An adaptive learning industrial endoscope control method and system

Abstract

The invention discloses an adaptive learning industrial endoscope control method and system, belonging to the field of industrial endoscopes, wherein the method comprises: setting a double-head industrial endoscope, the double-head industrial endoscope comprising a detection head and a detection head; generating a global detection path and adaptive detection parameters of the detection head based on a three-dimensional engine model, controlling the detection head to scan inside the target engine, obtaining preliminary scanning data, and marking the abnormal indication area inside the target engine; generating a fine detection area, obtaining a local detection path and adaptive detection parameters of the detection head, controlling the detection head to scan inside the target engine according to the local detection path, and obtaining fine detection data; fusing the fine detection data and the preliminary scanning data, and generating a target detection report of the target engine. Through the adaptive control of the industrial endoscope, the technical effect of improving the detection efficiency and accuracy of the industrial endoscope is achieved.

Description

Technical Field

[0001] The present invention relates to the field of industrial endoscopes, and in particular to an adaptive learning industrial endoscope control method and system. Background Art

[0002] In the existing technology, due to the lack of adaptive control capabilities of industrial endoscopes, when using industrial endoscopes to detect internal defects in automobile engines, there are problems with low detection efficiency and insufficient accuracy. Specifically, traditional industrial endoscopes use fixed detection paths and detection parameters, and cannot be adjusted autonomously according to the structural characteristics and potential defect locations of different engines, resulting in blind detection and repeated detection, wasting a lot of time and resources; at the same time, due to the lack of adaptive capabilities to the complex internal environment of the engine, the detection accuracy of existing industrial endoscopes is also limited, making it difficult to accurately locate and identify tiny defects. Summary of the invention

[0003] The present application provides an adaptive learning industrial endoscope control method and system, aiming to solve the technical problem in the prior art that industrial endoscopes lack adaptive control, resulting in low detection efficiency and insufficient accuracy of industrial endoscopes.

[0004] In view of the above problems, the present application provides an adaptive learning industrial endoscope control method and system.

[0005] The first aspect disclosed in the present application provides an adaptive learning industrial endoscope control method, the method comprising: setting a dual-head industrial endoscope, the dual-head industrial endoscope comprising a detection head and a detection head, the detection head being controlled by a detection control unit, the detection head being controlled by a detection control unit, the detection control unit and the detection control unit being respectively connected to a central analysis unit in communication; obtaining a three-dimensional engine model of a target engine, transmitting the three-dimensional engine model to the central analysis unit, generating a global detection path and adaptive detection parameters of the detection head based on the three-dimensional engine model; sending the global detection path and adaptive detection parameters to the detection control unit, controlling the detection head to scan inside the target engine, and obtaining preliminary scanning data; and transmitting the preliminary scanning data back to the central analysis unit. The detection data is transmitted to the central analysis unit, which analyzes the preliminary scanning data in combination with the three-dimensional engine model to identify the abnormal indication area inside the target engine; a fine detection area is generated based on the abnormal indication area, and a local detection path and adaptive detection parameters of the detection head are generated according to the fine detection area and the three-dimensional engine model, and sent to the detection control unit; the detection control unit configures the detection head based on the adaptive detection parameters, and controls the detection head to scan inside the target engine according to the local detection path to obtain fine detection data; the fine detection data is transmitted back to the central analysis unit, and the fine detection data and the preliminary scanning data are fused based on the three-dimensional engine model, and a target detection report for the target engine is generated according to the fusion processing result.

[0006] Another aspect disclosed in the present application provides an adaptive learning industrial endoscope control system, the system comprising: a dual-head setting module, used to set up a dual-head industrial endoscope, the dual-head industrial endoscope comprising a detection head and a detection head, the detection head is controlled by a detection control unit, the detection head is controlled by a detection control unit, the detection control unit and the detection control unit are respectively connected to a central analysis unit for communication; a detection planning module, used to obtain a three-dimensional engine model of a target engine, transmit the three-dimensional engine model to the central analysis unit, and generate a global detection path and adaptive detection parameters of the detection head based on the three-dimensional engine model; a detection control module, used to send the global detection path and adaptive detection parameters to the detection control unit, control the detection head to scan inside the target engine, and obtain preliminary scanning data; an abnormality recognition module, used to transmit the preliminary detection path and adaptive detection parameters to the detection control unit, and control the detection head to scan inside the target engine to obtain preliminary scanning data; and an abnormality recognition module, used to transmit the preliminary detection path and adaptive detection parameters to the detection control unit. The scanning data is transmitted back to the central analysis unit, which analyzes the preliminary scanning data in combination with the three-dimensional engine model to identify the abnormal indication area inside the target engine; the detection planning module is used to generate a fine detection area based on the abnormal indication area, and generate a local detection path and adaptive detection parameters of the detection head according to the fine detection area and the three-dimensional engine model, and send them to the detection control unit; the fine inspection control module is used for the detection control unit to configure the detection head based on the adaptive detection parameters, and control the detection head to scan inside the target engine according to the local detection path to obtain fine detection data; the data fusion module is used to transmit the fine detection data back to the central analysis unit, and fuse the fine detection data and the preliminary scanning data based on the three-dimensional engine model, and generate a target detection report for the target engine according to the fusion processing result.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] Due to the use of a double-headed industrial endoscope, including a detection head and a detection head, which are respectively controlled by a detection control unit and a detection control unit, the two control units are connected in communication with the central analysis unit, and the double-headed design is used to realize the division of labor and cooperation in detection and inspection, thereby improving the detection efficiency; the three-dimensional engine model of the target engine is obtained and transmitted to the central analysis unit, and the global detection path and adaptive detection parameters of the detection head are generated based on the three-dimensional model to fully understand the internal structure of the engine, thereby generating a global detection path and adaptive detection parameters to avoid blind detection; the global detection path and parameters are sent to the detection control unit to control the detection head to scan inside the engine to obtain preliminary scanning data, thereby dynamically adjusting according to the characteristics of the engine to improve the scanning quality; the preliminary scanning data is transmitted back to the central analysis unit, and combined with the three-dimensional model analysis data, the abnormal indication area is marked to provide guidance for subsequent detection; based on the A fine detection area is generated according to the abnormal indication area, and based on the fine detection area and the three-dimensional engine model, a local detection path and adaptive detection parameters of the detection head are generated, which are sent to the detection control unit to focus on the abnormal area for detection, and improve the detection accuracy through adaptive detection parameter adjustment; the detection control unit configures the detection head and controls it to scan in the target area according to the local detection path and adaptive detection parameters to obtain fine detection data and more detailed information; the fine detection data is transmitted back to the central analysis unit, and is fused with the preliminary scanning data and the three-dimensional engine model to generate a technical solution for the target detection report of the engine, which solves the technical problem of the lack of adaptive control of industrial endoscopes in the prior art, resulting in low detection efficiency and insufficient accuracy of industrial endoscopes, and achieves the technical effect of improving the detection efficiency and accuracy of industrial endoscopes through adaptive control of industrial endoscopes.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flowchart of an adaptive learning industrial endoscope control method is provided for an embodiment of the present application.

[0011] Figure 2 A structural schematic diagram of an adaptive learning industrial endoscope control system is provided for an embodiment of the present application.

[0012] Explanation of the reference numerals: dual-head setting module 11 , detection planning module 12 , detection control module 13 , abnormality identification module 14 , detection planning module 15 , precision inspection control module 16 , data fusion module 17 . DETAILED DESCRIPTION

[0013] The overall idea of ​​the technical solution provided by this application is as follows:

[0014] The embodiments of the present application provide an adaptive learning industrial endoscope control method and system, which realizes adaptive control of the endoscope through a dual-head industrial endoscope, a three-dimensional engine model and a central analysis unit, thereby improving the efficiency and accuracy of internal defect detection of the engine.

[0015] Specifically, first, the detection head and inspection head of the dual-head industrial endoscope are used to achieve division of labor and cooperation to improve detection efficiency; then, the three-dimensional engine model of the target engine is obtained, and the global detection path and adaptive detection parameters of the detection head are generated based on the model to guide the detection head to perform intelligent and full-coverage preliminary scanning; then, the preliminary scanning data is analyzed to identify the abnormal indication area, and the local detection path and adaptive detection parameters of the inspection head are generated to guide the inspection head to perform detailed detection of suspicious areas; then, the preliminary scanning data and detailed detection data are fused with the three-dimensional engine model to generate a comprehensive and accurate engine defect detection report.

[0016] To summarize, this application utilizes a three-dimensional engine model and a dual-head industrial endoscope to dynamically optimize the detection and inspection process of the industrial endoscope, thereby maximizing the inspection efficiency and accuracy, while realizing the fusion analysis of the inspection data, providing a reliable basis for engine defect diagnosis.

[0017] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0018] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides an adaptive learning industrial endoscope control method, the method comprising:

[0019] S1: Setting up a dual-head industrial endoscope, the dual-head industrial endoscope comprises a detection head and a detection head, the detection head is controlled by a detection control unit, the detection head is controlled by a detection control unit, the detection control unit and the detection control unit are respectively connected to the central analysis unit for communication.

[0020] Specifically, a double-head industrial endoscope is provided, which includes a detection head and a detection head, which are respectively controlled by a detection control unit and a detection control unit. Both the detection control unit and the detection control unit establish a communication connection with the central analysis unit, such as Bluetooth, WIFI, etc., to realize the transmission of data and control instructions.

[0021] By setting up independently controlled detection heads and inspection heads, the dual-head industrial endoscope can achieve refined detection and inspection in stages and regions. The detection head is responsible for performing a preliminary global scan of the target engine and obtaining global preliminary scanning data; the inspection head performs fine inspection of abnormal areas based on the preliminary scanning data obtained by the detection head and obtains local fine inspection data. Through the coordinated control of the dual-head design and the central analysis unit, the adaptive learning and intelligent inspection of the industrial endoscope can be realized, effectively improving the inspection efficiency and accuracy of the endoscope.

[0022] Compared with the existing single-control industrial endoscope, the dual-head design and multi-level control can complete the comprehensive detection and fine inspection tasks inside the engine more flexibly and efficiently, give full play to the performance of the industrial endoscope, and reduce the risk of missed detection and false detection.

[0023] S2: Acquire a three-dimensional engine model of the target engine, transmit the three-dimensional engine model to the central analysis unit, and generate a global detection path and adaptive detection parameters of the detection head based on the three-dimensional engine model.

[0024] Specifically, first, obtain the three-dimensional engine model of the target engine, for example, obtain the CAD model file of the target engine from the engine manufacturer, or use the three-dimensional scanning technology to scan and model the target engine. After obtaining the three-dimensional engine model, it is transmitted to the central analysis unit. After receiving the three-dimensional engine model, the central analysis unit generates the global detection path and adaptive detection parameters of the detection head based on the three-dimensional engine model. The purpose of generating the global detection path is to guide the detection head to perform a comprehensive preliminary scan inside the engine to cover all areas that need to be detected. The central analysis unit first analyzes the three-dimensional engine model and extracts the key structural features inside the engine, such as the air intake, combustion chamber, turbine blades and other key parts. Then, the central analysis unit uses a path planning algorithm, such as the Dijkstra algorithm, the A* algorithm, etc., to calculate the optimal detection path covering these key structural features. This path takes into account both detection efficiency and detection coverage, minimizes the moving distance and number of turns of the detection head, and ensures that no area that needs to be detected is missed.

[0025] While generating the global detection path, the central analysis unit determines the adaptive detection parameters of the detection head when executing the path, including the detection angle, detection distance and detection speed of the detection head. By analyzing the structural characteristics and complexity of different areas in the 3D engine model, the central analysis unit dynamically adjusts these parameters so that the detection head can adaptively optimize the detection process. For example, in narrow or complex areas, the detection angle is set to be smaller, the detection distance is shorter, and the detection speed is slower to improve the precision of the detection; while in open or simple areas, a larger detection angle, longer detection distance and faster detection speed are used to improve detection efficiency.

[0026] By acquiring the three-dimensional engine model of the target engine and using the model to adaptively generate the global detection path and adaptive detection parameters of the detection head, key data and control basis are provided for subsequent intelligent detection.

[0027] S3: Sending the global detection path and the adaptive detection parameters to the detection control unit, controlling the detection head to scan inside the target engine, and acquiring preliminary scanning data.

[0028] Specifically, the central analysis unit sends the generated global detection path and adaptive detection parameters to the detection control unit. After receiving the global detection path and adaptive detection parameters, the detection control unit converts the global detection path into a series of control instructions, such as moving direction, moving distance, rotation angle, etc., and sends them to the detection head for execution in sequence. At the same time, the detection control unit dynamically adjusts the detection angle, detection distance, detection speed and other parameters of the detection head according to the adaptive detection parameters, so that the detection head can adaptively optimize the detection process according to the characteristics of different areas inside the engine. During the scanning process of the detection head, the detection control unit receives the scanning data collected by the detection head in real time, and performs preliminary processing and analysis to obtain preliminary scanning data. The preliminary scanning data includes raw data in the form of images of the internal structure of the engine, point clouds, etc., reflecting the basic conditions inside the engine.

[0029] Through the control of the detection control unit and the scanning of the detection head, the preliminary detection of the target engine is completed, and the global scanning data is obtained, which provides an important information basis for subsequent abnormality detection and fine detection. Compared with the existing technology, through adaptive detection parameters, the detection head can dynamically adjust the detection strategy according to the characteristics of different areas inside the engine, improving the pertinence and effectiveness of the detection.

[0030] S4: The preliminary scanning data is transmitted back to the central analysis unit, and the preliminary scanning data is analyzed in combination with the three-dimensional engine model to identify abnormal indication areas inside the target engine.

[0031] Specifically, the detection control unit transmits the preliminary scanning data acquired by the detection head back to the central analysis unit. After receiving the preliminary scanning data, the central analysis unit analyzes and processes the preliminary scanning data in combination with the three-dimensional engine model of the target engine. First, the central analysis unit aligns the preliminary scanning data with the three-dimensional engine model, maps the initial scanning data to the coordinate space of the three-dimensional engine model, establishes the corresponding relationship between the preliminary scanning data and the internal physical structure of the engine, and provides a spatial reference for subsequent data analysis. After the alignment is completed, the central analysis unit performs various processing and analysis on the preliminary scanning data to identify the areas where abnormalities may exist inside the engine. For example, the central analysis unit performs image enhancement, edge detection and other operations on the scanned image to highlight the abnormal features in the image; for point cloud data, the central analysis unit performs point cloud filtering, normal vector estimation and other processing to extract the geometric features in the point cloud. After completing the data processing, the central analysis unit uses a variety of abnormality detection algorithms, such as threshold detection, pattern recognition, machine learning and other methods, to identify the abnormal indication area from the processed data. The abnormal indication area is manifested as damage, deformation, foreign matter and other problems of the internal structure of the engine, which are significantly different from the characteristics under normal conditions. The central analysis unit grades and marks the abnormal indication areas according to the degree of abnormality, providing a reference for subsequent fine detection.

[0032] By analyzing the preliminary scan data, the abnormal area inside the engine is automatically identified, which reduces the workload of manual detection and improves the efficiency and accuracy of abnormal detection. At the same time, through the three-dimensional engine model, the abnormal detection results can be associated with the specific internal structure of the engine, providing richer information for the analysis of the abnormal cause and subsequent maintenance decisions.

[0033] S5: Generate a fine detection area based on the abnormal indication area, generate a local detection path and adaptive detection parameters of the detection head according to the fine detection area and the three-dimensional engine model, and send them to the detection control unit.

[0034] Specifically, first, the central analysis unit further analyzes and evaluates the identified abnormal indication area to determine the geometric features such as the position, size, shape, and attribute features such as the abnormality degree and abnormality type of each abnormal indication area. Then, the central analysis unit generates a fine detection area based on the abnormal indication area according to these attribute features. The fine detection area includes the abnormal indication area itself, as well as the normal area within a certain range around it, to ensure that the abnormal area can be comprehensively and carefully detected. When delineating the fine detection area, the central analysis unit fully considers the structural layout and spatial constraints inside the engine to ensure that the detection head can actually reach and cover these areas.

[0035] After the fine detection area is generated, the central analysis unit begins to plan the local detection path and adaptive detection parameters of the detection head. Different from the global detection path of the detection head, the local detection path focuses on the detail coverage and key detection in the fine detection area. According to the specific characteristics and detection requirements of each area, the motion trajectory and sampling strategy of the detection head are designed. At the same time, the central analysis unit dynamically generates adaptive detection parameters such as the focal length, light intensity, and scanning density of the detection head according to the degree of abnormality and complexity of the fine detection area to optimize the detection effect. For example, for areas with a high degree of abnormality or complex structure, the detection head uses a smaller focal length, stronger lighting, and higher scanning density to obtain clearer and more complete detection data. After that, the central analysis unit sends the generated local detection path and adaptive detection parameters to the detection control unit. The detection control unit is responsible for controlling the detection head to scan and collect data in the fine detection area according to the planned path and parameters, providing support for subsequent refined detection.

[0036] By obtaining local path planning and adaptive parameter control, the detection head is guided to conduct targeted and refined detection of abnormal areas, thereby improving the accuracy and reliability of industrial endoscope detection. The results of preliminary scanning data analysis are fully utilized, and combined with the prior information of the three-dimensional engine model, intelligent scheduling and optimization of the detection process are achieved.

[0037] S6: The detection control unit configures the detection head based on the adaptive detection parameters, and controls the detection head to scan inside the target engine according to the local detection path to obtain fine detection data.

[0038] Specifically, after the detection control unit receives the local detection path and adaptive detection parameters sent by the central analysis unit, it starts to configure and control the detection head, guiding the detection head to scan the fine detection area of ​​the target engine according to the planned path and parameters, and obtain high-quality fine detection data.

[0039] First, the detection control unit sets and adjusts the various performance parameters of the detection head, such as focal length, light intensity, scanning density, etc., according to the adaptive detection parameters. By dynamically configuring these parameters, the detection control unit can match the detection performance of the detection head with the characteristics and requirements of each fine detection area, so as to obtain the best detection effect. For example, for areas with a high degree of abnormality or complex structure, the detection control unit will reduce the focal length of the detection head, increase the light intensity, and increase the scanning density to obtain clearer and more complete detection images and data. After the parameter configuration is completed, the detection control unit begins to control the detection head to scan and collect data in the fine detection area inside the target engine according to the planned local detection path. Unlike the global detection of the detection head, the local detection of the detection head is more focused and meticulous, and each fine detection area is completely and deeply scanned and covered, without missing any suspicious details.

[0040] During the scanning process, the detection head dynamically adjusts the detection state and data acquisition strategy according to the setting of adaptive detection parameters to adapt to the detection needs of different areas. At the same time, the detection head transmits the collected high-resolution images, videos, point clouds and other fine detection data to the detection control unit in real time. The detection control unit performs preliminary processing and analysis on these data, such as image enhancement, noise filtering, etc., to improve the quality and availability of the data. After the fine scanning of the detection head and the optimization control of the detection control unit, the fine detection data reflecting the detailed conditions of the abnormal area inside the target engine is obtained. Compared with conventional endoscopic detection, the intelligent adaptation of the detection head to the internal environment of the engine is realized through the strategy of adaptive detection parameters and local path planning, which improves the pertinence and effectiveness of the detection, and obtains more comprehensive and accurate internal abnormal information of the engine through refined and intelligent endoscopic detection, providing strong support for subsequent target detection reports.

[0041] S7: The fine detection data is transmitted back to the central analysis unit, and the fine detection data and the preliminary scanning data are fused based on the three-dimensional engine model, and a target detection report of the target engine is generated according to the fusion processing result.

[0042] Specifically, first, the central analysis unit registers and fuses the received fine detection data with the previous preliminary scanning data. By spatially aligning and associating the data acquired at different stages and different perspectives, the central analysis unit obtains the fusion processing results. After the data fusion is completed, the central analysis unit begins to combine the three-dimensional engine model to conduct in-depth mining and analysis of the fused data. Using intelligent algorithms such as anomaly detection, fault diagnosis, and pattern recognition, the central analysis unit extracts key features and indicators reflecting the health status of the engine from multi-source heterogeneous data, such as structural integrity, surface wear, and hot spot distribution. Through a comprehensive evaluation of these features and indicators, the central analysis unit accurately identifies and locates various anomalies inside the engine, and determines their severity and potential impact. After completing the anomaly analysis and evaluation, the central analysis unit will generate a target detection report for the target engine. The report is based on the three-dimensional engine model and intuitively displays the overall condition and local anomalies inside the engine, including the location, range, type, and other information of the abnormal area, thereby achieving efficient and accurate detection of the target engine.

[0043] Furthermore, the embodiment of the present application also includes:

[0044] Extracting internal structural features of the target engine from the three-dimensional engine model, and determining a plurality of key detection points of the detection head based on the internal structural features;

[0045] Using a path planning algorithm to generate the global detection path according to the multiple key detection points;

[0046] Based on the three-dimensional engine model and the global detection path, the detection angle, detection distance and detection speed of the detection head executing the global detection path movement are determined to generate the adaptive detection parameters.

[0047] In a feasible implementation, first, the central analysis unit extracts the internal structural features of the target engine from the acquired three-dimensional engine model, including the geometric shape, size, topological relationship and spatial position of key components inside the engine, so as to determine the key positions and path constraints when the probe head detects inside the engine. Based on the extracted internal structural features, the central analysis unit further determines multiple key detection points of the probe head, that is, the important positions that the probe head must reach and cover, such as the key parts such as the air intake, combustion chamber, and turbine blades of the engine. After determining the key detection points, the central analysis unit uses a path planning algorithm to generate a global detection path of the probe head according to these key detection points. Among them, the path planning algorithm can select a classic algorithm, such as the Dijkstra algorithm, the A* algorithm, etc., or a special path optimization algorithm can be designed according to the characteristics of the internal environment of the engine. Through the path planning algorithm, under the premise of meeting the coverage of key detection points, a detection path with the shortest total length, the least number of turns, and the best motion coherence is generated to improve the detection efficiency and feasibility, and obtain a global detection path.

[0048] After generating the global detection path, the central analysis unit further combines the three-dimensional engine model and the global detection path to determine the adaptive detection parameters for the detection head to perform the detection task, including detection angle, detection distance and detection speed. Among them, the detection angle refers to the inclination angle of the detection head relative to the inner surface of the engine, so that the detection head can obtain a clear and comprehensive field of view; the detection distance refers to the distance between the detection head and the inner surface of the engine, the safety of detection and the quality of data; the detection speed refers to the speed at which the detection head moves along the detection path, while ensuring the integrity of data collection, improving the detection efficiency. The adaptive detection parameters are dynamically adjusted according to the complexity and detection difficulty of different areas inside the engine to achieve adaptive optimization control of the detection process. Since the detection head mainly performs preliminary detection and has relatively low requirements for data accuracy, the values ​​of the adaptive detection parameters can be relatively loose, and the focus is on ensuring the comprehensive coverage and path accessibility of the detection.

[0049] By extracting information from the 3D engine model and applying the path planning algorithm, the global detection path and detection parameters of the detection head are adaptively generated, providing a basis for the intelligent deployment and motion control of the detection head. Compared with the conventional manual planning method, the digital advantages of the 3D model are combined with the optimization capabilities of the intelligent algorithm to automatically generate a safe, efficient and feasible detection plan, reduce the blindness and uncertainty of manual planning, and improve the intelligence level of the detection process.

[0050] Furthermore, the embodiment of the present application also includes:

[0051] Acquiring fault diagnosis information of the target engine;

[0052] According to the fault diagnosis information, a potential fault area is identified in the three-dimensional engine model, a plurality of key detection points are generated based on the potential fault area and the internal structure characteristics, and a fault correlation between the key detection points and the potential fault area is identified;

[0053] Generating detection priority weights for a plurality of key detection points based on the fault correlation, and generating the global detection path according to the priority weights and the plurality of key detection points;

[0054] The adaptive detection parameters are generated according to the three-dimensional engine model, the global detection path and the detection priority weight.

[0055] In a preferred embodiment, first, the central analysis unit obtains the fault diagnosis information of the target engine, which comes from multiple channels such as the online monitoring system and the expert diagnosis system, reflecting various faults, abnormal conditions and suspicious parts of the target engine during operation. Subsequently, the central analysis unit combines the fault diagnosis information with the three-dimensional engine model and identifies potential fault areas in the model. These areas are the parts that are most likely to have problems, such as high temperature areas, high pressure areas, high wear areas, etc., based on historical fault data and expert experience. While identifying the potential fault areas, the central analysis unit combines the internal structural characteristics of the engine to generate multiple key detection points, including key parts such as the engine's air intake, combustion chamber, turbine blades, etc., and also covers the potential fault areas. In addition, the central analysis unit calculates the correlation between each key detection point and the potential fault area to describe the importance of the key detection point to fault diagnosis. For example, for each key detection point, the number of historical fault cases near it is counted. If the number of fault cases near a key detection point is large, and the fault cases are concentrated near certain specific potential fault areas, it is considered that the correlation between the key detection point and the specific potential fault area is very high, and the importance to fault diagnosis is higher.

[0056] With the key detection points and fault correlation, the central analysis unit assigns a detection priority weight to each key detection point. The higher the detection priority weight, the greater the correlation between the key detection point and the potential fault area, and it should be given priority and covered in the subsequent detection path planning. Based on the detection priority weight, the central analysis unit generates a global detection path that takes into account both global coverage and local emphasis. It can not only cover all key detection points, but also arrange the order and time allocation of detection according to the detection priority, making the detection process more efficient and targeted. After that, the central analysis unit comprehensively considers the three-dimensional engine model, the optimized global detection path and the detection priority weight, and generates adaptive detection parameters for the detection head. Specifically, for areas with high fault correlation and high detection priority, the detection head adopts a smaller detection angle, a closer detection distance and a slower detection speed to obtain more detailed information; while for areas with low fault correlation and low detection priority, the detection head relaxes these parameters accordingly to improve detection efficiency. Through adaptive adjustment of detection parameters, the detection head can flexibly respond to the detection needs of different areas under the guidance of the global detection path, while ensuring the detection quality and taking into account the detection efficiency.

[0057] Through fault diagnosis information and detection priority, the detection path planning and parameter optimization are further enhanced in terms of pertinence and intelligence. Compared with the three-dimensional engine model alone, the integration of historical data and expert experience can more accurately and efficiently guide the movement and data collection of the detection head, making the detection process more targeted and more in line with actual detection needs.

[0058] Furthermore, the embodiment of the present application also includes:

[0059] Registering the preliminary scan data with the three-dimensional engine model to obtain registered scan data, and performing feature extraction on the registered scan data to obtain a feature extraction result;

[0060] Comparing the feature extraction result with a normal engine internal feature template, identifying a feature deviation area, and marking the feature deviation area as the abnormality indication area;

[0061] Extracting the characteristic deviation amount of the characteristic deviation area, and determining the abnormality degree of the abnormal indication area according to the characteristic deviation amount;

[0062] The abnormality indication area is marked in the three-dimensional engine model, and an abnormality level is set for the abnormality indication area according to the degree of abnormality.

[0063] In a feasible implementation, first, the central analysis unit registers the preliminary scanning data sent back by the detection head with the three-dimensional engine model, and performs spatial alignment and coordinate conversion on the scanning data and the three-dimensional model, so that each data point in the scanning data can correspond to a specific position in the three-dimensional model. The registered scanning data obtained after registration not only contains the original measurement information, but also integrates the position and structure information provided by the three-dimensional model, laying a data foundation for subsequent feature extraction. Then, the central analysis unit performs feature extraction on the registered scanning data. Feature extraction refers to identifying and extracting key features that can reflect the internal state of the engine from the data, such as surface geometry, material properties, thermal distribution, etc., to obtain feature extraction results. Through feature extraction, the original preliminary scanning data is converted into a feature vector representation with clear physical and engineering meanings, which is convenient for subsequent abnormal analysis and judgment.

[0064] Then, the central analysis unit compares the extracted feature extraction results with the pre-established normal engine internal feature template. Among them, the normal engine internal feature template refers to the standard feature representation summarized based on the detection data of a large number of engines of the same model under normal working conditions, reflecting the characteristic attributes that the engine should have under ideal conditions. By comparing the feature extraction results with the normal engine internal feature template item by item, the central analysis unit finds the differences and deviations between the two, thereby identifying the feature deviation area. The feature deviation area is usually manifested as a deviation exceeding the threshold between the feature extraction result and the standard value in the normal engine internal feature template, indicating that there may be an abnormality inside the engine. Therefore, the central analysis unit identifies the feature deviation area as an abnormal indication area as the focus of subsequent analysis. In order to further quantify the degree of abnormality in the abnormal indication area, the central analysis unit extracts the feature deviation amount in each abnormal indication area. The feature deviation amount refers to the difference or distance between the feature extraction result and the standard value in the normal engine internal feature template, reflecting the severity of the actual state deviating from the normal state. Among them, the larger the feature deviation amount, the more serious the problem in the abnormal indication area, and the greater the threat to the operation of the engine. Therefore, by calculating and analyzing the feature deviation, the degree of abnormality of each abnormal indication area is determined, providing a quantitative basis for subsequent detection decisions.

[0065] Afterwards, the central analysis unit associates the identified abnormal indication area and its abnormality degree information with the three-dimensional engine model. Specifically, abnormal marks and annotations are added at the corresponding positions of the three-dimensional engine model to intuitively display the spatial distribution and severity of the abnormal area. At the same time, according to the different degrees of abnormality, the central analysis unit sets different abnormality levels for each abnormal indication area, such as mild abnormality, moderate abnormality, severe abnormality, etc., to facilitate subsequent detection adaptive control.

[0066] Furthermore, the embodiment of the present application also includes:

[0067] Acquire historical detection data of engines of the same model as the target engine, and acquire abnormal area correlation information according to the historical detection data;

[0068] Based on the abnormal area association information and the abnormality degree of the abnormal indication area, analyzing the associated position set of the abnormal indication area, and determining the associated indication area in combination with the three-dimensional engine model, and setting the abnormality degree of the associated indication area to 0;

[0069] The associated indication area and the abnormal indication area are aggregated to generate the fine detection area, each fine detection area having a corresponding abnormality degree.

[0070] In a preferred embodiment, first, the central analysis unit obtains the historical detection data of engines of the same model as the target engine, records the various abnormal conditions found in the past detection of the same model engine, and the spatial distribution and correlation relationship of these abnormal conditions inside the engine. By analyzing and mining the historical detection data, the central analysis unit summarizes the correlation rules between abnormal areas and forms abnormal area correlation information. For example, some abnormal areas often appear in pairs or groups at the same time, indicating that these areas are connected or adjacent in physical location; for another example, after some abnormal areas appear, they often induce other abnormal areas within a certain range near them, indicating that these areas have a causal relationship in fault propagation. Abnormal area correlation information provides an important basis for predicting and inferring the spatial distribution of abnormal areas.

[0071] Then, the central analysis unit further analyzes and determines the associated position set of the abnormal indication area based on the abnormal area association information and the abnormal degree of the abnormal indication area. Specifically, for each abnormal indication area, the central analysis unit first determines whether the area is a serious abnormality or a minor abnormality according to its abnormal degree. Then, based on the abnormal area association information, analyze other locations that may be associated with the abnormal indication area in space. If the association information indicates that other abnormal areas often appear near a serious abnormal area, the central analysis unit will identify the parts adjacent to the serious abnormal area as associated positions; if the association information indicates that some minor abnormal areas are usually caused by the propagation of other abnormal areas, the central analysis unit will identify the possible abnormal propagation source as an associated position. Through association analysis, the central analysis unit starts from the known abnormal indication area and infers more potential abnormal distribution areas. After completing the identification of the associated position set, the central analysis unit combines the three-dimensional engine model to mark these associated positions in the model to form a set of associated indication areas. Unlike the abnormal indication area, the associated indication area is only an area where there may be abnormalities inferred based on experience and association information, and has not been directly verified in the preliminary scanning data. Therefore, its abnormal degree will be initially set to 0, indicating that further detection and confirmation are required.

[0072] Afterwards, the central analysis unit will summarize and merge the obtained associated indication area with the original abnormal indication area to generate a fine detection area, which includes not only the abnormal area directly identified from the initial scanning data, but also the potential abnormal area inferred based on the associated analysis, which can cover the parts of the engine that may have problems more comprehensively and systematically. At the same time, since the abnormality degree information of different areas is also recorded and transmitted together, each fine detection area will be assigned a corresponding abnormality degree, providing an important priority reference for the subsequent detection task planning.

[0073] By utilizing historical detection data and correlation analysis, the local abnormality information discovered in the initial scan is expanded into a systematic abnormality prediction for the entire internal area of ​​the engine, forming a more comprehensive and wider coverage of a detailed detection area, improving the accuracy of abnormality detection, reducing the risk of missed detection, and providing information for subsequent detection path planning.

[0074] Furthermore, the embodiment of the present application also includes:

[0075] Prioritize the fine detection areas according to the abnormality degree, and obtain the area priority ranking result;

[0076] Using the three-dimensional engine model, obtaining a set of shortest paths between regions in the fine detection region;

[0077] Generate an initial local detection path based on the regional priority sorting results and the shortest path set;

[0078] In combination with obstacle information in the three-dimensional engine model, the initial local detection path is optimized to obtain the local detection path;

[0079] According to the abnormality degree of the fine detection area, the three-dimensional engine model and the local detection path, the detection angle, detection distance and detection speed of the detection head are determined to generate the adaptive detection parameters.

[0080] In a feasible implementation, first, the central analysis unit prioritizes the fine detection areas according to the degree of abnormality. Since different fine detection areas have different degrees of abnormality, the potential impact and harm on engine operation are also different, so they are treated differently when planning the detection task. Through priority sorting, areas with higher abnormality can be arranged for detection first, while areas with lower abnormality can be placed at the back for detection, thereby obtaining the regional priority sorting result, which is a regional priority list that records the detection priority order of all fine detection areas. Then, the central analysis unit uses the three-dimensional engine model to calculate the shortest path set between each area in the fine detection area, including the shortest path between each area in the fine detection area, wherein the shortest path refers to the optimal route connecting two fine detection areas in three-dimensional space. Since the fine detection areas are distributed in different positions inside the engine, the detection head needs to move and switch frequently between these areas, so the shortest path can help the detection head to complete the jump between areas more efficiently and reduce unnecessary repeated movements.

[0081] After obtaining the regional priority sorting results and the shortest path set, the central analysis unit generates an initial local detection path. The local detection path refers to the optimal route covering all fine detection areas planned for the detection head based on the distribution and priority of the fine detection area. Specifically, the central analysis unit first includes the areas with high abnormality into the path according to the regional priority sorting results, and then connects the fine detection areas in the most efficient way through the shortest path set to form a complete closed path. Although the initial local detection path generated in this way is optimized in macroscopic terms, it ignores local obstacles and travel restrictions and requires further fine-tuning and optimization. Therefore, the central analysis unit optimizes and improves the initial local detection path in combination with the obstacle information in the three-dimensional engine model. Among them, the obstacle information includes various structural parts, partitions, rotors, etc. inside the engine. These parts will hinder the straight-line travel of the detection head and limit its range of activities in certain areas. By introducing obstacle information in path planning, the central analysis unit makes local adjustments and detours to the blocked and restricted sections on the initial local detection road, and obtains a local detection path that meets both detection requirements and spatial constraints.

[0082] Afterwards, the central analysis unit generates adaptive detection parameters for the detection head according to the degree of abnormality in the fine detection area, the three-dimensional engine model, and the local detection path. Adaptive detection parameters also include multiple dimensions such as detection angle, detection distance, and detection speed, which match the characteristics of each fine detection area. For example, for areas with a higher degree of abnormality, the detection head needs to use a smaller detection distance and a slower detection speed to obtain more detailed information; while for areas with a lower degree of abnormality and a wider space, the detection head appropriately increases the detection distance and speed to save time. In addition, in narrow spaces or spaces with many obstacles, the detection angle also needs to be appropriately adjusted to ensure the flexibility and accessibility of the detection head.

[0083] By generating a local detection path for fine detection for the detection head and matching it with adaptive detection parameters. Compared with the conventional detection path planning method, it makes full use of the existing abnormality degree information, engine structure information and inter-regional correlation information, while ensuring the detection efficiency and accuracy, it maximizes the pertinence and feasibility of the path. The resulting local detection scheme can not only efficiently and comprehensively cover all fine detection areas, but also can be flexibly and dynamically adjusted according to the characteristics and needs of each area, thereby improving the overall intelligence level and actual effect of industrial endoscope detection.

[0084] Furthermore, the embodiment of the present application also includes:

[0085] Based on the regional priority ranking result, selecting a starting detection area and determining an initial detection order;

[0086] The initial detection order is adjusted according to the spatial relationship between the regions in the fine detection area in the three-dimensional engine model to obtain a target detection order;

[0087] A path point sequence connecting each area is generated through the starting detection area, the target detection sequence and the shortest path set, and an initial local detection path is generated in combination with the three-dimensional engine model.

[0088] In a preferred embodiment, first, the central analysis unit selects the fine detection area with the highest abnormality as the starting detection area based on the regional priority sorting result, and as the starting point for the detection head to start performing the detection task. At the same time, the central analysis unit numbers and arranges all fine detection areas according to the regional priority sorting result to form an initial detection order, following the principle of regional priority from high to low, to ensure that the area with higher abnormality can be detected as early as possible. However, the initial detection order does not take into account the relationship between different fine detection areas in spatial position. Therefore, the central analysis unit further adjusts and optimizes the initial detection order according to the spatial relationship of the fine detection area in the three-dimensional engine model. By analyzing the specific position of each fine detection area inside the engine, adjacent areas, obstacle distribution and other spatial information, the central analysis unit finds the regional correlation that is ignored in the initial detection order. For example, although two areas have different priorities, they are close to each other in space. Detection according to the initial order will lead to the problem of repeated movement back and forth; for example, although some areas have lower priorities, they happen to be located on the only way to the high-priority area. Detecting them in advance can save the total distance. The central analysis unit adjusts and optimizes the initial detection order through spatial relationships to obtain a more reasonable and efficient target detection order.

[0089] After obtaining the starting detection area and the target detection order, the central analysis unit generates an initial local detection path based on the shortest path set. The specific approach is to take the starting detection area as the starting point, connect each fine detection area in sequence according to the target detection order, and select the shortest path in the shortest path set as the transition section between two adjacent detection areas, thereby forming a complete closed path, and extract the key position points on the path to form a discrete path point sequence. The path point sequence has largely characterized the overall motion trajectory of the detection head, but the local details and continuity of the path still need to be further optimized. To this end, the central analysis unit will once again use the three-dimensional engine model, and according to the specific internal structure of the engine, perform local fine-tuning, interpolation smoothing and other processing on the path point sequence, and eliminate possible singular points and unreachable points in the sequence to generate an initial local detection path.

[0090] In summary, the adaptive learning industrial endoscope control method provided in the embodiment of the present application has the following technical effects:

[0091] A dual-head industrial endoscope is set up, which includes a detection head and a detection head. The detection head is controlled by a detection control unit, and the detection head is controlled by a detection control unit. The detection control unit and the detection control unit are respectively connected to the central analysis unit in communication. The dual-head design realizes the division of labor and cooperation in detection and inspection, thereby improving the inspection efficiency. The three-dimensional engine model of the target engine is obtained, and the three-dimensional engine model is transmitted to the central analysis unit. The global detection path and adaptive detection parameters of the detection head are generated based on the three-dimensional engine model. Through the three-dimensional engine model, the internal structure information of the engine is fully grasped. The global detection path and adaptive detection parameters are generated using the model to guide the detection head to perform full coverage and blind spot-free scanning; the global detection path and adaptive detection parameters are sent to the detection control unit to control the detection head to scan inside the target engine, obtain preliminary scanning data, and provide a data basis for subsequent inspections. The preliminary scanning data is transmitted back to the central analysis unit, and the preliminary scanning data is analyzed in combination with the three-dimensional engine model to identify the abnormal indication area inside the target engine. By analyzing the preliminary scanning data, the suspected defect area is preliminarily determined, the inspection range is narrowed, and the inspection efficiency is improved. Generate a fine detection area based on the abnormal indication area, generate a local detection path and adaptive detection parameters of the detection head based on the fine detection area and the three-dimensional engine model, and send them to the detection control unit to generate a local detection path and adaptive detection parameters in a targeted manner, guide the detection head to focus on detecting suspicious areas, and improve the defect detection rate. The detection control unit configures the detection head based on the adaptive detection parameters, and controls the detection head to scan inside the target engine according to the local detection path, obtain fine detection data, perform high-precision scanning in the suspicious area, and obtain detailed information such as the shape and location of the defect. The fine detection data is transmitted back to the central analysis unit, and the fine detection data and preliminary scanning data are fused based on the three-dimensional engine model, and a target detection report of the target engine is generated based on the fusion processing results, so as to comprehensively and accurately evaluate the engine defect status and improve the detection efficiency and accuracy of the industrial endoscope.

[0092] Embodiment 2, based on the same inventive concept as the adaptive learning industrial endoscope control method in the above embodiment, Figure 2 As shown, the embodiment of the present application provides an adaptive learning industrial endoscope control system, the system comprising:

[0093] A dual-head setting module 11 is used to set a dual-head industrial endoscope, wherein the dual-head industrial endoscope includes a detection head and a detection head, wherein the detection head is controlled by a detection control unit, and the detection head is controlled by a detection control unit, and the detection control unit and the detection control unit are respectively connected to the central analysis unit for communication;

[0094] A detection planning module 12, used for acquiring a three-dimensional engine model of a target engine, transmitting the three-dimensional engine model to the central analysis unit, and generating a global detection path and adaptive detection parameters of the detection head based on the three-dimensional engine model;

[0095] A detection control module 13, used for sending the global detection path and the adaptive detection parameters to the detection control unit, controlling the detection head to scan inside the target engine, and acquiring preliminary scanning data;

[0096] The abnormality identification module 14 is used to transmit the preliminary scanning data back to the central analysis unit, analyze the preliminary scanning data in combination with the three-dimensional engine model, and identify the abnormal indication area inside the target engine;

[0097] A detection planning module 15, for generating a fine detection area based on the abnormal indication area, generating a local detection path and adaptive detection parameters of the detection head according to the fine detection area and the three-dimensional engine model, and sending the local detection path and adaptive detection parameters to the detection control unit;

[0098] A fine inspection control module 16, configured for the detection control unit to configure the detection head based on the adaptive detection parameters, and to control the detection head to scan inside the target engine according to the local detection path to obtain fine detection data;

[0099] The data fusion module 17 is used to transmit the fine detection data back to the central analysis unit, and fuse the fine detection data and the preliminary scanning data based on the three-dimensional engine model, and generate a target detection report of the target engine according to the fusion processing result.

[0100] Furthermore, the detection planning module 12 includes the following execution steps:

[0101] Extracting internal structural features of the target engine from the three-dimensional engine model, and determining a plurality of key detection points of the detection head based on the internal structural features;

[0102] Using a path planning algorithm to generate the global detection path according to the multiple key detection points;

[0103] Based on the three-dimensional engine model and the global detection path, the detection angle, detection distance and detection speed of the detection head executing the global detection path movement are determined to generate the adaptive detection parameters.

[0104] Furthermore, the detection planning module 12 further includes the following execution steps:

[0105] Acquiring fault diagnosis information of the target engine;

[0106] According to the fault diagnosis information, a potential fault area is identified in the three-dimensional engine model, a plurality of key detection points are generated based on the potential fault area and the internal structure characteristics, and a fault correlation between the key detection points and the potential fault area is identified;

[0107] Generating detection priority weights for a plurality of key detection points based on the fault correlation, and generating the global detection path according to the priority weights and the plurality of key detection points;

[0108] The adaptive detection parameters are generated according to the three-dimensional engine model, the global detection path and the detection priority weight.

[0109] Furthermore, the abnormality identification module 14 includes the following execution steps:

[0110] Registering the preliminary scan data with the three-dimensional engine model to obtain registered scan data, and performing feature extraction on the registered scan data to obtain a feature extraction result;

[0111] Comparing the feature extraction result with a normal engine internal feature template, identifying a feature deviation area, and marking the feature deviation area as the abnormality indication area;

[0112] Extracting the characteristic deviation amount of the characteristic deviation area, and determining the abnormality degree of the abnormal indication area according to the characteristic deviation amount;

[0113] The abnormality indication area is marked in the three-dimensional engine model, and an abnormality level is set for the abnormality indication area according to the degree of abnormality.

[0114] Furthermore, the detection planning module 15 includes the following execution steps:

[0115] Acquire historical detection data of engines of the same model as the target engine, and acquire abnormal area correlation information according to the historical detection data;

[0116] Based on the abnormal area association information and the abnormality degree of the abnormal indication area, analyzing the associated position set of the abnormal indication area, and determining the associated indication area in combination with the three-dimensional engine model, and setting the abnormality degree of the associated indication area to 0;

[0117] The associated indication area and the abnormal indication area are aggregated to generate the fine detection area, each fine detection area having a corresponding abnormality degree.

[0118] Furthermore, the detection planning module 15 also includes the following execution steps:

[0119] Prioritize the fine detection areas according to the abnormality degree, and obtain the area priority ranking result;

[0120] Using the three-dimensional engine model, obtaining a set of shortest paths between regions in the fine detection region;

[0121] Generate an initial local detection path based on the regional priority sorting results and the shortest path set;

[0122] In combination with obstacle information in the three-dimensional engine model, the initial local detection path is optimized to obtain the local detection path;

[0123] According to the abnormality degree of the fine detection area, the three-dimensional engine model and the local detection path, the detection angle, detection distance and detection speed of the detection head are determined to generate the adaptive detection parameters.

[0124] Furthermore, the detection planning module 15 also includes the following execution steps:

[0125] Based on the regional priority ranking result, selecting a starting detection area and determining an initial detection order;

[0126] The initial detection order is adjusted according to the spatial relationship between the regions in the fine detection area in the three-dimensional engine model to obtain a target detection order;

[0127] A path point sequence connecting each area is generated through the starting detection area, the target detection sequence and the shortest path set, and an initial local detection path is generated in combination with the three-dimensional engine model.

[0128] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0129] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. An adaptive learning industrial endoscope control method, characterized in that: The method comprises: A double-head industrial endoscope is provided, wherein the double-head industrial endoscope comprises a detection head and a detection head, wherein the detection head is controlled by a detection control unit, and the detection head is controlled by a detection control unit, and the detection control unit and the detection control unit are respectively connected to the central analysis unit for communication; Acquire a three-dimensional engine model of a target engine, transmit the three-dimensional engine model to the central analysis unit, and generate a global detection path and adaptive detection parameters of the detection head based on the three-dimensional engine model; Sending the global detection path and the adaptive detection parameters to the detection control unit, controlling the detection head to scan inside the target engine, and acquiring preliminary scanning data; The preliminary scanning data is transmitted back to the central analysis unit, and the preliminary scanning data is analyzed in combination with the three-dimensional engine model to identify abnormal indication areas inside the target engine; Generate a fine detection area based on the abnormal indication area, generate a local detection path and adaptive detection parameters of the detection head according to the fine detection area and the three-dimensional engine model, and send them to the detection control unit; The detection control unit configures the detection head based on the adaptive detection parameters, and controls the detection head to scan inside the target engine according to the local detection path to obtain fine detection data; The fine detection data is transmitted back to the central analysis unit, and the fine detection data and the preliminary scanning data are fused based on the three-dimensional engine model, and a target detection report of the target engine is generated according to the fusion processing result; Generating a global detection path and adaptive detection parameters of the detection head based on the three-dimensional engine model includes: Extracting internal structural features of the target engine from the three-dimensional engine model, and determining a plurality of key detection points of the detection head based on the internal structural features; Using a path planning algorithm to generate the global detection path according to the multiple key detection points; Based on the three-dimensional engine model and the global detection path, determining the detection angle, detection distance and detection speed of the detection head executing the global detection path movement, and generating the adaptive detection parameters; The method further comprises: Acquiring fault diagnosis information of the target engine; According to the fault diagnosis information, a potential fault area is identified in the three-dimensional engine model, a plurality of key detection points are generated based on the potential fault area and the internal structure characteristics, and a fault correlation between the key detection points and the potential fault area is identified; Generating detection priority weights for a plurality of key detection points based on the fault correlation, and generating the global detection path according to the priority weights and the plurality of key detection points; The adaptive detection parameters are generated according to the three-dimensional engine model, the global detection path and the detection priority weight.

2. The adaptive learning industrial endoscope control method according to claim 1, characterized in that: Analyzing the preliminary scan data in combination with the three-dimensional engine model to identify abnormal indication areas inside the target engine, including: Registering the preliminary scan data with the three-dimensional engine model to obtain registered scan data, and performing feature extraction on the registered scan data to obtain a feature extraction result; Comparing the feature extraction result with a normal engine internal feature template, identifying a feature deviation area, and marking the feature deviation area as the abnormality indication area; Extracting the characteristic deviation amount of the characteristic deviation area, and determining the abnormality degree of the abnormal indication area according to the characteristic deviation amount; The abnormality indication area is marked in the three-dimensional engine model, and an abnormality level is set for the abnormality indication area according to the degree of abnormality.

3. The adaptive learning industrial endoscope control method according to claim 2, characterized in that: Generating a fine detection area based on the abnormal indication area includes: Acquire historical detection data of engines of the same model as the target engine, and acquire abnormal area correlation information according to the historical detection data; Based on the abnormal area association information and the abnormality degree of the abnormal indication area, analyzing the associated position set of the abnormal indication area, and determining the associated indication area in combination with the three-dimensional engine model, and setting the abnormality degree of the associated indication area to 0; The associated indication area and the abnormal indication area are aggregated to generate the fine detection area, each fine detection area having a corresponding abnormality degree.

4. The adaptive learning industrial endoscope control method according to claim 3 is characterized in that: Generating a local detection path and adaptive detection parameters of the detection head according to the fine detection area and the three-dimensional engine model, including: Prioritize the fine detection areas according to the abnormality degree, and obtain the area priority ranking result; Using the three-dimensional engine model, obtaining a set of shortest paths between regions in the fine detection region; Generate an initial local detection path based on the regional priority sorting results and the shortest path set; In combination with obstacle information in the three-dimensional engine model, the initial local detection path is optimized to obtain the local detection path; According to the abnormality degree of the fine detection area, the three-dimensional engine model and the local detection path, the detection angle, detection distance and detection speed of the detection head are determined to generate the adaptive detection parameters.

5. The adaptive learning industrial endoscope control method according to claim 4, characterized in that: According to the regional priority sorting result and the shortest path set, an initial local detection path is generated, and the initial local detection path is optimized, including: Based on the regional priority ranking result, selecting a starting detection area and determining an initial detection order; The initial detection order is adjusted according to the spatial relationship between the regions in the fine detection area in the three-dimensional engine model to obtain a target detection order; A path point sequence connecting each area is generated through the starting detection area, the target detection sequence and the shortest path set, and an initial local detection path is generated in combination with the three-dimensional engine model.

6. An adaptive learning industrial endoscope control system, characterized in that: A method for controlling an adaptive learning industrial endoscope according to any one of claims 1 to 5, the system comprising: A dual-head setting module, the dual-head setting module is used to set a dual-head industrial endoscope, the dual-head industrial endoscope includes a detection head and a detection head, the detection head is controlled by a detection control unit, the detection head is controlled by a detection control unit, and the detection control unit and the detection control unit are respectively connected to the central analysis unit for communication; a detection planning module, the detection planning module being used to obtain a three-dimensional engine model of a target engine, transmit the three-dimensional engine model to the central analysis unit, and generate a global detection path and adaptive detection parameters of the detection head based on the three-dimensional engine model; A detection control module, the detection control module is used to send the global detection path and the adaptive detection parameters to the detection control unit, control the detection head to scan inside the target engine, and obtain preliminary scanning data; An abnormality identification module, the abnormality identification module is used to transmit the preliminary scanning data back to the central analysis unit, analyze the preliminary scanning data in combination with the three-dimensional engine model, and identify the abnormal indication area inside the target engine; A detection planning module, the detection planning module is used to generate a fine detection area based on the abnormal indication area, generate a local detection path and adaptive detection parameters of the detection head according to the fine detection area and the three-dimensional engine model, and send them to the detection control unit; A fine inspection control module, wherein the fine inspection control module is used for the detection control unit to configure the detection head based on the adaptive detection parameters, and control the detection head to scan inside the target engine according to the local detection path to obtain fine detection data; A data fusion module, wherein the data fusion module is used to transmit the fine detection data back to the central analysis unit, and to fuse the fine detection data and the preliminary scanning data based on the three-dimensional engine model, and to generate a target detection report of the target engine according to the fusion processing result.

Images