A finished product detection system, method, storage medium and electronic equipment for wind turbine blades

Abstract

The present invention provides a finished product inspection system, method, storage medium, and electronic equipment for wind turbine blades, relating to the technical field of wind turbine blade material testing. The system includes: a detection module for inspecting multiple wind turbine blades in a finished product area based on an artificial intelligence inspection body network; an analysis module for performing a real-time work rationality analysis on the artificial intelligence inspection body network; an optimization module for continuously optimizing and iterating the artificial intelligence inspection body network based on the work rationality analysis results; and a relay module for relaying inspection of each wind turbine blade based on the optimized artificial intelligence inspection body network after each optimization iteration. When conducting material inspection on finished wind turbine blades, damage and defects in the wind turbine blades are automatically identified through artificial intelligence control, thereby improving inspection efficiency. Continuously optimizing and iterating the artificial intelligence inspection body network optimizes its work capability, further improving the system's data processing level for material testing of finished wind turbine blades.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine blade material testing, and in particular to a finished product detection system, method, storage medium and electronic equipment for wind turbine blades. Background Art

[0002] Currently, wind turbine blades, as core components of wind turbines, directly impact their operational stability and power generation efficiency. Wind turbine blades often must withstand strong winds, extreme weather conditions, and high-load environments. Manufacturing quality is crucial; any defects can not only impact the long-term operation of the wind turbine but can also pose safety risks. Therefore, finished wind turbine blade inspection is particularly important, and material testing is a key step in ensuring proper wind turbine operation.

[0003] At present, the inspection of finished wind turbine blades still relies mainly on manual operation, and workers conduct inspections through visual inspection, manual measurement and traditional instruments. Although this traditional manual inspection method can detect surface defects and problems to a certain extent, it has some significant shortcomings. First, manual inspection relies on a large number of workers to work together, which is inefficient and cannot conduct comprehensive inspections of mass-produced wind turbine blades in a short period of time. Secondly, manual operation is easily affected by human factors, such as improper operation and fatigue, which will lead to large errors in the inspection results and even miss some potential defects. Finally, traditional inspection methods cannot adapt to the large-scale, rapid and dynamic changes in wind turbine blade production scenarios, and lack real-time and flexibility. Overall, the existing technology is insufficient in data processing for material testing of finished wind turbine blades.

[0004] Therefore, there is an urgent need for a more efficient, accurate and intelligent solution for material testing of finished wind turbine blades. Summary of the Invention

[0005] One of the purposes of the present invention is to provide a finished product inspection system for wind turbine blades. When conducting material inspection on finished wind turbine blades, an artificial intelligence inspection body network is introduced. Multiple inspection bodies, such as robots, drones, and AGV carts, are set up to automatically identify damage and defects in wind turbine blades through artificial intelligence control, greatly improving inspection efficiency and enabling comprehensive inspection of large-scale wind turbine blades in a short period of time. At the same time, the system continuously optimizes and iterates the artificial intelligence inspection body network through real-time work rational analysis, continuously optimizing the working capacity of the artificial intelligence inspection body network, greatly improving the accuracy of finished wind turbine blade inspection, better adapting to large-scale, rapid, and dynamically changing production needs, improving the real-time and flexibility of the system, and further improving the system's data processing level for material testing of finished wind turbine blades.

[0006] An embodiment of the present invention provides a finished product inspection system for wind turbine blades, comprising:

[0007] The detection module is used to inspect multiple wind turbine blades in the finished product area based on the network of artificial intelligence detection bodies;

[0008] Analysis module, used to perform real-time rational analysis of the working state of the artificial intelligence detection network;

[0009] The optimization module is used to continuously optimize and iterate the AI ​​detection body network based on the results of work rational analysis;

[0010] The relay module is used to relay the detection of each wind turbine blade after each optimization iteration based on the network of artificial intelligence detection bodies after the optimization iteration.

[0011] Optionally, the analysis module performs a real-time rational analysis of the artificial intelligence detection object network, including:

[0012] Select a benchmark divergence point from a sequence of decision divergence points generated within a recent preset time period by the artificial intelligence detection body network;

[0013] Retroactively generate a historical working map of the target detection object corresponding to the benchmark divergence point within the target period; wherein the first and last moments of the target period are the generation moments of the decision divergence point before and after the benchmark divergence point in the decision divergence point sequence respectively;

[0014] Based on the traversal order corresponding to the bifurcation point type of the reference bifurcation point, each map feature in the historical working map is traversed in sequence; wherein the map features include at least: the minimum bounding box movement change of the target detection body and the work interaction record, and the global change trend of the historical working map;

[0015] Each time the traversal is completed, when the traversed map feature matches the trigger map feature, the traversal to the next map feature is stopped, and multiple sets of one-to-one corresponding replay mechanisms and verification rules corresponding to the trigger map features that match the traversed map feature are obtained;

[0016] Based on any replay mechanism, the historical work map is controlled to be replayed, and the corresponding verification rules are executed during the replay of the historical work map, and the execution results of the corresponding verification rules are used as the work rationality analysis results.

[0017] Optionally, the step of selecting a reference divergence point from a sequence of decision divergence points generated within a recent preset time period by the artificial intelligence detection body network includes:

[0018] Determine the first goal and the second goal respectively from the decision divergence point sequence;

[0019] The first target or the second target, whichever is generated first, is used as the reference divergence point;

[0020] The steps for determining the first target are as follows:

[0021] The j-th decision divergence point in the decision divergence point sequence is taken as the first target; the sum of the cost values ​​of the first j-1 decision divergence points in the decision divergence point sequence is closest to the cost value and the threshold;

[0022] The steps for determining the second target are as follows:

[0023] The first decision divergence point in the decision divergence point sequence whose type is the same as the standard divergence point type is taken as the second target.

[0024] Optionally, the optimization module continuously optimizes and iterates the artificial intelligence detection body network based on the work rationality analysis results, including:

[0025] Describe the results of the work rationality analysis to obtain a feature description vector;

[0026] Determine the optimization iteration strategy corresponding to the feature description vector from the optimization iteration strategy library;

[0027] Based on the optimization iteration strategy, the artificial intelligence detection body network is continuously optimized and iterated.

[0028] Optional wind turbine blade finished product inspection system also includes:

[0029] Visual monitoring module for:

[0030] Generate an optimized and iterative visual monitoring model for the AI ​​detection body network;

[0031] Based on the visual monitoring model, management personnel are assisted in monitoring the work of the optimized and iterated artificial intelligence detection network.

[0032] Optionally, the visual monitoring module generates an optimized and iterative visual monitoring model of the artificial intelligence detection object network, including:

[0033] Based on the visualization template corresponding to the personnel portrait of the manager, a visualization monitoring model is generated according to the multimodal work information of the artificial intelligence detection body network after optimization and iteration.

[0034] Optionally, the visual monitoring module assists management personnel in monitoring the operation of the optimized and iterated artificial intelligence detection entity network based on the visual monitoring model, including:

[0035] Detecting multiple work monitoring timing events that occur in a visual monitoring model;

[0036] Generate a quick selection table of monitoring tasks based on each work monitoring opportunity event;

[0037] Assist management personnel to quickly select monitoring tasks from the monitoring task quick selection table and execute them accordingly.

[0038] An embodiment of the present invention provides a method for detecting finished wind turbine blades, comprising:

[0039] Based on the network of artificial intelligence detection bodies, multiple wind turbine blades in the finished product area are inspected;

[0040] Real-time rational analysis of the work of artificial intelligence detection body networking;

[0041] Based on the results of rational work analysis, the network of artificial intelligence detection bodies is continuously optimized and iterated;

[0042] After each optimization iteration, the artificial intelligence detection body based on the optimization iteration is networked to relay the detection of each wind turbine blade.

[0043] Optionally, the real-time rational analysis of the artificial intelligence detection object network includes:

[0044] Select a benchmark divergence point from a sequence of decision divergence points generated within a recent preset time period by the artificial intelligence detection body network;

[0045] Retroactively generate a historical working map of the target detection object corresponding to the benchmark divergence point within the target period; wherein the first and last moments of the target period are the generation moments of the decision divergence point before and after the benchmark divergence point in the decision divergence point sequence respectively;

[0046] Based on the traversal order corresponding to the bifurcation point type of the reference bifurcation point, each map feature in the historical working map is traversed in sequence; wherein the map features include at least: the minimum bounding box movement change of the target detection body and the work interaction record, and the global change trend of the historical working map;

[0047] Each time the traversal is completed, when the traversed map feature matches the trigger map feature, the traversal to the next map feature is stopped, and multiple sets of one-to-one corresponding replay mechanisms and verification rules corresponding to the trigger map features that match the traversed map feature are obtained;

[0048] Based on any replay mechanism, the historical work map is controlled to be replayed, and the corresponding verification rules are executed during the replay of the historical work map, and the execution results of the corresponding verification rules are used as the work rationality analysis results.

[0049] Optionally, the step of selecting a reference divergence point from a sequence of decision divergence points generated within a recent preset time period by the artificial intelligence detection body network includes:

[0050] Determine the first goal and the second goal respectively from the decision divergence point sequence;

[0051] The first target or the second target, whichever is generated first, is used as the reference divergence point;

[0052] The steps for determining the first target are as follows:

[0053] The j-th decision divergence point in the decision divergence point sequence is taken as the first target; the sum of the cost values ​​of the first j-1 decision divergence points in the decision divergence point sequence is closest to the cost value and the threshold;

[0054] The steps for determining the second target are as follows:

[0055] The first decision divergence point in the decision divergence point sequence whose type is the same as the standard divergence point type is taken as the second target.

[0056] Optionally, the optimization module continuously optimizes and iterates the artificial intelligence detection body network based on the work rationality analysis results, including:

[0057] Describe the results of the work rationality analysis to obtain a feature description vector;

[0058] Determine the optimization iteration strategy corresponding to the feature description vector from the optimization iteration strategy library;

[0059] Based on the optimization iteration strategy, the artificial intelligence detection body network is continuously optimized and iterated.

[0060] Optionally, the finished product inspection method of wind turbine blades further includes:

[0061] Generate an optimized and iterative visual monitoring model for the AI ​​detection body network;

[0062] Based on the visual monitoring model, management personnel are assisted in monitoring the work of the optimized and iterated artificial intelligence detection network.

[0063] Optionally, generating a visual monitoring model of the artificial intelligence detection body network after optimization and iteration includes:

[0064] Based on the visualization template corresponding to the personnel portrait of the manager, a visualization monitoring model is generated according to the multimodal work information of the artificial intelligence detection body network after optimization and iteration.

[0065] Optionally, the visual monitoring model is used to assist management personnel in monitoring the operation of the artificial intelligence detection body network after optimization and iteration, including:

[0066] Detecting multiple work monitoring timing events that occur in a visual monitoring model;

[0067] Generate a quick selection table of monitoring tasks based on each work monitoring opportunity event;

[0068] Assist management personnel to quickly select monitoring tasks from the monitoring task quick selection table and execute them accordingly.

[0069] An embodiment of the present invention provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and a processor executes the computer program to implement the above method.

[0070] An embodiment of the present invention provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above method.

[0071] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0072] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0074] Figure 1 Schematic diagram of a finished product inspection system for wind turbine blades according to an embodiment of the present invention;

[0075] Figure 2 This is a schematic diagram of a specific implementation of a finished product inspection system for wind turbine blades according to an embodiment of the present invention;

[0076] Figure 3 Schematic diagram of a finished product inspection method for wind turbine blades according to an embodiment of the present invention. DETAILED DESCRIPTION

[0077] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0078] The embodiment of the present invention provides a finished product detection system for wind turbine blades, such as Figure 1 As shown, including:

[0079] Detection module 1, used to detect multiple wind turbine blades in the finished product area based on artificial intelligence detection body networking;

[0080] Analysis module 2 is used to perform real-time rational analysis of the artificial intelligence detection body network;

[0081] Optimization module 3 is used to continuously optimize and iterate the AI ​​detection body network based on the results of work rational analysis;

[0082] The relay module 4 is used to carry out relay detection on each wind turbine blade based on the artificial intelligence detection body network after each optimization iteration.

[0083] The artificial intelligence detection body network is a network of multiple detection bodies that perform wind turbine blade inspection operations based on artificial intelligence technology. The detection bodies include at least: robots, drones, AGV vehicles, etc., which independently perform inspection task planning, operation trajectory planning, on-site obstacle avoidance control, etc. based on artificial intelligence technology. Specifically, each detection body is equipped with an advanced machine vision camera, combined with a deep learning algorithm, which can automatically identify damage or defects in wind turbine blades; when planning inspection task tasks, based on the environmental layout in the finished product area and the distribution of wind turbine blades, future inspection objects are intelligently planned; the detection body uses environmental sensor data to identify obstacles in the environment in real time, and automatically adjusts the obstacle avoidance path to ensure operation safety.

[0084] Specifically, such as Figure 2 As shown, in the specific implementation, the detection robot 1.1, the bottom detection device 1.2 and the upper detection device 1.3 in the detection module 1 work together to detect the wind turbine blades, and the detection results are uploaded to the analysis module 2 for analysis. Further analysis results are transmitted to the optimization module 3, and the artificial intelligence detection body network is continuously optimized and iterated. Finally, it is handed over to the relay module 4 based on the optimized and iterated artificial intelligence detection body network, and the relay control detection module 1 is used to detect each wind turbine blade.

[0085] The finished product area is where the finished products of wind turbine blades are stacked after they are manufactured.

[0086] During the process of the artificial intelligence detection body networking to inspect multiple wind turbine blades in the finished product area, a rational analysis of their work is conducted to determine whether they need continuous optimization and iteration.

[0087] Then, based on the results of rational work analysis, the artificial intelligence detection body network is continuously optimized and iterated.

[0088] Finally, after each optimization iteration, the artificial intelligence detection body after the optimization iteration is networked to relay the detection of each wind turbine blade.

[0089] This application introduces an artificial intelligence detection body network when conducting material testing on finished wind turbine blades. It sets up a combination of robots, drones, AGV carts and other detection bodies, and automatically identifies damage and defects in wind turbine blades through artificial intelligence control, greatly improving the detection efficiency. It can conduct comprehensive inspections on large-scale wind turbine blades in a short period of time. At the same time, the system continuously optimizes and iterates the artificial intelligence detection body network through real-time work rational analysis, continuously optimizing the working capacity of the artificial intelligence detection body network, greatly improving the accuracy of wind turbine blade finished product inspection, and better adapting to large-scale, rapid, and dynamically changing production needs. It improves the real-time and flexibility of the system and further improves the system's data processing level for material testing of finished wind turbine blades.

[0090] In one embodiment, the analysis module performs real-time rational analysis of the artificial intelligence detection object network, including:

[0091] Select a benchmark divergence point from a sequence of decision divergence points generated within a recent preset time period by the artificial intelligence detection body network;

[0092] Retroactively generate a historical working map of the target detection object corresponding to the benchmark divergence point within the target period; wherein the first and last moments of the target period are the generation moments of the decision divergence point before and after the benchmark divergence point in the decision divergence point sequence respectively;

[0093] Based on the traversal order corresponding to the bifurcation point type of the reference bifurcation point, each map feature in the historical working map is traversed in sequence; wherein the map features include at least: the minimum bounding box movement change of the target detection body and the work interaction record, and the global change trend of the historical working map;

[0094] Each time the traversal is completed, when the traversed map feature matches the trigger map feature, the traversal to the next map feature is stopped, and multiple sets of one-to-one corresponding replay mechanisms and verification rules corresponding to the trigger map features that match the traversed map feature are obtained;

[0095] Based on any replay mechanism, the historical work map is controlled to be replayed, and the corresponding verification rules are executed during the replay of the historical work map, and the execution results of the corresponding verification rules are used as the work rationality analysis results.

[0096] When the artificial intelligence detection body is networked to inspect finished wind turbine blades, each detection body needs to continuously make task planning decisions, path planning decisions, and obstacle avoidance control decisions. Every time a decision disagreement occurs, a decision divergence point will be formed (for example: multiple planning tasks are decided, and then one of them is selected for final execution, then the multiple planning tasks decided and one of the planning tasks finally executed form a decision divergence point). The decision divergence points generated by the artificial intelligence detection body network within the most recent preset time (for example: 100 seconds) are sorted in time sequence to obtain a decision divergence point sequence.

[0097] First, a benchmark divergence point is selected from the decision divergence point sequence. This serves as the benchmark for work rationality analysis. Next, a historical work map is generated retroactively. This map reflects the work status of the target detection object during wind turbine blade inspection within a target period, reflecting the spatial and temporal dimensions. The benchmark divergence point corresponds to the generation of the target detection object. The benchmark divergence point serves as the benchmark for work rationality analysis. The target time can be the period between the generation times of the previous and next decision divergence points. That is, the first and last moments of the target period are the generation times of the previous and next decision divergence points, respectively. Then, each map feature in the historical work map is traversed sequentially. Each time a traversal occurs, if the traversed map feature matches the trigger map feature, it indicates that the current replay verification can determine the ideal work analysis result. Finally, the replay mechanism and verification rules are indicated by the matching trigger map features. The mechanism is executed and verified in sequence, and the execution results are used as the work rationality analysis results.

[0098] Specifically, the divergence point type of the reference divergence point corresponds to a traversal order, in which there is a traversal priority order corresponding to different map features. In the traversal order, the more likely it is to reflect the current replay verification to determine the ideal analysis result of the work, the higher it is ranked. For example: if the divergence point type of the reference divergence point is the obstacle avoidance path decision divergence, then in the corresponding traversal order, the map feature with the minimum bounding box movement change can determine whether it is ultimately successful in obstacle avoidance, and the map feature will be traversed first. In addition, the triggering map feature can be, for example, the unstable movement change of the minimum bounding box of the target detection body. Correspondingly, the replay mechanism and verification rules can be, for example, controlling the historical working map to perform obstacle avoidance replay to verify whether the obstacle avoidance process of the target detection body is smooth.

[0099] When performing work rationality analysis on the network of artificial intelligence detection bodies, the embodiment of the present invention introduces a decision divergence point sequence, from which a benchmark divergence point is selected for work rationality analysis, thereby improving the efficiency of work rationality analysis; retrospectively generates a historical work map, and uses the generation time of the decision divergence point before and after the benchmark divergence point in the decision divergence point sequence as the first and last moments to accurately generate the target time period, thereby improving the accuracy of the retrospective generation of the historical work map, reducing the retrospective generation resources, and improving the retrospective generation efficiency; based on the traversal order, traverse each map feature in the historical work map in turn, and the more likely it is to reflect the current replay verification to determine the ideal work analysis result, the more priority it will have in traversal, thereby improving the efficiency of determining the timing of replay verification; match the trigger map feature indication replay mechanism and verification rules, execute them one by one, and use the execution results as the work rationality analysis results, further improving the efficiency of work rationality analysis, and improving the accuracy and comprehensiveness of work rationality analysis as a whole.

[0100] In one embodiment, selecting a reference bifurcation point from a sequence of decision bifurcation points generated within a recent preset time period by the artificial intelligence detection body network includes:

[0101] Determine the first goal and the second goal respectively from the decision divergence point sequence;

[0102] The first target or the second target, whichever is generated first, is used as the reference divergence point;

[0103] The steps for determining the first target are as follows:

[0104] The j-th decision divergence point in the decision divergence point sequence is taken as the first target; the sum of the cost values ​​of the first j-1 decision divergence points in the decision divergence point sequence is closest to the cost value and the threshold; j is a positive integer greater than 1;

[0105] The steps for determining the second target are as follows:

[0106] The first decision divergence point in the decision divergence point sequence whose type is the same as the standard divergence point type is taken as the second target.

[0107] The cost value of the decision divergence point represents the extent to which the generation of the decision divergence point will affect the work efficiency of the test object; the cost value and the threshold are pre-set by the technical personnel; setting the constraint that the sum of the cost values ​​of the first j-1 decision divergence points in the decision divergence point sequence is closest to the cost value and the threshold can make the value of j unique, and the jth decision divergence point selected as the first target can be used as an alternative benchmark for rational analysis of work.

[0108] The standard branch point type is a type of decision branch point that can be used as a benchmark for work rational analysis, for example, a branch point where it is difficult to decide between more than 10 planning tasks.

[0109] In order to improve the accuracy of the benchmark selection used for work rationality analysis, the first one of the first goal and the second goal is selected as the benchmark divergence point.

[0110] In one embodiment, the optimization module continuously optimizes and iterates the artificial intelligence detection body network based on the work rationality analysis results, including:

[0111] Describe the results of the work rationality analysis to obtain a feature description vector;

[0112] Determine the optimization iteration strategy corresponding to the feature description vector from the optimization iteration strategy library;

[0113] Based on the optimization iteration strategy, the artificial intelligence detection body network is continuously optimized and iterated.

[0114] The results of the work rationality analysis are characterized as feature description vectors. The optimization iteration strategy library contains optimization iteration strategies corresponding to different feature description vectors. The corresponding optimization iteration strategy is determined by searching the library, and the AI ​​detection object network is continuously optimized and iterated based on the optimization iteration strategy. The optimization iteration strategy indicates how to continuously optimize and iterate the AI ​​detection object network. For example, if the results of the work rationality analysis indicate that a certain detection object's obstacle avoidance accuracy is not suitable for the current complex finished product area, the optimization iteration strategy corresponding to its feature description as a feature description vector is to improve obstacle avoidance accuracy.

[0115] In one embodiment, the finished product inspection system for wind turbine blades further includes:

[0116] Visual monitoring module for:

[0117] Generate an optimized and iterative visual monitoring model for the AI ​​detection body network;

[0118] Based on the visual monitoring model, it assists management personnel in monitoring the work of the optimized and iterated artificial intelligence detection network;

[0119] The visual monitoring module generates an optimized and iterative visual monitoring model of the artificial intelligence detection body network, including:

[0120] Based on the visualization template corresponding to the personnel portrait of the manager, a visualization monitoring model is generated according to the multimodal work information of the artificial intelligence detection body network after optimization and iteration.

[0121] When generating a visual monitoring model, the manager's personnel portrait corresponds to a visual template (the visual template will sort the content that the manager needs to view first based on the manager's personnel portrait). The visual monitoring model is generated based on the multimodal work information of the optimized and iterative artificial intelligence detection body network. The targeted generation of the visual monitoring model improves the efficiency of managers in monitoring their work based on the optimized and iterative artificial intelligence detection body network.

[0122] In one embodiment, the visual monitoring module assists management personnel in monitoring the operation of the optimized and iterated artificial intelligence detection entity network based on the visual monitoring model, including:

[0123] Detecting multiple work monitoring timing events that occur in a visual monitoring model;

[0124] Generate a quick selection table of monitoring tasks based on each work monitoring opportunity event;

[0125] Assist management personnel to quickly select monitoring tasks from the monitoring task quick selection table and execute them accordingly.

[0126] Work monitoring timing events are those that require management intervention, such as a downtime on a test unit requiring maintenance. The monitoring task quick-select table contains multiple monitoring tasks generated based on each work monitoring timing event, allowing managers to quickly select and execute them.

[0127] The embodiment of the present invention provides a method for detecting finished products of wind turbine blades, such as Figure 3 As shown, including:

[0128] S1. Based on the network of artificial intelligence detection bodies, multiple wind turbine blades in the finished product area are inspected;

[0129] S2. Real-time analysis of the working principle of the artificial intelligence detection network;

[0130] S3. Based on the results of rational work analysis, continuously optimize and iterate the AI ​​detection body network;

[0131] S4. After each optimization iteration, the artificial intelligence detection body after the optimization iteration is networked and relayed to detect each wind turbine blade.

[0132] The real-time rational analysis of the artificial intelligence detection body network includes:

[0133] Select a benchmark divergence point from a sequence of decision divergence points generated within a recent preset time period by the artificial intelligence detection body network;

[0134] Retroactively generate a historical working map of the target detection object corresponding to the benchmark divergence point within the target period; wherein the first and last moments of the target period are the generation moments of the decision divergence point before and after the benchmark divergence point in the decision divergence point sequence respectively;

[0135] Based on the traversal order corresponding to the bifurcation point type of the reference bifurcation point, each map feature in the historical working map is traversed in sequence; wherein the map features include at least: the minimum bounding box movement change of the target detection body and the work interaction record, and the global change trend of the historical working map;

[0136] Each time the traversal is completed, when the traversed map feature matches the trigger map feature, the traversal to the next map feature is stopped, and multiple sets of one-to-one corresponding replay mechanisms and verification rules corresponding to the trigger map features that match the traversed map feature are obtained;

[0137] Based on any replay mechanism, the historical work map is controlled to be replayed, and the corresponding verification rules are executed during the replay of the historical work map, and the execution results of the corresponding verification rules are used as the work rationality analysis results.

[0138] The step of selecting a reference divergence point from a sequence of decision divergence points generated within a recent preset time period of the artificial intelligence detection body network includes:

[0139] Determine the first goal and the second goal respectively from the decision divergence point sequence;

[0140] The first target or the second target, whichever is generated first, is used as the reference divergence point;

[0141] The steps for determining the first target are as follows:

[0142] The j-th decision divergence point in the decision divergence point sequence is taken as the first target; the sum of the cost values ​​of the first j-1 decision divergence points in the decision divergence point sequence is closest to the cost value and the threshold;

[0143] The steps for determining the second target are as follows:

[0144] The first decision divergence point in the decision divergence point sequence whose type is the same as the standard divergence point type is taken as the second target.

[0145] The optimization module continuously optimizes and iterates the AI ​​detection body network based on the results of work rational analysis, including:

[0146] Describe the results of the work rationality analysis to obtain a feature description vector;

[0147] Determine the optimization iteration strategy corresponding to the feature description vector from the optimization iteration strategy library;

[0148] Based on the optimization iteration strategy, the artificial intelligence detection body network is continuously optimized and iterated.

[0149] The finished product inspection method of wind turbine blades also includes:

[0150] Generate an optimized and iterative visual monitoring model for the AI ​​detection body network;

[0151] Based on the visual monitoring model, management personnel are assisted in monitoring the work of the optimized and iterated artificial intelligence detection network.

[0152] The generation of the optimized and iteratively optimized visual monitoring model of the artificial intelligence detection body network includes:

[0153] Based on the visualization template corresponding to the personnel portrait of the manager, a visualization monitoring model is generated according to the multimodal work information of the artificial intelligence detection body network after optimization and iteration.

[0154] The visualization monitoring model assists management personnel in monitoring the work of the optimized and iterative artificial intelligence detection network, including:

[0155] Detecting multiple work monitoring timing events that occur in a visual monitoring model;

[0156] Generate a quick selection table of monitoring tasks based on each work monitoring opportunity event;

[0157] Assist management personnel to quickly select monitoring tasks from the monitoring task quick selection table and execute them accordingly.

[0158] An embodiment of the present invention provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and a processor executes the computer program to implement the above method.

[0159] An embodiment of the present invention provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above method.

[0160] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A finished product inspection system for wind turbine blades, characterized in that: include: The detection module is used to inspect multiple wind turbine blades in the finished product area based on the network of artificial intelligence detection bodies; Analysis module, used to perform real-time rational analysis of the working state of the artificial intelligence detection network; The optimization module is used to continuously optimize and iterate the AI ​​detection body network based on the results of work rational analysis; The relay module is used to relay the inspection of each wind turbine blade after each optimization iteration based on the network of artificial intelligence detection bodies after optimization iteration; The analysis module performs a real-time rational analysis of the AI ​​detection network, including: Select a benchmark divergence point from a sequence of decision divergence points generated within a recent preset time period by the artificial intelligence detection body network; Retroactively generate a historical working map of the target detection object corresponding to the benchmark divergence point within the target period; wherein the first and last moments of the target period are the generation moments of the decision divergence point before and after the benchmark divergence point in the decision divergence point sequence respectively; Based on the traversal order corresponding to the bifurcation point type of the reference bifurcation point, each map feature in the historical working map is traversed in sequence; wherein the map features include at least: the minimum bounding box movement change of the target detection body and the work interaction record, and the global change trend of the historical working map; Each time the traversal is completed, when the traversed map feature matches the trigger map feature, the traversal to the next map feature is stopped, and multiple sets of one-to-one corresponding replay mechanisms and verification rules corresponding to the trigger map features that match the traversed map feature are obtained; Based on any replay mechanism, the historical work map is controlled to be replayed, and the corresponding verification rules are executed during the replay of the historical work map, and the execution results of the corresponding verification rules are used as the work rationality analysis results.

2. The finished product inspection system for wind turbine blades according to claim 1, characterized in that: The step of selecting a reference divergence point from a sequence of decision divergence points generated within a recent preset time period of the artificial intelligence detection body network includes: Determine the first goal and the second goal respectively from the decision divergence point sequence; The first target or the second target, whichever is generated first, is used as the reference divergence point; The steps for determining the first target are as follows: The j-th decision divergence point in the decision divergence point sequence is taken as the first target; the sum of the cost values ​​of the first j-1 decision divergence points in the decision divergence point sequence is closest to the cost value and the threshold; The steps for determining the second target are as follows: The first decision divergence point in the decision divergence point sequence whose type is the same as the standard divergence point type is taken as the second target.

3. The finished product inspection system for wind turbine blades according to claim 1, characterized in that: The optimization module continuously optimizes and iterates the AI ​​detection body network based on the results of work rational analysis, including: Describe the results of the work rationality analysis to obtain a feature description vector; Determine the optimization iteration strategy corresponding to the feature description vector from the optimization iteration strategy library; Based on the optimization iteration strategy, the artificial intelligence detection body network is continuously optimized and iterated.

4. The finished product inspection system for wind turbine blades according to claim 1, characterized in that: Also includes: Visual monitoring module for: Generate an optimized and iterative visual monitoring model for the AI ​​detection body network; Based on the visual monitoring model, management personnel are assisted in monitoring the work of the optimized and iterated artificial intelligence detection network.

5. The finished product inspection system for wind turbine blades according to claim 4, characterized in that: The visual monitoring module generates an optimized and iterative visual monitoring model of the artificial intelligence detection body network, including: Based on the visualization template corresponding to the personnel portrait of the manager, a visualization monitoring model is generated according to the multimodal work information of the artificial intelligence detection body network after optimization and iteration.

6. The finished product inspection system for wind turbine blades according to claim 4, characterized in that: The visual monitoring module, based on the visual monitoring model, assists management personnel in monitoring the work of the optimized and iterated artificial intelligence detection network, including: Detecting multiple work monitoring timing events that occur in a visual monitoring model; Generate a quick selection table of monitoring tasks based on each work monitoring opportunity event; Assist management personnel to quickly select monitoring tasks from the monitoring task quick selection table and execute them accordingly.

7. A method for detecting finished products of wind turbine blades, characterized in that: include: Based on the network of artificial intelligence detection bodies, multiple wind turbine blades in the finished product area are inspected; Real-time rational analysis of the work of artificial intelligence detection body networking; Based on the results of rational work analysis, the network of artificial intelligence detection bodies is continuously optimized and iterated; After each optimization iteration, the AI ​​detection body after optimization iteration is networked and relayed to detect each wind turbine blade; The rational analysis of the artificial intelligence detection body network includes: Select a benchmark divergence point from a sequence of decision divergence points generated within a recent preset time period by the artificial intelligence detection body network; Retroactively generate a historical working map of the target detection object corresponding to the benchmark divergence point within the target period; wherein the first and last moments of the target period are the generation moments of the decision divergence point before and after the benchmark divergence point in the decision divergence point sequence respectively; Based on the traversal order corresponding to the bifurcation point type of the reference bifurcation point, each map feature in the historical working map is traversed in sequence; wherein the map features include at least: the minimum bounding box movement change of the target detection body and the work interaction record, and the global change trend of the historical working map; Each time the traversal is completed, when the traversed map feature matches the trigger map feature, the traversal to the next map feature is stopped, and multiple sets of one-to-one corresponding replay mechanisms and verification rules corresponding to the trigger map features that match the traversed map feature are obtained; Based on any replay mechanism, the historical work map is controlled to be replayed, and the corresponding verification rules are executed during the replay of the historical work map, and the execution results of the corresponding verification rules are used as the work rationality analysis results.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and the processor executes the computer program to implement the method according to claim 7.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method according to claim 7.

Images