Robotic inspection method, system, and medium for facility maintenance

Abstract

This invention discloses a robot inspection method, system, and medium for facility maintenance, belonging to the field of facility maintenance. The method includes: equipping a facility maintenance robot with a sensor array to collect multi-dimensional environmental datasets in real time; combining map data with the multi-dimensional environmental datasets to generate a three-dimensional environmental spatial model; identifying obstacles and abnormal areas in the three-dimensional spatial model to obtain environmental obstacle information and abnormal facility area information; performing fault diagnosis on the abnormal area information to obtain fault diagnosis results; and determining a global inspection path based on the fault diagnosis results, and performing a global facility inspection according to the global inspection path. This application solves the technical problem in the prior art where inspection robots cannot match actual operation and maintenance needs, making it difficult to simultaneously ensure passage safety, fault identification accuracy, and operational efficiency. It achieves the technical effect of improving fault identification accuracy and the safety and efficiency of inspection operations, effectively adapting to actual operation and maintenance needs.

Description

Technical Field

[0001] This invention relates to the field of facility maintenance, and more specifically to robotic inspection methods, systems, and media for facility maintenance. Background Technology

[0002] In the full lifecycle operation and maintenance management of various fixed facilities such as municipal integrated pipe corridors, industrial plants, transportation hubs, and large parks, facility inspection is a core link in preventing safety hazards and ensuring the safe and stable operation of facilities. As the scale of facilities under maintenance continues to expand, their structural complexity increases, and the requirements for operation and maintenance safety rise, the traditional manual inspection model suffers from numerous shortcomings, including high risks in high-risk scenarios, low inspection efficiency, delayed detection of potential faults, and the susceptibility of inspection results to subjective human factors, leading to missed or false inspections. It can no longer meet the intelligent and refined management needs of modern facility operation and maintenance. While some robotic solutions for facility inspection have emerged in existing technologies, most of these solutions adopt a pre-set fixed path operation mode and cannot be based on the actual conditions of the facility. The multi-source environmental data collected in real time on site enables dynamic 3D spatial modeling and accurate perception of the entire scene. On the one hand, it is difficult to simultaneously achieve accurate identification of dynamic obstacles and effective positioning of abnormal areas in the inspection scene. It is impossible to complete the full-dimensional qualitative and quantitative diagnosis from abnormal appearance to fault type, hazard level and precise location. On the other hand, it is impossible to carry out global optimal path planning by combining the passage constraints of on-site obstacles and the inspection priority corresponding to the urgency of the fault. This results in a serious disconnect between inspection operations and the actual operation and maintenance needs of the facilities. It is impossible to prioritize the coverage of high-risk fault points and achieve early detection and early handling of safety hazards. It is also impossible to ensure the passage safety of robot inspections throughout the process. At the same time, there are also redundancy in inspection paths, high energy consumption and low overall inspection efficiency. Summary of the Invention

[0003] This application provides a robotic inspection method, system, and medium for facility maintenance, aiming to solve the technical problems in the prior art where facility maintenance inspection robots cannot match actual operation and maintenance needs, and it is difficult to simultaneously ensure traffic safety, fault identification accuracy, and operation efficiency.

[0004] In view of the above problems, this application provides a robotic inspection method, system and medium for facility maintenance.

[0005] The first aspect disclosed in this application provides a robotic inspection method for facility maintenance, the method comprising:

[0006] A sensor array is mounted on a facility maintenance robot to collect multi-dimensional environmental datasets of the facility in real time. This dataset is then combined with facility map data to create a dynamic 3D spatial model of the facility environment. Obstacles and abnormal areas are identified within the 3D spatial model to obtain obstacle information and abnormal facility area information. Fault diagnosis is performed on the abnormal facility area information to obtain facility fault diagnosis results. Global path planning is then performed based on the obstacle information and the facility fault diagnosis results to determine the robot's global inspection path. The facility maintenance robot is then controlled to perform a global facility inspection according to this global inspection path.

[0007] A second aspect of this application provides a robotic inspection system for facility maintenance, the system comprising:

[0008] The data acquisition module is used to mount a sensor group on the facility maintenance robot. The sensor group collects a multi-dimensional environmental dataset of the facility in real time, and combines it with facility map data to perform dynamic three-dimensional spatial modeling, generating a three-dimensional spatial model of the facility environment. The anomaly identification module identifies obstacles and abnormal areas in the three-dimensional spatial model of the facility environment, obtaining environmental obstacle information and abnormal facility area information. The fault diagnosis module diagnoses faults in the abnormal facility areas, obtaining facility fault diagnosis results. The inspection path module performs global path planning based on the environmental obstacle information and the facility fault diagnosis results, determines the robot's global inspection path, and controls the facility maintenance robot to perform a global facility inspection according to the robot's global inspection path.

[0009] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the robotic inspection method for facility maintenance provided in this application.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] By equipping a facility maintenance robot with a sensor array, multi-dimensional environmental datasets are collected in real time. These datasets are then combined with map data to generate a 3D spatial model of the environment. Obstacles and abnormal areas are identified within the 3D model, yielding information on environmental obstacles and abnormal facility areas. Fault diagnosis is performed on the abnormal areas, providing diagnostic results. Based on these results, a global inspection path is determined, and a global facility inspection is conducted according to this path. This approach solves the technical problems of existing inspection robots, where inspection operations cannot match actual maintenance needs, and it is difficult to simultaneously ensure traffic safety, fault identification accuracy, and operational efficiency. It achieves the technical effect of improving fault identification accuracy and the safety and efficiency of inspection operations, effectively adapting to actual maintenance requirements.

[0012] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0013] Figure 1 This application provides a flowchart illustrating a robotic inspection method for facility maintenance;

[0014] Figure 2 This application provides a structural schematic diagram of a robotic inspection system for facility maintenance;

[0015] Figure labeling: Data acquisition module 11, anomaly identification module 12, fault diagnosis module 13, inspection path module 14. Detailed Implementation

[0016] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0017] The overall concept of the technical solution provided in this application is as follows:

[0018] This application provides a robotic inspection method, system, and medium for facility maintenance. It addresses the technical problem in existing technologies where inspection robots cannot match actual maintenance needs, making it difficult to simultaneously ensure traffic safety, fault identification accuracy, and operational efficiency. By equipping the facility maintenance robot with a sensor array, it collects multi-dimensional environmental datasets in real time. This dataset is then combined with map data to generate a three-dimensional environmental spatial model. The model is then used to identify obstacles and abnormal areas, obtaining information on environmental obstacles and abnormal facility areas. Fault diagnosis is performed on the abnormal areas to obtain fault diagnosis results. Based on the fault diagnosis results, a global inspection path is determined, and a global facility inspection is conducted according to this path. This achieves the technical effect of improving fault identification accuracy and the safety and efficiency of inspection operations, effectively adapting to actual maintenance needs.

[0019] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0020] Example 1, as Figure 1 As shown in the embodiment of this application, a robot inspection method for facility maintenance is provided, the method comprising:

[0021] S100: A sensor group is mounted on the facility maintenance robot. The sensor group collects multi-dimensional environmental data of the facility in real time. The data is combined with the facility map data and the multi-dimensional environmental data of the facility to perform dynamic three-dimensional spatial modeling and generate a three-dimensional spatial model of the facility environment.

[0022] Specifically, a sensor suite is first installed on the facility maintenance robot. This wheeled inspection robot is equipped with a sensor suite including: a 16-line mechanical LiDAR with a ranging accuracy of ±2cm, a ranging range of 0.1-100m, and a sampling frequency of 10Hz; an RGB-D depth camera with a resolution of 1920×1080, a field of view of 85°×65°, and a sampling frequency of 30Hz; a 6-axis IMU with an update frequency of 200Hz and zero-bias stability ≤0.5° / h; and a 2-megapixel visible light camera with a sampling frequency of 25Hz. This sensor suite is a multi-source sensing hardware collection adapted to facility inspection scenarios, typically including LiDAR, depth cameras, inertial measurement units, visible light cameras, and other sensing devices. It uses an "eye-on-hand" installation method, with the LiDAR and depth camera optical axes parallel and a baseline distance ≤5cm. Hardware triggering enables microsecond-level time synchronization of multiple sensors. The facility map data is comprehensive. The BIM as-built model of the utility tunnel's power compartment includes static spatial information of fixed structures such as tunnel walls, supports, high-voltage switchgear, cables, inspection passages, and fire doors. A total of 120 pre-set inspection points are used to synchronously collect multi-dimensional information such as geometric contours, spatial locations, visual textures, and posture changes of the facility environment. Subsequently, a multi-source heterogeneous data set covering the facility to be inspected and its surrounding environment is synchronously collected by the sensor group during the robot's inspection movement. This data includes multi-dimensional information such as laser point cloud data, image frame data, robot pose data, and environmental depth data. Combined with facility map data, which refers to the prior basic map information of the facility to be inspected, including static basic information such as the facility's fixed building structure, standard equipment layout, pre-set inspection points, and safety passage boundaries, it serves as the spatial benchmark for 3D modeling. This data is then used to conduct dynamic 3D spatial modeling with the facility's multi-dimensional environmental dataset to generate a 3D spatial model of the facility environment.

[0023] S200: Identify obstacles and abnormal areas in the three-dimensional spatial model of the facility environment to obtain environmental obstacle information and abnormal facility area information.

[0024] Specifically, for obstacle recognition, the original point cloud data in the 3D spatial model of the facility environment is first downsampled. This involves reducing the dimensionality of the high-density, highly redundant original point cloud data while preserving the core geometric contours and key spatial boundary features of the scene. Duplicate and noisy invalid point clouds are removed, significantly reducing the computational load of subsequent algorithms without sacrificing recognition accuracy. This generates standardized 3D point cloud data for the facility environment. Then, the DBSCAN algorithm, a density-based spatial clustering algorithm with noise, is used. This algorithm does not require pre-setting the number of clusters and can effectively identify spatial clusters of arbitrary shapes while isolating discrete noise data. The aforementioned 3D point cloud data of the facility environment is then segmented into multiple independent point clouds corresponding to different spatial entities based on spatial density characteristics. Clustering is used to obtain the cluster segmentation results of facility environment point cloud data, realize the spatial isolation and individual division of different entities in the scene, and then, based on the pre-set obstacle judgment criteria, that is, the judgment rules formulated by combining the inspection robot body size, passage parameters, and facility inspection safety operation specifications, covering core judgment dimensions such as object spatial size, position boundary, overlap with the preset inspection channel, and whether it belongs to the inherent static structure of the facility, it is used to distinguish the fixed facility structure in the scene from the obstacle object that affects the robot's passage. Each group of point cloud clusters in the facility environment point cloud data cluster segmentation results is judged and identified one by one, and finally the environmental obstacle information is extracted. This information includes the precise spatial coordinates, three-dimensional contour size, type attributes, and the range of influence on the robot's passage, which clarifies the impassable areas and obstacle avoidance constraint boundaries in the inspection scene.

[0025] While advancing obstacle identification, the identification of abnormal facility areas is carried out simultaneously. Based on facility safety application standards—namely, official or enterprise-level standardized specifications such as equipment operation and maintenance safety regulations, facility structural safety standards, and normal operating status requirements of the industry to which the facility to be inspected belongs—a standard feature database of facility objects is constructed. This database integrates benchmark information such as the standard appearance features, geometric dimensions, fixed spatial locations, and characteristic parameters under normal operating conditions of all equipment and structural components within the facility to be inspected. It serves as the core reference benchmark for determining whether anomalies exist in the facility area. Subsequently, using this database as a benchmark, the corresponding facilities in the three-dimensional spatial model of the facility environment are analyzed. The system aligns and matches the areas of the fixed structure one by one to identify and extract abnormal areas. This involves comparing the point cloud features and visual texture features of the area to be detected in the model with the corresponding standard features in the database. The system extracts feature data that deviates significantly from the standard features and identifies abnormal states that do not meet safety standards, such as structural deformation, missing parts, positional displacement, surface damage, and foreign object attachment. This allows for the precise delineation of the spatial range of the abnormality and the final determination of the abnormal facility area information. This information includes the precise spatial location of the abnormal area, the unique identifier of the facility or equipment to which it belongs, the type and degree of deviation from the standard features, and other details, thus completing the precise location of the abnormality of the facility itself.

[0026] S300: Perform fault diagnosis on the abnormal facility area information to obtain facility fault diagnosis results.

[0027] Specifically, based on historical facility failure cases—that is, the complete failure and maintenance records accumulated during the entire lifecycle operation and maintenance of the facility to be inspected and similar standardized facilities—covering full-dimensional experience data such as the scenario characteristics of the failure, abnormal appearance, corresponding equipment type, failure mechanism, hazard impact level, and supporting disposal plan, fault diagnosis reasoning is carried out. This involves sorting out and refining the mapping relationship between abnormal appearance characteristics and the essential attributes of the failure, constructing a standardized reasoning logic chain of abnormal characteristics-failure type-hazard level, and ultimately forming a facility failure knowledge base. This is an intelligent diagnostic database that integrates full failure feature tags, standardized failure reasoning rules, failure type classification standards, and precise failure location logic, serving as the core judgment benchmark and reasoning support for fault diagnosis in this step. Subsequently, this facility failure knowledge base is called, and the input abnormal facility area information is matched and intelligently reasoned with the failure feature tags and reasoning rules in the knowledge base in multiple dimensions. This is simultaneously combined with the unified spatial coordinate system of the facility environment three-dimensional spatial model generated in the previous step. The system achieves hierarchical and precise fault location from the regional level to the component level, ultimately outputting three core diagnostic elements: facility fault mode (identifying the specific type and mechanism of the fault, such as structural cracking, component detachment, equipment leakage, and aging wiring), accurately defining the essential attributes of the fault; facility fault severity (classifying the severity level based on facility safety application standards and the impact of the fault, specifying the scope of impact and the degree of interference with normal facility operation), quantifying the fault hazard; and facility fault location (the precise spatial coordinates of the fault in the 3D spatial model of the facility environment, the unique identifier of the associated facility equipment / structural component, and the specific component-level location information), achieving precise spatial anchoring of the fault. Finally, the three core elements of facility fault mode, facility fault severity, and facility fault location are uniquely bound and integrated to form standardized and structured facility fault diagnosis results where each fault item corresponds to a unique spatial location, clearly defines the fault attributes, and clearly defines the hazard level, thus completing the entire fault diagnosis process.

[0028] This step transforms the superficial anomaly identification results from the front end, which can only locate the anomalies, into diagnostic conclusions that clearly define what the anomaly is, how serious it is, and its specific location. This solves the problem that simple anomaly identification cannot qualitatively determine the fault or support maintenance decisions. On the other hand, the standardized facility fault diagnosis results it outputs are the core inputs and constraints for subsequent inspection priority analysis and global path planning. This directly determines the pertinence and rationality of subsequent inspection path planning, ensuring that the robot can prioritize covering high-risk fault areas according to the severity of the fault, improving the efficiency of inspection operations and the actual maintenance value. At the same time, this step's intelligent reasoning diagnosis mode based on the fault knowledge base not only ensures the standardization and consistency of fault diagnosis and avoids the subjective errors of manual judgment, but also continuously iterates and optimizes the knowledge base through continuously added fault cases, realizing the self-iteration and self-upgrading of the entire inspection system's diagnostic capabilities. This supports the upgrade of the entire solution from automated environmental inspection to intelligent facility maintenance, directly determining the accuracy, effectiveness, and practical value of the entire facility maintenance solution.

[0029] S400: Based on the environmental obstacle information and the facility fault diagnosis results, perform global path planning, determine the robot's global inspection path, and control the facility maintenance robot to perform global facility inspection according to the robot's global inspection path.

[0030] Specifically, inspection priority analysis is conducted based on facility fault diagnosis results. This involves constructing standardized scoring rules by combining core dimensions such as the severity of the fault, its impact on the safe operation of the facility, and the urgency of the fault handling. All diagnosed fault inspection points are then graded and ranked to determine the facility fault inspection priority sequence. This clarifies the order in which fault points need to be prioritized and checked during robot inspections, resolving the problem of prioritizing inspection targets in scenarios with multiple fault points and ensuring that high-risk and high-urgency fault points are inspected first. Subsequently, the aforementioned environmental obstacle information is used as a hard spatial constraint for robot passage, and the facility fault inspection priority sequence is used as a temporal constraint for inspection targets. The 3D spatial model of the facility environment generated in the pre-process is loaded into a high-precision digital twin scene model that integrates the prior static structure of the facility with real-time dynamic environmental information and has a unified spatial coordinate system. This model serves as the global spatial basis for path planning. Global path planning is carried out within the entire spatial range of the facility to be inspected. Under the premise of fully covering all fault inspection points, avoiding all obstacles and inaccessible areas, and meeting the kinematic constraints of the facility maintenance robot, a global path planning algorithm adapted to indoor and outdoor facility inspection scenarios is adopted to generate multiple facility environment inspection paths that meet the basic constraints. Each candidate path achieves complete coverage of all fault points to be inspected and effective avoidance of all obstacles.

[0031] Next, the generated multiple facility environment inspection paths are optimized. Based on facility maintenance and inspection requirements, including core operation and maintenance optimization objectives such as shortest total inspection path length, least total inspection operation time, earliest arrival time at high-priority fault points, lowest robot driving energy consumption, and lowest inspection point overlap coverage, a multi-objective function for facility inspection is constructed. This function is a quantitative evaluation function that integrates multiple inspection optimization objectives and can assign corresponding weights to different objectives according to actual operation and maintenance needs. It is used to perform standardized and quantifiable comprehensive scoring on all candidate paths. Subsequently, based on this multi-objective function, all candidate facility environment inspection paths are quantitatively calculated, comprehensively evaluated, and ranked one by one. The path with the best comprehensive score is selected, and the robot's global inspection path is finally determined. This global path is adapted to the actual on-site environment. A standardized inspection execution path for the entire domain is designed to meet the priority requirements of facility operation and maintenance and achieve optimal overall performance. This global inspection path is then transmitted to the robot's underlying motion controller via Ethernet TCP / IP protocol. Model predictive control (MPC) algorithm is used for path tracking, with the lateral error controlled within ±5cm. After reaching a fault inspection point, the robot stops, triggering the sensor array to complete detailed data verification and collection for that fault point. After completing the full inspection, the robot returns to its docking point. The process strictly adheres to the preset parameters, including point sequence, travel speed, stopping requirements at inspection points, and data collection rules. This precise control ensures the facility maintenance robot completes a full-scene global inspection of the facilities along the planned path, fully covering all fault points to be checked, and simultaneously completing data verification and collection during the inspection process.

[0032] This step, on the one hand, transforms the obstacle risks identified at the front end and the fault results diagnosed by the faults into standardized actions that the robot can execute through priority ranking and multi-constraint path planning, ensuring the feasibility of the entire inspection plan. On the other hand, through a multi-objective function path optimization mechanism, it achieves a multi-dimensional balance of inspection efficiency, maintenance priority, and operation cost while ensuring the robot's safe passage and collision-free risk. It not only ensures that high-risk fault points are inspected first, maximizing the facility safety guarantee value of the inspection operation, but also reduces the robot's operation energy consumption and time consumption through path optimization, thereby improving the overall inspection operation efficiency.

[0033] Further, in step S100, the facility map data and the facility multidimensional environment dataset are combined to perform three-dimensional spatial dynamic modeling to generate a three-dimensional spatial model of the facility environment, including:

[0034] The facility multidimensional environment dataset is standardized according to the data collection source to obtain a standard facility multidimensional environment dataset; the standard facility multidimensional environment dataset is spatiotemporally aligned and fused to obtain a facility environment fused dataset; the facility map data is loaded to perform three-dimensional spatial modeling to generate a facility map base model; the facility environment fused dataset is matched and located to the facility map base model for dynamic updating to generate a facility environment three-dimensional spatial model.

[0035] Specifically, the facility maintenance robot, equipped with a sensor array, collects a multi-dimensional environmental dataset of the facility in real time. This dataset covers the facility to be inspected and its surrounding environment, consisting of multi-source heterogeneous raw sensor data, including LiDAR point clouds, depth camera environmental data, IMU pose data, and visible light images. It also includes facility map data, which is the prior static basic spatial information of the facility to be inspected, including the facility's fixed building structure, standard equipment layout, preset inspection points, and safety passage boundaries. Following a four-step logical sequence, a dynamic 3D spatial model is created. First, the multi-dimensional environmental dataset is standardized according to its data acquisition source. This standardization refers to a unified preprocessing operation categorized by acquisition source for heterogeneous data from different types of sensors due to varying acquisition principles and output rules. This addresses the incompatibility issues between different sensor data formats. The core issues of inconsistent sampling frequencies and the presence of acquisition noise and invalid values ​​in the raw data are addressed by performing noise filtering, format standardization, dimension normalization, and missing timestamp completion on data from different acquisition sources such as LiDAR, depth cameras, IMUs, and visible light cameras. For LiDAR point clouds, statistical filtering is used to remove outliers, with filter parameters set to 50 neighboring points and a standard deviation factor of 2. Depth camera images are denoised using a 3×3 Gaussian kernel, while lens distortion correction using the Zhang Zhengyou calibration method is performed. For IMU data, extended Kalman filtering is used to remove motion jitter noise, and linear interpolation is used to complete missing timestamp data. All data is unified to the UTM global coordinate system, with the dimension standardized to meters. Finally, all data is unified to the same format, spatial dimension, and time reference, resulting in a standardized multidimensional environmental dataset for facilities.

[0036] Subsequently, the standard facility multidimensional environment dataset is spatiotemporally aligned and fused. This process involves accurately matching the standardized multi-source data in both time and space dimensions, followed by complementary feature integration. Specifically, the extrinsic parameter transformation matrix between the LiDAR, depth camera, and IMU is solved through hand-eye calibration, unifying all sensor data to the robot's base coordinate system. Using the LiDAR's 10Hz timestamp as a benchmark, linear interpolation is used to align the IMU, depth camera, and visible light camera data within the same time window (100ms), achieving time synchronization. Feature-level fusion is then performed on the spatiotemporally aligned data, matching and fusing the 3D geometric information of the LiDAR point cloud with the texture information of the image to obtain the facility environment fused dataset. Spatiotemporal alignment eliminates the spatiotemporal misalignment of data caused by differences in installation position and sampling frequency between different sensors, while spatial alignment is achieved through... By using hand-eye calibration to solve the extrinsic parameter transformation matrix between each sensor, all data collected by the sensors are uniformly mapped to the same three-dimensional world coordinate system, solving the problem of spatial coordinate deviation between different sensors. In the time alignment dimension, the sampling timestamp of the LiDAR is used as the reference anchor point. Through the time synchronization strategy, all sensor data within the same time window are aligned to ensure that each set of output data corresponds to the perception result of the robot at the same moment and in the same spatial pose, avoiding the deviation of data spatiotemporal misalignment. In the data fusion dimension, the multi-source data that has completed spatiotemporal alignment is integrated with feature-level complementarity. The high-precision three-dimensional geometric information of the LiDAR point cloud, the visual texture details of the image, and the pose dynamic information of the IMU are fused and complemented. Redundant and duplicate data are eliminated, and the core spatial and visual features of the scene are enhanced, resulting in a facility environment fusion dataset that has both spatial geometric accuracy and visual detail integrity, and is completely unified in spatiotemporal dimensions.

[0037] Next, the facility map data is loaded for 3D spatial modeling. Based on prior static facility information, the overall spatial structure of the facility, the layout of fixed equipment, and the inspection benchmark framework are reconstructed, generating a facility map base model with a unified global coordinate system. This provides a stable spatial reference benchmark for subsequent real-time environmental data matching and updating. The fused facility environment dataset is matched and located to the facility map base model for dynamic updating. Here, dynamic updating refers to the incremental iterative optimization of the model after accurately anchoring the real-time collected and fused environmental data to the static base model. Specifically, the utility tunnel BIM model is loaded to generate the facility map base model. The Cartographer laser SLAM algorithm is used to achieve real-time robot localization. The ICP algorithm is used to match the fused dataset to the base model. The corresponding point distance threshold of the ICP algorithm is set to 0.05m, the maximum number of iterations is 50, and the mean square error convergence threshold is [not specified]. 1e-6; The model is dynamically updated every 500ms to identify dynamic changes in the scene and correct the model, generating a 3D spatial model of the facility environment. Through simultaneous localization and mapping (SLAM) technology and ICP iterative nearest point registration algorithm, the facility environment fusion dataset is accurately matched and positioned in the global coordinate system of the facility map base model, solving the positioning drift problem during robot movement and ensuring that the real-time collected environmental data completely corresponds to the spatial position of the prior base model. Then, the real-time fusion data is differentially compared with the facility map base model at a fixed short period to identify the environmental elements that have changed in the scene, including temporary obstacles, positional offsets of facilities and equipment, structural deformations in the scene, and newly added environmental objects. These changed elements are supplemented and corrected into the base model in real time, and finally a 3D spatial model of the facility environment that combines prior static reference information of the facility with real-time dynamic environmental information of the inspection process is generated.

[0038] Further, in step S200, obstacles and abnormal areas are identified in the three-dimensional spatial model of the facility environment to obtain environmental obstacle information and abnormal facility area information, including:

[0039] Clustering and segmentation, and obstacle identification are performed on the point cloud data in the three-dimensional spatial model of the facility environment to obtain environmental obstacle information; a standard feature database of facility objects is constructed according to the facility safety application standards; abnormal area identification and feature extraction are performed on the three-dimensional spatial model of the facility environment according to the standard feature database of facility objects to obtain abnormal facility area information.

[0040] Specifically, clustering and segmentation are performed on the point cloud data in the 3D spatial model of the facility environment. Clustering and segmentation refers to the point cloud preprocessing operation that groups and classifies the discretely distributed global point cloud data in 3D space based on spatial density characteristics. Its function is to split the originally mixed scene point cloud data into multiple point cloud clusters corresponding to different independent spatial entities in the scene, realizing the spatial isolation and individual division of different objects in the scene, and providing independent and accurately analyzable minimum units for subsequent obstacle recognition. Specifically, the original point cloud data in the model is first downsampled to simplify redundant point clouds and reduce computational load while retaining the core geometric contours. Then, the DBSCAN density clustering algorithm with noise is used to divide spatially adjacent point clouds into mutually independent point cloud clusters according to preset spatial distance thresholds and density thresholds, thus completing the clustering and segmentation processing of the global point cloud.

[0041] Subsequently, obstacle recognition is carried out based on the point cloud clusters obtained from clustering and segmentation. Here, obstacle recognition refers to the core operation of determining the attributes, distinguishing the types, and marking the boundaries of each group of point cloud clusters after clustering and segmentation, in combination with the inspection robot's own passage parameters and facility inspection safety operation specifications. The goal is to distinguish between fixed facility structures in the scene and obstacles that may affect the robot's passage safety. Specifically, obstacle judgment criteria are pre-set, including the object's spatial size, position boundary, spatial overlap with the preset inspection channel, and whether it belongs to the inherent static structure of the facility. Each group of point cloud clusters is judged one by one to identify non-fixed obstacle entities such as temporary stacked materials, temporary safety fences, relocated maintenance equipment, and fallen parts. Finally, environmental obstacle information is obtained. Here, environmental obstacle information refers to structured spatial constraint data including the obstacle's precise spatial coordinates, three-dimensional contour size, type attributes, the range of influence on the robot's passage, and the boundary of the impassable area.

[0042] While conducting obstacle identification, the facility safety application standards for abnormal area identification are simultaneously promoted. These standards include official or enterprise-level standardized specifications such as equipment operation and maintenance safety regulations, facility structural safety standards, and normal operating status requirements for the industry to which the facility to be inspected belongs. This integrates benchmark information such as the standard appearance characteristics, geometric dimensions, fixed spatial positions, and characteristic parameters under normal operating conditions of all equipment and structural components within the facility to be inspected, constructing a standard feature database of facility objects. This database serves as the sole reference benchmark for subsequent anomaly determination. Subsequently, based on this database, abnormal area identification and feature extraction are performed on the 3D spatial model of the facility environment. Abnormal area identification and feature extraction refer to the detailed comparison of the actual features of the facility area to be inspected in the 3D spatial model with the corresponding standard features in the database. This identifies facility areas with significant deviations from the standard features and extracts the corresponding deviation features. The core operation aims to pinpoint abnormal areas that do not meet safety operation standards from the entire facility scenario, clarifying the operation and maintenance risk points of the facility itself. Specifically, based on... A standard feature database for facilities and objects is constructed, containing baseline information such as standard geometric dimensions, appearance features, and installation locations of facilities such as pipe rack supports, high-voltage switchgear, and cables. The standard features adopt FPFH point cloud features and ORB image features. The features of the areas to be detected in the model are compared with the standard features in the database. The feature similarity threshold is set to 0.75. Areas with similarity below the threshold are identified as abnormal areas. Information such as the location, deviation type, and deviation degree of abnormal areas are extracted. The point cloud geometric features and visual texture features of the areas to be detected in the model are matched and aligned one by one with the standard features in the database to identify abnormal states that do not meet the standards, such as structural deformation, missing parts, positional offset, surface damage, and foreign object attachment. The spatial range of the abnormality is accurately delineated. At the same time, core features such as deviation type, deviation degree, unique identifier of the facility or equipment, and precise spatial location of the abnormal area are extracted. Finally, information on abnormal facility areas is obtained. At the same time, newly added obstacles caused by the detachment of facility parts and structural collapse can be identified to supplement and improve the environmental obstacle information.

[0043] Furthermore, the point cloud data in the three-dimensional spatial model of the facility environment is clustered and segmented, and obstacles are identified to obtain environmental obstacle information, including:

[0044] The point cloud data in the three-dimensional spatial model of the facility environment is downsampled to generate three-dimensional point cloud data of the facility environment; DBSCAN is used to perform clustering and segmentation processing on the three-dimensional point cloud data of the facility environment to obtain the cluster segmentation results of the facility environment point cloud data; obstacle judgment criteria are defined to identify obstacles in the cluster segmentation results of the facility environment point cloud data to obtain environmental obstacle information.

[0045] Specifically, the original point cloud data in the three-dimensional spatial model of the facility environment is downsampled, and the voxel size is set to 5cm×5cm×5cm; the DBSCAN algorithm is used for clustering and segmentation, the neighborhood distance threshold eps is set to 0.1m, and the minimum number of neighborhood points MinPts of the core point is set to 20; the obstacle judgment criteria are set as follows: not belonging to the inherent static structure in the pipe gallery BIM model; spatial overlap with the preset inspection channel ≥30%; three-dimensional contour size ≥10cm×10cm×10cm; point cloud clusters that meet the above criteria are judged as obstacles, and the spatial coordinates, contour size, and boundary of the impassable area of ​​the obstacle are output. Downsampling is a preprocessing operation for high-density, highly redundant raw point cloud data generated by multiple sensors in a 3D spatial model. It involves dimensionality reduction, noise removal, and redundant data cleaning through standardized methods such as voxel grid filtering, while preserving the core geometric contours, key spatial boundary features, and core spatial information required for obstacle identification in the inspection scene. Specifically, based on the accuracy requirements of the inspection scene, the voxel grid size is set. Without losing obstacle contour features, the original millimeter-level high-density point cloud is reduced to a centimeter-level standardized point cloud. Simultaneously, discrete outlier noise points caused by environmental reflections and sensor jitter are removed, thus generating 3D point cloud data for the facility environment. Here, 3D point cloud data for the facility environment refers to a standardized 3D point cloud dataset that, after downsampling, has significantly reduced data volume, completely removed noise, fully preserved spatial features, and a completely unified format and coordinate system.

[0046] Subsequently, DBSCAN was used to perform clustering and segmentation processing on the 3D point cloud data of the facility environment. DBSCAN stands for Density-Based Spatial Clustering Algorithm with Noise. It is an unsupervised clustering algorithm that does not require a pre-set number of clusters, can effectively identify spatial clusters of arbitrary shapes, and can isolate discrete noise data. It can perfectly adapt to the recognition needs of obstacles with variable shapes and random noise interference in the environment. The clustering and segmentation processing refers to grouping and classifying the 3D point cloud data of the facility environment and dividing its boundaries based on the algorithm according to a preset neighborhood distance threshold and core point density threshold. Specifically... In this process, spatially adjacent point clouds with density meeting the threshold requirements are divided into a group of independent point cloud clusters. Discrete point clouds that do not meet the density requirements are marked as noise points and removed. Ultimately, the originally mixed global scene point cloud is split into multiple independent point cloud clusters corresponding to different independent spatial entities in the inspection scene, such as fixed walls, equipment cabinets, temporary materials, and fallen parts. This yields the facility environment point cloud data cluster segmentation results, realizing spatial isolation and individual division of different entities in the scene, and providing independent and accurately analyzable minimum units for subsequent obstacle attribute determination.

[0047] Finally, based on the inspection robot's body size, passage parameters, and facility inspection safety operation specifications, obstacle judgment criteria are predefined, including object spatial dimensions, positional boundaries, spatial overlap with the preset inspection channel, and whether it belongs to the inherent static structure of the facility. Each independent point cloud cluster in the obtained facility environment point cloud data cluster segmentation result is individually attributed and boundary-calibrated to distinguish between fixed facility structures and obstacles that may affect the robot's passage safety. Non-fixed obstacle entities such as temporarily piled maintenance materials, temporary safety fences, moved maintenance equipment, and fallen facility parts are accurately identified. Finally, structured environmental obstacle information including the obstacle's precise spatial coordinates, three-dimensional contour dimensions, type attributes, range of influence on robot passage, and boundaries of impassable areas is obtained.

[0048] Further, in step S300, fault diagnosis is performed on the abnormal facility area information to obtain facility fault diagnosis results, including:

[0049] Based on historical facility failure cases, fault diagnosis reasoning is performed to construct a facility failure knowledge base; the facility failure knowledge base is used to diagnose and locate the abnormal facility area information to obtain the facility failure mode, facility failure degree, and facility failure location; the facility failure mode, facility failure degree, and facility failure location are correlated and integrated to obtain the facility failure diagnosis result.

[0050] Specifically, based on historical facility failure cases—that is, the complete set of failure operation and maintenance records accumulated during the entire life cycle operation and maintenance of facilities to be inspected and similar standardized facilities—covering full-dimensional experience data such as the scenario characteristics of the failure, abnormal appearance, corresponding equipment type, failure mechanism, hazard impact level, supporting disposal plan, and historical maintenance records, fault diagnosis reasoning is carried out. Through data mining, rule induction, and logical association, the mapping relationship between abnormal appearance characteristics and the essential attributes of the failure is sorted out and extracted. A standard facility failure knowledge base of abnormal characteristics-failure type-hazard level-location logic is constructed. That is, a structured and callable intelligent diagnostic database that integrates full set of failure feature tags, standardized failure reasoning rules, failure type classification standards, component-level failure location logic, and failure hazard assessment system.

[0051] Subsequently, the previously constructed facility fault knowledge base is invoked to perform fault diagnosis and localization on the input abnormal facility area information. This involves using feature matching, rule-based reasoning, and logical verification to perform multi-dimensional, hierarchical matching and intelligent reasoning between the abnormal features in the abnormal facility area information and the fault feature tags and reasoning rules in the knowledge base. Simultaneously, this is combined with the unified global coordinate system of the facility environment 3D spatial model generated in the previous stage to achieve hierarchical and precise fault localization from the area level to the equipment level and then to the component level. Finally, three diagnostic elements are output: the facility fault mode, which clarifies the specific type, mechanism, and essential attributes of the fault, such as structural cracking or component detachment. This system accurately identifies the nature of anomalies such as equipment leakage, aging wiring, and cabinet deformation, addressing the question of what the fault is, the severity of the facility fault (based on facility safety application standards, the scope of the fault's impact on normal facility operation, and the risk of fault spread), clarifies the fault's impact boundaries and emergency response priorities, and quantifies the severity of the fault's hazard. It also identifies the location of the facility fault (the precise spatial coordinates of the fault in the facility's 3D spatial model, the unique identifier of the associated facility equipment / structural component, and the location information at the specific component level), achieving precise spatial and component-level anchoring of the fault and addressing the question of where the fault is specifically located.

[0052] Finally, the three core diagnostic elements obtained above—facility failure mode, facility failure severity, and facility failure location—are linked and integrated. That is, through a unique fault ID code, the type attribute, severity level, and precise location of the same fault are bound and structured one by one, ensuring that each fault entry has complete, unique, and traceable multi-dimensional information, avoiding misalignment and omission of fault information, and ultimately forming a standardized, structured facility fault diagnosis result that can directly support subsequent operation and maintenance decisions.

[0053] Further, in step S400, global path planning is performed based on the environmental obstacle information and the facility fault diagnosis results to determine the robot's global inspection path, including:

[0054] Based on the facility fault diagnosis results, inspection priority analysis is performed to determine the facility fault inspection priority sequence; using the environmental obstacle information and the facility fault inspection priority sequence as constraint information, global path planning is performed on the three-dimensional spatial model of the facility environment to obtain multiple facility environment inspection paths; the multiple facility environment inspection paths are optimized to determine the robot's global inspection path.

[0055] Specifically, the first step is to conduct an inspection priority analysis based on the facility fault diagnosis results. This inspection priority analysis involves constructing standardized quantitative scoring rules by combining core dimensions identified in the facility fault diagnosis results, such as fault severity, fault mode, impact on facility safe operation, fault propagation risk, and emergency response requirements. This is the core operation of hierarchically ranking all diagnosed fault inspection points to address the inspection sequence planning problem in multi-fault scenarios, ensuring that high-risk, high-urgency faults are prioritized for inspection. This analysis process ultimately determines the priority sequence, which is referred to as the facility fault... The obstacle inspection priority sequence refers to a list of fault inspection points ranked from highest to lowest according to quantitative scores. The list clearly defines the inspection order, precise spatial coordinates, and verification requirements for each fault point. High-priority fault points must be visited and verified first, serving as the core objective for subsequent route planning. The environmental obstacle information and the facility fault inspection priority sequence are then used as constraint information. These constraints refer to the hard boundary conditions and core objective requirements that must be strictly adhered to during the global route planning process. They are divided into two insurmountable categories: one is the hard spatial constraint ensuring passage safety, namely, environmental obstacle information... The information clearly defines impassable areas, obstacle 3D contour boundaries, and safe obstacle avoidance distances. The robot's planned path must absolutely not enter these areas to fundamentally avoid collision risks. Another type of constraint is the core objective constraint to ensure operational value, namely, the requirement for complete coverage of all fault points in the priority sequence and the order of inspections. The planned path must completely cover all fault points to be inspected and strictly follow the priority sequence. Based on this, using the aforementioned 3D spatial model of the facility environment as the global spatial basis for path planning, a global path planning algorithm adapted to indoor and outdoor facility inspection scenarios is used to solve the global path problem, ultimately obtaining... There are multiple facility environment inspection paths. Here, multiple facility environment inspection paths refer to all candidate inspection paths that meet the aforementioned constraint information requirements. Each candidate path achieves complete coverage of all fault points to be inspected and effective avoidance of all obstacles throughout the entire process. At the same time, it adapts to the kinematic constraints of the facility maintenance robot, including the robot's own driving boundary conditions such as minimum turning radius, maximum climbing angle, and driving speed limit. The differences only exist in the dimensions of total path length, total inspection time, arrival time of high priority points, robot driving energy consumption, and repeated path coverage, providing sufficient compliant candidate samples for subsequent path optimization.

[0056] Next, the multiple facility environment inspection paths are optimized. This optimization refers to constructing a multi-objective quantitative evaluation function based on the core requirements of facility maintenance and inspection. All candidate inspection paths are quantitatively calculated, comprehensively scored, and ranked one by one, ultimately selecting the inspection path with the best overall performance. Specifically, this involves combining core optimization objectives such as the shortest total inspection path length, the least total inspection operation time, the earliest arrival time of high-priority fault points, the lowest robot driving energy consumption, and the lowest inspection point overlap rate. Different weights are assigned to different objectives according to the actual operation and maintenance scenario requirements. For example, in high-priority operation and maintenance scenarios such as core power facilities and integrated utility tunnels, the highest weight is assigned to the arrival time of high-priority fault points. Then, based on this multi-objective evaluation function, the comprehensive score of each candidate path is calculated, and the candidate path with the highest comprehensive score is finally determined as the robot's global inspection path.

[0057] Furthermore, the multiple facility environment inspection paths are optimized to determine the robot's global inspection path, including:

[0058] Based on the facility maintenance and inspection requirements, a multi-objective function for facility inspection is constructed; based on the multi-objective function for facility inspection, the multiple facility environment inspection paths are evaluated and optimized to determine the global inspection path for the robot.

[0059] Specifically, firstly, a multi-objective function for facility inspection is constructed based on the needs of facility maintenance and inspection. This construction refers to transforming multiple interrelated, and even mutually constraining, inspection optimization dimensions into a standardized function that is quantifiable and uniformly assessable through mathematical modeling. This provides a unified, objective, and reproducible evaluation benchmark for all candidate inspection paths, avoiding subjective arbitrariness in path selection. This function is constructed using an industry-standard linear weighted summation form, and can be represented by the following weighted multi-objective function:

[0060] First, four core optimization sub-objectives were identified, namely, minimizing the arrival time of high-priority fault locations. The minimum total time required for inspection operations is The shortest total path length is The lowest path repetition rate is The corresponding general expression for the constructed multi-objective function is:

[0061] ;

[0062] in , , , To assign weight coefficients to the corresponding sub-objectives, the normalization constraint condition that the sum of all weight coefficients is 1 must be satisfied. Considering the requirement to prioritize high-risk fault handling, the weight allocation can be set as follows: =0.4、 =0.25、 =0.2、 =0.15, to These are the quantized values ​​of each sub-objective after min-max normalization, thereby eliminating the dimensional differences between different sub-objectives. Normalized values ​​of arrival times for high-priority fault locations The normalized value of the total inspection time The normalized value of the total path length This is the normalized value of the path repetition rate. The smaller the value of all sub-targets, the better the performance of the corresponding dimension. The smaller the final function output value F, the better the overall operation and maintenance performance of the candidate path.

[0063] After completing the construction of the multi-objective function, the multiple facility environment inspection paths are evaluated and optimized based on the facility inspection multi-objective function. Evaluation and optimization refers to using the aforementioned constructed facility inspection multi-objective function as the sole evaluation benchmark to quantitatively calculate, verify compliance, and sort each compliant candidate facility environment inspection path. Specifically, the core parameters of each candidate path, such as the total path length, arrival time of each fault point, total travel time, and percentage of duplicate paths, are first extracted and substituted into the multi-objective function to complete the normalization calculation, obtaining the comprehensive score value F corresponding to each path. Then, it is simultaneously verified whether the path meets the supplementary constraints such as robot range and maximum continuous travel time. After eliminating paths that do not meet the constraints, the paths are sorted from smallest to largest based on their comprehensive scores. Finally, the candidate path with the lowest comprehensive score, i.e. the one with the best overall performance, is selected to determine the robot's global inspection path. This global inspection path is a standardized inspection execution path that is finally determined after multi-objective function evaluation and optimization. It is adapted to the entire scene of the facility to be inspected, meets all spatial and operational constraints, and has the best overall operational performance. This path includes the robot's full-domain spatial coordinate sequence, the arrival time of each inspection point, the stopping and verification requirements, the driving speed planning, and other full execution parameters. It is the sole instruction basis for the robot's underlying motion control unit to perform the global inspection operation of the facility.

[0064] This step standardizes and objectifies path evaluation, flexibly adapting to inspection scenarios with different facility types and maintenance priorities. At the same time, through a quantitative evaluation and optimization process, it ensures that the final output global inspection path achieves an optimal balance of maintenance priority, inspection efficiency, and operational energy consumption while meeting all hard constraints. This not only ensures priority inspection of high-risk fault points, maximizing the facility safety guarantee value of inspection operations, but also reduces the robot's operational energy consumption and time through path optimization, thereby improving the overall efficiency of the inspection operation.

[0065] In summary, the robot inspection method for facility maintenance provided in this application has the following technical effects:

[0066] By equipping a facility maintenance robot with a sensor array, multi-dimensional environmental datasets are collected in real time. These datasets are then combined with map data to generate a 3D spatial model of the environment. Obstacles and abnormal areas are identified within the 3D model, yielding information on environmental obstacles and abnormal facility areas. Fault diagnosis is performed on the abnormal areas, providing diagnostic results. Based on these results, a global inspection path is determined, and a global facility inspection is conducted according to this path. This approach solves the technical problems of existing inspection robots, where inspection operations cannot match actual maintenance needs, and it is difficult to simultaneously ensure traffic safety, fault identification accuracy, and operational efficiency. It achieves the technical effect of improving fault identification accuracy and the safety and efficiency of inspection operations, effectively adapting to actual maintenance requirements.

[0067] Example 2, based on the same inventive concept as the robot inspection method for facility maintenance in the foregoing examples, such as... Figure 2 As shown in the figure, this application provides a robotic inspection system for facility maintenance, the system comprising:

[0068] The data acquisition module 11 is used to mount a sensor group on the facility maintenance robot, and to collect a multi-dimensional environmental dataset of the facility in real time through the sensor group. It combines the facility map data with the multi-dimensional environmental dataset to perform dynamic three-dimensional spatial modeling and generate a three-dimensional spatial model of the facility environment. The anomaly identification module 12 is used to identify obstacles and abnormal areas in the three-dimensional spatial model of the facility environment to obtain environmental obstacle information and abnormal facility area information. The fault diagnosis module 13 is used to diagnose the abnormal facility area information to obtain facility fault diagnosis results. The inspection path module 14 is used to perform global path planning based on the environmental obstacle information and the facility fault diagnosis results to determine the robot's global inspection path and control the facility maintenance robot to perform global facility inspection according to the robot's global inspection path.

[0069] Furthermore, the data acquisition module 11 is also used for:

[0070] The facility multidimensional environment dataset is standardized according to the data collection source to obtain a standard facility multidimensional environment dataset; the standard facility multidimensional environment dataset is spatiotemporally aligned and fused to obtain a facility environment fused dataset; the facility map data is loaded to perform three-dimensional spatial modeling to generate a facility map base model; the facility environment fused dataset is matched and located to the facility map base model for dynamic updating to generate a facility environment three-dimensional spatial model.

[0071] Furthermore, the anomaly identification module 12 is also used for:

[0072] Clustering and segmentation, and obstacle identification are performed on the point cloud data in the three-dimensional spatial model of the facility environment to obtain environmental obstacle information; a standard feature database of facility objects is constructed according to the facility safety application standards; abnormal area identification and feature extraction are performed on the three-dimensional spatial model of the facility environment according to the standard feature database of facility objects to obtain abnormal facility area information.

[0073] Furthermore, the anomaly identification module 12 is also used for:

[0074] The point cloud data in the three-dimensional spatial model of the facility environment is downsampled to generate three-dimensional point cloud data of the facility environment; DBSCAN is used to perform clustering and segmentation processing on the three-dimensional point cloud data of the facility environment to obtain the cluster segmentation results of the facility environment point cloud data; obstacle judgment criteria are defined to identify obstacles in the cluster segmentation results of the facility environment point cloud data to obtain environmental obstacle information.

[0075] Furthermore, the fault diagnosis module 13 is also used for:

[0076] Based on historical facility failure cases, fault diagnosis reasoning is performed to construct a facility failure knowledge base; the facility failure knowledge base is used to diagnose and locate the abnormal facility area information to obtain the facility failure mode, facility failure degree, and facility failure location; the facility failure mode, facility failure degree, and facility failure location are correlated and integrated to obtain the facility failure diagnosis result.

[0077] Furthermore, the inspection path module 14 is also used for:

[0078] Based on the facility fault diagnosis results, inspection priority analysis is performed to determine the facility fault inspection priority sequence; using the environmental obstacle information and the facility fault inspection priority sequence as constraint information, global path planning is performed on the three-dimensional spatial model of the facility environment to obtain multiple facility environment inspection paths; the multiple facility environment inspection paths are optimized to determine the robot's global inspection path.

[0079] Furthermore, the inspection path module 14 is also used for:

[0080] Based on the facility maintenance and inspection requirements, a multi-objective function for facility inspection is constructed; based on the multi-objective function for facility inspection, the multiple facility environment inspection paths are evaluated and optimized to determine the global inspection path for the robot.

[0081] In Embodiment 3, based on the same inventive concept as the robot inspection method for facility maintenance in the foregoing embodiments, this application also provides a computer-readable storage medium storing a computer program that, when executed, implements the steps of any one of the methods described in Embodiment 1.

[0082] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0083] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A robotic inspection method for facility maintenance, characterized in that, The method includes: A sensor group is mounted on the facility maintenance robot. The sensor group collects multi-dimensional environmental datasets of the facility in real time. The data is then combined with the facility map data and the multi-dimensional environmental datasets of the facility to perform dynamic three-dimensional spatial modeling and generate a three-dimensional spatial model of the facility environment. Obstacles and abnormal areas are identified in the three-dimensional spatial model of the facility environment to obtain environmental obstacle information and abnormal facility area information; Perform fault diagnosis on the abnormal facility area information to obtain facility fault diagnosis results; Based on the environmental obstacle information and the facility fault diagnosis results, global path planning is performed to determine the robot's global inspection path. The facility maintenance robot is then controlled to perform a global facility inspection according to the robot's global inspection path.

2. The robot inspection method for facility maintenance as described in claim 1, characterized in that, By combining facility map data with the facility multidimensional environment dataset, a three-dimensional spatial dynamic model is generated to produce a three-dimensional spatial model of the facility environment, including: The facility multidimensional environment dataset is standardized according to the data collection source to obtain a standard facility multidimensional environment dataset. The standard facility multidimensional environment dataset is spatiotemporally aligned and fused to obtain a facility environment fused dataset. Load the facility map data to perform three-dimensional spatial modeling and generate a basic model of the facility map; The facility environment fusion dataset is matched and located to the facility map base model for dynamic updating, generating a three-dimensional spatial model of the facility environment.

3. The robot inspection method for facility maintenance as described in claim 1, characterized in that, Obstacle and abnormal area identification is performed on the three-dimensional spatial model of the facility environment to obtain environmental obstacle information and abnormal facility area information, including: Clustering and segmentation, and obstacle identification are performed on the point cloud data in the three-dimensional spatial model of the facility environment to obtain environmental obstacle information; Construct a database of standard features for facility objects based on facility safety application standards; The abnormal area information of the abnormal facility area is obtained by identifying and extracting features from the three-dimensional spatial model of the facility environment according to the standard feature database of the facility object.

4. The robot inspection method for facility maintenance as described in claim 3, characterized in that, Clustering and segmentation, and obstacle identification are performed on the point cloud data in the three-dimensional spatial model of the facility environment to obtain environmental obstacle information, including: The point cloud data in the three-dimensional spatial model of the facility environment is downsampled to generate three-dimensional point cloud data of the facility environment. DBSCAN was used to perform clustering and segmentation processing on the three-dimensional point cloud data of the facility environment to obtain the cluster segmentation results of the facility environment point cloud data. By defining obstacle judgment criteria, obstacle identification is performed on the cluster segmentation results of the facility environment point cloud data to obtain environmental obstacle information.

5. The robotic inspection method for facility maintenance as described in claim 1, characterized in that, Fault diagnosis is performed on the abnormal facility area information to obtain facility fault diagnosis results, including: Based on historical failure cases of facilities, fault diagnosis reasoning is carried out to build a facility failure knowledge base; The facility failure knowledge base is used to diagnose and locate the abnormal facility area information to obtain the facility failure mode, facility failure degree and facility failure location. By associating and integrating the facility failure modes, facility failure severity, and facility failure location, facility failure diagnosis results are obtained.

6. The robot inspection method for facility maintenance as described in claim 1, characterized in that, Based on the environmental obstacle information and the facility fault diagnosis results, global path planning is performed to determine the robot's global inspection path, including: Based on the facility fault diagnosis results, an inspection priority analysis is performed to determine the facility fault inspection priority sequence. Using the environmental obstacle information and the facility fault inspection priority sequence as constraint information, global path planning is performed on the three-dimensional spatial model of the facility environment to obtain multiple facility environment inspection paths. The multiple facility environment inspection paths are optimized to determine the robot's global inspection path.

7. The robot inspection method for facility maintenance as described in claim 6, characterized in that, The optimal inspection path for the robot is determined by selecting the best path from the multiple facility environment inspection paths, including: Based on the needs of facility maintenance and inspection, construct a multi-objective function for facility inspection; The multiple facility environment inspection paths are evaluated and optimized based on the facility inspection multi-objective function to determine the robot's global inspection path.

8. A robotic inspection system for facility maintenance, characterized in that, The system is used to perform the robotic inspection method for facility maintenance as described in any one of claims 1-7, the system comprising: The data acquisition module is used to mount a sensor group on the facility maintenance robot. The sensor group collects a multi-dimensional environmental dataset of the facility in real time. The data is then combined with the facility map data and the multi-dimensional environmental dataset to perform dynamic three-dimensional spatial modeling and generate a three-dimensional spatial model of the facility environment. Anomaly identification module is used to identify obstacles and abnormal areas in the three-dimensional spatial model of the facility environment, and obtain environmental obstacle information and abnormal facility area information; The fault diagnosis module is used to diagnose faults in the abnormal facility area information and obtain facility fault diagnosis results. The inspection path module is used to perform global path planning based on the environmental obstacle information and the facility fault diagnosis results, determine the robot's global inspection path, and control the facility maintenance robot to perform global facility inspection according to the robot's global inspection path.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the robotic inspection method for facility maintenance as described in any one of claims 1-7.

Images