A method, device and medium for detecting cracks in a runner blade of a hydraulic turbine

Abstract

The embodiment of the application provides a kind of water turbine runner blade crack detection method, equipment and medium, method includes: by unmanned underwater vehicle obtains underwater real-time detection information;According to underwater real-time detection information, unmanned underwater vehicle is navigated to the target area where water turbine runner blade is located;By unmanned underwater vehicle, obtain the blade image information of target area;According to blade image information, obtain the crack information of water turbine runner blade.By unmanned underwater vehicle, underwater real-time detection information is automatically obtained, and according to underwater real-time detection information, unmanned underwater vehicle is navigated to the crack detection of water turbine runner blade, and the crack detection efficiency of water turbine runner blade is high;And without setting up tail water maintenance platform, emptying pressure steel pipe and tail water pipe, by unmanned underwater vehicle, it can be driven to the crack detection of water turbine runner blade from tail water pipe, and the detection cost is low.

Description

Technical Field

[0001] This invention relates to the field of crack detection technology for turbine runner blades, and particularly to a method, equipment, and medium for crack detection of turbine runner blades. Background Technology

[0002] Frequent start-ups and shutdowns of hydroelectric turbine units in hydropower stations cause drastic changes in pressure pulsations and flow rates, leading to frequent vibrations as the turbines pass through vibration zones during operation. This can result in cracks in the turbine blades. The presence of cracks reduces blade stiffness or increases localized flexibility. If these cracks are not monitored and repaired in a timely manner, they can propagate from surface cracks into penetrating cracks, causing devastating damage to the turbine unit. Therefore, effective detection of cracks in hydroelectric turbine units can reduce the occurrence of accidents and economic losses.

[0003] Currently, the conventional method for inspecting turbine runner blade cracks is to drain the pressure steel pipe and tailrace pipe during unit maintenance, open the tailrace inlet door, set up a tailrace maintenance platform, and have personnel enter the runner chamber to perform non-destructive testing on the runner blades. This method is costly and inefficient. Summary of the Invention

[0004] The following is an overview of the subject matter described in detail herein. This overview is not intended to limit the scope of the claims.

[0005] The main objective of this invention is to provide a method, equipment, and medium for detecting cracks in turbine runner blades, which can improve the efficiency of crack detection and reduce the detection cost.

[0006] In a first aspect, embodiments of the present invention provide a method for detecting cracks in turbine runner blades, including: The underwater real-time detection information obtained by the unmanned underwater vehicle represents the real-time image information, real-time sonar information, real-time magnetic field information, real-time pose information and real-time hydrological information acquired by the unmanned underwater vehicle in the process from the tailrace pipe to the turbine runner blades. The unmanned underwater vehicle is guided to the target area where the turbine runner blades are located based on the underwater real-time detection information. The unmanned underwater vehicle acquires blade image information of the target area; Crack information of the turbine runner blades is obtained based on the blade image information.

[0007] In some optional embodiments, navigating the unmanned underwater vehicle to the target area where the turbine runner blades are located based on the underwater real-time detection information includes: Obtain a weight table, which indicates the parameter weights corresponding to different spatial locations; The real-time spatial position of the unmanned underwater vehicle is determined based on the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, and the weight table. A first path is generated based on the real-time spatial location and the preset three-dimensional model, wherein the preset three-dimensional model represents the BIM model corresponding to the tailrace pipe and the turbine. Based on the real-time sonar information, the spatial contour information and the first obstacle information of the unmanned underwater vehicle's location are obtained; The second obstacle information is obtained by correcting the first obstacle information based on the real-time image information; The second path is obtained by correcting the first path based on the second obstacle information, the hydrological information, and the spatial contour information; Control the unmanned underwater vehicle to reach the target area along the second path.

[0008] In some optional embodiments, determining the real-time spatial position of the unmanned underwater vehicle based on the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, and the weight table includes: The first position is determined based on the real-time image information and the first corresponding data, wherein the first corresponding data represents the correspondence between different spatial positions and image information; The second position is determined based on the real-time sonar information and the second corresponding data, wherein the second corresponding data represents the correspondence between different spatial positions and sonar information. The third position is determined based on the real-time pose information and the third corresponding data, wherein the third corresponding data represents the correspondence between different spatial positions and pose information. The fourth position is determined based on the real-time magnetic field information and the fourth corresponding data, wherein the fourth corresponding data represents the correspondence between different spatial positions and magnetic field information. The first weight coefficient corresponding to the first position, the second weight coefficient corresponding to the second position, the third weight coefficient corresponding to the third position, and the fourth weight coefficient corresponding to the fourth position are obtained according to the weight table. The real-time spatial position is determined based on the first position, the first weight coefficient, the second position, the second weight coefficient, the third position, the third weight coefficient, the fourth position, and the fourth weight coefficient.

[0009] In some optional embodiments, the weight calculation method indicated by the weight table includes: The relative distance between the unmanned underwater vehicle and the rotor is determined based on the first position, the second position, the third position, and the fourth position. The distance factor corresponding to the Nth type of information is determined according to the relative distance and the preset distance parameter table. The preset distance parameter table indicates the correspondence between different relative distances and the distance factor. The Nth type of information represents any one of the real-time image information, the real-time sonar information, the real-time pose information, and the real-time magnetic field information. The information quality factor is determined based on the signal-to-noise ratio of the Nth type of information and the similarity between the Nth type of information and the corresponding Mth data at the same location. The corresponding Mth data represents the correspondence between different spatial locations and the Nth type of information. Environmental factors are determined based on underwater visibility and electromagnetic interference intensity at the location of the unmanned underwater vehicle. The stage calculation weights of the distance factor, the information quality factor, and the environmental factor are obtained based on the navigation stage of the unmanned underwater vehicle. The first weight coefficient, the second weight coefficient, the third weight coefficient, or the fourth weight coefficient corresponding to the Nth type of information are calculated based on the distance factor, the information quality factor, the environmental factor, and the stage calculation weight.

[0010] In some optional embodiments, the generation of the second path further includes: In the case of historical detection records for turbine runner blades, the historical path and historical obstacle information corresponding to the historical detection records are obtained. The historical path represents the travel path of the unmanned underwater vehicle when detecting turbine runner blades, and the historical obstacle information represents the obstacles on the historical path. The first path is determined based on the historical path, historical obstacle information, and the first information obtained by the unmanned underwater vehicle; In the absence of historical inspection records for the turbine runner blades, obstacle map information is obtained from the BIM model. The obstacle map information indicates the location, type, and size of obstacles in each area of ​​the BIM model. The second path is determined based on the obstacle map information and the second information obtained by the unmanned underwater vehicle. The first information and the second information both include the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, the real-time hydrological information, and the weight table.

[0011] In some optional embodiments, determining the first path based on the historical path, historical obstacle information, and the first information obtained by the unmanned underwater vehicle includes: Based on the first information, determine the real-time obstacle information and the real-time hydrological information acquired by the unmanned underwater vehicle at various locations; The obstacle difference information is obtained by comparing the historical obstacle information with the real-time obstacle information; If the historical obstacle that was avoided on the first segment of the historical path, as represented by the obstacle difference information, is eliminated, the first segment is changed to a straight segment, which passes through the area where the eliminated historical obstacle was located. If an obstacle is added to the second segment of the historical path as represented by the obstacle difference information, the minimum detour segment at both ends of the second segment is determined according to the type and size of the added obstacle, and the second segment is changed to the minimum detour segment. The second path is obtained by correcting the historical path based on the water flow velocity and direction represented by the real-time hydrological information.

[0012] In some optional embodiments, acquiring leaf image information of the target area via the unmanned underwater vehicle includes: The unmanned underwater vehicle is controlled to move along the lower ring water exit edge of the Kth blade to the upper crown water exit edge, and the first image information of the Kth blade is captured during the movement, where K represents a positive integer greater than or equal to 1; The unmanned underwater vehicle is controlled to move along the upper crown exit edge of the K+1th blade to the lower ring exit edge, and during the movement, a second image of the K+1th blade is captured. The blade image information includes the first image information and the second image information.

[0013] In some optional embodiments, obtaining the crack information of the turbine runner blades based on the blade image information includes: The preprocessed image information is obtained by filtering and denoising the leaf image information; After performing feature point extraction, camera pose calculation, and feature point coordinate transformation on the preprocessed image information, sparse point cloud data is obtained. The sparse point cloud data is processed by a dense matching algorithm to generate dense point cloud data. A three-dimensional blade model is obtained by constructing a three-dimensional mesh and mapping a texture based on the dense point cloud data. The crack information is extracted from the three-dimensional blade model.

[0014] In a second aspect, embodiments of the present invention provide a crack detection device for turbine runner blades, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the crack detection method for turbine runner blades described in the first aspect.

[0015] Thirdly, embodiments of the present invention provide a computer storage medium storing computer-executable instructions, which are used to execute the crack detection method for turbine runner blades described in the first aspect.

[0016] The beneficial effects of this invention include: acquiring real-time underwater detection information via an unmanned underwater vehicle (UUV), wherein the real-time underwater detection information represents real-time image information, real-time sonar information, real-time magnetic field information, real-time attitude information, and real-time hydrological information acquired by the UUV during its journey from the tailrace pipe to the turbine runner blades; navigating the UUV to the target area where the turbine runner blades are located based on the real-time underwater detection information; acquiring blade image information of the target area via the UUV; and acquiring crack information of the turbine runner blades based on the blade image information. In this embodiment, the UUV automatically acquires real-time underwater detection information and navigates to the turbine runner blades for crack detection based on this information, resulting in high efficiency in crack detection of the turbine runner blades; and eliminating the need for a tailrace maintenance platform, venting pressure steel pipe, and tailrace pipe, allowing the UUV to travel from the tailrace pipe to the turbine runner blades for crack detection, thus reducing detection costs.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a system platform architecture for performing a crack detection method for turbine runner blades according to an embodiment of the present invention; Figure 2 This is a flowchart of a method for detecting cracks in turbine runner blades according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the generation of a first curve path provided in an embodiment of the present invention.

[0019] Figure label: The system platform architecture is 1000, the processor is 1100, and the memory is 1200. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0021] It should be noted that although functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, or the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0022] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0023] like Figure 1 As shown, Figure 1 This is a schematic diagram of a system platform architecture for performing a crack detection method for turbine runner blades according to an embodiment of the present invention.

[0024] exist Figure 1 In the example, the system platform architecture 1000 includes a processor 1100 and a memory 1200, which can be connected via a bus or other means. Figure 1 Taking the example of a connection between China and Israel via a bus.

[0025] Memory 1200, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory 1200 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 1200 may optionally include memory remotely located relative to processor 1100, and these remote memories can be connected to the solid-state device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0026] Those skilled in the art will understand that the system platform architecture 1000 can be applied to 5G communication network systems and subsequent evolved mobile communication network systems, etc., and this embodiment does not specifically limit it.

[0027] It will be understood by those skilled in the art that Figure 1 The system platform architecture 1000 shown does not constitute a limitation on the embodiments of the present invention. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0028] like Figure 2 As shown, this embodiment of the invention provides a method for detecting cracks in turbine runner blades, including steps S100, S200, S300, and S400.

[0029] Step S100: Obtain real-time underwater detection information through the unmanned underwater vehicle. The real-time underwater detection information represents the real-time image information, real-time sonar information, real-time magnetic field information, real-time pose information and real-time hydrological information obtained by the unmanned underwater vehicle in real time during the process from the tailrace pipe to the turbine runner blades. Step S200: Navigate the unmanned underwater vehicle to the target area where the turbine runner blades are located based on the underwater real-time detection information; Step S300: Obtain blade image information of the target area using the unmanned underwater vehicle; Step S400: Obtain the crack information of the turbine runner blades based on the blade image information.

[0030] Specifically, the unmanned underwater vehicle of this application is equipped with sonar equipment, video shooting device, and hydrological information acquisition device (temperature sensor, water flow sensor, and salinity sensor, etc.). The sonar equipment can acquire surrounding sonar information in real time, thereby generating corresponding sonar information; the video shooting device can acquire surrounding video information in real time, and then identify obstacles and the tailrace pipe wall based on the video information and sonar information.

[0031] The underwater real-time detection information collected by the unmanned underwater vehicle (UUV) consists of multi-dimensional environmental and self-state information continuously and synchronously acquired by the UUV as it travels from the tailrace pipe to the turbine runner blades. Specifically, it includes real-time image information for visual capture of the detection scene, real-time sonar information for detecting the surrounding space and obstacles, real-time magnetic field information for sensing the magnetic field characteristics of the detection area, real-time pose information reflecting its own three-dimensional position and attitude, and real-time hydrological information characterizing the water flow velocity and direction in the detection area.

[0032] Based on the aforementioned multi-dimensional underwater real-time detection information, the unmanned underwater vehicle's (UUV) travel path, attitude, and motion state are adjusted in real time through a multi-source information fusion navigation algorithm, guiding the UUV to navigate precisely and without collision to the target detection area where the turbine runner blades are located.

[0033] After the unmanned underwater vehicle arrives at the target area, it completes multi-angle, high-overlap, and high-definition visual acquisition of the turbine runner blades, obtaining blade image information covering key areas of the blade surface.

[0034] The collected blade image information is preprocessed and feature analyzed. Through image modeling, point cloud generation, feature extraction and quantitative analysis, crack-related information such as the location, length, width, depth, morphology and distribution scale of cracks on the turbine runner blades are extracted from the blade image information.

[0035] In some optional embodiments, navigating the unmanned underwater vehicle to the target area where the turbine runner blades are located based on the underwater real-time detection information includes: S210. Obtain a weight table, wherein the weight table is used to indicate the parameter weights corresponding to different spatial locations; Specifically, a pre-generated weight table is retrieved. This table is based on the different spatial positions along the unmanned underwater vehicle's path from the tailrace to the turbine runner blades, clearly defining the navigation parameter weights corresponding to real-time image information, real-time sonar information, real-time pose information, and real-time magnetic field information at each spatial position. The specific navigation parameter weights are also related to the information quality of the real-time image information, real-time sonar information, real-time pose information, and real-time magnetic field information, as well as the surrounding environment.

[0036] S220. Determine the real-time spatial position of the unmanned underwater vehicle based on the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, and the weight table. Specifically, real-time image information, real-time sonar information, real-time pose information, and real-time magnetic field information are substituted into the weight table, and the real-time spatial position of the unmanned underwater vehicle in the tailrace-turbine spatial coordinate system is determined through multi-source information fusion calculation.

[0037] S230. Generate a first path based on the real-time spatial location and the preset three-dimensional model, wherein the preset three-dimensional model represents the BIM model corresponding to the tailrace pipe and the turbine. Specifically, a pre-set BIM 3D model representing the overall spatial structure of the tailrace pipe and turbine is retrieved. Combined with the real-time spatial location, the initial travel path of the unmanned underwater vehicle from its current position to the target area of ​​the turbine runner blades is planned in the model, which is denoted as the first path.

[0038] S240. Obtain the spatial contour information and first obstacle information of the location of the unmanned underwater vehicle based on the real-time sonar information; Specifically, based on real-time sonar information (forward-looking sonar + side-scan sonar collaborative detection data), the surrounding spatial contour information and first obstacle information of the current location of the unmanned underwater vehicle are extracted. The spatial contour information mainly includes the three-dimensional contour, cross-sectional dimensions, and spatial orientation of the tailrace pipe wall in the current area, as well as the preliminary contour features of the outer perimeter of the rotor, used to verify the adaptability of the path to the actual space. The first obstacle information is obtained by identifying abnormal signals in the sonar echo (areas where echo intensity or distance changes exceed the normal range), filtering out potential obstacles within the path range, including temporary obstacles not marked in the BIM model (such as suspended debris, blade fragments), and suspected obstacles detected by sonar. Simultaneously, the preliminary three-dimensional coordinates, size range, and real-time distance to the unmanned underwater vehicle of each obstacle are recorded.

[0039] S250. After correcting the first obstacle information based on the real-time image information, the second obstacle information is obtained; Specifically, the first obstacle information is verified, supplemented, and corrected by combining real-time image information to obtain the second obstacle information. The specific correction process is as follows: using pre-processed real-time images (high definition, no reflection, no fog), obstacles in the image are identified by target detection algorithms (such as the YOLO algorithm), and compared one by one with various obstacles in the first obstacle information. False obstacles generated by sonar detection (such as sonar echo anomalies caused by bubbles or suspended particles) are eliminated, while obstacle information in sonar blind spots (such as tailpipe corners, blade gaps, and other areas not covered by sonar) is supplemented. For suspected obstacles detected by sonar, their authenticity is verified by image visual features (obstacle shape, texture, color). If confirmed as obstacles, their three-dimensional coordinates and size parameters are corrected (the size deviation of sonar detection is calibrated by image visual measurement). Small obstacles (such as small debris) identified in the image but not detected by sonar are added to the obstacle information, ultimately forming accurate and complete second obstacle information, providing reliable obstacle avoidance constraints for path correction.

[0040] S260. The second path is obtained by correcting the first path based on the second obstacle information, the hydrological information, and the spatial contour information. Specifically, using the information of the second obstacle as an obstacle avoidance constraint, combined with real-time hydrological information (water flow velocity, flow direction vector data) and extracted spatial contour information, the first path is dynamically adjusted and optimized to obtain the corrected path (the second path). The correction process mainly includes three dimensions: First, obstacle avoidance correction. Based on the second obstacle information, the road segments in the first path that are less than the safety threshold (e.g., 5cm) from the obstacle are replanned. By adjusting the path turning points and detouring, the path is ensured to avoid all obstacles throughout, while avoiding excessively long detours. Second, flow field adaptation correction. Combining the water flow velocity and direction in real-time hydrological information, the path's travel direction and velocity parameters are adjusted. For example, in high-velocity disturbance areas, the path direction is adjusted to reduce flow field resistance. In low-velocity stable areas, the path smoothness is optimized, and flow field disturbance compensation parameters are added to ensure that the unmanned underwater vehicle can still travel stably along the path under water flow disturbances. Third, spatial adaptation correction. Combining the spatial contour information of the current area, the adaptability of the path to the actual spatial boundary of the tailpipe is verified to avoid the path exceeding the tailpipe range or being too close to the wall, ensuring that the unmanned underwater vehicle has sufficient movement space, while adapting to the curved surface features of the outer edge of the rotor.

[0041] S270, Control the unmanned underwater vehicle to reach the target area along the second path.

[0042] Specifically, the parameters of the second path (three-dimensional coordinates of path inflection points, travel speed of each segment, and attitude requirements) are imported into the propulsion and attitude control system of the unmanned underwater vehicle (UUV). Combined with real-time pose information and real-time hydrological information, control commands (speed, direction, and attitude adjustment parameters) are issued to the thrusters in real time through Model Predictive Control (MPC) + Sliding Mode Variable Structure Control (SMC) algorithms. During travel, the actual travel position and attitude of the UUV are collected in real time and compared with the preset parameters of the second path. If a deviation is detected (e.g., position deviation ≥2mm, attitude deviation ≥0.5°), the control commands are immediately dynamically adjusted to correct the travel direction and attitude. At the same time, real-time sonar information and real-time image information are monitored. If a new obstacle is found within the path range, a temporary path adjustment is immediately triggered to ensure that there is no collision during the travel process. The navigation process continues until the UUV reaches the target area where the turbine runner blades are located, triggering a position lock command to complete the navigation process.

[0043] In some optional embodiments, determining the real-time spatial position of the unmanned underwater vehicle based on the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, and the weight table includes: S221. Determine a first position based on the real-time image information and the first corresponding data, wherein the first corresponding data represents the correspondence between different spatial positions and image information; Specifically, the first corresponding data is a pre-generated dataset representing the one-to-one correspondence between different spatial locations from the tailrace pipe to the turbine area and image visual features. It includes a typical visual feature library for each spatial location (such as tailrace pipe wall joints, runner crown / lower ring contours, blade numbers and outlet edge feature points, spillway cone morphology, etc.), visual positioning calculation parameters, and a mapping model between image features and spatial coordinates. It is constructed based on the association between visual acquisition data and precise positioning data from the early mapping stage. First, underwater-specific preprocessing is performed on the real-time image information to complete defogging, de-reflection, noise suppression, and feature enhancement. Using the ORB / SIFT hybrid feature extraction algorithm, effective visual feature points and feature contours are extracted from the preprocessed image. Then, the extracted visual features are precisely matched with the feature library in the first corresponding data. Combined with the visual positioning calculation parameters and mapping model in the dataset, the three-dimensional spatial position and attitude parameters of the unmanned underwater vehicle based on the real-time image information are obtained through visual odometry (VIO), which is denoted as the first position.

[0044] S222. Determine the second position based on the real-time sonar information and the second corresponding data, wherein the second corresponding data represents the correspondence between different spatial positions and sonar information; Specifically, the second corresponding data is a pre-generated dataset representing a one-to-one correspondence between different spatial locations from the tailrace pipe to the turbine area and sonar detection features. It includes sonar point cloud contours, echo intensity features, sonar odometry parameters, and a mapping model between sonar features and spatial coordinates for each spatial location. This dataset is constructed by linking sonar detection data from the initial mapping phase with precise positioning data. First, the real-time sonar information is preprocessed to perform reverberation suppression, noise removal, point cloud registration, and echo feature extraction. The spatial point cloud contours and target echo features of the current area are extracted from the sonar detection data. Then, the extracted sonar features are matched with the feature library in the second corresponding data. Combined with the sonar odometry parameters and mapping model in the dataset, the three-dimensional spatial position and attitude parameters of the unmanned underwater vehicle based on real-time sonar information are calculated and denoted as the second position. This position represents the sonar-dimensional positioning result and can achieve effective coarse positioning in areas lacking visual features (such as the large space at the tailrace pipe inlet).

[0045] S223. Determine the third position based on the real-time pose information and the third corresponding data, wherein the third corresponding data represents the correspondence between different spatial positions and pose information; Specifically, the third corresponding data is a pre-generated dataset representing the one-to-one correspondence between different spatial positions and attitude motion parameters from the tailrace pipe to the turbine area. It includes a dead reckoning (DR) kinematic model, IMU (Inertial Measurement Unit) error compensation parameters, wheel speedometer slippage correction coefficients, and correlation models between attitude parameters (velocity, angle) and spatial coordinates at different spatial positions. This dataset is constructed by combining the kinematic characteristics of the unmanned underwater vehicle (UUV) with previous trajectory test data. First, the real-time attitude information (angular velocity and acceleration collected by the IMU, and travel speed collected by the wheel speedometer, etc.) is preprocessed to complete zero-drift correction, filtering, smoothing, and error compensation. Then, the preprocessed attitude parameters are substituted into the DR kinematic model and correlation model in the third corresponding data. Based on the UUV's precise spatial position at the previous moment, the current three-dimensional spatial position and attitude parameters based on the real-time attitude information are calculated and denoted as the third position. This position serves as the basic positioning result for the attitude dimension, enabling continuous positioning throughout the entire journey and providing a benchmark for positioning in other dimensions.

[0046] S224. Determine the fourth position based on the real-time magnetic field information and the fourth corresponding data, wherein the fourth corresponding data represents the correspondence between different spatial positions and magnetic field information; Specifically, the fourth corresponding data is a pre-generated dataset representing the one-to-one correspondence between different spatial locations and magnetic field characteristics from the tailrace pipe to the turbine area. This is a high-precision 3D magnetic feature map of the area, containing magnetic flux, magnetic gradient values, magnetic fingerprint feature vectors, and precise mapping relationships between magnetic field characteristics and spatial coordinates at each spatial location. This is constructed by accurately collecting data from magnetic sensors and associating it with spatial coordinates. First, the real-time magnetic field information is preprocessed to perform electromagnetic interference deduction, temperature drift compensation, magnetic gradient extraction, and magnetic feature vector construction for the UUV (Unmanned Underwater Vehicle). Then, the extracted real-time magnetic feature vectors are precisely matched with the magnetic feature map in the fourth corresponding data (e.g., using an improved ICP algorithm or NCC template matching). After matching the optimal magnetic feature unit, its bound spatial coordinates and attitude reference values ​​are retrieved to obtain the 3D spatial position and attitude parameters of the UUV based on the real-time magnetic field information, denoted as the fourth position. This position represents the positioning result in the magnetic field dimension, achieving millimeter-level precision positioning around the runner's metal components (near the blade tip, upper crown / lower ring area), effectively correcting the cumulative errors of other positioning methods.

[0047] S225. Obtain the first weight coefficient corresponding to the first position, the second weight coefficient corresponding to the second position, the third weight coefficient corresponding to the third position, and the fourth weight coefficient corresponding to the fourth position according to the weight table; Specifically, based on the calculated first, second, third, and fourth positions, the current spatial region of the unmanned underwater vehicle (UUV) is comprehensively determined (such as the tailpipe inlet, elbow section, outer edge of the rotor, near end of the blades, etc.). Then, according to the calculation formula indicated by the pre-generated weight table, the weight coefficients of the four types of positioning results corresponding to this spatial region are calculated (or the weight coefficients of the UUV under the corresponding spatial position, information quality, and environmental characteristics are retrieved from the weight table), namely, the first weight coefficient corresponding to the first position (image positioning), the second weight coefficient corresponding to the second position (sonar positioning), the third weight coefficient corresponding to the third position (pose positioning), and the fourth weight coefficient corresponding to the fourth position (magnetic field positioning). It should be noted that the values ​​of the four weighting coefficients are based on the environmental characteristics of the current area, and the sum of the values ​​of the four weighting coefficients is 1. For example, in the near-end area of ​​the blade, the first and fourth weighting coefficients are relatively high, while the second weighting coefficient is relatively low, giving full play to the precise positioning advantages of images and magnetic fields; in the tailrace pipe inlet area, the second and third weighting coefficients are relatively high, while the first and fourth weighting coefficients are relatively low, relying on sonar and pose to achieve effective global coarse positioning.

[0048] S226. Determine the real-time spatial position based on the first position, the first weight coefficient, the second position, the second weight coefficient, the third position, the third weight coefficient, the fourth position, and the fourth weight coefficient.

[0049] Specifically, the positioning results of the first, second, third, and fourth positions are uniformly transformed to the same spatial coordinate system to ensure that the three-dimensional coordinates (x, y, z) of each position are consistent with the attitude parameters (roll). , looking up ,yaw By using the same coordinate reference, positioning deviations caused by coordinate system differences are eliminated. Then, for each position data in the global coordinate system, weighted fusion calculations are performed on the three-dimensional coordinate components and attitude parameter components separately. Specifically, the three-dimensional coordinate components corresponding to the first position are multiplied by the first weight coefficient, the three-dimensional coordinate components corresponding to the second position by the second weight coefficient, the three-dimensional coordinate components corresponding to the third position by the third weight coefficient, and the three-dimensional coordinate components corresponding to the fourth position by the fourth weight coefficient, and then superimposed to obtain the real-time three-dimensional coordinates. Similarly, the attitude parameter components corresponding to the first position are multiplied by the first weight coefficient, the second position by the second weight coefficient, the third position by the third weight coefficient, and the fourth position by the fourth weight coefficient, and then superimposed to obtain the real-time attitude parameters. Integrating the real-time three-dimensional coordinates and real-time attitude parameters yields the final real-time spatial position of the unmanned underwater vehicle. By leveraging the positioning advantages of fusing detection information from various dimensions, the positioning error of a single piece of information is effectively reduced, thereby achieving precise navigation of the unmanned underwater vehicle.

[0050] In some optional embodiments, the weight calculation method indicated by the weight table includes: S2251. Determine the relative distance between the unmanned underwater vehicle and the rotor based on the first position, the second position, the third position, and the fourth position; Specifically, first extract the first position. Second position Third position Fourth position The three-dimensional coordinates in the global coordinate system are then extracted; then the preset wheel reference points in the BIM model are extracted. The three-dimensional coordinates are then calculated; subsequently, the Euclidean distances from the four positions to the wheel reference point are calculated respectively. , , , The formula is:

[0051] in =1,2,3,4.

[0052] The weighted average of the four Euclidean distances is taken as the final relative distance. (The weights can be preset to equal weights, or a basic weight can be assigned to each location based on the accuracy of the initial mapping.) .

[0053] S2252. Determine the distance factor corresponding to the Nth type of information according to the relative distance and the preset distance parameter table, wherein the preset distance parameter table indicates the correspondence between different relative distances and the distance factor, and the Nth type of information represents any one of the real-time image information, the real-time sonar information, the real-time pose information, and the real-time magnetic field information; Specifically, the preset distance parameter table is a table of one-to-one correspondence between relative distance intervals and distance factors, calibrated in the early stage based on the navigation characteristics of the entire area from the tailrace pipe to the runner. It is formulated according to the variation law of the positioning effectiveness of various information with the relative distance to the runner, and the distance factors... The larger the value, the higher the positioning reference value of this type of information at the current distance.

[0054] Distance ranges are defined by combining the positioning characteristics of various types of information. For example, the closer the real-time image / magnetic field information is to the wheel, the higher the positioning effectiveness and the closer the distance factor is to 1; the farther the distance, the lower the positioning effectiveness and the closer the distance factor is to 0. The farther the real-time sonar / pose information is to the wheel, the higher the positioning effectiveness and the closer the distance factor is to 1; the closer the distance, the lower the positioning effectiveness and the moderately lower the distance factor. Based on the calculated relative distance Match the corresponding distance interval in the preset distance parameter table, and directly retrieve the distance factor for the Nth type of information within that interval. This completes the quantification of the effectiveness of information positioning based on the distance dimension.

[0055] S2253. Determine the information quality factor based on the signal-to-noise ratio of the Nth type of information and the similarity between the Nth type of information and the corresponding Mth data at the same position, wherein the corresponding Mth data represents the correspondence between different spatial positions and the Nth type of information; Specifically, information quality factor This factor is used to quantify the data quality of the Nth type of real-time detection information and its matching degree with pre-built corresponding data. A higher value indicates higher information quality, better matching with preset features, and higher reliability of the localization result. This factor is derived from the signal-to-noise ratio normalized value. and similarity normalized value The weighted fusion is obtained, and the weight coefficients of the two types of indicators satisfy... (It can be pre-calibrated according to the information type, such as the partial signal-to-noise ratio value of sonar / magnetic field information, and the partial similarity value of image information), the calculation formula is:

[0056] in, The calculation is as follows: For the raw data of the Nth type of real-time detection information, calculate the ratio of effective signal strength to noise signal strength (signal-to-noise ratio SNR), and then normalize it to the [0,1] interval to obtain the result. .

[0057] The Nth type of real-time detection information is matched with the pre-built Mth corresponding data (the dataset corresponding to the spatial location features of this type of information) at the current location, and the matching similarity is calculated and normalized to the [0,1] interval. .

[0058] S2254. Determine environmental factors based on the underwater visibility and electromagnetic interference intensity at the location of the unmanned underwater vehicle; Specifically, environmental factors This quantifies the degree of interference from the underwater environment at the current location of the unmanned underwater vehicle (UUV) on the acquisition and positioning of type N information. A higher value indicates less environmental interference and higher positioning effectiveness for this type of information. Environmental factors are normalized values ​​based on underwater clarity. and electromagnetic interference intensity normalized value It consists of two dimensions, and differentiated weights are assigned to the two dimensions based on the environmental sensitivity of the Nth type of information. , (satisfy + =1), the calculation formula is:

[0059] Among these methods, the underwater clarity of the current area is detected in real time by the image acquisition module of the unmanned underwater vehicle (e.g., using image edge gradient method or gray-level variance method), and the detection results are compared with a preset clarity threshold range. , Normalized to [0,1], a larger value indicates clearer underwater conditions and less interference with visual detection information. The electromagnetic interference intensity of the current area is detected in real time by the unmanned underwater vehicle's magnetic field detection module (e.g., detecting fluctuations in the background magnetic field), and the detection results are compared with a preset electromagnetic interference threshold range. , Normalized to [0,1], a larger value indicates stronger electromagnetic interference and greater interference with magnetic field and electronic detection information; therefore, in the calculation, (1) is taken as [0,1]. ).

[0060] Weight , Based on the differentiated allocation of information according to its environmental sensitivity characteristics, for example: real-time image information is highly sensitive to underwater clarity but not sensitive to electromagnetic interference, therefore... , Real-time magnetic field information is highly sensitive to electromagnetic interference but not to underwater clarity, therefore... , Real-time sonar information is not sensitive to either of them, therefore , Real-time pose information is least affected by environmental interference, therefore , .

[0061] S2255. Obtain the stage calculation weights of the distance factor, the information quality factor, and the environmental factor based on the navigation stage of the unmanned underwater vehicle; Specifically, the stage calculation weights are used for the unmanned underwater vehicle in different navigation stages, and are represented by distance factors. Information quality factor Environmental factors The assigned fusion calculation weights, the three weights satisfy... Based on the core positioning requirements for different navigation stages, the navigation stages are divided into the tailrace inlet section (stage 1), the runner peripheral section (stage 2), and the blade proximal section (stage 3). The emphasis on these three factors varies in each stage. For example: Phase 1 (Tailpipe Inlet) ≥3m): Based on global coarse positioning, the effectiveness of information positioning is mainly determined by the distance to the rotating wheel, hence the distance factor weight. Maximum (e.g., 0.5), quality factor Secondly (e.g., 0.3), environmental factors Minimum (e.g., 0.2).

[0062] Phase 2 (outer perimeter of the rotating wheel, 1m < <3m: With transitional fine positioning as the core, distance, mass, and environment are all important factors, so the weights of the three factors are evenly distributed (e.g., 0.33 for each).

[0063] Stage 3 (proximal end of blade, ≤1m): Based on close-fitting micro-positioning, the quality of the information itself is the key to accurate positioning; therefore, the quality factor weighting is crucial. Maximum (e.g., 0.5), distance factor Secondly (e.g., 0.3), environmental factors Minimum (e.g., 0.2).

[0064] Based on the current navigation phase of the unmanned underwater vehicle, the pre-calibrated phase weights are directly retrieved for calculation. , , .

[0065] S2256. Calculate the first weight coefficient, the second weight coefficient, the third weight coefficient, or the fourth weight coefficient corresponding to the Nth type of information based on the distance factor, the information quality factor, the environmental factor, and the stage calculation weight.

[0066] Specifically, the distance factor obtained by combining the above steps Information quality factor Environmental factors And the stage calculation weight of the current navigation stage. , , The initial weight coefficients of the Nth type of information are obtained by weighted summation, and then the initial weight coefficients of the four types of information are normalized to obtain the final weight coefficients of the Nth type of information. (i.e., the first / second / third / fourth weighting coefficients), the calculation formula is: Initial weighting coefficients for the Nth type of information :

[0067] Calculate the real-time image information respectively Real-time sonar information Real-time pose information Real-time magnetic field information After determining the initial weight coefficients, they are converted into final weight coefficients using a normalization formula, ensuring that the sum of the four types of coefficients is 1. The final weight coefficients are then... :

[0068] in, represents the final weighting coefficient for the Nth type of information.

[0069] In some optional embodiments, the generation of the second path further includes: S261. In the case of historical detection records for turbine runner blades, obtain the historical path and historical obstacle information corresponding to the historical detection records. The historical path represents the travel path of the unmanned underwater vehicle when detecting turbine runner blades, and the historical obstacle information represents the obstacles on the historical path. S262. Determine the first path based on the historical path, historical obstacle information, and the first information obtained by the unmanned underwater vehicle; Specifically, when the target turbine runner blades have undergone underwater inspection and have retained valid historical inspection records, the path and obstacle data accumulated from historical inspections are used as a priority to perform initial path planning, combined with the first information collected in real time by the unmanned underwater vehicle. This determines the first path, making full use of the scenario adaptability of historical data, reducing path exploration, and improving planning efficiency.

[0070] Historical inspection records corresponding to the target runner blades are retrieved from the turbine equipment inspection database. Historical path and obstacle information are extracted, and data preprocessing is performed: coordinate system calibration is performed on the historical paths, and the spatial coordinates of the historical paths are uniformly transformed to the spatial coordinate system used in the current navigation, eliminating path deviations caused by differences in coordinate system calibration between different inspection cycles. At the same time, core parameters of the historical paths are extracted, including the three-dimensional coordinates of path inflection points, the travel speed of each segment, the safe distance from the blade / tailrace tube wall, and the navigation stage division of the historical inspection. The validity of historical obstacle information is verified and features are extracted. Historical obstacles that have been eliminated through maintenance are removed, and historical obstacles that may still exist are marked (such as tailrace tube fixed support components, permanent structural protrusions of runner blades, etc.). The three-dimensional coordinates, size, type, and path segment of the valid historical obstacles are extracted to form a standardized historical obstacle feature set.

[0071] Based on the calibrated historical paths, and integrating various real-time detection data and weight tables contained in the first information, the historical paths are adjusted in real time to ultimately determine the first path that conforms to the current detection scenario. The specific adjustment logic is as follows: Based on the weight table in the first information and various real-time information, the current real-time spatial position of the unmanned underwater vehicle is calculated according to the above method. The historical path is then accurately matched with the real-time spatial position to ensure the continuity between the path starting point and the current position of the unmanned underwater vehicle.

[0072] By integrating real-time sonar and image information from the first information, the historical obstacle feature set is verified on-site. If the real-time detection data shows that the historical obstacle still exists, the corresponding obstacle avoidance plan in the historical path is retained; if the historical obstacle has disappeared, the corresponding obstacle avoidance section is deleted and the historical path is smoothed and optimized; if the real-time detection finds new temporary obstacles within the historical path range, obstacle avoidance space is reserved on the basis of the historical path first, and precise obstacle avoidance correction is made in subsequent steps.

[0073] Combining the real-time hydrological information in the first information, and based on the water flow velocity and direction distribution in the current detection scenario, the travel speed and path of the historical path are adjusted to adapt the flow field. For example, in the current high-velocity disturbance area, the direction of the historical path is appropriately adjusted to reduce flow field resistance, and in the low-velocity area, the smoothness of the historical path is maintained.

[0074] Based on the planning logic of historical paths, starting from real-time spatial location, with flow field adaptation as a constraint, and historical and real-time obstacles as obstacle avoidance references, the first path adapted to the current detection scenario is obtained after integration and optimization. Subsequently, spatial contour extraction, obstacle correction, and hydrological and spatial contour adaptation correction are completed according to the predetermined steps to finally obtain the second path.

[0075] S263. In the absence of historical inspection records for the turbine runner blades, obtain obstacle map information on the BIM model. The obstacle map information indicates the location, type, and size of obstacles in each area of ​​the BIM model. S264. Determine the second path based on the obstacle map information and the second information obtained by the unmanned underwater vehicle. The first information and the second information both include the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, the real-time hydrological information, and the weight table.

[0076] Specifically, when there are no historical inspection records available for the target turbine runner blades, the BIM model representing the overall spatial structure of the tailrace pipe and turbine is used as the basis. The obstacle map information built into the model is retrieved, and combined with the second information collected in real time by the unmanned underwater vehicle, path planning is performed to determine a second path suitable for the current scenario, ensuring the rationality and accuracy of the path planning. The specific steps are as follows: Based on the obstacle map information from the BIM model, retrieve the location, type, and size of obstacles in each area from the tailrace pipe to the runner blades: Obstacle locations: Using the draft tube-turbine global spatial coordinate system as a reference, mark the three-dimensional coordinates and spatial regions of all obstacles (such as the draft tube bend, the outer perimeter of the runner, the near end of the blades, etc.). Obstacle types: Distinguish between fixed structural obstacles (such as concrete support components for the tailrace pipe, protrusions connecting the upper crown / lower ring of the runner, and spill cones) and potential risk obstacles (such as blade gaps and narrow passages between the runner and the tailrace pipe). Obstacle size: Mark the three-dimensional dimensions and outline range of each obstacle to provide accurate dimensional reference for obstacle avoidance path planning.

[0077] Based on the various real-time detection data and weight tables included in the second information, the current real-time spatial position of the unmanned underwater vehicle (UUV) is calculated using the aforementioned method. Starting from this current real-time spatial position and ending at the detection area of ​​the target rotor blade, an initial path is planned in the BIM model following the principles of shortest path, sufficient safety distance, and adaptation to UUV motion constraints. Based on the obstacle map information in the BIM model, basic obstacle avoidance planning is performed on the initial path. For all fixed structural obstacles within the path range, a safety avoidance distance of ≥5cm is reserved. For potentially risky obstacle areas, the turning radius of the path is appropriately increased and the planned travel speed is reduced to ensure the basic feasibility of the path.

[0078] Combining the real-time sonar information and real-time image information in the second information, the obstacle map information of the BIM model is verified and corrected on-site. Since the BIM model is a theoretical spatial structure, there are deviations from the actual scene. Through real-time sonar spatial contour detection and real-time image visual recognition, the position and size information of obstacles in the map that do not match the actual situation are corrected, temporary obstacles (such as floating debris) that are not marked in the BIM model are supplemented, and the initial path is corrected accordingly for obstacle avoidance.

[0079] By integrating real-time hydrological information from the second information, and based on the real-time water flow velocity and flow direction vector distribution of the current detection scenario, the corrected initial path is dynamically adapted to the flow field. In areas with high flow velocity and complex flow direction, the path direction is adjusted to reduce flow field resistance. Suggested travel speeds adapted to the current water flow are marked on each segment of the path. At the same time, the impact of flow field disturbances on the attitude of the unmanned underwater vehicle is considered to ensure the feasibility of the path.

[0080] After obstacle avoidance planning, real-world correction, and flow field adaptation, the path is finally smoothed and optimized. The path is then checked to see if it meets the motion constraints of the unmanned underwater vehicle (such as minimum turning radius, maximum travel speed, etc.). The optimized path is then the second path.

[0081] In some optional embodiments, determining the first path based on the historical path, historical obstacle information, and the first information obtained by the unmanned underwater vehicle includes: S2621. Determine the real-time obstacle information and real-time hydrological information acquired by the unmanned underwater vehicle at various locations based on the first information; Specifically, the first information includes real-time image information, real-time sonar information, real-time pose information, real-time magnetic field information, real-time hydrological information, and a weight table. Based on the first information, the real-time spatial position calculation of the unmanned underwater vehicle and the extraction of real-time obstacle and hydrological information are completed. Following the aforementioned method for calculating real-time spatial position, the real-time spatial position of the unmanned underwater vehicle (UUV) is calculated, establishing a correlation between the position coordinates and detection data within the global spatial coordinate system. Using the real-time spatial position as a benchmark, real-time sonar information (point cloud contours, echo features) and real-time image information (visual recognition, feature matching) are fused to extract real-time obstacle information within each location range. This includes the obstacle's three-dimensional coordinates, type, actual size, real-time distance to the UUV, and the path segment it belongs to, forming a real-time obstacle information set with location labels. From the real-time hydrological information of the first set of data, the corresponding water flow velocity and direction vector data are extracted for each location. This data is then combined with the real-time spatial position for coordinate binding, forming a real-time hydrological information set with location labels, clarifying the flow field distribution characteristics of each region from the tailrace pipe to the runner blades.

[0082] Based on the segmentation rules of the historical path, real-time obstacle information and real-time hydrological information are divided into segments to ensure that the real-time data corresponds one-to-one with each segment of the historical path.

[0083] S2622. Obtain obstacle difference information by comparing the historical obstacle information with the real-time obstacle information; Specifically, the 3D coordinates and road segments of historical obstacles are precisely matched with those of real-time obstacles to ensure effective comparison of obstacle information within the same area and road segment. Differences are categorized into three types based on the comparison results: 1) Historical obstacle removal (historical obstacles show no detection signal in real-time monitoring, confirming their removal through maintenance or other means); 2) New obstacles (new obstacles with no historical records are detected within the historical path range); 3) Changes in obstacle characteristics (historical obstacles still exist, but their location and size have slightly changed). Each type of difference information is linked to a specific road segment of the historical path, clearly defining the first road segment corresponding to historical obstacle removal, the second road segment corresponding to new obstacles, and the associated road segment corresponding to changes in obstacle characteristics. The characteristics of the differing obstacles are also labeled (e.g., the original location of the removed obstacle, the type / size of the new obstacle), forming standardized obstacle difference information of "road segment + difference type + obstacle characteristic".

[0084] S2623. If the historical obstacle that was avoided on the first road segment of the historical path represented by the obstacle difference information is eliminated, the first road segment is changed to a straight road segment, and the straight road segment passes through the area where the eliminated historical obstacle is located. Specifically, if the obstacle difference information shows that the historical obstacle to be avoided on the first segment of the historical route has been eliminated, the original first segment will be directly adjusted to a straight segment. This straight segment passes through the area where the eliminated historical obstacle was located, and there is no need to retain the turning points or detour curves set up in the original segment to avoid the historical obstacle, so as to shorten the travel distance and improve traffic efficiency.

[0085] S2624. If an obstacle is added to the second road segment representing the historical path by the obstacle difference information, determine the minimum detour segment at both ends of the second road segment to bypass the new obstacle according to the type and size of the new obstacle, and change the second road segment to the minimum detour segment. Specifically, when obstacle difference information shows that there is a new obstacle on the second segment of the historical path, the specific type of the new obstacle (e.g., fixed obstacle: gravel; floating obstacle: branches, fishing nets; biological obstacle: schools of fish) and its actual size (e.g., diameter, length, volume, etc.) are first determined. Then, based on this information, the minimum detour segment that can bypass the new obstacle is calculated and determined. The minimum detour segment must meet the following principles: maintaining a safe distance from the new obstacle (set according to the size of the obstacle, e.g., small obstacles ≥ 0.3m, large obstacles ≥ 0.5m), gentle turning angle, and shortest path length. Finally, the original second segment is replaced with the minimum detour segment to ensure that the new obstacle is avoided without significantly deviating from the original driving direction.

[0086] S2625. The second path is obtained by correcting the historical path based on the water flow velocity and direction represented by the real-time hydrological information.

[0087] Specifically, after adjusting the path based on obstacle differences, the path is further optimized and corrected by combining real-time hydrological information representing water flow velocity and direction. This multi-dimensional correction, incorporating obstacle differences and hydrological conditions, ultimately forms a second path, ensuring both safety and efficiency in driving.

[0088] In some optional embodiments, the step of correcting the historical path based on the water flow velocity and direction represented by the real-time hydrological information to obtain the second path includes: S2626. If the water flow velocity is less than the first velocity threshold, configure the historical path as the second path; Specifically, if the water flow speed is less than the first speed threshold (which is a preset critical speed at which the water flow affects the movement of the unmanned underwater vehicle, such as 0.3 m / s), it means that the current water flow conditions can negligibly interfere with the unmanned underwater vehicle's movement along the original path. There is no need to adjust the path, and the historical path can be directly determined as the second path.

[0089] S2057. When the water flow velocity is greater than a first velocity threshold and the first angle between the water flow direction and the historical path is not zero, a first curved path is set according to the water flow direction, the water flow velocity and the first angle, and the first curved path is configured as the first path so that the path trajectory of the unmanned underwater vehicle when traveling along the first curved path coincides with the historical path.

[0090] Specifically, if the water flow velocity is greater than the first velocity threshold, and the first angle between the water flow direction and the historical path is not zero (i.e., the water flow direction has an angle with the original direction of travel of the submersible), then a first curved path is planned and set based on the specific water flow direction, water flow velocity, and the value of the first angle. (Refer to...) Figure 3 The core design of this first curved path is to offset the lateral thrust of the water flow by using the curve shape, so as to ensure that the actual trajectory of the submersible coincides with the preset trajectory of the historical path. This not only helps to improve the driving efficiency with the help of the water flow, but also avoids deviation from the route due to the impact of the water flow.

[0091] In some optional embodiments, acquiring leaf image information of the target area via the unmanned underwater vehicle includes: S310. Control the unmanned underwater vehicle to move along the lower ring water outlet edge of the Kth blade to the upper crown water outlet edge, and capture the first image information of the Kth blade during the movement, where K represents a positive integer greater than or equal to 1. Specifically, the unmanned underwater vehicle (UUV) is controlled to move to a preset starting point outside the lower ring water outlet edge of the Kth blade, maintaining a safe distance from the lower ring water outlet edge (avoiding collision with the blade while ensuring the shooting angle is close to the blade surface); based on the preset BIM model and real-time pose information, the UUV's movement trajectory is calibrated, the trajectory conforms to the curved contour of the Kth blade, starting from the lower ring water outlet edge, extending radially along the blade to the upper crown water outlet edge, the trajectory covers the water outlet edge, working surface, back surface of the blade, and the connection with the upper crown / lower ring, with no visual acquisition blind spots.

[0092] Throughout the entire process of the unmanned underwater vehicle moving from the lower ring of the Kth blade to the upper crown of the blade, all captured images are transmitted in real time to the shore-based workstation for temporary storage, while redundant storage is performed locally on the UUV. After the acquisition is completed, the shore-based workstation performs a preliminary quality screening of the first image information, eliminating invalid images that are blurry, overexposed, or have insufficient overlap, to ensure that the first image information completely covers the Kth blade.

[0093] S320. Control the unmanned underwater vehicle to move along the upper crown water exit edge of the K+1th blade to the lower ring water exit edge, and capture the second image information of the K+1th blade during the movement. The blade image information includes the first image information and the second image information.

[0094] Specifically, after acquiring the data for the Kth blade, the unmanned underwater vehicle (UUV) performs a collision-free transition, moving from the upper crown exit edge of the Kth blade to the starting position of the upper crown exit edge of the (K+1)th blade, and then moving along the reverse trajectory from the upper crown exit edge to the lower ring exit edge to complete the second image acquisition for the (K+1)th blade, ensuring the continuity of the detection. This process is repeated for the (K+2)th blade, proceeding sequentially from the lower ring exit edge to the upper crown exit edge. The detection of subsequent blades is not detailed here.

[0095] The unmanned underwater vehicle is also equipped with a multi-degree-of-freedom robotic arm, which is equipped with a camera. When the unmanned underwater vehicle cannot capture the back of the blade or a hidden corner, the multi-degree-of-freedom robotic arm can extend to the corresponding hidden area and capture video images of the blade through the camera on the multi-degree-of-freedom robotic arm, thereby completing the blade inspection without blind spots and achieving good inspection results.

[0096] In some optional embodiments, obtaining the crack information of the turbine runner blades based on the blade image information includes: S410. After filtering and denoising the leaf image information, preprocessed image information is obtained. Specifically, the original blade images are batch-screened based on validity, clarity, and overlap, eliminating invalid images. Blurry, overexposed, underexposed, and motion-blurred images are removed, while clear, evenly lit images are retained. Images containing numerous air bubbles or suspended particles are also removed to prevent them from creating false feature points. Image overlap is verified, ensuring that adjacent images have an overlap of ≥80%, and images in high-crisis areas (water outlet, root, and junction) have an overlap of ≥90%. Areas with insufficient overlap are supplemented with corresponding images. Valid images are then structurally categorized by blade number and shooting area.

[0097] The selected valid images are processed sequentially using geometric correction, dehazing and deglare removal, noise suppression, and feature enhancement. Geometric correction: Importing previously calibrated underwater camera intrinsic parameters (including radial / tangential distortion parameters, considering water refractive index) corrects lens distortion and ensures the spatial accuracy of image pixels. Dehazing and deglare removal: A dark channel prior algorithm removes the fog effect caused by water scattering, and polarization image fusion eliminates high-gloss reflections on the blade's metal surface, restoring the detailed texture of the blade surface and cracks. Noise suppression: A wavelet transform and median filtering combination algorithm removes random noise and impulse noise generated by suspended particles in the water, camera shake, and ensures the continuity of image texture. Feature enhancement: Laplacian sharpening and adaptive histogram equalization are applied to areas with high crack incidence to improve the contrast between crack edges and the normal blade surface, enhancing key features such as cracks, water-exit edges, and blade contours.

[0098] All processed images are stored in a structured manner according to leaf number and shooting area. Each image is bound with metadata such as shooting pose coordinates, leaf number, area label, and timestamp to form standardized pre-processed image information for sparse point cloud generation.

[0099] S420. After extracting feature points, calculating camera pose, and transforming feature point coordinates on the preprocessed image information, sparse point cloud data is obtained. Specifically, to address the issue of limited natural feature points on the metal surface of blades, an ORB-SIFT hybrid feature extraction algorithm (balancing accuracy and speed) is employed to extract and match corresponding feature points between preprocessed images. SIFT (Scale-Invariant Feature Transform) is used to extract high-discrimination feature points such as blade cracks, water outlet edges, root junctions, and contour lines, preserving their rotation and scale invariance to improve matching accuracy. ORB (OrientedFAST and Rotated BRIEF) is used to extract micro-texture feature points on the smooth surface of the blade, compensating for the insufficient number of feature points in SIFT. False points are removed from the extracted feature points (through feature descriptor similarity and spatial distance constraints), retaining valid pairs of corresponding feature points between images while ensuring that the number of valid feature points in each image is ≥20,000 to meet the requirements of subsequent pose calculation.

[0100] Using bundle adjustment as the core algorithm, combined with the real-time pose data of the unmanned underwater vehicle (UUV) as initial constraints, the exterior orientation elements (3D position and attitude) of the camera in each image are calculated. Simultaneously, the camera intrinsic parameters are optimized to ensure the accuracy of the pose calculation: The intrinsic parameters calibrated by the underwater camera in the previous stage are imported as initial values, and the real-time pose coordinates of the UUV during shooting are used as the initial constraints for the camera exterior orientation elements (compensating for the lack of GPS underwater); based on pairs of corresponding feature points, multiple rounds of iterative bundle adjustment are performed, gradually eliminating feature point pairs with large calculation errors and optimizing the camera exterior and interior orientation elements; the calculation accuracy is verified, with a reprojection error ≤ 0.5 pixels (if the error is too large, feature points are re-extracted and recalculated), resulting in accurate camera pose data for all images after the calculation is completed.

[0101] Two-dimensional image feature points are converted into three-dimensional spatial coordinates through triangulation. Then, the coordinate system is unified and filtered for optimization to obtain sparse point cloud data. Based on the solved camera pose data, the three-dimensional spatial coordinates of each pair of corresponding feature points are calculated using a triangulation algorithm. The three-dimensional coordinates of all feature points are transformed to the local coordinate system of the runner (with the runner center as the origin and the blade normal direction as the z-axis) to eliminate the deviation caused by the coordinate system difference. The three-dimensional coordinates of all feature points are integrated to generate the initial sparse point cloud of the blade. Statistical filtering and radius filtering are used to remove outliers and floating points caused by noise in the water and false feature points, retaining the effective sparse point cloud that fits the actual shape of the blade, and finally obtaining standardized sparse point cloud data.

[0102] S430. Dense point cloud data is generated by processing the sparse point cloud data through a dense matching algorithm. Specifically, sparse point clouds only contain key feature points of the blade and cannot reproduce fine features such as cracks and micro-surfaces. Therefore, based on the sparse point cloud framework, a dense matching algorithm is used to refine the feature point matching granularity to each pixel, calculate the three-dimensional coordinates of all pixels, and generate a high-density point cloud covering the entire surface of the blade. At the same time, the differentiated density settings of underwater blades are incorporated to ensure the detailed reproduction of areas with high crack incidence.

[0103] The dense matching algorithm uses semi-global matching (SGM) (balancing matching accuracy and computational efficiency), combined with PMVS (patch multi-view stereo matching) algorithm to supplement the fine details of the blade surface and cracks. Differentiated dense matching parameters are set for different areas of the blade: high matching density is set in areas with high crack incidence, with a point spacing ≤0.001m to maximize the restoration of the micromorphology of cracks; normal matching density is set in other areas of the blade, with a point spacing ≤0.01m, to control the file size while ensuring detail; depth filtering and texture constraints are enabled to remove false point clouds caused by water background and suspended particles, ensuring the fit between the matched point cloud and the blade surface.

[0104] The generation and initial optimization of dense point clouds are based on sparse point clouds. A dense matching algorithm is used to calculate the three-dimensional coordinates of all pixels in the preprocessed image information to generate the initial dense point cloud of the blade. Then, noise reduction and completion optimization are performed: first, false points and outliers far away from the main body of the blade are removed by statistical filtering and radius filtering. Then, feature-preserving filtering is used to smooth the dense point cloud, eliminating point cloud noise while strictly preserving key features such as crack edges and water-exit contours (avoiding the loss of crack information due to filtering). For point cloud gaps caused by blind spots such as blade gaps and the inner side of the lower ring, real-time sonar point cloud data from the unmanned underwater vehicle is fused, and point cloud stitching is performed using the ICP algorithm to ensure the integrity of the dense point cloud. Uniform resampling is performed on overly dense point cloud areas to unify the point cloud spacing and improve the efficiency of three-dimensional mesh construction.

[0105] The accuracy verification of dense point cloud involves coarsely registering the optimized dense point cloud with the design CAD model of the turbine runner blade. The overall deviation between the point cloud and the CAD model is verified to be ≤0.005m, and the deviation in high-incidence crack areas is ≤0.001m. For areas with deviations exceeding the standard, dense matching is re-performed and optimized to finally obtain high-precision, full-coverage dense point cloud data without false points.

[0106] S440. A three-dimensional blade model is obtained by constructing a three-dimensional mesh and mapping a texture based on the dense point cloud data. Specifically, based on dense point cloud data, the three-dimensional curved surface shape of the blade is restored by constructing a three-dimensional mesh, and then the texture of the pre-processed image information is attached to the mesh surface through texture mapping to form a three-dimensional blade model with high-precision texture, which has both the three-dimensional spatial shape of the blade and the surface visual details.

[0107] A 3D mesh is constructed using the Poisson reconstruction algorithm (suitable for reconstructing curved components, accurately restoring the surface morphology of blades and the 3D contours of cracks). Simultaneously, partitioning parameters are set, holes are filled, and the mesh is optimized to ensure accuracy and integrity: the mesh resolution of different partitions is matched to the density of the dense point cloud, with a mesh resolution ≤0.001m in high-crack areas and ≤0.01m in other areas. Small holes (diameter ≤0.005m) caused by missing dense point clouds are automatically filled, while large holes (diameter >0.005m, such as blade gaps or blind spots) are repaired using 3D point cloud processing software (Meshlab / CloudCompare). Crack areas are pre-marked to avoid misclassifying cracked holes as ordinary holes, thus preventing the loss of crack information. A feature-preserving mesh simplification algorithm is employed to reduce the number of mesh faces and improve subsequent processing efficiency while retaining key features such as cracks and water outlet edges; issues such as inverted faces and non-manifold edges in the mesh are repaired to prevent abnormal texture mapping in subsequent steps. The constructed 3D mesh was precisely registered with the blade CAD model, and the overall deviation was verified to be ≤0.005m, ensuring that the mesh shape was consistent with the actual blade shape.

[0108] The surface texture of the blades is extracted from preprocessed images and attached to a 3D mesh surface. Simultaneously, lighting differences and stitching artifacts are eliminated to ensure precise texture-mesh fit. Leaf textures for corresponding regions are extracted from preprocessed images, and a multi-image texture fusion algorithm is used to eliminate texture color differences and stitching artifacts caused by uneven underwater lighting and different shooting angles, restoring the true texture of the blade surface. The texture resolution in areas with high crack incidence is consistent with the original image, while other areas can have their resolution appropriately reduced to balance detail and file size. The fused texture is precisely attached to the 3D mesh surface. For areas with metallic reflection and blurred textures, multiple low-reflectivity images from different angles are used to regenerate textures, correcting texture stretching and misalignment issues. The crack-prone areas are magnified to verify the fit between the texture and the mesh outline, ensuring no texture shift or stretching, and guaranteeing that crack details are clearly discernible in the texture model.

[0109] S450. Extract the crack information from the three-dimensional blade model.

[0110] Specifically, in the texture view of the 3D blade model, suspected crack areas are marked, generating crack regions of interest (ROIs) and associated with the blade number and detection area (lower ring water outlet edge / upper crown water outlet edge / blade root, etc.); based on the point cloud data of the 3D blade model, the elevation difference, curvature, and slope features of the point cloud are extracted, and thresholds are set (such as elevation difference ≥ 0.001m, curvature ≥ 5°) to automatically filter out suspected crack areas, which complement the marking results and reduce missed detections; for each crack area, the 3D coordinates of the crack center and the 3D coordinates of the crack start and end points (in the local coordinate system of the rotor) are extracted to complete the precise spatial positioning of the crack.

[0111] Based on the texture view and point cloud data of the 3D blade model, the length, width, and depth of the crack are precisely quantified, with the quantification accuracy adapted to engineering maintenance needs: along the natural direction of the crack, a curved surface arc length measurement tool (adapted to the curved surface shape of the blade to avoid errors caused by planar measurements) is used on the 3D blade model to measure the actual curved surface length of the crack. In the direction perpendicular to the crack direction, a measurement section is taken every 0.01m to measure the maximum crack width at each section, and the average width and maximum width are calculated. Based on the 3D point cloud data, the elevation difference between the cracked area and the surrounding normal blade surface is extracted, and the average value is taken as the average crack depth; simultaneously, the depth is corrected by combining phased array ultrasonic detection data from unmanned underwater vehicle inspections.

[0112] After locating and quantifying the cracks, morphological classification and scale statistics are performed: based on the crack direction and distribution characteristics, cracks are classified into linear cracks, bifurcation cracks, and network cracks. The direction of crack extension (e.g., radial along the outlet edge, circumferential along the blade surface, etc.) is also recorded. The actual crack coverage area is calculated based on 3D point cloud data, and the overall scale of the cracks is statistically analyzed by combining length, width, and depth. Referring to the maintenance standards for hydropower turbine runners, cracks are classified into grades (e.g., micro-cracks, slight cracks, moderate cracks, severe cracks) based on their length, depth, width, and scale, thereby determining the maintenance priority.

[0113] The extracted crack location information, size information, morphological scale information, and grade information are integrated with metadata such as blade number, detection area, detection time, and UUV detection parameters to form a structured crack information set, including: blade number, crack number, detection area, three-dimensional coordinates of crack center, length, average width, maximum width, average depth, morphological type, extension direction, coverage area, crack grade, and maintenance suggestions. Finally, the resulting information on turbine runner blade cracks can be directly used for database establishment and maintenance decisions.

[0114] The embodiments of this invention have the following beneficial effects: Underwater real-time detection information is acquired via an unmanned underwater vehicle (UUV), which represents real-time image information, real-time sonar information, real-time magnetic field information, real-time attitude information, and real-time hydrological information acquired by the UUV during its journey from the tailrace pipe to the turbine runner blades; the UUV is navigated to the target area where the turbine runner blades are located based on the underwater real-time detection information; blade image information of the target area is acquired via the UUV; and crack information of the turbine runner blades is obtained based on the blade image information. In this embodiment, the UUV automatically acquires underwater real-time detection information and navigates to the turbine runner blades for crack detection, resulting in high efficiency in crack detection of the turbine runner blades; and there is no need to construct a tailrace maintenance platform, empty pressure steel pipe, or tailrace pipe, as the UUV can travel from the tailrace pipe to the turbine runner blades for crack detection, thus reducing detection costs.

[0115] In addition, one embodiment of the present invention provides a crack detection device for turbine runner blades, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0116] The processor and memory can be connected via a bus or other means.

[0117] It should be noted that the computer in this embodiment may correspond to, for example, including, Figure 1 The memory and processor in the illustrated embodiment can constitute Figure 1 The system architecture platform shown in the embodiment belongs to the same inventive concept, and therefore has the same implementation principle and beneficial effects, which will not be described in detail here.

[0118] The non-transient software program and instructions required to implement the methods of the above embodiments are stored in memory. When executed by a processor, the crack detection method for turbine runner blades of the above embodiments is executed, for example, the method described above is executed. Figure 2 Method steps S100 to S400.

[0119] Furthermore, one embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are used to execute the aforementioned method for detecting cracks in turbine runner blades, for example, to execute the methods described above... Figure 2 Method steps S100 to S400.

[0120] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as processors, such as central processing units, digital signal processors, or microprocessors executing software, or as hardware, or as integrated circuits, such as application-specific integrated circuits. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically include computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0121] The above provides a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of the present invention.

Claims

1. A method for detecting cracks in turbine runner blades, characterized in that, include: The underwater real-time detection information obtained by the unmanned underwater vehicle represents the real-time image information, real-time sonar information, real-time magnetic field information, real-time pose information and real-time hydrological information acquired by the unmanned underwater vehicle in the process from the tailrace pipe to the turbine runner blades. The unmanned underwater vehicle is guided to the target area where the turbine runner blades are located based on the underwater real-time detection information. The unmanned underwater vehicle acquires blade image information of the target area; Crack information of the turbine runner blades is obtained based on the blade image information.

2. The method for detecting cracks in turbine runner blades according to claim 1, characterized in that, The step of navigating the unmanned underwater vehicle to the target area where the turbine runner blades are located based on the underwater real-time detection information includes: Obtain a weight table, which indicates the parameter weights corresponding to different spatial locations; The real-time spatial position of the unmanned underwater vehicle is determined based on the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, and the weight table. A first path is generated based on the real-time spatial location and the preset three-dimensional model, wherein the preset three-dimensional model represents the BIM model corresponding to the tailrace pipe and the turbine. Based on the real-time sonar information, the spatial contour information and the first obstacle information of the unmanned underwater vehicle's location are obtained; The second obstacle information is obtained by correcting the first obstacle information based on the real-time image information; The second path is obtained by correcting the first path based on the second obstacle information, the hydrological information, and the spatial contour information; Control the unmanned underwater vehicle to reach the target area along the second path.

3. The method for detecting cracks in turbine runner blades according to claim 2, characterized in that, The step of determining the real-time spatial position of the unmanned underwater vehicle based on the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, and the weight table includes: The first position is determined based on the real-time image information and the first corresponding data, wherein the first corresponding data represents the correspondence between different spatial positions and image information; The second position is determined based on the real-time sonar information and the second corresponding data, wherein the second corresponding data represents the correspondence between different spatial positions and sonar information. The third position is determined based on the real-time pose information and the third corresponding data, wherein the third corresponding data represents the correspondence between different spatial positions and pose information. The fourth position is determined based on the real-time magnetic field information and the fourth corresponding data, wherein the fourth corresponding data represents the correspondence between different spatial positions and magnetic field information. The first weight coefficient corresponding to the first position, the second weight coefficient corresponding to the second position, the third weight coefficient corresponding to the third position, and the fourth weight coefficient corresponding to the fourth position are obtained according to the weight table. The real-time spatial position is determined based on the first position, the first weight coefficient, the second position, the second weight coefficient, the third position, the third weight coefficient, the fourth position, and the fourth weight coefficient.

4. The method for detecting cracks in turbine runner blades according to claim 3, characterized in that, The weight calculation method indicated by the weight table includes: The relative distance between the unmanned underwater vehicle and the rotor is determined based on the first position, the second position, the third position, and the fourth position. The distance factor corresponding to the Nth type of information is determined according to the relative distance and the preset distance parameter table. The preset distance parameter table indicates the correspondence between different relative distances and the distance factor. The Nth type of information represents any one of the real-time image information, the real-time sonar information, the real-time pose information, and the real-time magnetic field information. The information quality factor is determined based on the signal-to-noise ratio of the Nth type of information and the similarity between the Nth type of information and the corresponding Mth data at the same location. The corresponding Mth data represents the correspondence between different spatial locations and the Nth type of information. Environmental factors are determined based on underwater visibility and electromagnetic interference intensity at the location of the unmanned underwater vehicle. The stage calculation weights of the distance factor, the information quality factor, and the environmental factor are obtained based on the navigation stage of the unmanned underwater vehicle. The first weight coefficient, the second weight coefficient, the third weight coefficient, or the fourth weight coefficient corresponding to the Nth type of information are calculated based on the distance factor, the information quality factor, the environmental factor, and the stage calculation weight.

5. The method for detecting cracks in turbine runner blades according to claim 2, characterized in that, The generation of the second path also includes: In the case of historical detection records for turbine runner blades, the historical path and historical obstacle information corresponding to the historical detection records are obtained. The historical path represents the travel path of the unmanned underwater vehicle when detecting turbine runner blades, and the historical obstacle information represents the obstacles on the historical path. The first path is determined based on the historical path, historical obstacle information, and the first information obtained by the unmanned underwater vehicle; In the absence of historical inspection records for the turbine runner blades, obstacle map information is obtained from the BIM model. The obstacle map information indicates the location, type, and size of obstacles in each area of ​​the BIM model. The second path is determined based on the obstacle map information and the second information obtained by the unmanned underwater vehicle. The first information and the second information both include the real-time image information, the real-time sonar information, the real-time pose information, the real-time magnetic field information, the real-time hydrological information, and the weight table.

6. The method for detecting cracks in turbine runner blades according to claim 5, characterized in that, Determining the first path based on the historical path, historical obstacle information, and the first information obtained by the unmanned underwater vehicle includes: Based on the first information, determine the real-time obstacle information and the real-time hydrological information acquired by the unmanned underwater vehicle at various locations; The obstacle difference information is obtained by comparing the historical obstacle information with the real-time obstacle information; If the historical obstacle that was avoided on the first segment of the historical path, as represented by the obstacle difference information, is eliminated, the first segment is changed to a straight segment, which passes through the area where the eliminated historical obstacle was located. If an obstacle is added to the second segment of the historical path as represented by the obstacle difference information, the minimum detour segment at both ends of the second segment is determined according to the type and size of the added obstacle, and the second segment is changed to the minimum detour segment. The second path is obtained by correcting the historical path based on the water flow velocity and direction represented by the real-time hydrological information.

7. The method for detecting cracks in turbine runner blades according to claim 1, characterized in that, The step of acquiring blade image information of the target area through the unmanned underwater vehicle includes: The unmanned underwater vehicle is controlled to move along the lower ring water exit edge of the Kth blade to the upper crown water exit edge, and the first image information of the Kth blade is captured during the movement, where K represents a positive integer greater than or equal to 1; The unmanned underwater vehicle is controlled to move along the upper crown exit edge of the K+1th blade to the lower ring exit edge, and during the movement, a second image of the K+1th blade is captured. The blade image information includes the first image information and the second image information.

8. The method for detecting cracks in turbine runner blades according to claim 1, characterized in that, The step of obtaining crack information of the turbine runner blades based on the blade image information includes: The preprocessed image information is obtained by filtering and denoising the leaf image information; After performing feature point extraction, camera pose calculation, and feature point coordinate transformation on the preprocessed image information, sparse point cloud data is obtained. The sparse point cloud data is processed by a dense matching algorithm to generate dense point cloud data. A three-dimensional blade model is obtained by constructing a three-dimensional mesh and mapping a texture based on the dense point cloud data. The crack information is extracted from the three-dimensional blade model.

9. A crack detection device for turbine runner blades, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method for detecting cracks in turbine runner blades as described in any one of claims 1-8.

10. A computer storage medium, characterized in that, The computer storage medium stores computer-executable instructions, which are used to execute the method for detecting cracks in turbine runner blades according to any one of claims 1-8.

Images