Crane metal structure detection planning method and system and application thereof

Abstract

The invention discloses a crane metal structure detection planning method and system and application of the crane metal structure detection planning method and system. According to an outdoor portal crane, multi-factor information is introduced through a three-dimensional structure model based on the crane, candidate detection points are generated on a metal surface, and risk weights are established; obtaining a detection point set meeting key component coverage through constraint optimization; constructing a flight-wall-climbing joint graph model by combining the attachment / detachment intermediate point, and planning a detection route by integrating the wall-climbing distance and the mode switching cost; then the flying and climbing robot is controlled to collect visual and ultrasonic detection data according to a route, and finally the comprehensive risk of each detection point is calculated through a damage identification model; and according to the risk result, a secondary detection point is adaptively generated in the neighborhood of the high-risk point, and a secondary detection route is re-planned, so that local encryption reinspection and risk updating are realized, and the coverage rate, the quantitative evaluation capability and the operation safety of crane metal structure detection are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of crane detection, in particular to a crane metal structure detection planning method and system and application thereof. BACKGROUND

[0002] Crane, especially portal crane deployed in outdoor port, wharf and construction site, its main beam, end beam, leg and connecting node bear alternating load, wind load and corrosion environment for a long time, which is easy to produce cracks, thinning and pitting and other damages at the intersection of weld, the root of reinforcing plate, track support and other parts. In order to ensure the safety of the structure, the following detection methods are mainly used in the current engineering: First, relying on manual high-altitude operation to cooperate with handheld ultrasonic, magnetic powder and other non-destructive testing equipment to carry out segmented sampling inspection. This method needs to set up scaffolding or use aerial work platform, the working environment is dangerous and the labor intensity is large, the selection of detection points is seriously dependent on the experience of detection personnel, it is difficult to quantitatively plan the detection range from the overall structure, and the detection results are difficult to close loop with the subsequent life assessment model.

[0003] Second, using ordinary unmanned aerial vehicle to carry out visual inspection, through shooting the surface image of main beam, leg and other parts, manually or using image algorithm to identify cracks, rust and other defects. Although this method improves the inspection efficiency and reduces the personnel risk, the unmanned aerial vehicle usually works in free flight state, there is a certain distance and attitude fluctuation between the unmanned aerial vehicle and the metal structure surface, it is difficult to carry out contact detection such as ultrasonic thickness measurement, and the detection path is mainly manually planned or simply scanned according to the rules, without optimizing the selection of detection points combined with the stress characteristics of the equipment and historical fault information.

[0004] Third, some wall-climbing robots or magnetic detection trolleys are used for close-range detection of steel structure surface. Such equipment can carry ultrasonic, eddy current and other sensors for wall detection, but usually needs manual arrangement at high altitude or local reinspection through simple trajectory control, lacks global path planning mechanism combined with flight capability, and lacks the ability to automatically generate secondary detection area and encrypted detection path based on the first round of detection results.

[0005] The flying and climbing robot is an innovative robot integrating flight and climbing (or ground movement) capabilities. It overcomes the limitations of single motion form robots through unique structural design and technology integration to cope with complex and variable environments and task requirements. However, there is still little research on the application of flying and climbing robots in crane detection.

[0006] In summary, the existing crane metal structure detection technology generally has the following problems: (1) The selection of detection points lacks a unified three-dimensional structure model and risk weight modeling, which makes it difficult to balance key component coverage and high-risk area encryption under the constraints of limited endurance and operation time; (2) Flight and wall climbing detection are often implemented separately, and the detection path planning does not uniformly model the wall climbing path, flight transition and mode switching cost, and cannot perform integrated optimization for a flying and climbing robot which has two motion modes; (3) There is a lack of quantitative risk indicators based on multi-source detection data, which cannot adaptively generate a secondary detection point set and a detection route based on the first round of detection results, and it is difficult to form a closed-loop detection planning mechanism of "first round screening-local encryption-risk update".

[0007] Therefore, it is necessary to provide a metal structure detection planning method and system for gantry cranes and other large equipment, which realizes the organic integration of detection point selection, joint graph model route planning, multi-source data risk assessment and adaptive secondary detection under the constraints of three-dimensional structure model and flying and climbing robot motion ability, so as to improve the systematicness, accuracy and safety of detection. SUMMARY

[0008] Therefore, it is necessary to provide a metal structure detection planning method and system for gantry cranes and other large equipment, which realizes the organic integration of detection point selection, joint graph model route planning, multi-source data risk assessment and adaptive secondary detection under the constraints of three-dimensional structure model and flying and climbing robot motion ability, so as to improve the systematicness, accuracy and safety of detection.

[0009] In order to achieve the above technical purpose, the technical scheme adopted by the present application is: A crane metal structure detection planning method applied to an external detection execution unit, comprising: S1, establishing a three-dimensional digital structure model of a crane, determining candidate detection points according to a preset condition, and then selecting planned detection points from the candidate detection points; S2, taking the planned detection points and a designated position point on the crane which can be attached or detached by the external detection execution unit as nodes, constructing a joint graph model A, the joint graph model A comprising a node set A and an edge set A, the node set A comprising the planned detection points and the position point, the edge set A comprising connection edges A of two-by-two virtual connections between the planned detection points and between the planned detection points and the position point, the connection edges A being assigned with comprehensive cost values A according to a preset rule; and then generating a detection route planning based on the joint graph model A and the comprehensive cost values A; S3, the external detection execution unit performs metal structure detection on the planned detection points on the crane according to the detection route planning, and obtains detection data, wherein the detection data is associated with the position information of the corresponding planned detection point.

[0010] As a possible implementation, further, the present scheme S1 comprises: S11, based on the design model of the crane, the BIM model and / or the three-dimensional scanning data, a three-dimensional digital structure model including the main girder, the end beam, the outrigger, the trolley track and the connecting nodes thereof is established, and then the three-dimensional digital structure model is divided into a plurality of detection area units in a unified structure coordinate system; S12, candidate detection points are generated on the metal structure surface of each detection area unit according to a preset grid step, and the candidate detection points located at the intersection of the weld, the root of the reinforcing plate, the track connecting plate area and the historical failure concentrated area are marked as key candidate points; S13, according to the stress level, the environmental exposure level and the historical failure record information of the detection area to which the candidate detection point belongs, an initial risk weight is assigned to each candidate detection point according to a preset condition, then the planned detection points are selected from the candidate detection points under the constraint premise that the key components of the crane are covered, and a planned detection point set is formed; wherein, when the planned detection points are selected from the candidate detection points, the candidate detection points marked as key candidate points are directly selected as planned detection points, and the remaining candidate detection points are selected as planned detection points when the initial risk weight is greater than a preset weight threshold.

[0011] As a possible implementation, further, in the scheme S3, the detection data includes one or more of the following: Visible light image, used for detecting metal surface cracks, corrosion patches and coating peeling; Infrared image, used for detecting temperature anomalies under the coating or near the weld; Ultrasonic thickness measurement data or guided wave detection data, used for detecting plate thickness thinning and internal defects.

[0012] As a relatively preferred implementation, preferably, the scheme further comprises: S4, inputting the detection data obtained by the external detection execution unit into the trained metal damage recognition model to obtain the damage type, damage level and / or risk score of the crane metal at the position of each planned detection point, then performing comprehensive risk assessment to obtain a comprehensive risk index, and then dividing each planned detection point into a safe point, a suspicious point or a high-risk point according to the comprehensive risk index; S5, selecting all high-risk points and part of suspicious points with a comprehensive risk index in a preset suspicious interval as secondary detection points from the planned detection points, and at the same time, constructing a preset spatial neighborhood on the metal structure surface at each high-risk point, and generating new detection points in the preset spatial neighborhood according to the planned detection point coverage, which are set as supplementary detection points; S6, taking the secondary detection points and the supplementary detection points as re-inspection nodes, constructing a joint graph model B, the joint graph model B comprising a node set B and an edge set B, the node set B comprising the secondary detection points and the supplementary detection points, the edge set B comprising connecting edges B of two-by-two virtual connections between the secondary detection points and between the secondary detection points and the supplementary detection points, the connecting edges B being assigned a comprehensive generation value B according to a preset rule; and then generating a re-inspection route plan based on the joint graph model B, the comprehensive generation value B and the priority of the re-inspection nodes.

[0013] As a preferred selection implementation, preferably, the scheme further comprises: S7, performing metal structure detection on the secondary detection points and the supplementary detection points on the crane according to the re-inspection route plan, to obtain re-inspection data, wherein the re-inspection data is associated with position information of the corresponding re-inspection nodes; S8, performing damage identification and risk re-evaluation on the re-inspection nodes based on the re-inspection data, and combining the evaluation results with the comprehensive risk indicators in S4 to judge and output a crane metal structure detection result.

[0014] Based on the above, the scheme further proposes an application method of a flying and climbing robot in crane metal structure detection, which comprises the crane metal structure detection planning method described above; the external detection execution unit is a flying and climbing robot with a detection component; and the flying and climbing robot has a climbing mode and a flying mode.

[0015] Based on the above, as a preferred selection implementation, preferably, in the scheme S1, the three-dimensional digital structure model is divided into a plurality of detection area units, and each detection area unit is pre-assigned with an importance coefficient.

[0016] As a preferred selection implementation, preferably, the scheme S2 comprises: S21, taking the planned detection points and a position point on the crane designated for the external detection execution unit to attach or detach as nodes, constructing a joint graph model A, the joint graph model A comprising a node set A and an edge set A, the node set A comprising the planned detection points and the position point, and the edge set A comprising connecting edges A of two-by-two virtual connections between the planned detection points and between the planned detection points and the position point, the connecting edges A comprising wall-climbing edges and flying edges, wherein the wall-climbing edges represent node pairs that can be reached by the flying and climbing robot in a wall-climbing mode on the same metal structure surface, and the flying edges represent node pairs that can be transferred between different components in a flying mode; S22, assigning a comprehensive generation value A to each connecting edge A in the edge set A of the joint graph model A, and the calculation of the comprehensive generation value A taking into account at least a wall-climbing cost item, a flying cost item and / or a mode switching cost item; S23, based on the joint graph model A and the comprehensive generation value A, planning a flight transfer sequence at a regional cluster level, and planning a wall-climbing detection sub-path within each regional cluster to obtain a detection route plan covering all planned detection points.

[0017] As a preferred selection embodiment, preferably, in the scheme S6, the connection edge B includes a wall-climbing edge and a flight edge, wherein the wall-climbing edge represents a node pair reachable by the wall-climbing robot on the same metal structure surface through the wall-climbing mode, and the flight edge represents a node pair transferred between different components through the flight mode.

[0018] As a preferred selection embodiment, preferably, in the scheme S7, when the external detection execution unit detects the secondary detection points and the supplementary detection points on the crane according to the re-inspection route plan, the precision parameter configuration of the detection is higher than that in the scheme S3.

[0019] As a preferred selection embodiment, preferably, in the scheme S23, when planning a flight transfer sequence at a regional cluster level, each detection regional cluster is regarded as a regional node, a regional-level graph model is constructed according to a flight cost item between the regional nodes, and a traveling salesman problem solving algorithm is used to determine a regional node access sequence, and then a wall-climbing detection sub-path is planned within each regional cluster to obtain a detection route plan covering all planned detection points.

[0020] As a preferred selection embodiment, preferably, in the scheme S3, the wall-climbing robot with a detection assembly as an external detection execution unit moves to the vicinity of a target position point for attachment on the crane in the flight mode according to the detection route plan, then realizes coarse positioning through visual and structural feature matching, subsequently attaches to the metal structure surface of the crane and switches to the wall-climbing mode, enters the first planned detection point in the detection route plan for detection in the wall-climbing mode, then after the detection of a single planned detection point is completed, obtains detection sub-data, associates the detection sub-data with the position information of the planned detection point, and then sequentially enters the next planned detection point, which moves in the flight mode or the wall-climbing mode according to the connection edge A between the planned detection points; finally, all detection sub-data are collected to form detection data.

[0021] As a preferred selection embodiment, preferably, in the scheme S4, the trained metal damage identification model is an image recognition model based on a deep convolutional neural network and / or a signal recognition model based on time-frequency features. The comprehensive risk indicator is obtained by comprehensive evaluation and calculation according to the risk score, the initial risk weight of the planned detection point, and the importance coefficient of the structure where the planned detection point is located, and includes: ; wherein, This is the number of the planned testing site. For planned testing sites Comprehensive risk indicators For planned testing sites The initial risk weight, The output of the trained metal damage recognition model regarding the planned detection points Risk score, For planned testing sites The importance coefficient corresponding to the structure of the unit in the detection area, where, , All are preset values. , , These are the normalized weighting coefficients.

[0022] As a preferred implementation method, preferably, in scheme S4, based on the comprehensive risk index... The definitions for classifying each planned testing point as a safe point, a suspicious point, or a high-risk point are as follows: ; in, , These are the risk classification thresholds.

[0023] Based on the above, this solution also proposes a crane metal structure inspection planning system, which is applied to the inspection route planning of an external inspection execution unit. The external inspection execution unit is a climbing robot equipped with inspection components. The climbing robot has a crawling mode and a flight mode. The system applies the crane metal structure inspection planning method described above or the application method of the climbing robot in crane metal structure inspection described above, including: The digital modeling unit is used to create a three-dimensional digital structural model of the crane, determine candidate inspection points according to preset conditions, and then select the planned inspection points from them. The route planning unit is used to construct a joint graph model A, using planned detection points and designated locations on the crane that can be attached to or detached by external detection execution units as nodes. The joint graph model A includes a set of nodes A and a set of edges A. The set of nodes A includes the planned detection points and the designated locations, and the set of edges A includes virtual connecting edges A between the planned detection points and between the planned detection points and the designated locations. Each connecting edge A is assigned a comprehensive value A according to a preset rule. Then, a detection route plan is generated based on the joint graph model A and the comprehensive value A. The detection execution unit is used to mobilize the external detection execution unit to perform metal structure detection on the planned detection points on the crane according to the detection route plan, and obtain detection data, wherein the detection data is associated with the location information of the corresponding planned detection points; The data processing unit is configured to input detection data obtained by the external detection execution unit into the trained metal damage identification model to obtain a damage type, a damage level and / or a risk score of the crane metal at a position of each planned detection point, then perform comprehensive risk assessment to obtain a comprehensive risk index, and then divide each planned detection point into a safe point, a suspicious point or a high-risk point according to the comprehensive risk index. The route planning unit is further configured to select all high-risk points and suspicious points with a comprehensive risk index in a preset suspicious interval as secondary detection points from the planned detection points, construct a preset spatial neighborhood on a surface of a metal structure at each high-risk point, supplement new detection points in the preset spatial neighborhood according to a planned detection point coverage, and set the new detection points as supplemented detection points; and construct a joint graph model B by taking the secondary detection points and the supplemented detection points as re-inspection nodes, wherein the joint graph model B includes a node set B and an edge set B, the node set B includes the secondary detection points and the supplemented detection points, the edge set B includes connection edges B between the secondary detection points and between the secondary detection points and the supplemented detection points, and the connection edges B are assigned with a comprehensive generation value B according to a preset rule; and generate a re-inspection route plan based on the joint graph model B, the comprehensive generation value B and priorities of the re-inspection nodes. The detection execution unit is further configured to mobilize the external detection execution unit to perform metal structure detection on the secondary detection points and the supplemented detection points on the crane according to the re-inspection route plan to obtain re-inspection data, wherein the re-inspection data is associated with position information of the corresponding re-inspection nodes. The system further includes a data evaluation unit configured to perform damage identification and risk re-evaluation on the re-inspection nodes based on the re-inspection data, combine the evaluation results with the comprehensive risk index output by the data processing unit to make a judgment, and output a crane metal structure detection result.

[0024] Compared with the prior art, the technical scheme has the beneficial effects that the scheme introduces stress level, environmental exposure, historical failure and other multi-source information on the crane three-dimensional structure model, builds a risk weight for the candidate detection points, models the detection point selection problem as a constrained optimization problem under the constraint of key component coverage, and obtains a planned detection point set by solving. Compared with the detection point arrangement relying on experience or uniform sampling, the scheme can preferentially cover high-risk areas under limited detection resources and ensure that key components are not missed, thereby providing a more reasonable spatial sampling basis for subsequent route planning and risk assessment.

[0025] In the planning of the detection route, the scheme uses the planned detection points, the attached / detached intermediate points and the three-dimensional structural surface information to construct a joint graph model, determines the connectivity between nodes through the wall-climbing accessibility and flight accessibility functions, and abstracts the wall-climbing distance, mode switching and other factors into an edge cost function, and solves the detection route by combining the regional cluster hierarchical planning. Compared with the simple manual path planning or the scheme considering only the flight path, the scheme can globally compromise between the two motion modes of the wall-climbing robot, reduce invalid transitions and frequent mode switching, improve the detection efficiency and reduce the energy consumption and instability risk.

[0026] In addition, the scheme also divides the safe points, suspicious points and high-risk points through the comprehensive risk indicators obtained by the first round of detection, constructs a spatial neighborhood in the neighborhood of the metal surface where the high-risk point is located, selects the points not covered or with insufficient resolution in the first round as new detection points to form a second detection point set, and then re-plans the second detection route on the joint graph model, and adjusts the crawling speed, sampling resolution and detection parameters for the second detection points to realize local encryption and re-inspection. By comparing the risk indicators of the first round and the second detection, the risk is updated and finally classified, which can assist maintenance personnel to construct a detailed maintenance scheme based on the detection results, such as repair priority points, key monitoring points and temporary safe points, so as to build a closed-loop planning and dynamic evaluation mechanism for crane metal structure detection.

[0027] The scheme improves the coverage, quantitative evaluation capability and operation safety of crane metal structure detection. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0029] Figure 1 is one of the implementation structures of the portal crane mentioned in the scheme, which is a prior art product; Figure 2 is one of the implementation flow schematic examples of the crane metal structure detection planning method of the scheme; Figure 3 is the second implementation flow schematic example of the crane metal structure detection planning method of the scheme; Figure 4 is the third implementation flow schematic example of the crane metal structure detection planning method of the scheme; Figure 5 is the unit module connection schematic of the crane metal structure detection planning system of the scheme. DETAILED DESCRIPTION

[0030] The application will be described in further detail below with reference to the drawings and embodiments. It is particularly pointed out that the following embodiments are only for illustrating the application, but not for limiting the scope of the application. Similarly, the following embodiments are only part of the embodiments of the application, not all embodiments, and all other embodiments obtained by those skilled in the art without creative labor are within the scope of the application.

[0031] As shown in Figure 1 , as an example, the application scenario of the embodiment is a metal structure detection scenario of a gantry crane, further combined with Figure 2 , the embodiment is a crane metal structure detection planning method, which is applied to route planning when an external detection execution unit detects a crane metal structure, and includes: S1, a three-dimensional digital structure model of the crane is established, candidate detection points are determined according to preset conditions, and then planned detection points are selected from the candidate detection points; S2, a joint graph model A is constructed with the planned detection points and position points on the crane that can be attached or detached by the external detection execution unit as nodes, the joint graph model A includes a node set A and an edge set A, the node set A includes the planned detection points and the position points, the edge set A includes connection edges A of virtual connections between the planned detection points and between the planned detection points and the position points, and the connection edges A are assigned with comprehensive generation values A according to preset rules; and then a detection route planning is generated based on the joint graph model A and the comprehensive generation values A; S3, the external detection execution unit detects the planned detection points on the crane according to the detection route planning to obtain detection data, wherein the detection data is associated with position information of the corresponding planned detection point.

[0032] Among them, in terms of model construction and detection point determination, as a possible implementation manner, further, the embodiment S1 includes: S11, based on the design model, BIM model and / or three-dimensional scanning data of the crane, a three-dimensional digital structure model including a main beam, an end beam, a leg, a trolley track and a connecting node thereof is established (refer to Figure 1 ), and then the three-dimensional digital structure model is divided into a plurality of detection area units under a unified structure coordinate system; S12, candidate detection points are generated on the metal structure surface of each detection area unit according to a preset grid step, and candidate detection points located at weld intersections, reinforcement plate roots, track connecting plate regions and historical fault concentrated regions are marked as key candidate points; S13, according to the stress level, environmental exposure level, historical failure record information of the detection area to which the candidate detection point belongs, the initial risk weight of each candidate detection point is allocated according to the preset condition, and then the planned detection point is selected from the candidate detection point under the constraint premise that the key components of the crane are covered, and the planned detection point set is formed by collecting; wherein, when the planned detection point is selected from the candidate detection point, the candidate detection point marked as the key candidate point is directly selected as the planned detection point, and the initial risk weight of the remaining candidate detection point is greater than the preset weight threshold, which is selected as the planned detection point.

[0033] The scheme introduces stress level, environmental exposure, historical failure and component importance and other multi-source information on the three-dimensional structure model of the crane, builds a risk weight model of the candidate detection point, and models the detection point selection problem as a constraint optimization problem under the constraint of key component coverage, and obtains the planned detection point set by solving. Compared with the detection point arrangement depending on experience or uniform sampling, the scheme can preferentially cover high-risk areas under limited detection resources and ensure that key components are not missed, providing a more reasonable spatial sampling basis for subsequent route planning and risk assessment.

[0034] As an example, based on the BIM model, a three-dimensional digital structure model of the crane is established, including the main beam, end beam, leg, trolley track and their connecting nodes as key components, and the metal surface is divided into grids to obtain the candidate point set defined as follows: ; Wherein, each candidate detection point is associated with one or more of the following normalized characteristic quantities of risk factors: (1) , which is the normalized stress level of the area where the candidate detection point is located, and the value is 0-1; (2) , which is the environmental exposure level of the area where the candidate detection point is located, which can be analyzed by finite element or experience classification, and the value is 0-1; (3) , which is the historical maintenance record intensity of the area where the candidate detection point is located, which is defined according to the frequency or number of cracks or corrosion in historical maintenance, and the value is 0-1; (4) , which is the component importance of the area where the candidate detection point is located (for example, the importance of the main beam, end beam, leg, trolley track and their connecting nodes is different), and the value is 0-1; The characteristic quantities of (1)-(4) above constitute the initial risk weight Each of the values can be manually pre-set and then manually updated subsequently, or automatically generated and updated by other fitting algorithms. This is a relatively basic means, and will not be described here.

[0035] For the fitting calculation of the risk weight , the definition is as follows: ; Wherein, is the initial risk weight of the candidate detection point , k is the risk factor number, and in the embodiment, it takes the value of 1-4 positive integer; is the normalized value of the candidate detection point in the kth risk factor, is the weight coefficient of the kth risk factor, , and . .

[0036] For the constraint condition that the key components of the crane are covered, it can include the following: Set the candidate point determination function, which is defined as follows: ; When the point on the crane is the candidate detection point , , otherwise 0.

[0037] Let the set of key components of the crane be , and define the matrix : ; When the candidate detection point is on the key component of the crane, , otherwise 0.

[0038] In order to ensure that each key component has at least one detection point, the following constraint is set: ; is an indication parameter of whether the key component is covered by the candidate detection point , is the number of key components.

[0039] For some candidate detection points located at the intersection of welds, the root of the reinforcing plate, the track connecting plate area, and the historical failure concentrated area, they can be marked as key candidate points to facilitate subsequent direct selection as planned detection points.

[0040] Through the above scheme content and constraint conditions, the planned detection points are selected from the candidate detection points, and are collected to form a planned detection point set.

[0041] Based on the above, the embodiment scheme is applied to crane metal structure detection by introducing a flying and climbing robot, which includes the crane metal structure detection planning method described above; the external detection execution unit is a flying and climbing robot with a detection component; the flying and climbing robot has a climbing mode and a flying mode.

[0042] In the scheme S1, the three-dimensional digital structure model is divided into a plurality of detection area units, and each detection area unit is pre-assigned with an importance coefficient.

[0043] Based on the above, as a preferred selection implementation, preferably, the scheme S2 includes: S21, taking the planned detection points and the position points on the crane that can be attached or detached by the external detection execution unit as nodes, a joint graph model A is constructed, the joint graph model A includes a node set A and an edge set A, the node set A includes the planned detection points and the position points, and the edge set A includes connecting edges A of virtual connections between two nodes, the connecting edges A include wall climbing edges and flight edges, wherein the wall climbing edges represent node pairs that can be reached by the flying and climbing robot on the same metal structure surface through the wall climbing mode, and the flight edges represent node pairs that are transferred between different components through the flight mode; The node set A of the joint graph model A can be defined as follows: ; Wherein, is the starting point of the flying and climbing robot, is the number of nodes in the node set A, and is defined as , represents a detection point or an intermediate attachment or detachment position point on the crane, if the node , can be reached by climbing or flying, a connecting edge A connecting the nodes , is established in the graph of the joint graph model A, and the connecting edge A can be specifically defined as a wall climbing edge or a flight edge.

[0044] In this case, the indicator for the climbing mode moving through the wall climbing edge can be defined as , and the indicator for the flight mode moving through the flight edge can be defined as .

[0045] The working mode of the flying and climbing robot is defined by binary method, and Meanwhile, .

[0046] S22, assign a comprehensive cost value A to each connection edge A in the edge set A of the joint graph model A, and the calculation of the comprehensive cost value A considers at least a wall-climbing cost term, a flight cost term and / or a mode switching cost term; The function definition of the comprehensive cost value A can be as follows: ; Wherein, is the comprehensive cost value corresponding to the connection edge A from the node to , is the wall-climbing cost when the node to adopts the wall-climbing mode, is the flight cost when the node to adopts the flight mode, is the cost term of whether mode switching is needed before and after the connection edge A from the node to corresponds to the path, i.e. the switching cost term, which can be a preset fixed value.

[0047] For the wall-climbing cost , it can be defined as follows: ; Wherein, is a wall-climbing weight coefficient, is the wall-climbing distance from the node to .

[0048] For the flight cost , it can be defined as follows: ;

[0049] Wherein, , is a flight weight coefficient, is the spatial straight-line distance from the node to , , are the spatial heights of the nodes , respectively.

[0050] S23, based on the joint graph model A and the comprehensive cost value A, plan a flight transfer sequence at the regional cluster level, and plan a wall-climbing detection sub-path within each regional cluster to obtain a detection route plan covering all planned detection points; The scheme S23 aims to visit all planned detection points when detecting route planning, and can regard each detection region cluster as a region node when planning flight transfer sequence at the region cluster level, construct a region-level graph model according to the flight cost items between region nodes, and determine the region node visiting sequence by using a traveling salesman problem solving algorithm. Then, wall-climbing detection sub-paths are planned in each region cluster to obtain detection route planning covering all planned detection points.

[0051] As an example, when planning a route, the following decision variables are set to determine whether a node is planned: ; The edge set A formed by the connection edge A is defined as , which represents all possible movement paths between nodes, and the objective function of the detection route planning can be defined as follows: The constraint condition of the objective function is as follows: ; 1. Start / end point constraint Let ,

[0052] wherein is the movement path of the flying and climbing robot from the starting point to one of the planned detection points, is the movement path of the flying and climbing robot from one of the planned detection points back to the starting point.

[0053] 2. Each planned detection point is visited once Let , .

[0054] The objective function of the detection route planning is solved based on the constraint condition, and the optimal path is obtained, which is used as the movement route plan in the detection route planning.

[0055] ​In the scheme S3, the flying and climbing robot with a detection assembly as an external detection execution unit moves to the vicinity of the target position point for attachment on the crane in flight mode according to the detection route planning, and then can be attached to the crane manually by manual control or can be coarsely positioned by vision and structure feature matching, and then is attached to the metal structure surface of the crane and switched to the wall climbing mode to detect the first planned detection point in the detection route planning, and then after the detection of a single planned detection point is completed, detection sub-data is obtained, which is associated with the position information of the planned detection point, and then the next planned detection point is entered in turn, which moves in flight mode or climbing mode according to the connection edge A between the planned detection points; finally, all detection sub-data is collected to form detection data.

[0056] As a possible implementation, further, in the scheme S3, the detection data includes one or more of the following: A visible light image for detecting metal surface cracks, corrosion patches and coating peeling; An infrared image for detecting temperature anomalies under the coating or near the weld; Ultrasonic thickness measurement data or guided wave detection data for detecting plate thickness thinning and internal defects.

[0057] The scheme uses a joint graph model constructed by using planned detection points, attachment / detachment intermediate points and three-dimensional structure surface information, determines the connection relationship between nodes through climbing accessibility and flight accessibility functions, and abstracts factors such as climbing distance and mode switching into edge cost functions, and solves the detection route by combining regional cluster hierarchical planning. Compared with simple manual path planning or schemes considering only flight paths, the scheme can globally compromise between the two motion modes of the flying and climbing robot, reduce invalid transitions and frequent mode switching, improve detection efficiency and reduce energy consumption and instability risk.

[0058] Due to the large size and structural span of the gantry crane, if each planned detection point is detected in a precise manner, a large amount of time will be consumed, especially for normalized detection, which will have the problem of low efficiency, therefore, rough detection to precise detection helps to improve detection efficiency and optimize work quality. As a preferred selection implementation, preferably, in combination with Figure 3 As shown in the scheme S3, the scheme further includes: S4, inputting the detection data obtained by the external detection execution unit into the trained metal damage recognition model to obtain the damage type, damage level and / or risk score of the crane metal at the position of each planned detection point, then performing comprehensive risk assessment to obtain a comprehensive risk index, and then dividing each planned detection point into a safe point, a suspicious point or a high-risk point according to the comprehensive risk index; S5. Select all high-risk points and some suspicious points whose comprehensive risk indicators are in the preset suspicious range from the planned detection points as secondary detection points. At the same time, construct a preset spatial neighborhood on the surface of the metal structure where each high-risk point is located. In the preset spatial neighborhood, generate new detection points according to the coverage of the planned detection points and set them as supplementary detection points. S6. Using the secondary detection points and supplementary detection points as re-inspection nodes, construct a joint graph model B. The joint graph model B includes a node set B and an edge set B. The node set B includes the secondary detection points and the supplementary detection points. The edge set B includes the connecting edges B of the virtual lines between the secondary detection points and between the secondary detection points and the supplementary detection points. The connecting edges B are assigned a comprehensive value B according to a preset rule. Then, based on the joint graph model B, the comprehensive value B, and the priority of the re-inspection nodes, generate a re-inspection route plan.

[0059] Similar to the above, in this scheme S6, the connecting edge B includes a climbing edge and a flying edge. The climbing edge represents a pair of nodes that the flying robot can reach on the same metal structure surface through the climbing mode, and the flying edge represents a pair of nodes that can transfer between different components through the flying mode.

[0060] In S4 of this scheme, the trained metal damage recognition model is an image recognition model based on a deep convolutional neural network and / or a signal recognition model based on time-frequency features. Specifically, a trained detection neural network can be used as the image recognition model to output the damage type, damage level, and / or risk score of the crane metal at each planned detection point.

[0061] The comprehensive risk index is calculated based on a comprehensive evaluation of the risk score, the initial risk weight of the planned monitoring point, and the importance coefficient of the structure where the planned monitoring point is located. It includes: in, This is the number of the planned testing site. For planned testing sites Comprehensive risk indicators For planned testing sites The initial risk weight, The output of the trained metal damage recognition model regarding the planned detection points Risk score, For planned testing sites The importance coefficient corresponding to the structure of the unit in the detection area, where, , All are preset values. , , These are the normalized weighting coefficients.

[0062] As a preferred selection implementation, preferably, in the scheme S4, according to the comprehensive risk index The definition of dividing each planned detection point into a safety point, a suspicious point or a high-risk point is as follows: ; Among them, 、 are risk classification threshold values respectively.

[0063] For step S05, all high-risk points and suspicious points with part of the comprehensive risk index in the preset suspicious interval (for example, ( , ]) can be selected from the planned detection points as secondary detection points, a preset spatial neighborhood (for example, a spatial neighborhood is obtained by expanding the high-risk point as the center with L as the neighborhood radius) is constructed on the surface of the metal structure at each high-risk point, and then new detection points are generated according to the planned detection point coverage in the preset spatial neighborhood (for example, a detection point is randomly selected, and then the planned detection points defined as safety points are excluded to realize detection deduplication), which are set as supplemented detection points.

[0064] As for the calculation of the comprehensive generation value B mentioned in step S6, the function definition of the comprehensive generation value A can be directly referred to; the difference lies in that the objective function of the re-inspection route planning is different from that of the detection route planning. The edge set B formed by the connection edge B is defined as , which represents all possible movement paths between nodes, and then is matched with of the edge set A, the difference items are removed, the intersection data is retained, and then the objective function of the re-inspection route planning is planned by referring to the objective function of the detection route planning, while increasing the priority weight item, which can be defined as follows: ; Among them, Part of the calculation can be directly calculated by referring to the objective function of the detection route planning, is the priority of the re-inspection node, wherein is the weight coefficient, is the detection point marked as a high-risk point, is the set formed by summarizing the high-risk points, is an indication parameter, which represents whether the high-risk point is planned by the route planning, when planned, , otherwise .

[0065] The constraint condition of the objective function is the same as that of the detection route planning objective function, which is as follows: 1. Start / End point constraint Let ,

[0066] wherein, is the moving path of the flying and climbing robot from the starting point to one of the planned detection points, is the moving path of the flying and climbing robot from one of the planned detection points back to the starting point.

[0067] 2. Each planned detection point is visited once Let , .

[0068] Solving the objective function of the re-inspection route planning based on the constraint conditions can obtain an optimal path, which is used as the moving route plan in the re-inspection route planning.

[0069] On the basis of the above, in combination with Figure 4 as shown, as a preferred selection implementation manner, preferably, the scheme further comprises: S7, the external detection execution unit performs metal structure detection on the secondary detection points and the supplemental detection points on the crane according to the re-inspection route planning, and obtains re-inspection data, wherein the re-inspection data is associated with the position information of the corresponding re-inspection node; S8, based on the re-inspection data, damage identification and risk re-evaluation are performed on the re-inspection node, and then the evaluation result is combined with the comprehensive risk index in S4 for judgment, and the crane metal structure detection result is output.

[0070] In order to improve the quality of detection, as a preferred selection implementation manner, preferably, in the scheme S7, when the external detection execution unit performs metal structure detection on the secondary detection points and the supplemental detection points on the crane according to the re-inspection route planning, the detection precision parameter configuration is higher than that in S3.

[0071] In addition, in the scheme S8, when damage identification and risk re-evaluation are performed on the re-inspection node based on the re-inspection data, the trained metal damage identification model can also be used, which is still a deep convolutional neural network-based image recognition model and / or a time-frequency feature-based signal recognition model. Specifically, the trained detection neural network can be used as an image recognition model to output the damage type, damage level and / or risk score of the crane metal at the position of each re-inspection detection point (re-inspection node).

[0072] As an embodiment of S8, after matching the results of two detections, the consistent detection result is defined as a determination item, and then the crane metal structure detection result is output.

[0073] In the first round of detection, the flying and climbing robot is controlled to collect multi-source data such as visible light images, infrared images and ultrasonic thickness measurement signals at the planned detection points along the detection route. The trained recognition model outputs the damage type, damage level and / or risk score of each detection point. This quantitative indicator facilitates the formation of a risk distribution map on the three-dimensional model of the structure, providing directly usable data support for operation and maintenance decision-making and reliability assessment.

[0074] In addition, the comprehensive risk indicators obtained by the first round of detection are used to divide safe points, suspicious points and high-risk points. A spatial neighborhood is constructed in the neighborhood of the metal surface where the high-risk point is located. Points that are not covered or have insufficient resolution in the first round are selected as new detection points to form a secondary detection point set. Then, the secondary detection route is re-planned on the joint graph model, and the sampling resolution and detection parameters can be adjusted for local encryption and re-inspection at the secondary detection points. By comparing the risk indicators of the first round and the secondary detection, the risk is updated and finally classified, which can assist maintenance personnel in constructing a detailed maintenance plan based on the detection results, such as repair priority points, key monitoring points and temporary safe points, thereby establishing a closed-loop planning and dynamic evaluation mechanism for crane metal structure detection.

[0075] Based on the three-dimensional digital structure model of the crane, the detection points, detection paths and risk results are managed in a unified structure coordinate system. The spatial distribution risk data of multiple detection periods can be directly mapped into the crane digital twin model, providing high-quality observation data for non-probabilistic reliability assessment, fatigue life prediction and maintenance strategy optimization. The crane metal structure state assessment evolves from empirical and discrete to data-driven and full-life-cycle management.

[0076] In combination with Figure 5 Based on the above, the present application also proposes a crane metal structure detection planning system, which is applied to the detection route planning of an external detection execution unit, the external detection execution unit being a flying and climbing robot with a detection component; the flying and climbing robot has a climbing mode and a flying mode; the system applies the crane metal structure detection planning method described above or the application method of the flying and climbing robot in crane metal structure detection described above, which includes: a digital modeling unit for establishing a three-dimensional digital structure model of the crane, determining candidate detection points according to predetermined conditions, and then selecting planned detection points from the candidate detection points; a route planning unit, configured to take the planned detection points and the position points on the crane at which the external detection execution unit can be attached or detached as nodes, construct a joint graph model A, the joint graph model A comprising a node set A and an edge set A, the node set A comprising the planned detection points and the position points, the edge set A comprising connecting edges A between two of the planned detection points, between two of the planned detection points and the position points, the connecting edges A being assigned with a comprehensive generation value A according to a preset rule; and generate a detection route plan based on the joint graph model A and the comprehensive generation value A; a detection execution unit, configured to mobilize the external detection execution unit to perform metal structure detection on the planned detection points on the crane according to the detection route plan, and obtain detection data, wherein the detection data is associated with position information of the corresponding planned detection point; a data processing unit, configured to input the detection data obtained by the external detection execution unit into a trained metal damage identification model, to obtain a damage type, a damage level and / or a risk score of the metal of the crane at the position of each planned detection point, and then perform comprehensive risk assessment to obtain a comprehensive risk index, and divide each planned detection point into a safe point, a suspicious point or a high-risk point according to the comprehensive risk index; The route planning unit is further configured to select all high-risk points and some suspicious points with a comprehensive risk index in a preset suspicious interval as secondary detection points from the planned detection points, construct a preset spatial neighborhood on a metal structure surface at each high-risk point, supplement new detection points in the preset spatial neighborhood according to a planned detection point coverage, and set the new detection points as additional detection points; and construct a joint graph model B taking the secondary detection points and the additional detection points as recheck nodes, the joint graph model B comprising a node set B and an edge set B, the node set B comprising the secondary detection points and the additional detection points, the edge set B comprising connecting edges B between two of the secondary detection points, between two of the secondary detection points and the additional detection points, the connecting edges B being assigned with a comprehensive generation value B according to a preset rule; and generate a recheck route plan based on the joint graph model B, the comprehensive generation value B and a priority of the recheck nodes; The detection execution unit is further configured to mobilize the external detection execution unit to perform metal structure detection on the secondary detection points and the additional detection points on the crane according to the recheck route plan, and obtain recheck data, wherein the recheck data is associated with position information of the corresponding recheck node; The system further comprises a data evaluation unit, configured to perform damage identification and risk reevaluation on the recheck nodes based on the recheck data, combine the evaluation results with the comprehensive risk index output by the data processing unit, and output a crane metal structure detection result.

[0077] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0078] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0079] The above description is only a part of the embodiments of the present invention and does not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made based on the content of the present invention specification and drawings, or direct or indirect application in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A crane metal structure inspection planning method applied to an external inspection execution unit, characterized by, It comprises: S1, establish a three-dimensional digital structure model of the crane, determine candidate detection points under preset conditions, and then select planned detection points from them; S2, taking the planned detection points and the designated position points on the crane that can be attached or detached by the external detection execution unit as nodes, construct a joint graph model A, the joint graph model A includes a node set A and an edge set A, the node set A includes the planned detection points and the position points, the edge set A includes the connection edges A of the virtual connection between each two of the planned detection points and the position points, and the connection edges A are assigned with comprehensive generation values A according to preset rules; then generate a detection route planning based on the joint graph model A and the comprehensive generation values A; S3, the external detection execution unit performs metal structure detection on the planned detection points on the crane according to the detection route planning, and obtains detection data, wherein the detection data is associated with the position information of the corresponding planned detection point.

2. Crane metal structure inspection planning method according to claim 1, characterized in that, S1 comprises: S11, based on the design model, BIM model and / or three-dimensional scanning data of the crane, a three-dimensional digital structure model containing the main girder, end beam, leg, trolley track and their connecting nodes is established, and then the three-dimensional digital structure model is divided into multiple detection area units in a unified structure coordinate system; S12, generate candidate detection points on the metal structure surface of each detection area unit according to a preset grid step, and mark the candidate detection points located at the intersection of welds, the root of reinforcing plates, the track connecting plate area and the historical fault concentrated area as key candidate points; S13, according to the stress level, environmental exposure level and historical fault record information of the detection area to which the candidate detection point belongs, assign an initial risk weight to each candidate detection point according to preset conditions, then select planned detection points from the candidate detection points under the constraint premise that the key components of the crane are covered, and then collect to form a planned detection point set; wherein when selecting planned detection points from candidate detection points, the candidate detection points marked as key candidate points are directly selected as planned detection points, and the remaining candidate detection points are selected as planned detection points when their initial risk weight is greater than a preset weight threshold.

3. The crane metal structure inspection planning method of claim 1, wherein, In S3, the detection data includes one or more of the following: Visible light image, used for detecting metal surface cracks, corrosion patches and coating peeling; Infrared image, used for detecting temperature anomalies under the coating or near the welds; Ultrasonic thickness measurement data or guided wave detection data, used for detecting plate thickness thinning and internal defects.

4. Crane metal structure inspection planning method according to one of claims 1 to 3, characterized in that, It further comprises: S4, input the detection data obtained by the external detection execution unit into the trained metal damage recognition model to obtain the damage type, damage level and / or risk score of the crane metal at the position of each planned detection point, then perform comprehensive risk assessment to obtain a comprehensive risk index, and then divide each planned detection point into a safe point, a suspicious point or a high-risk point according to the comprehensive risk index; S5, select all high-risk points and suspicious points with part of the comprehensive risk indicators in the preset suspicious interval from the planned detection points as secondary detection points, and at the same time, construct a preset spatial neighborhood on the surface of the metal structure at each high-risk point, and supplement new detection points in the preset spatial neighborhood according to the coverage of the planned detection points, and set the supplemented detection points as supplemented detection points; S6, the secondary detection points and the supplemented detection points are used as reinspection nodes, a joint graph model B is constructed, the joint graph model B includes a node set B and an edge set B, the node set B includes the secondary detection points and the supplemented detection points, the edge set B includes connection edges B between the secondary detection points and between the secondary detection points and the supplemented detection points, and the connection edges B are assigned with a comprehensive generation value B according to a preset rule; then, a reinspection route planning is generated based on the joint graph model B, the comprehensive generation value B and the priority of the reinspection nodes.

5. Crane metal structure inspection planning method according to claim 4, characterized in that, It also includes: S7, the external detection execution unit performs metal structure detection on the secondary detection points and the supplemented detection points on the crane according to the reinspection route planning, and obtains reinspection data, wherein the reinspection data is associated with the position information of the corresponding reinspection nodes; S8, damage identification and risk reevaluation are performed on the reinspection nodes based on the reinspection data, and the evaluation results are combined with the comprehensive risk indicators in S4 to judge and output the crane metal structure detection results.

6. The method for the application of the flying climbing robot in the detection of the metal structure of the crane, characterized in that, It includes the crane metal structure detection planning method of claim 4 or 5; the external detection execution unit is a flying and climbing robot with a detection component; the flying and climbing robot has a climbing mode and a flying mode; In S1, the three-dimensional digital structure model is divided into a plurality of detection area units, and each detection area unit is pre-assigned with an importance coefficient; S2 includes: S21, the planned detection points and the position points on the crane designated for the external detection execution unit to attach or detach are used as nodes, a joint graph model A is constructed, the joint graph model A includes a node set A and an edge set A, the node set A includes the planned detection points and the position points, and the edge set A includes connection edges A between the planned detection points and between the planned detection points and the position points, the connection edges A include wall climbing edges and flight edges, wherein the wall climbing edges represent node pairs that can be reached by the flying and climbing robot on the same metal structure surface through the wall climbing mode, and the flight edges represent node pairs that are transferred between different components through the flight mode; S22, a comprehensive generation value A is assigned to each connection edge A in the edge set A of the joint graph model A, and the calculation of the comprehensive generation value A considers at least a wall climbing cost item, a flight cost item and / or a mode switching cost item; S23, based on the joint graph model A and the comprehensive generation value A, a flight transfer sequence is planned at the regional cluster level, and a wall climbing detection sub-path is planned within each regional cluster, to obtain a detection route planning covering all the planned detection points; In S6, the connection edges B include wall climbing edges and flight edges, wherein the wall climbing edges represent node pairs that can be reached by the flying and climbing robot on the same metal structure surface through the wall climbing mode, and the flight edges represent node pairs that are transferred between different components through the flight mode; In S7, when the external detection execution unit performs metal structure detection on the secondary detection points and the additional detection points on the crane according to the re-inspection route planning, the precision parameter configured by the external detection execution unit is higher than the precision in S3.

7. The method for applying the flying and climbing robot in the detection of crane metal structure according to claim 6, characterized in that, In S23, when the flight transfer sequence is planned at the cluster level, each detection region cluster is regarded as a region node, a region-level graph model is constructed according to the flight cost items between the region nodes, and a traveling salesman problem solving algorithm is used to determine the region node access sequence, and then a wall-climbing detection sub-path is planned within each region cluster to obtain a detection route planning covering all the planned detection points.

8. The method for applying the flying and climbing robot in the detection of crane metal structure according to claim 6, characterized in that, In S3, the flying and climbing robot with a detection assembly as the external detection execution unit moves to the vicinity of the target position point for attachment on the crane in the flight mode according to the detection route planning, then realizes coarse positioning through visual and structural feature matching, and then is attached to the surface of the crane metal structure and switches to the wall-climbing mode, enters the first planned detection point in the detection route planning for detection, and then obtains detection sub-data after the detection of a single planned detection point is completed, associates the detection sub-data with the position information of the planned detection point, and then enters the next planned detection point in turn, which moves in the flight mode or the climbing mode according to the connection edge A between the planned detection points. Finally, all the detection sub-data are collected to form detection data.

9. The method for applying the flying and climbing robot in the detection of crane metal structure according to claim 6, characterized in that, In S4, the trained metal damage recognition model is an image recognition model based on a deep convolutional neural network and / or a signal recognition model based on time-frequency features. The comprehensive risk index is calculated by comprehensively evaluating the risk score, the initial risk weight of the planned detection point, and the importance coefficient of the structure where the planned detection point is located, and includes: ； wherein, is the number of the planned detection point, is the comprehensive risk index of the planned detection point , is the initial risk weight of the planned detection point , is the risk score output by the trained metal damage recognition model about the planned detection point , is the importance coefficient corresponding to the structure of the detection area unit where the planned detection point is located, wherein, , are preset values, , , is the normalization weight coefficient; In S4, the comprehensive risk index is used to determine the risk level of each planned inspection point The definitions of safe, suspicious and high-risk points for each planned inspection point are as follows: ； wherein, , are risk classification threshold values, respectively.

10. A crane metal structure detection planning system applied to detection route planning of an external detection execution unit, the external detection execution unit being a flying and climbing robot with a detection assembly; the flying and climbing robot having a climbing mode and a flying mode; the system being applied with the crane metal structure detection planning method of any one of claims 1 to 5 or the application method of the flying and climbing robot in crane metal structure detection of any one of claims 6 to 9, characterized in that, It includes: A digital modeling unit for establishing a three-dimensional digital structure model of the crane, determining candidate detection points according to preset conditions, and then selecting planned detection points from the candidate detection points; A route planning unit for constructing a joint graph model A with the planned detection points and the position points on the crane designated for the external detection execution unit to attach or detach as nodes, the joint graph model A including a node set A and an edge set A, the node set A including the planned detection points and the position points, and the edge set A including connection edges A between the planned detection points and between the planned detection points and the position points, the connection edges A being assigned with comprehensive cost values A according to preset rules; and generating a detection route planning based on the joint graph model A and the comprehensive cost values A; A detection execution unit for mobilizing the external detection execution unit to perform metal structure detection on the planned detection points on the crane according to the detection route planning, and obtaining detection data associated with the position information of the corresponding planned detection point. A data processing unit for inputting the detection data obtained by the external detection execution unit into the trained metal damage recognition model to obtain the damage type, damage level and / or risk score of the crane metal at the position of each planned detection point, then performing comprehensive risk assessment to obtain a comprehensive risk index, and then dividing each planned detection point into a safe point, a suspicious point or a high-risk point according to the comprehensive risk index. The route planning unit is further configured to select all high-risk points and suspicious points with a part of the comprehensive risk indicators in a preset suspicious interval as secondary detection points from the planned detection points, and to construct a preset spatial neighborhood on a surface of a metal structure at each high-risk point, and to generate new detection points in the preset spatial neighborhood according to a coverage condition of the planned detection points, and to set the new detection points as supplemented detection points; and to construct a joint graph model B by taking the secondary detection points and the supplemented detection points as re-inspection nodes, the joint graph model B including a node set B and an edge set B, the node set B including the secondary detection points and the supplemented detection points, and the edge set B including connection edges B of virtual connections between the secondary detection points and between the secondary detection points and the supplemented detection points, the connection edges B being assigned with a comprehensive generation value B according to a preset rule; and to generate a re-inspection route planning based on the joint graph model B, the comprehensive generation value B, and priorities of the re-inspection nodes; The detection execution unit is further configured to mobilize an external detection execution unit to perform metal structure detection on the secondary detection points and the supplemented detection points on the crane according to the re-inspection route planning, and to obtain re-inspection data, wherein the re-inspection data is associated with position information of corresponding re-inspection nodes; The system further includes a data evaluation unit configured to perform damage identification and risk re-evaluation on the re-inspection nodes based on the re-inspection data, to combine an evaluation result with the comprehensive risk indicators output by the data processing unit to make a judgment, and to output a crane metal structure detection result.

Images