Method and device for measuring the thickness of a steam turbine blade

Abstract

This invention provides a method and apparatus for measuring the thickness of steam turbine blades. The method includes: scanning the steam turbine blade according to preset scanning parameters to obtain first point cloud data; performing point cloud preprocessing on the first point cloud data to obtain second point cloud data; performing coarse registration processing on the second point cloud data to obtain third point cloud data; performing fine registration processing on the third point cloud data to obtain fourth point cloud data; performing trajectory conversion based on the fourth point cloud data to obtain robot trajectory data; and having a robot equipped with a detection component measure the thickness of the steam turbine blade according to the robot trajectory data to obtain the thickness measurement result of the steam turbine blade. This invention can improve the accuracy and efficiency of steam turbine blade thickness measurement.

Description

Technical Field

[0001] This invention relates to the field of turbine blade thickness measurement technology, and also to a method and apparatus for measuring turbine blade thickness. Background Technology

[0002] Turbine blades are core load-bearing components subjected to high temperature, high pressure, and high speed rotation. Their wall thickness uniformity and accuracy directly determine the unit's efficiency, vibration characteristics, and operational safety. Current mainstream technologies for blade thickness measurement fall into three categories: contact measurement: using micrometers, dedicated templates, and coordinate measuring machines for point-by-point measurement; this is inefficient, easily scratches the blade profile, and struggles to cover narrow areas such as the inlet / outlet edges; semi-automatic ultrasonic testing: relying on manual alignment and teaching programming; the probe cannot always be perpendicular to the curved surface, water immersion coupling is unstable, and measurement repeatability is poor; conventional point cloud detection: only performs surface comparison, without linkage with ultrasonic thickness measurement, lacking true thickness verification, and suffers from high noise in highly reflective areas, registration drift, and insufficient local accuracy. Existing automated solutions generally suffer from inaccurate positioning, non-adaptive trajectory, insufficient registration accuracy, and unclosed-loop system errors, making it difficult to meet the requirements for full-area, high-speed, and high-precision blade thickness measurement. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method and device for measuring the thickness of steam turbine blades, so as to improve the accuracy and efficiency of steam turbine blade thickness measurement.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0005] A first aspect of the present invention provides a method for measuring the thickness of a steam turbine blade, comprising:

[0006] The turbine blades are scanned according to the preset scanning parameters to obtain the first point cloud data;

[0007] Perform point cloud preprocessing on the first point cloud data to obtain the second point cloud data;

[0008] The second point cloud data is coarsely registered to obtain the third point cloud data;

[0009] The third point cloud data is subjected to fine registration processing to obtain the fourth point cloud data;

[0010] The trajectory data of the robot is obtained by performing trajectory conversion based on the fourth point cloud data;

[0011] The robot, equipped with a detection component, measures the thickness of the turbine blade according to the robot's trajectory data, and obtains the thickness measurement result of the turbine blade.

[0012] Optionally, the turbine blades are scanned according to preset scanning parameters to obtain the first point cloud data, including:

[0013] Obtain preset scanning parameters; the preset scanning parameters include a scanning distance of 50 to 100 mm, a point cloud density covering the crown, root, leading edge, and trailing edge of the turbine blade, and a scanning range covering the entire area of ​​the turbine blade;

[0014] The turbine blades fixed at a preset position are scanned according to the preset scanning parameters to obtain the first point cloud data.

[0015] Optionally, point cloud preprocessing is performed on the first point cloud data to obtain second point cloud data, including:

[0016] The first point cloud data is denoised to obtain denoised point cloud data;

[0017] The denoised point cloud data is simplified to obtain the second point cloud data.

[0018] Optionally, coarse registration processing is performed on the second point cloud data to obtain third point cloud data, including:

[0019] Descriptive data is extracted from the second point cloud data and the target point cloud data to obtain descriptive data;

[0020] The second point cloud data and the target point cloud data are matched based on the similarity of the described data to obtain the third point cloud data.

[0021] Optionally, the third point cloud data is subjected to fine registration processing to obtain the fourth point cloud data, including:

[0022] Obtain the objective function; the objective function is Where R is the rotation matrix, p is any point on the turbine blade in the third point cloud data, q is the target point cloud data corresponding to p in the third point cloud data, and t is the translation vector. Let q be the normal vector;

[0023] The objective function is iteratively optimized until the preset convergence condition is met, and the fourth point cloud data is obtained.

[0024] Optionally, trajectory transformation is performed based on the fourth point cloud data to obtain robot trajectory data, including:

[0025] The fourth point cloud data is subjected to coordinate system transformation to obtain transformed point cloud data;

[0026] Based on the transformed point cloud data, determine the pose correction data;

[0027] The preset detection path is corrected based on the pose correction data to obtain robot trajectory data.

[0028] Optionally, the thickness of the turbine blade is measured according to the robot trajectory data to obtain the thickness measurement result of the turbine blade, including:

[0029] The cloud data from the second point is sliced ​​to obtain sliced ​​data;

[0030] Based on the slice data, the initial thickness data is obtained;

[0031] The turbine blades are ultrasonically thickened according to the robot trajectory data to obtain ultrasonic thickness data.

[0032] Based on the initial thickness data and the ultrasonic thickness data, the thickness measurement results of the turbine blades are obtained.

[0033] A second aspect of the present invention provides a device for measuring the thickness of a steam turbine blade, comprising:

[0034] The scanning module is used to scan the turbine blades according to preset scanning parameters to obtain the first point cloud data;

[0035] The processing module is used to perform point cloud preprocessing on the first point cloud data to obtain second point cloud data; perform coarse registration processing on the second point cloud data to obtain third point cloud data; perform fine registration processing on the third point cloud data to obtain fourth point cloud data; perform trajectory conversion based on the fourth point cloud data to obtain robot trajectory data; and have the robot equipped with a detection component perform thickness measurement on the turbine blade according to the robot trajectory data to obtain the thickness measurement result of the turbine blade.

[0036] A third aspect of the present invention provides a computing device, comprising: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described in the first aspect.

[0037] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method as described in the first aspect.

[0038] The above-described solution of the present invention has at least the following beneficial effects:

[0039] The above-described solution of the present invention obtains first point cloud data by scanning the turbine blade according to preset scanning parameters, performs point cloud preprocessing on the first point cloud data to obtain second point cloud data, performs coarse registration processing on the second point cloud data to obtain third point cloud data, performs fine registration processing on the third point cloud data to obtain fourth point cloud data, performs trajectory conversion based on the fourth point cloud data to obtain robot trajectory data, and finally uses a robot equipped with a detection component to measure the thickness of the turbine blade according to the robot trajectory data to obtain the thickness measurement result of the turbine blade. This can improve the accuracy and efficiency of turbine blade thickness measurement. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating the method for measuring the thickness of steam turbine blades in an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram of the structure of the turbine blade thickness measuring device in an embodiment of the present invention. Detailed Implementation

[0042] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0043] like Figure 1 As shown, an embodiment of the present invention proposes a method for measuring the thickness of a steam turbine blade, comprising the following steps:

[0044] Step 101: Scan the turbine blades according to the preset scanning parameters to obtain the first point cloud data;

[0045] Step 102: Perform point cloud preprocessing on the first point cloud data to obtain the second point cloud data;

[0046] Step 103: Perform coarse registration processing on the second point cloud data to obtain the third point cloud data;

[0047] Step 104: Perform fine registration processing on the third point cloud data to obtain the fourth point cloud data;

[0048] Step 105: Perform trajectory conversion based on the fourth point cloud data to obtain robot trajectory data;

[0049] Step 106: The robot, equipped with a detection component, measures the thickness of the turbine blade according to the robot trajectory data to obtain the thickness measurement result of the turbine blade.

[0050] The turbine blade thickness measurement method of this invention involves scanning the turbine blade according to preset scanning parameters to obtain first point cloud data, performing point cloud preprocessing on the first point cloud data to obtain second point cloud data, performing coarse registration on the second point cloud data to obtain third point cloud data, performing fine registration on the third point cloud data to obtain fourth point cloud data, performing trajectory conversion based on the fourth point cloud data to obtain robot trajectory data, and finally using a robot equipped with a detection component to measure the thickness of the turbine blade according to the robot trajectory data to obtain the turbine blade thickness measurement result. This method can improve the accuracy and efficiency of turbine blade thickness measurement.

[0051] In an optional embodiment of the present invention, step 101, scanning the turbine blades according to preset scanning parameters to obtain first point cloud data, may include:

[0052] Step 1011: Obtain preset scanning parameters; the preset scanning parameters include a scanning distance of 50 to 100 mm, a point cloud density covering the crown, root, leading edge, and trailing edge of the turbine blade, and a scanning range covering the entire area of ​​the turbine blade.

[0053] Step 1012: Scan the turbine blades fixed at the preset position according to the preset scanning parameters to obtain the first point cloud data.

[0054] Specifically, a line laser scanner (or structured light camera) is used as a 3D sensor, installed at the end effector of the robot, to perform a full-area scan of the turbine blades. Before scanning, the blades need to be fixed (positioned using specialized fixtures to ensure no blade movement). Following preset scanning parameters, the distance between the sensor and the blades is adjusted (maintaining 50 to 100 mm to ensure scan clarity), covering the entire turbine blade area for scanning to obtain the first point cloud data. Here, the preset scanning parameters can be selected and modified according to actual conditions. For example, the scanning resolution and angle interval can be adjusted based on the blade size and complexity. For complex curved surfaces, the scanning density can be increased to capture details. The preset position can be a turntable. The turbine blades are fixed on the turntable and rotated in stages at fixed angles (e.g., 10 to 30 degrees). The laser scanner acquires local point clouds sequentially. For example, rotating the turntable 12 times (30 degrees each time) can achieve full coverage.

[0055] In an optional embodiment of the present invention, step 102, which involves performing point cloud preprocessing on the first point cloud data to obtain second point cloud data, may include:

[0056] Step 1021: Denoise the first point cloud data to obtain denoised point cloud data;

[0057] Specifically, a checkerboard calibration board (accuracy ±0.01mm) is used. A robot drives the calibration board to capture images in different postures (at least 15 postures, covering all angles within the camera's scanning range). Zhang Zhengyou's calibration method is used to solve for the camera's intrinsic parameters. Then, combined with the robot's end-effector pose data, the transformation matrix from the camera coordinate system to the robot's base coordinate system is calculated. After calibration, multiple repeated calibration verifications ensure that the calibration error is controlled within 0.05mm. Then, based on the target size data of the blade, boundary ranges for the X, Y, and Z axes are set (example: X±300mm, Y±150mm, Z±50mm). Points in the first point cloud data that exceed these boundaries are directly cropped (considered invalid background points, such as point clouds generated by tooling or environmental debris). The boundary ranges can be flexibly adjusted according to the target size data of different blade models.

[0058] After the above calibration and filtering, valid point cloud data is obtained, and then denoising is performed on the valid point cloud data. Based on the neighborhood distance distribution characteristics of the point cloud, point-by-point analysis is performed on the cropped valid point cloud data; the number of neighborhood points for each point is set to k=15 (k=15 can effectively identify outliers while avoiding the accidental deletion of valid edge points), the average distance between each point and its k nearest neighbors is calculated, and then the global mean and global standard deviation of the average distance of all points are calculated. If the average distance of a point exceeds the range of the global mean ± 3 times the global standard deviation, it is identified as an outlier and removed. The purpose of denoising the first point cloud data is to accurately identify and remove outlier noise points generated during the scanning process (such as false points in highly reflective areas and abnormal points caused by sensor errors), while retaining the normal surface point cloud of the blade; the selection of the 3 times standard deviation principle can achieve a balance between denoising effect and point cloud retention, avoiding the accidental deletion of key feature points such as blade edges.

[0059] Step 1022: Simplify the denoised point cloud data to obtain the second point cloud data.

[0060] Specifically, the 3D space containing the clean point cloud is divided into a uniform voxel grid (the voxel side length is set to 0.5mm, which achieves an optimal balance between reducing the point cloud size and preserving features). For each voxel grid, the centroid coordinates of all points within the grid are calculated, and this centroid is used as the representative point of that voxel, replacing all the original points within the grid, thus reducing the point cloud size. The purpose of this simplification process is to significantly reduce the point cloud size while preserving key blade features (profile, edge, root, and crown), thereby reducing the computational complexity of subsequent registration and feature extraction and improving processing efficiency. Without point cloud simplification, an excessively large point cloud size (reaching millions of points) would lead to excessively long registration and trajectory planning steps, making it unsuitable for production line cycles.

[0061] In an optional embodiment of the present invention, step 103, which involves coarsely registering the second point cloud data to obtain the third point cloud data, may include:

[0062] Step 1031: Extract descriptive data from the second point cloud data and the target point cloud data to obtain descriptive data;

[0063] Specifically, based on the normal vector and curvature features of the point cloud, fast point feature histogram description data is extracted from the second point cloud data and the target point cloud data (the target point cloud data is the point cloud data of the target steam turbine blade model). The fast point feature histogram description data can effectively describe the local geometric features (normal vector direction, curvature magnitude) of each point, and has rotation invariance and scale invariance, which can effectively improve the robustness of matching.

[0064] Here, the steps for extracting fast point feature histogram description data from the second point cloud data and the target point cloud data are the same. We will take extracting fast point feature histogram description data from the second point cloud data as an example:

[0065] For the target point in the second point cloud data Select its k nearest neighbors (k=15) to obtain a local point set S; for the local point set S, through Obtain the unit normal vectors of the target point p and its neighboring points, where v is the i-th target point. and the unit normal vector of each neighboring point, For the i-th target point in the second point cloud data, for The j-th neighborhood point; through The angle between the normal vectors is calculated, where, The angle between the normal vectors, For target point The normal vector; through The distance weight d is calculated; through The angle difference is calculated, where, Let w be the angle difference, and w be the cross product of the normal vectors. , for The normal vector is divided into intervals (11 levels for the normal vector angle, 11 levels for the distance weight, and 11 levels for the angle difference, for a total of 33 levels). The frequency of the feature quantity in each interval is counted to form a 33-dimensional fast point feature histogram describing the data.

[0066] Step 1032: Match the second point cloud data and the target point cloud data based on the similarity of the description data to obtain the third point cloud data.

[0067] Specifically, through The similarity between the second point cloud data and the target point cloud data is obtained, where, To determine the similarity between the second point cloud data and the target point cloud data, This is the descriptive data for the second point of cloud data in the descriptive data. The description data for the target point cloud data in the description data is p and q, which are feature point pairs. Feature point pairs with similarity higher than a preset screening threshold (e.g., the preset screening threshold is 0.8, balancing matching efficiency and accuracy, and avoiding too many mismatched point pairs) are selected to form an initial point pair set. The sampling consistency algorithm is used to randomly sample 4 groups (each group has 4 non-coplanar point pairs to ensure that the transformation matrix can be uniquely solved) of non-coplanar feature point pairs. The temporary transformation matrix corresponding to each point pair is calculated, and the number of interior points corresponding to each group is counted (point pairs that meet the transformation error < 0.5 mm are selected as effective matching point pairs to improve the accuracy of the optimal initial transformation matrix). The temporary transformation matrix with the most interior points is retained as the candidate matrix. Based on the interior point set corresponding to the candidate matrix, the optimal initial transformation matrix (4×4 homogeneous transformation matrix) is obtained by optimization using the least squares method. The optimal initial transformation matrix includes the initial rotation matrix and the initial translation vector.

[0068] Based on the similarity of the data described by the fast point feature histogram, several sets of feature point pairs are randomly sampled from the second point cloud data and the target point cloud data. Mismatched point pairs are eliminated, and the initial transformation matrix of the second point cloud data and the target point cloud data is calculated to achieve preliminary alignment. Furthermore, a rotation-invariant local coordinate system can be established for each key point (a point in a preset key region of the turbine blade) to eliminate the influence of point cloud rotation attitude on matching, further improving the robustness of matching and avoiding misjudgments caused by slight changes in the blade installation attitude. The third point cloud data includes the matched point pairs, namely the source point cloud data (i.e., the turbine blade point cloud obtained by scanning) and the corresponding target point cloud data.

[0069] In an optional embodiment of the present invention, step 104, performing fine registration processing on the third point cloud data to obtain the fourth point cloud data, may include:

[0070] Step 1041, obtain the objective function; the objective function is Where R is the rotation matrix, p is any point in the source point cloud data, q is the target point cloud data corresponding to p in the source point cloud data, and t is the translation vector. Let q be the normal vector;

[0071] Specifically, R is the rotation matrix of the source point cloud data relative to the target point cloud data, and t is the translation vector of the source point cloud data relative to the target point cloud data. It is the projection of the position deviation vector onto the direction of the target point's normal vector (normal deviation). Squaring, summing, and minimizing this projection achieves optimal alignment of the source and target point cloud data in the normal direction. By minimizing the normal distance between the source and target point cloud data, precise alignment is achieved. Minimizing the Euclidean distance from point to surface better suits the measurement scenario of blade curved surfaces, effectively improving registration accuracy, and is especially suitable for registering complex curved surfaces such as blades.

[0072] Step 1042: Iteratively optimize the objective function until the preset convergence condition is met to obtain the fourth point cloud data.

[0073] Specifically, the preset convergence condition is that the maximum number of iterations is 100 (100 iterations can meet the accuracy requirements while avoiding inefficiency due to excessive iterations), or the change in the transformation matrix (R, t) between two consecutive iterations is <0.01mm (i.e., the change in position adjustment is <0.01mm, and the change in attitude adjustment is <0.01°), indicating that the iteration has stabilized and further iterations will not significantly improve accuracy. Here, the steps for iterative optimization of the objective function include:

[0074] Initialization: Input the initial transformation matrix obtained from coarse registration, and extract the initial rotation matrix from it. and the initial translation vector , as the initial value for iteration;

[0075] Point-to-point matching: based on the current rotation matrix Translation vector ,pass Convert point p in the source point cloud data into a point in the target coordinate system. Search for points in each target coordinate system The nearest point q in the target point cloud is used to form a point pair (p,q);

[0076] Objective function solution: based on point pair (p, q) and the normal vector of q Calculate the objective function value, and then use the least squares method to find the optimal rotation matrix that minimizes the objective function. Translation vector (Core solution logic: Take the partial derivative of the objective function and set it to 0 to solve for the rotation matrix R and the translation vector t).

[0077] Convergence criterion: Calculate the deviation of the transformation matrix between two adjacent iterations (including...) and Rotational deviation, and (Translation deviation), if the deviation is <0.01mm, or the number of iterations reaches 100, stop the iteration; otherwise, return to the step point pair matching and continue the iteration;

[0078] Output the optimal solution: After the iteration stops, the final rotation matrix R and translation vector t constitute the optimal transformation matrix. At this time, the source point cloud data and the target point cloud data are finely registered, and the fourth point cloud data is obtained.

[0079] Here, precise registration processing can achieve accurate alignment between the source point cloud data and the target turbine blade model (i.e., theoretical model), eliminating pose errors caused by installation deviations and scanning deviations, providing a unified benchmark for subsequent coordinate correction, trajectory planning, and thickness calculation, and ensuring the accuracy of thickness calculation.

[0080] In an optional embodiment of the present invention, step 105, performing trajectory conversion based on the fourth point cloud data to obtain robot trajectory data, may include:

[0081] Step 1051: Perform coordinate system transformation on the fourth point cloud data to obtain transformed point cloud data;

[0082] Specifically, combining the transformation matrix obtained in step 1021 and the optimal transformation matrix obtained in step 1042, the precisely registered source point cloud data is transformed back to the robot's base coordinate system. The transformation process is as follows: first, the point cloud is transformed from the theoretical model coordinate system to the camera coordinate system using the inverse of the optimal transformation matrix; then, it is transformed back to the robot's base coordinate system using the transformation matrix. Finally, the actual point cloud, precisely aligned with the theoretical model in the robot's base coordinate system, is obtained—this is the transformed point cloud data. This ensures that the point cloud coordinates are consistent with the robot's motion coordinates, providing accurate coordinate references for subsequent robot trajectory planning.

[0083] Step 1052: Determine pose correction data based on the converted point cloud data;

[0084] Specifically, based on the source point cloud data after coordinate system transformation, the deviation between the actual pose and the theoretical pose of the blade is calculated, including translational deviation ( X, Y, Z (unit: mm), and rotational deviation ( Roll, Pitch Yaw (unit: °), pose correction data includes translational and rotational deviations. , , , ( X1, Y1, Z1) is the centroid coordinate of the source point cloud data, ( X2, Y2, Z2) represents the centroid coordinates of the target point cloud data. , , ,in, The Euler angles (roll, pitch, and yaw) of the source point cloud data pose. Euler angles for the pose of the target point cloud data.

[0085] Step 1053: Correct the preset detection path according to the pose correction data to obtain robot trajectory data.

[0086] Specifically, the translational and rotational deviations in the pose correction data are used as robot pose correction values ​​and input into the robot control system to correct the robot's motion posture in real time. Additionally, based on the surface features of the source point cloud in the converted point cloud data, linear interpolation (interpolation step size 0.1 mm) is performed on the water immersion detection path points in the preset detection path. This ensures that the thickness probe's trajectory perfectly matches the blade surface, and that the probe remains perpendicular to the blade surface (normal deviation < 1°), ensuring effective coupling and avoiding thickness measurement errors caused by probe tilt. The resulting linear interpolation yields the robot trajectory data.

[0087] In an optional embodiment of the present invention, step 106, in which the robot, equipped with a detection component, measures the thickness of the turbine blade according to the robot trajectory data to obtain the thickness measurement result of the turbine blade, may include:

[0088] Step 1061: Perform slicing processing on the second point cloud data to obtain slice data;

[0089] Specifically, the Poisson reconstruction algorithm or moving least squares method can be used to fit the final actual point cloud in the robot's base coordinate system to obtain a 3D surface model of the blade. Based on the fitted surface, the edge point clouds of the leading and trailing edges of the blade are extracted (these edges are the key reference lines for blade thickness calculation, which needs to be performed along the edge normal). Using the leading and trailing edge point clouds as a reference, slices are made along the normal direction of the edges (perpendicular to the blade surface). Each slice corresponds to a cross-section of the blade, with a slice spacing of 0.1 mm (to ensure full coverage and no missed areas), thus obtaining the slice data.

[0090] Step 1062: Obtain initial thickness data based on the slice data;

[0091] Specifically, for each slice in the slice data, extract the inner arc surface points and outer arc surface points on the slice, calculate the minimum distance between the two points in the normal direction, which is the initial thickness value at that location. The initial thickness data includes the initial thickness value.

[0092] Step 1063: The robot, equipped with a detection component, performs ultrasonic thickness measurement on the turbine blade according to the robot trajectory data to obtain ultrasonic thickness data;

[0093] Specifically, the robot, carrying a water immersion ultrasonic probe and a laser emitter, moves along an adaptive thickness measurement trajectory (i.e., robot trajectory data) to perform ultrasonic thickness measurement at each slice location, obtaining the ultrasonic thickness value. The principle of ultrasonic thickness measurement is to calculate the ultrasonic thickness data (ultrasonic thickness data is the speed of ultrasonic waves in the metal multiplied by the propagation time divided by 2) using the propagation time of ultrasonic waves in the blade metal. Here, the turbine blade is the target of inspection. The robot is equipped with inspection components (including a laser emitter and a water immersion ultrasonic probe), which move according to the robot trajectory data to measure the thickness of the turbine blade.

[0094] Step 1064: Based on the initial thickness data and the ultrasonic thickness data, obtain the thickness measurement results of the turbine blade.

[0095] Specifically, the initial thickness data from point cloud computing is fused with ultrasonic thickness data (weighting: 70% point cloud thickness, 30% ultrasonic thickness) to obtain the fused thickness value. Then, Gaussian filtering is used to smooth the fused thickness data, eliminating random errors (such as those caused by ultrasonic coupling fluctuations and point cloud noise), resulting in the final turbine blade thickness measurement result. This linkage between point cloud and ultrasonic measurements ensures the accuracy and reliability of the thickness measurement results, providing core data support for blade quality inspection.

[0096] A specific embodiment of the method for measuring the thickness of steam turbine blades according to the present invention includes:

[0097] Step 111: Scan the turbine blades according to the preset scanning parameters to obtain the first point cloud data;

[0098] The turbine blades are fixed on a turntable and rotated at fixed angles in stages. Local point clouds are acquired one by one by a laser scanner according to the preset scanning parameters, and finally the first point cloud data covering the entire blade area is formed.

[0099] Step 112: Perform point cloud preprocessing on the first point cloud data to obtain the second point cloud data;

[0100] The first point cloud data is denoised and simplified to improve the efficiency of subsequent processing.

[0101] Step 113: Perform coarse registration processing on the second point cloud data to obtain the third point cloud data;

[0102] Target point cloud data is extracted based on the theoretical model of turbine blades. Descriptive data is then extracted from the target point cloud data and the second point cloud data (the actual point cloud data acquired, which can be referred to as the source point cloud). The source point cloud and the target point cloud are matched based on the descriptive data.

[0103] Step 114: Perform fine registration processing on the third point cloud data to obtain the fourth point cloud data;

[0104] By iteratively optimizing the acquired objective function, the source point cloud and the target point cloud are precisely registered to obtain the fourth point cloud data.

[0105] Step 115: Perform trajectory conversion based on the fourth point cloud data to obtain robot trajectory data;

[0106] The precisely registered source point cloud data is converted back to the robot's base coordinate system, and pose correction data is calculated based on this.

[0107] Step 116: The robot, equipped with a detection component, measures the thickness of the turbine blade according to the robot trajectory data to obtain the thickness measurement result of the turbine blade.

[0108] Initial thickness data is obtained based on the precisely registered source point cloud data, and the thickness measurement results of the turbine blade are obtained by combining the ultrasonic thickness data.

[0109] The turbine blade thickness measurement method of this invention improves the reliability of the measurement results through a two-level registration, point cloud and ultrasonic linkage thickness measurement method, and can cover the thickness measurement needs of all key areas of the blade (blade root, steam inlet edge, etc.).

[0110] like Figure 2 As shown, an embodiment of the present invention provides a turbine blade thickness measuring device 200, comprising:

[0111] The scanning module 201 is used to scan the turbine blades according to preset scanning parameters to obtain the first point cloud data;

[0112] The processing module 202 is used to perform point cloud preprocessing on the first point cloud data to obtain second point cloud data; perform coarse registration processing on the second point cloud data to obtain third point cloud data; perform fine registration processing on the third point cloud data to obtain fourth point cloud data; perform trajectory conversion based on the fourth point cloud data to obtain robot trajectory data; and have the robot equipped with a detection component perform thickness measurement on the turbine blade according to the robot trajectory data to obtain the thickness measurement result of the turbine blade.

[0113] Optionally, the turbine blades are scanned according to preset scanning parameters to obtain the first point cloud data, including:

[0114] Obtain preset scanning parameters; the preset scanning parameters include a scanning distance of 50 to 100 mm, a point cloud density covering the crown, root, leading edge, and trailing edge of the turbine blade, and a scanning range covering the entire area of ​​the turbine blade;

[0115] The turbine blades fixed at a preset position are scanned according to the preset scanning parameters to obtain the first point cloud data.

[0116] Optionally, point cloud preprocessing is performed on the first point cloud data to obtain second point cloud data, including:

[0117] The first point cloud data is denoised to obtain denoised point cloud data;

[0118] The denoised point cloud data is simplified to obtain the second point cloud data.

[0119] Optionally, coarse registration processing is performed on the second point cloud data to obtain third point cloud data, including:

[0120] Descriptive data is extracted from the second point cloud data and the target point cloud data to obtain descriptive data;

[0121] The second point cloud data and the target point cloud data are matched based on the similarity of the described data to obtain the third point cloud data.

[0122] Optionally, the third point cloud data is subjected to fine registration processing to obtain the fourth point cloud data, including:

[0123] Obtain the objective function; the objective function is Where R is the rotation matrix, p is any point on the turbine blade in the third point cloud data, q is the target point cloud data corresponding to p in the third point cloud data, and t is the translation vector. Let q be the normal vector;

[0124] The objective function is iteratively optimized until the preset convergence condition is met, and the fourth point cloud data is obtained.

[0125] Optionally, trajectory transformation is performed based on the fourth point cloud data to obtain robot trajectory data, including:

[0126] The fourth point cloud data is subjected to coordinate system transformation to obtain transformed point cloud data;

[0127] Based on the transformed point cloud data, determine the pose correction data;

[0128] The preset detection path is corrected based on the pose correction data to obtain robot trajectory data.

[0129] Optionally, the thickness of the turbine blade is measured according to the robot trajectory data to obtain the thickness measurement result of the turbine blade, including:

[0130] The cloud data from the second point is sliced ​​to obtain sliced ​​data;

[0131] Based on the slice data, the initial thickness data is obtained;

[0132] The turbine blades are ultrasonically thickened according to the robot trajectory data to obtain ultrasonic thickness data.

[0133] Based on the initial thickness data and the ultrasonic thickness data, the thickness measurement results of the turbine blades are obtained.

[0134] The turbine blade thickness measuring device of this invention scans the turbine blade according to preset scanning parameters to obtain first point cloud data, performs point cloud preprocessing on the first point cloud data to obtain second point cloud data, performs coarse registration on the second point cloud data to obtain third point cloud data, performs fine registration on the third point cloud data to obtain fourth point cloud data, performs trajectory conversion based on the fourth point cloud data to obtain robot trajectory data, and finally uses a robot equipped with a detection component to measure the thickness of the turbine blade according to the robot trajectory data to obtain the turbine blade thickness measurement result. This method can improve the accuracy and efficiency of turbine blade thickness measurement.

[0135] It should be noted that this device corresponds to the method described above, and all implementations in the method embodiments described above are applicable to the embodiments of this device and can achieve the same technical effect. Further details are omitted in this embodiment.

[0136] This invention also provides a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method as described in any of the above embodiments. All implementations in the above method embodiments are applicable to the embodiments of this device and can achieve the same technical effects. Further details are omitted in this embodiment.

[0137] This invention also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method as described in any of the above embodiments. All implementations in the above method embodiments are applicable to the embodiments of this device and can achieve the same technical effects. Further details are omitted in this embodiment.

[0138] It should be noted that in the apparatus and method of the present invention, the components or steps can obviously be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent solutions of the present invention. Furthermore, the steps for performing the above series of processes can naturally be performed in the order described and in chronological order, but are not necessarily required to be performed in chronological order. Some steps can be performed in parallel, overlapping, or independently of each other.

[0139] It should be noted that in the above embodiments, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments described above is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0140] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for measuring the thickness of steam turbine blades, characterized in that, include: The turbine blades are scanned according to the preset scanning parameters to obtain the first point cloud data; Perform point cloud preprocessing on the first point cloud data to obtain the second point cloud data; The second point cloud data is coarsely registered to obtain the third point cloud data; The third point cloud data is subjected to fine registration processing to obtain the fourth point cloud data; The trajectory data of the robot is obtained by performing trajectory conversion based on the fourth point cloud data; The robot, equipped with a detection component, measures the thickness of the turbine blade according to the robot trajectory data, and obtains the thickness measurement result of the turbine blade. The process involves scanning the turbine blades according to preset scanning parameters to obtain the first point cloud data, including: Obtain preset scanning parameters; the preset scanning parameters include a scanning distance of 50 to 100 mm, a point cloud density covering the crown, root, leading edge, and trailing edge of the turbine blade, and a scanning range covering the entire area of ​​the turbine blade; The turbine blades fixed at a preset position are scanned according to the preset scanning parameters to obtain the first point cloud data. The preset position is a turntable. The turbine blades are fixed on the turntable. The turntable rotates 12 times, 30 degrees each time. Local point clouds are obtained one by one by a laser scanner. Specifically, point cloud preprocessing is performed on the first point cloud data to obtain the second point cloud data, including: The first point cloud data is denoised to obtain denoised point cloud data. A checkerboard calibration board is used, and the robot drives the calibration board to take pictures in different postures. The Zhang Zhengyou calibration method is used to solve the camera intrinsic parameters. Combined with the robot end-effector pose data, the transformation matrix from the camera coordinate system to the robot base coordinate system is solved. According to the target size data of the blade, the boundary range of the three axes XYZ is set, and points in the first point cloud data that exceed the boundary are clipped. The clipped effective point cloud data is analyzed point by point. The number of neighboring points of each point is set to k=15. The average distance between each point and its k nearest neighbors is calculated. The global mean and global standard deviation of the average distance of all points are calculated. If the average distance of a point exceeds the range of the global mean ± 3 times the global standard deviation, it is identified as an outlier and removed. The denoised point cloud data is simplified to obtain the second point cloud data; the three-dimensional space where the clean point cloud is located is divided into a uniform voxel grid; for each voxel grid, the centroid coordinates of all points in the grid are calculated, and the centroid is used as the representative point of the voxel to replace all the original points in the grid. Specifically, trajectory conversion is performed based on the fourth point cloud data to obtain robot trajectory data, including: The fourth point cloud data is subjected to coordinate system transformation to obtain transformed point cloud data; Based on the transformed point cloud data, pose correction data is determined; based on the source point cloud data after coordinate system transformation, the deviation between the actual and theoretical poses of the blade is calculated, including translational and rotational deviations. The pose correction data includes translational and rotational deviations. , , , ( X1, Y1, Z1) is the centroid coordinate of the source point cloud data, ( X2, Y2, Z2) represents the centroid coordinates of the target point cloud data; , , ,in, The Euler angles of the pose of the source point cloud data. The Euler angles of the target point cloud pose; The preset detection path is corrected based on the pose correction data to obtain robot trajectory data; the translational and rotational deviations in the pose correction data are used as robot pose correction values ​​and input into the robot control system to correct the robot's motion posture; based on the surface features of the source point cloud in the converted point cloud data, linear interpolation is performed on the water immersion detection path points in the preset detection path to obtain robot trajectory data. The robot, equipped with a detection component, measures the thickness of the turbine blades according to the robot's trajectory data, obtaining the thickness measurement results of the turbine blades, including: The second point cloud data is sliced ​​to obtain slice data; the second point cloud data in the robot base coordinate system is fitted to obtain a three-dimensional surface model of the blade; based on the three-dimensional surface model of the blade, the edge point clouds of the leading and trailing edges of the blade are extracted; with the edge point clouds of the leading and trailing edges of the blade as the reference, slices are made along the normal direction of the edge, each slice corresponds to a cross section of the blade, and the slice spacing is 0.1 mm to obtain slice data. Based on the slice data, the initial thickness data is obtained; for each slice in the slice data, the inner arc surface points and outer arc surface points on the slice are extracted, and the minimum distance between the two points in the normal direction is calculated to obtain the initial thickness data; The robot, equipped with a detection component, performs ultrasonic thickness measurement on the turbine blade according to the robot trajectory data to obtain ultrasonic thickness data. Based on the initial thickness data and the ultrasonic thickness data, the thickness measurement result of the turbine blade is obtained; the initial thickness data and the ultrasonic thickness data are fused to obtain the fused thickness value; the fused thickness data is smoothed by Gaussian filtering to obtain the thickness measurement result of the turbine blade.

2. The method for measuring the thickness of turbine blades according to claim 1, characterized in that, The second point cloud data is coarsely registered to obtain the third point cloud data, including: Descriptive data is extracted from the second point cloud data and the target point cloud data to obtain descriptive data; The second point cloud data and the target point cloud data are matched based on the similarity of the described data to obtain the third point cloud data.

3. The method for measuring the thickness of turbine blades according to claim 1, characterized in that, The third point cloud data is subjected to fine registration processing to obtain the fourth point cloud data, including: Obtain the objective function; the objective function is Where R is the rotation matrix, p is any point on the turbine blade in the third point cloud data, q is the target point cloud data corresponding to p in the third point cloud data, and t is the translation vector. Let q be the normal vector; The objective function is iteratively optimized until the preset convergence condition is met, and the fourth point cloud data is obtained.

4. A device for measuring the thickness of steam turbine blades, characterized in that, include: The scanning module is used to scan the turbine blades according to preset scanning parameters to obtain the first point cloud data; The processing module is used to perform point cloud preprocessing on the first point cloud data to obtain second point cloud data; perform coarse registration processing on the second point cloud data to obtain third point cloud data; perform fine registration processing on the third point cloud data to obtain fourth point cloud data; perform trajectory conversion based on the fourth point cloud data to obtain robot trajectory data; and have the robot equipped with a detection component perform thickness measurement on the turbine blade according to the robot trajectory data to obtain the thickness measurement result of the turbine blade. The process involves scanning the turbine blades according to preset scanning parameters to obtain the first point cloud data, including: Obtain preset scanning parameters; the preset scanning parameters include a scanning distance of 50 to 100 mm, a point cloud density covering the crown, root, leading edge, and trailing edge of the turbine blade, and a scanning range covering the entire area of ​​the turbine blade; The turbine blades fixed at a preset position are scanned according to the preset scanning parameters to obtain the first point cloud data. The preset position is a turntable. The turbine blades are fixed on the turntable. The turntable rotates 12 times, 30 degrees each time. Local point clouds are obtained one by one by a laser scanner. Specifically, point cloud preprocessing is performed on the first point cloud data to obtain the second point cloud data, including: The first point cloud data is denoised to obtain denoised point cloud data. A checkerboard calibration board is used, and the robot drives the calibration board to take pictures in different postures. The Zhang Zhengyou calibration method is used to solve the camera intrinsic parameters. Combined with the robot end-effector pose data, the transformation matrix from the camera coordinate system to the robot base coordinate system is solved. According to the target size data of the blade, the boundary range of the three axes XYZ is set, and points in the first point cloud data that exceed the boundary are clipped. The clipped effective point cloud data is analyzed point by point. The number of neighboring points of each point is set to k=15. The average distance between each point and its k nearest neighbors is calculated. The global mean and global standard deviation of the average distance of all points are calculated. If the average distance of a point exceeds the range of the global mean ± 3 times the global standard deviation, it is identified as an outlier and removed. The denoised point cloud data is simplified to obtain the second point cloud data; the three-dimensional space where the clean point cloud is located is divided into a uniform voxel grid; for each voxel grid, the centroid coordinates of all points in the grid are calculated, and the centroid is used as the representative point of the voxel to replace all the original points in the grid. Specifically, trajectory conversion is performed based on the fourth point cloud data to obtain robot trajectory data, including: The fourth point cloud data is subjected to coordinate system transformation to obtain transformed point cloud data; Based on the transformed point cloud data, pose correction data is determined; based on the source point cloud data after coordinate system transformation, the deviation between the actual and theoretical poses of the blade is calculated, including translational and rotational deviations. The pose correction data includes translational and rotational deviations. , , , ( X1, Y1, Z1) is the centroid coordinate of the source point cloud data, ( X2, Y2, Z2) represents the centroid coordinates of the target point cloud data; , , ,in, The Euler angles of the pose of the source point cloud data. The Euler angles of the target point cloud pose; The preset detection path is corrected based on the pose correction data to obtain robot trajectory data; the translational and rotational deviations in the pose correction data are used as robot pose correction values ​​and input into the robot control system to correct the robot's motion posture; based on the surface features of the source point cloud in the converted point cloud data, linear interpolation is performed on the water immersion detection path points in the preset detection path to obtain robot trajectory data. The robot, equipped with a detection component, measures the thickness of the turbine blades according to the robot's trajectory data, obtaining the thickness measurement results of the turbine blades, including: The second point cloud data is sliced ​​to obtain slice data; the second point cloud data in the robot base coordinate system is fitted to obtain a three-dimensional surface model of the blade; based on the three-dimensional surface model of the blade, the edge point clouds of the leading and trailing edges of the blade are extracted; with the edge point clouds of the leading and trailing edges of the blade as the reference, slices are made along the normal direction of the edge, each slice corresponds to a cross section of the blade, and the slice spacing is 0.1 mm to obtain slice data. Based on the slice data, the initial thickness data is obtained; for each slice in the slice data, the inner arc surface points and outer arc surface points on the slice are extracted, and the minimum distance between the two points in the normal direction is calculated to obtain the initial thickness data; The robot, equipped with a detection component, performs ultrasonic thickness measurement on the turbine blade according to the robot trajectory data to obtain ultrasonic thickness data. Based on the initial thickness data and the ultrasonic thickness data, the thickness measurement result of the turbine blade is obtained; the initial thickness data and the ultrasonic thickness data are fused to obtain the fused thickness value; the fused thickness data is smoothed by Gaussian filtering to obtain the thickness measurement result of the turbine blade.

5. A computing device, characterized in that, include: A processor, a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The system stores instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 3.

Images