conflicts and resource waste.

[0045] In practical implementation, the scenario of multi-robot collaborative scheduling in substations is as follows: The substation is pre-set with 10 inspection points, of which #3 and #7 are critical equipment inspection points (highest priority), and #1-#2 and #8-#10 are ordinary inspection points (medium priority).

[0046] At a certain moment: Robot A (80% remaining power, located at inspection point #1), Robot B (40% remaining power, located at inspection point #5), Robot C (90% remaining power, located at inspection point #9).

[0047] The intelligent scheduling engine dynamically allocates tasks: it assigns key inspection points #3 and #7 to robots A and C, which have sufficient power and are close by; it assigns ordinary inspection points #2 and #4 to robot B and adjusts its path to avoid conflict with the inspection routes of A and C; after robot B completes its task, it is scheduled to the charging area to replenish its power, thus achieving optimized resource allocation.

[0048] In a specific example, the front-end application layer is built on Qt 5.15 and above, and is compatible with Windows and Linux desktop operating systems and embedded Linux touch devices. It provides a unified visual monitoring interface that integrates electronic maps, data charts, video monitoring windows and alarm information lists, including panoramic visual monitoring, multi-dimensional data cockpit, task and asset configuration, alarm center and emergency response function modules.

[0049] It should be noted that the front-end application layer is developed using Qt 5.15 or later, and it includes: Panoramic Visualization Monitoring Module: Uses QtLocation and QtWebEngine (optional) to load online or offline maps, receives robot locations pushed from the backend via QWebSocket and updates them dynamically. Uses Qt3D (optional) to display the device's 3D model.

[0050] Multidimensional Data Dashboard Module: Uses QtCharts components to draw line charts and bar charts, dynamically displaying equipment temperature trends, alarm statistics, and other information.

[0051] Task and Asset Configuration Module: Provides a graphical editor that allows users to draw inspection routes on a map.

[0052] Alarm Center and Emergency Response Module: Displays real-time alarms in list and pop-up formats, and provides a one-click function to retrieve on-site video.

[0053] In the specific implementation, see Figure 2 This demonstrates an exemplary layout of a Qt front-end visual monitoring interface. The interface includes: A main electronic map view is used to display the robot's location, trajectory, and alarm points.

[0054] A robot status list that displays key status information for all robots in real time.

[0055] A data chart area used to display historical data trends for a selected device or region.

[0056] A live video window is used to display the live feed from the selected robot or a fixed camera.

[0057] An alarm information bar displays the latest alarm information in a scrolling manner.

[0058] In a specific example, the unified permission and security management module adopts the RBAC model to achieve fine-grained permission control, meets the Level 3 security requirements of the Information Security Protection System, and provides JWTAPI authentication and authorization, operation log auditing, data transmission encryption (HTTPS / SSL), and access control functions.

[0059] To achieve the above objectives, see Figure 4 This embodiment also provides a method for controlling an intelligent inspection robot based on the above platform, including the following steps: S1. Task orchestration and cloud distribution: Managers configure inspection plans through the front-end application layer, and the task scheduling and collaboration module of the cloud platform layer decomposes the plans into specific tasks and distributes them asynchronously to the target inspection robot through a message queue. S2. Edge-side execution and data acquisition: After receiving the task, the inspection robot performs autonomous navigation and dynamic obstacle avoidance based on the environmental map constructed by SLAM technology, triggering the onboard high-definition visible light camera, infrared thermal imager, and acoustic sensor equipment to collect raw data. S3. Data Uplink and Edge Preprocessing: Edge computing nodes encode, compress, preliminarily filter, and cache the collected raw data. Key data and alarm information are uploaded to the cloud platform layer through standardized RESTful APIs. S4. Cloud-based intelligent analysis and fusion diagnosis: The data intelligent analysis engine at the cloud platform layer processes the uploaded multi-source data in parallel. Image data is used to identify defects through a deep learning model, sensor data is used for real-time calculation and trend analysis, and then the abnormal status of the equipment is confirmed through multi-source data fusion and cross-validation mechanism. S5. Real-time early warning and visualization: When an equipment malfunction is confirmed, alarm information is pushed to all online front-end application layers in real time via WebSocket service, accurately located and highlighted on the electronic map, and supports linkage to retrieve on-site video or dispatch nearby robots for verification. S6. Data closed loop and iterative optimization: All data in the inspection process is stored in the corresponding database. Equipment health reports are generated based on historical data, the inspection cycle and route are optimized, and the marked abnormal samples are fed back to the deep learning model to realize the model iterative upgrade.

[0060] In practical implementation, the entire substation inspection process is applied as follows: S1 (Task Orchestration and Cloud Deployment): Maintenance personnel set up a "Daily Substation Inspection Plan" through the front end, which includes 10 inspection points and the inspection items for each point (such as temperature measurement and appearance inspection). The cloud scheduling module decomposes the plan into 5 specific tasks and asynchronously deploys them to 5 robots through a message queue.

[0061] S2 (Edge-side execution and data acquisition): After receiving the task, the robot autonomously navigates based on the substation environment map built by SLAM technology, avoiding tool cabinets and temporary obstacles in the inspection channel, triggering high-definition cameras to capture the appearance of equipment, infrared thermal imagers to detect temperature, and acoustic sensors to collect operating noise.

[0062] S3 (Data Uplink and Edge Preprocessing): Edge computing nodes compress the acquired 4K images to 1080P, filter out blurry and occluded invalid data, and upload valid data and preliminarily identified suspected abnormal information (such as a device with an excessively high temperature) to the cloud via RESTful API.

[0063] S4 (Cloud-based Intelligent Analysis and Fusion Diagnosis): The cloud engine uses AI to recognize images and confirms that the device has no defects in appearance; it performs trend analysis on temperature data and finds that the device temperature is 15°C higher than the historical average. Combined with the fact that the acoustic data at the same time is normal, the device is judged to be "temperature abnormal" through the fusion verification mechanism.

[0064] S5 (Real-time Early Warning and Visualization): The system pushes temperature anomaly alarms to the front end via WebSocket, accurately locates the device on the electronic map and marks it with a flashing red mark, and simultaneously retrieves the device's real-time video. Maintenance personnel can dispatch nearby robots for verification.

[0065] S6 (Data Closed Loop and Iterative Optimization): Store all data from this inspection into the corresponding database, generate an equipment health report, and recommend shortening the inspection cycle of the equipment from 2 hours to 1 hour; label the abnormal temperature samples and feed them back to the deep learning model to optimize the model's accuracy in identifying abnormal temperatures in this type of equipment.

[0066] See a specific example. Figure 3 The specific process of the multi-source data fusion and cross-validation mechanism described in step S4 is as follows: S41. Receive image data, sensor data, and corresponding device ID; S42. Parallel execution of image AI recognition and sensor data trend analysis; S43. If the image AI recognition does not find any abnormalities, the process ends; if a suspected abnormality is found, retrieve the sensor data of the device during the same time period. S44. If the sensor data shows an abnormal trend, confirm the defect alarm and execute step S5; if the sensor data is not abnormal, mark it as a low confidence alarm and store it in the database for manual review.

[0067] In a specific cross-validation scenario (for transformer bushing defect detection): S41: Receive transformer bushing image data collected by the robot, concurrent partial discharge sensor data, and the transformer's device ID (#T-003).

[0068] S42: Parallel execution of image AI recognition (detecting whether there are cracks in the bushing) and sensor data trend analysis (whether the partial discharge signal exceeds the standard).

[0069] S43: Image AI recognition detected a suspected crack on the edge of the sleeve (confidence level 0.75), proceed to the next verification step; if no abnormality is found, the process ends.

[0070] S44: Retrieve the partial discharge sensor data of transformer #T-003 during the same time period. The partial discharge signal value is found to be 80pC (threshold 50pC, with an abnormal trend). Confirm the defect alarm and proceed to step S5. If there is no abnormality in the partial discharge data, mark it as a low confidence alarm and store it in the database for manual review.

[0071] Example 2 To achieve the above objectives, this embodiment also provides an intelligent inspection robot management and control system, including: the aforementioned intelligent inspection robot management and control platform; at least one inspection robot, which is equipped with a high-definition visible light camera, an infrared thermal imager, an acoustic sensor, a gas sensor, a lidar, and a navigation module for performing inspection tasks and data acquisition; an edge computing node, implemented using an industrial gateway or server, for data preprocessing, caching, and basic functional assurance in offline states; and a third-party system interface for integration with existing industrial systems such as MES and EAM to achieve data interoperability.

[0072] In practical implementation, the complete substation system consists of: Management and control platform: The above-mentioned platform is used and deployed on the substation's private cloud.

[0073] Inspection robots: 5 wheeled inspection robots, each equipped with a high-definition visible light camera (1080P), an infrared thermal imager (temperature measurement range -20℃~150℃), an acoustic sensor (detection range 20Hz~20kHz), an SF6 gas sensor, a lidar (range measurement range 0.1~50m) and a navigation module.

[0074] Edge computing node: 1 industrial gateway (CPU is Intel Core i5, memory is 8GB, storage is 1TB) to realize data preprocessing, caching and network outage keep-alive.

[0075] Third-party system interfaces: Through standardized APIs, the system integrates with the substation's MES (Manufacturing Execution System) and EAM (Enterprise Asset Management System) to synchronize inspection data to the MES system for production scheduling and equipment anomaly information to the EAM system for maintenance work order generation, thereby achieving data interoperability.

[0076] Example 3 To achieve the above objectives, this embodiment also provides a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the intelligent inspection robot control method described above.

[0077] In a specific implementation, a computing device that deploys the control method: Hardware configuration: 16GB DDR4 memory, 1TB SSD hard drive; Intel Xeon E3-1230v6 processor (4 cores, 8 threads, 3.5GHz).

[0078] Software configuration: The computer program stored on the memory is written in Go and integrates core logic such as task scheduling, data processing, and AI analysis.

[0079] Performance: When the processor executes this program, it can implement the above-mentioned control methods and steps, such as task assignment, data acquisition and preprocessing, multi-source data fusion diagnosis, and real-time early warning. The single task processing latency is ≤500ms, and it supports 5 robots working online collaboratively at the same time.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart inspection robot control platform, characterized in that, The cloud-edge collaborative architecture is adopted, including: Deployed on the edge side in the field, it includes at least one inspection robot for performing inspection tasks and collecting data, and edge computing nodes for data preprocessing, caching and network outage survival. The cloud platform layer adopts a front-end and back-end separation architecture, including: The backend service layer uses Go to build a microservice cluster, including a unified access and communication module, a task scheduling and collaboration module, a data intelligence analysis engine, a data resource management and service module, and a unified permission and security management module. The front-end application layer is built using the Qt framework to provide a cross-platform visual monitoring and operation interface; The edge side and the cloud platform layer interact with each other via a standardized RESTful API based on the HTTP / HTTPS protocol. The unified access and communication module integrates WebSocket service to establish a real-time communication link with the front-end application layer, enabling low-latency push of robot dynamics and alarm information.

2. The intelligent inspection robot control platform according to claim 1, characterized in that, The data intelligence analysis engine includes: The image recognition unit is used to identify instrument readings, determine status indicator lights, and detect appearance defects in equipment images collected by the robot using a pre-trained deep learning model. The sensor data analysis unit is used to perform real-time calculations, trend analysis, and abnormal fluctuation detection on the time-series data of acoustic, gas, and temperature sensors uploaded by the robot. The multi-source data fusion and cross-validation unit is used to combine the recognition results of the image recognition unit, the analysis results of the sensor data analysis unit, and the historical operating data of the equipment to make a comprehensive judgment to confirm the abnormal state of the equipment and reduce the false alarm rate.

3. The intelligent inspection robot control platform according to claim 1, characterized in that, The data resource management and service module adopts a layered hybrid storage architecture, including: a relational database for storing structured data such as device information, inspection tasks, and alarm records; a distributed object storage system for storing unstructured data such as images, videos, and infrared thermal images; a memory database for caching real-time robot status and frequently accessed data; and a distributed file system or time-series database for storing massive amounts of historical sensor data and supporting big data analysis.

4. The intelligent inspection robot control platform according to claim 1, characterized in that, The task scheduling and collaboration module has a built-in intelligent scheduling engine that can dynamically allocate and adjust inspection tasks for multiple robots based on the priority of inspection points, the real-time battery level of the robots, their health status and location information, and perform collaborative path planning to avoid task conflicts and resource waste.

5. The intelligent inspection robot control platform according to claim 1, characterized in that, The front-end application layer is built on Qt 5.15 and above, and is compatible with Windows and Linux desktop operating systems and embedded Linux touch devices. It provides a unified visual monitoring interface that integrates electronic maps, data charts, video monitoring windows and alarm information lists, including panoramic visual monitoring, multi-dimensional data cockpit, task and asset configuration, alarm center and emergency response function modules.

6. The intelligent inspection robot control platform according to claim 1, characterized in that, The unified access and security management module adopts the RBAC model to achieve fine-grained access control, meets the Level 3 security requirements of the Information Security Protection System, and provides JWTAPI authentication and authorization, operation log auditing, data transmission encryption and access control functions.

7. A method for controlling an intelligent inspection robot based on the platform described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Task orchestration and cloud distribution: Managers configure inspection plans through the front-end application layer, and the task scheduling and collaboration module of the cloud platform layer decomposes the plans into specific tasks and distributes them asynchronously to the target inspection robot through a message queue. S2. Edge-side execution and data acquisition: After receiving the task, the inspection robot performs autonomous navigation and dynamic obstacle avoidance based on the environmental map constructed by SLAM technology, triggering the onboard high-definition visible light camera, infrared thermal imager, acoustic sensor and other equipment to collect raw data. S3. Data Uplink and Edge Preprocessing: Edge computing nodes encode, compress, preliminarily filter, and cache the collected raw data. Key data and alarm information are uploaded to the cloud platform layer through a standardized RESTful API. S4. Cloud-based intelligent analysis and fusion diagnosis: The data intelligent analysis engine at the cloud platform layer processes the uploaded multi-source data in parallel. Image data is used to identify defects through a deep learning model, sensor data is used for real-time calculation and trend analysis, and then the abnormal status of the equipment is confirmed through multi-source data fusion and cross-validation mechanism. S5. Real-time early warning and visualization: When an equipment malfunction is confirmed, alarm information is pushed to all online front-end application layers in real time via WebSocket service, accurately located and highlighted on the electronic map, and supports linkage to retrieve on-site video or dispatch nearby robots for verification. S6. Data closed loop and iterative optimization: All data in the inspection process is stored in the corresponding database. Equipment health reports are generated based on historical data, the inspection cycle and route are optimized, and the marked abnormal samples are fed back to the deep learning model to realize the model iterative upgrade.

8. The intelligent inspection robot control method according to claim 7, characterized in that, The specific process of the multi-source data fusion and cross-validation mechanism described in step S4 is as follows: S41. Receive image data, sensor data, and corresponding device ID; S42. Parallel execution of image AI recognition and sensor data trend analysis; S43. If the image AI recognition does not detect any anomalies, the process ends; If a suspected anomaly is detected, retrieve the sensor data of the device from the same time period. S44. If the sensor data shows an abnormal trend, confirm the defect alarm and execute step S5; if the sensor data is not abnormal, mark it as a low confidence alarm and store it in the database for manual review.

9. A smart inspection robot control system, characterized in that, include: The intelligent inspection robot control platform according to any one of claims 1-6; At least one inspection robot, equipped with a high-definition visible light camera, an infrared thermal imager, an acoustic sensor, a gas sensor, a lidar, and a navigation module, is used to perform inspection tasks and data collection; an edge computing node, implemented using an industrial gateway or server, is used for data preprocessing, caching, and ensuring basic functions in offline states. Third-party system interfaces are used to integrate with existing industrial systems such as MES and EAM to achieve data interoperability.

10. A computing device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the intelligent inspection robot control method as described in claim 7 or 8.

Images