Method, device and equipment for three-dimensional measurement of power plant blade based on camber line constraint and medium

Abstract

The application discloses a kind of based on camber line constraint power device blade three-dimensional measurement method, device, equipment and medium, comprising: respectively the target blade model point cloud of the power device blade to be measured and the blade section point cloud of standard blade model point cloud are carried out part segmentation, based on blade basin point cloud and blade back point cloud extraction camber line point cloud, based on the front and rear edge part point cloud of target blade model point cloud and standard blade model point cloud and camber line point cloud constructs adaptive weight error function, based on the transformation matrix of adaptive weight error function solution pose conversion is carried out to target blade model point cloud, based on the target blade model point cloud after pose conversion and standard blade model point cloud calculates three-dimensional measurement precision, by to blade key part segmentation, using camber line to offset the error generated in registration of overall blade point cloud, accurately calculate the error of the power device blade to be measured, improve the registration precision of power device blade three-dimensional measurement.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional measurement of power plant blades based on mid-arc line constraints, and particularly to a method, apparatus, equipment, and medium for three-dimensional measurement of power plant blades based on mid-arc line constraints. Background Technology

[0002] Currently, with rising labor costs and enterprises' increasing demands for production efficiency, 3D vision measurement technology has been widely applied in the manufacturing processes of various parts. The measurement accuracy of parts determines the accuracy of subsequent processing, thus affecting product quality. While the measurement and processing of simple parts is already very mature, precise measurement of complex curved surfaces such as power unit blades still presents challenges.

[0003] Due to limitations imposed by scanning distance and equipment physical constraints, issues such as uneven point cloud distribution, blurred functional region boundaries, and missing key feature points can arise in critical geometric feature areas of power plant blades (e.g., blade base, blade back, leading edge, and trailing edge). Furthermore, slight deformations occur on the surface and leading and trailing edges during blade manufacturing, making it difficult to establish accurate point correspondences using traditional measurement-based registration methods. This is particularly problematic on the surfaces of weakly textured metal blades, where conventional texture- or curvature-based feature matching methods fail, resulting in a significant decrease in measurement accuracy. Summary of the Invention

[0004] The main objective of this invention is to provide a method, apparatus, equipment, and medium for three-dimensional measurement of power plant blades based on mid-arc line constraints, aiming to solve the technical problem that existing technologies cannot solve the problem of inaccurate segmentation and registration failure caused by feature loss during the measurement of weakly textured metal blades.

[0005] To achieve the above objectives, the present invention provides a three-dimensional measurement method for power plant blades based on mid-arc line constraints, the method comprising the following steps:

[0006] A line structured light scanner was used to scan and model the blade of the power unit under test, and the point cloud of the target blade model was obtained.

[0007] The point cloud of the target blade model and the point cloud of the standard blade model are respectively divided into the point cloud of the blade basin, the point cloud of the blade back, and the point cloud of the leading and trailing edges. The point cloud of the standard blade model is obtained by scanning and modeling a standard power unit blade in advance.

[0008] Based on the matching point pairs between the leaf basin point cloud and the leaf back point cloud, the mid-arc point cloud of the target leaf model point cloud and the standard leaf model point cloud are extracted respectively.

[0009] An adaptive weighted error function is constructed based on the point clouds of the leading and trailing edges and the mid-arc line of the target blade model point cloud and the standard blade model point cloud.

[0010] The adaptive weight error function is solved using the Gauss-Newton optimization algorithm to obtain the transformation matrix;

[0011] The target blade model point cloud is pose-transformed based on the transformation matrix, and the three-dimensional measurement accuracy of the power unit blade under test is calculated based on the pose-transformed target blade model point cloud and the standard blade model point cloud.

[0012] Optionally, the step of using a line structured light scanner to scan and model the blade of the power unit under test to obtain a point cloud of the target blade model includes:

[0013] The line structured light scanner is automatically calibrated using a calibration plate.

[0014] The position and angle of the blades of the power unit under test, as well as the brightness of the line structured light scanner, are adjusted.

[0015] In response to meeting the preset scanning conditions, the turntable stepping angle is set to the preset angle, and the line structured light scanner is started to perform multi-view scanning on the blade of the power device under test. The preset adjustment conditions include: the laser emitted by the line structured light scanner overlaps on the blade, and there is a preset number of identifiable marker points in the image preview field of view.

[0016] The point cloud images of the blade of the power unit under test obtained from multiple perspective scans are stitched together to obtain the point cloud of the target blade model.

[0017] Optionally, the step of segmenting the leaf cross-section point cloud of the target leaf model point cloud and the standard leaf model point cloud into the leaf basin point cloud, the leaf back point cloud, and the leading and trailing edge point cloud respectively includes:

[0018] Extract the key geometric features of each layer of the point cloud of the target blade model and the point cloud of the standard blade model. The key geometric features include the total number of points, the span in the X direction, and the span in the Y direction.

[0019] Based on the key geometric features, the variation trend characteristics between adjacent layers in the point cloud of the target blade model and the point cloud of the standard blade model are analyzed to identify the structural abrupt changes between the tenon and the blade body.

[0020] Based on the reference plane and search accuracy, the blade cross-section contour point cloud is extracted, and the blade cross-section contour point cloud is projected onto a two-dimensional coordinate system to obtain a two-dimensional point cloud.

[0021] The blade boundary contour is extracted from the two-dimensional point cloud based on the α-shape algorithm;

[0022] The leading edge endpoint and trailing edge endpoint are determined based on the change in the directional angle between adjacent line segments on the blade boundary profile.

[0023] The leading edge boundary point and the trailing edge boundary point are determined based on the chord connecting the leading edge endpoint and the trailing edge endpoint;

[0024] Based on the leading edge endpoint, the trailing edge endpoint, the leading edge boundary point, and the trailing edge boundary point, the leading and trailing edge point clouds of the target blade model point cloud and the standard blade model point cloud are constructed respectively, and the leading and trailing edge point clouds include the leading edge point cloud and the trailing edge point cloud.

[0025] Based on the point clouds of the leading and trailing edges, the point clouds of the target blade model and the standard blade model are segmented, and the remaining regions to be segmented are extracted.

[0026] The remaining regions to be segmented are fitted with a high-order parabola, and the point clouds of the target leaf model and the standard leaf model are segmented according to the opening orientation of the parabola fitting, to obtain the leaf basin point cloud and the leaf back point cloud.

[0027] Optionally, the step of extracting the mid-arc point clouds of the target leaf model point cloud and the standard leaf model point cloud based on the matching point pairs between the leaf basin point cloud and the leaf underside point cloud includes:

[0028] The leaf basin point cloud and the leaf underside point cloud are sorted to obtain the leaf basin point set and the leaf underside point set.

[0029] Search for the Euclidean distance between each point in the leaf basin point set and each point in the leaf underside point set to obtain the search results;

[0030] Based on the search results, a matching point pair between the leaf basin point set and the leaf underside point set is determined;

[0031] Construct the intersection of the perpendicular bisectors of each matching point pair, and extract the mid-arc point cloud of the target blade model point cloud and the standard blade model point cloud based on the intersection of the perpendicular bisectors.

[0032] Optionally, the step of constructing an adaptive weighted error function based on the point clouds of the leading and trailing edges and the mid-arc line of the target blade model point cloud and the standard blade model point cloud includes:

[0033] A point-to-point error term is constructed based on the point clouds of the leading and trailing edges of the target blade model point cloud and the point clouds of the leading and trailing edges of the standard blade model point cloud.

[0034] A point-to-surface error term is constructed based on the mid-arc point cloud of the target blade model point cloud and the mid-arc point cloud of the standard blade model point cloud.

[0035] The point-to-point error term and the point-to-surface error term are fused based on the error complementary weights to obtain an adaptive weighted error function.

[0036] Optionally, the step of using the Gauss-Newton optimization algorithm to solve the adaptive weight error function to obtain the transformation matrix includes:

[0037] The adaptive weight error function is expanded using the Gauss-Newton optimization algorithm, transforming it into the product of the residual vector and the corresponding Jacobian matrix. An iterative equation is then constructed based on the product of the residual vector and the Jacobian matrix, where the rigid body transformation parameters of the iterative equation are rotation vectors and translation vectors.

[0038] The rotation and translation matrices are solved iteratively for the iterative equation, and the rigid body transformation parameters are updated accordingly.

[0039] If the current iteration number of the iterative equation reaches the upper limit of iteration, or if the point-to-point error term and the point-to-surface error term simultaneously satisfy the convergence threshold condition, then the transformation matrix is ​​output.

[0040] Optionally, the step of performing pose transformation on the target blade model point cloud based on the transformation matrix, and calculating the three-dimensional measurement accuracy of the power unit blade under test based on the pose-transformed target blade model point cloud and the standard blade model point cloud, includes:

[0041] The pose transformation of the point cloud of the target blade model is performed based on the transformation matrix;

[0042] The matching points in the standard blade model point cloud corresponding to each point in the target blade model point cloud after pose transformation are determined by the nearest neighbor search algorithm, thereby obtaining the matching point pairs between the target blade model point cloud after pose transformation and the standard blade model point cloud.

[0043] The root mean square error between the point cloud of the target blade model after pose transformation and the point cloud of the standard blade model is calculated based on the Euclidean distance of each matching point pair, and the three-dimensional measurement accuracy of the blade of the power device under test is calculated based on the root mean square error.

[0044] Furthermore, to achieve the above objectives, the present invention also proposes a three-dimensional measurement device for power plant blades based on mid-arc line constraints, wherein the three-dimensional measurement device for power plant blades based on mid-arc line constraints includes:

[0045] The point cloud modeling module is used to scan and model the blades of the power unit under test using a line structured light scanner to obtain the point cloud of the target blade model.

[0046] The blade segmentation module is used to segment the blade cross-section point cloud of the target blade model point cloud and the standard blade model point cloud into the blade basin point cloud, the blade back point cloud, and the leading and trailing edge point cloud, respectively. The standard blade model point cloud is obtained by scanning and modeling a standard power unit blade in advance.

[0047] The mid-arc line extraction module is used to extract the mid-arc line point clouds of the target leaf model point cloud and the standard leaf model point cloud based on the matching point pairs between the leaf basin point cloud and the leaf back point cloud.

[0048] The error function construction module is used to construct an adaptive weighted error function based on the point clouds of the leading and trailing edges and the mid-arc line of the target blade model point cloud and the point cloud of the standard blade model point cloud.

[0049] The transformation matrix construction module is used to solve the adaptive weight error function using the Gauss-Newton optimization algorithm to obtain the transformation matrix;

[0050] The blade 3D registration module is used to perform pose transformation on the point cloud of the target blade model based on the transformation matrix, and to calculate the 3D measurement accuracy of the blade of the power device under test based on the point cloud of the target blade model after pose transformation and the point cloud of the standard blade model.

[0051] Furthermore, to achieve the above objectives, this application also proposes a three-dimensional measurement device for power plant blades based on mid-arc line constraints. The device includes: a memory, a processor, and a three-dimensional measurement program for power plant blades based on mid-arc line constraints stored in the memory. The processor is used to run the three-dimensional measurement program for power plant blades based on mid-arc line constraints. The computer program is configured to implement the steps of the three-dimensional measurement method for power plant blades based on mid-arc line constraints as described above.

[0052] In addition, to achieve the above objectives, this application also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the three-dimensional measurement method for power unit blades based on mid-arc constraints as described above.

[0053] This invention employs a line structured light scanner to scan and model the blade of a power unit under test, obtaining a target blade model point cloud. The blade cross-section point clouds of both the target and standard blade model point clouds are segmented into blade base point clouds, blade back point clouds, and leading and trailing edge point clouds. The standard blade model point cloud is obtained by pre-scanning and modeling a standard power unit blade. Based on the matching point pairs between the blade base and blade back point clouds, the mid-arc point clouds of both the target and standard blade model point clouds are extracted. An adaptive weighted error function is constructed based on the leading and trailing edge point clouds and mid-arc point clouds of both the target and standard blade model point clouds. The adaptive weight error function is solved using the Gauss-Newton optimization algorithm to obtain the transformation matrix. Based on the transformation matrix, the point cloud of the target blade model is pose-transformed. The 3D measurement accuracy of the power unit blade is calculated based on the pose-transformed point cloud of the target blade model and the point cloud of the standard blade model. Since this invention replaces the overall blade profile by segmenting and extracting the mid-arc line and leading and trailing edges at different heights, and uses the mid-arc line to offset the error generated by the overall blade point cloud in the registration, it effectively solves the problem of inaccurate segmentation and registration failure caused by feature loss in traditional measurement of weakly textured metal blades, and significantly improves the high-precision registration of 3D measurement of power unit blades. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of the structure of a three-dimensional measurement device for power unit blades based on mid-arc line constraints, which is part of the hardware operating environment of the embodiment of the present invention.

[0056] Figure 2 This is a flowchart illustrating the first embodiment of the three-dimensional measurement method for power unit blades based on mid-arc line constraints according to the present invention.

[0057] Figure 3 This is a schematic diagram of the process of error term complementarity fusion and iterative solution in one embodiment of the three-dimensional measurement method for power plant blades based on mid-arc line constraints of the present invention.

[0058] Figure 4 This is a flowchart illustrating the second embodiment of the three-dimensional measurement method for power unit blades based on mid-arc line constraints according to the present invention.

[0059] Figure 5This is a flowchart illustrating the third embodiment of the three-dimensional measurement method for power unit blades based on mid-arc line constraints according to the present invention.

[0060] Figure 6 This is a structural block diagram of the first embodiment of the three-dimensional measurement device for power unit blades based on mid-arc line constraints of the present invention.

[0061] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0062] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0063] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a three-dimensional measurement device for power unit blades based on mid-arc line constraints, which is part of the hardware operating environment of the embodiment of the present invention.

[0064] like Figure 1 As shown, the three-dimensional measurement device for power unit blades based on mid-arc constraints may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0065] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the three-dimensional measurement equipment for power plant blades based on mid-arc constraints, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0066] like Figure 1As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a three-dimensional measurement program for power unit blades based on mid-arc constraints.

[0067] exist Figure 1 In the three-dimensional measurement device for power plant blades based on mid-arc line constraints shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and memory 1005 in the three-dimensional measurement device for power plant blades based on mid-arc line constraints can be set in the device. The three-dimensional measurement device for power plant blades based on mid-arc line constraints calls the three-dimensional measurement program for power plant blades based on mid-arc line constraints stored in the memory 1005 through the processor 1001, and executes the three-dimensional measurement method for power plant blades based on mid-arc line constraints provided in the embodiment of the present invention.

[0068] This invention provides a three-dimensional measurement method for power unit blades based on mid-arc line constraints, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the three-dimensional measurement method for power unit blades based on mid-arc line constraints according to the present invention.

[0069] In this embodiment, the three-dimensional measurement method for power unit blades based on mid-arc line constraints includes the following steps:

[0070] Step S10: Use a line structured light scanner to scan and model the blade of the power unit under test to obtain the point cloud of the target blade model.

[0071] It should be noted that this embodiment is applied to the three-dimensional measurement of power plant blades. This embodiment is based on a high-precision segmentation algorithm for weak textures on the measurement surface and a three-dimensional measurement algorithm for blade profiles with mid-arc constraints, thereby achieving high-precision three-dimensional measurement of power plant blades. Traditional measurement methods include manual inspection such as scale detection and visual inspection with lenses, but these methods have certain drawbacks. Manual inspection methods are inefficient and difficult to measure some key parameters in complex curved surfaces. With the development of three-dimensional measurement technology, point cloud registration technology is mainly used in industrial production to measure complex curved surfaces. Existing point cloud registration technology mainly adopts rigid body transformation estimation methods based on iterative optimization. Among them, the Iterative Closest Point (ICP) and its improved algorithm that relies only on a single error quantity have high requirements for initial pose, are prone to getting trapped in local optima, and are easily affected by noise. To address the above problems, some embodiments use the generalized ICP (GICP) algorithm, which introduces local normal or covariance constraints of points and integrates point-to-point and point-to-surface errors, but lacks adaptive adjustment. Registration methods based on probabilistic models treat point clouds as probability distributions and then optimize them using expectation maximization or quasi-Newton methods. These methods perform well when processing sparse point clouds, but are slow when dealing with dense point cloud data with complex structures. This implementation optimizes and improves three aspects: key component extraction, registration, and measurement analysis, greatly enhancing measurement accuracy, efficiency, and robustness.

[0072] It should be understood that the executing entity of this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or a terminal electronic device capable of performing the above functions. The following description uses a three-dimensional measurement device for power unit blades based on mid-arc constraints (hereinafter referred to as the measurement device) as an example to illustrate this embodiment and the following embodiments.

[0073] It should be noted that a line structured light scanner is a non-contact 3D measurement device based on the principle of optical triangulation. It reconstructs the 3D shape by projecting a laser line and analyzing its deformation on the object's surface. The point cloud of the target blade model mentioned above can be a set of point clouds obtained after scanning the actual blade of the power unit under test. Each point contains three-dimensional coordinates (x, y, z) and represents the geometric shape of the blade surface.

[0074] In some embodiments, the measuring device uses a high-precision line structured light scanner to scan the blade from multiple angles along a preset trajectory, and achieves 360° blind-spot-free coverage by rotating the worktable; the laser intensity is automatically adjusted during the scanning process to adapt to the reflective characteristics of the metal blade; for example, the ICP (Iterative Closest Point) algorithm can be used to fuse multi-view point cloud data, and outlier noise points can be removed by statistical filtering to finally generate a complete and dense target blade point cloud model.

[0075] Furthermore, to construct a high-precision blade model point cloud, step S10 above may include:

[0076] Step S101: Automatically calibrate the line structured light scanner using a calibration plate;

[0077] Step S102: Adjust the blade pose and angle of the power device under test, as well as the brightness of the line structured light scanner;

[0078] Step S103: In response to meeting the preset scanning conditions, the turntable stepping angle is set to the preset angle, and the line structured light scanner is started to perform multi-view scanning on the blade of the power device under test.

[0079] Step S104: The point cloud images of the blade of the power device under test obtained from multiple perspective scans are stitched together to obtain the point cloud of the target blade model.

[0080] It should be noted that the preset adjustment conditions include: the laser emitted by the line structured light scanner overlaps on the blade, and there is a preset number of identifiable markers in the image preview field of view.

[0081] In practical implementation, the measuring equipment can use a calibration plate to automatically calibrate the line structured light scanner; adjust the pose and angle of the blade of the power unit under test so that the laser emitted by the line structured light scanner can overlap on the blade; based on the laser emitted by the line structured light scanner being able to overlap on the blade, adjust the brightness of the line structured light scanner to ensure that there are a preset number of identifiable marker points in the image preview field of view, and that the blade image is clear; based on the preset number of identifiable marker points and the clear blade image, set the turntable step angle to 8° (set to 3° at the leading and trailing edges), and start the line structured light scanner to scan the blade; stitch together the point cloud images scanned from multiple different perspectives to obtain the target blade model point cloud P1.

[0082] Step S20: Divide the point cloud of the target blade model and the point cloud of the standard blade model into the blade basin point cloud, the blade back point cloud and the point cloud of the leading and trailing edges, respectively. The point cloud of the standard blade model is obtained by scanning and modeling the standard power unit blade in advance.

[0083] It should be noted that the blade basin point cloud can be a collection of point clouds on the inner side of the blade (airflow pressure surface), typically exhibiting a concave surface shape. The blade back point cloud can be a collection of point clouds on the outer side of the blade (airflow suction surface), typically exhibiting a convex surface shape. The standard blade model point cloud can be a reference point cloud obtained by pre-scanning and modeling a known, qualified standard blade.

[0084] In some embodiments, the measuring device can set a reference plane based on the flatness of the blade bottom and perform equidistant layering along the normal vector direction; extract features such as the X / Y span in each layer of point cloud; identify structural abrupt changes in the tenon and blade by analyzing the changing trends of adjacent layers; project the blade cross-section point cloud onto the xOy plane; extract the contour using the α-shape algorithm; based on the characteristic that the curvature of the leading edge of the blade is greater than that of the trailing edge, select the two intersection points with the largest angle abrupt change as the leading edge endpoints, and the second largest as the trailing edge endpoints; finally, distinguish the blade base and blade back by fitting a high-order parabola, thereby solving the problem of region division caused by uneven point cloud distribution in weakly textured blades, achieving accurate segmentation of functional areas, and providing a reliable foundation for subsequent feature extraction. The xOy plane can be a horizontal reference plane with Z=0 in a three-dimensional Cartesian coordinate system, used to simplify contour analysis.

[0085] Step S30: Based on the matching point pairs between the leaf basin point cloud and the leaf back point cloud, extract the mid-arc point cloud of the target leaf model point cloud and the standard leaf model point cloud respectively.

[0086] It should be noted that the mid-arc point cloud can be the set of corresponding midpoints of the leaf base and the leaf back, representing the geometric center line of the blade cross section, and is the core point cloud representing the symmetry of the blade.

[0087] In some embodiments, the measuring device can calculate the midpoint coordinates of the matching point pairs in the leaf back point cloud for each point in the leaf basin point cloud using the nearest neighbor matching algorithm, and set all midpoints to form the mid-arc point cloud, thereby establishing a stable geometric benchmark that is not affected by slight surface deformation, overcoming the feature extraction failure problem caused by the lack of key points in traditional methods.

[0088] Step S40: Construct an adaptive weighted error function based on the point clouds of the leading and trailing edges and the mid-arc line of the target blade model point cloud and the standard blade model point cloud.

[0089] It should be noted that the adaptive weight error function can be an optimization objective function that dynamically adjusts the weights according to the importance of functional regions, ensuring that key regions dominate the registration process.

[0090] In some embodiments, the measurement device can use the leading and trailing edge point clouds and the mid-arc point cloud as feature point sets, and dynamically assign weights according to the importance of the regions to construct a weighted error function. This solves the problem of differences in importance between different regions, allowing the optimization process to focus on the key regions that have the greatest impact on aerodynamic performance, and significantly improves the registration reliability in weak texture scenes.

[0091] Furthermore, in order to effectively perform error complementarity fusion and improve error capture accuracy, step S40 above may include:

[0092] Step S401: Construct a point-to-point error term based on the point clouds of the leading and trailing edges of the target blade model point cloud and the point clouds of the leading and trailing edges of the standard blade model point cloud;

[0093] Step S402: Construct a point-to-surface error term based on the mid-arc point cloud of the target blade model point cloud and the mid-arc point cloud of the standard blade model point cloud;

[0094] Step S403: Based on the error complementary weights, the point-to-point error term and the point-to-surface error term are fused to obtain an adaptive weighted error function.

[0095] Understandably, referring to Figure 3 , Figure 3 This is a schematic diagram illustrating the process of complementary fusion and iterative solution of error terms in one embodiment. The measuring device acquires point clouds of the leading and trailing edges and the mid-arc at different heights, using them to represent the overall blade profile. The leading and trailing edges of the target blade point cloud and the standard blade point cloud are represented by A1 and B1, respectively, and the mid-arc portions of the target blade point cloud and the standard blade point cloud are represented by A2 and B2, respectively. Error functions are constructed using the leading and trailing edge point clouds and the mid-arc point clouds extracted from the target blade model point cloud P1 and the standard blade model point cloud Q at different heights. Specifically, the mid-arc point clouds are used to construct the point-to-surface error, and the leading and trailing edge point clouds are used to construct the point-to-point error.

[0096] In the specific implementation, the point-to-point error is constructed for the leading and trailing edges of both objects, in the form of:

[0097]

[0098] in Here is the transformed measurement point cloud, and here is the target point cloud.

[0099] The point-to-surface error term in the mid-arc point cloud construction for both is expressed as:

[0100]

[0101] in It is the estimated normal vector at the arc point in the target.

[0102] These two types of errors are merged into a single error function, and complementary weights are applied to each type of error based on the current error, constructing a complementary adaptive weighted error function. The relative weights of the point-to-point error and the point-to-surface error are adaptively calculated based on the error value in the current iteration.

[0103]

[0104] Thus, the overall adaptive weight error function is constructed:

[0105]

[0106] Step S50: Solve the adaptive weight error function using the Gauss-Newton optimization algorithm to obtain the transformation matrix.

[0107] It should be noted that the transformation matrix can be a rigid transformation matrix, containing rotation and translation parameters, used for coordinate system transformation. The Gauss-Newton optimization algorithm is an iterative algorithm for solving nonlinear least squares problems, achieving fast convergence of the adaptive weight error function by linearizing the error function.

[0108] In some embodiments, the measuring device constructs normal equations by parameterizing rigid body transformation into rotation and translation vectors through Taylor expansion, solves for parameter update amounts, and converts rotation vector increments into rotation matrix increments through exponential mapping; iterative optimization continues until convergence conditions are met.

[0109] Furthermore, in order to accurately align the target blade point cloud to the coordinate system of the standard blade point cloud, step S50 above may include:

[0110] Step S501: The adaptive weight error function is expanded using the Gauss-Newton optimization algorithm to transform it into the product of the residual vector and the corresponding Jacobian matrix. An iterative equation is constructed based on the product of the residual vector and the Jacobian matrix. The rigid body transformation parameters of the iterative equation are rotation vectors and translation vectors.

[0111] Step S502: Iteratively solve the rotation and translation matrices of the iterative equation and update the rigid body transformation parameters;

[0112] Step S503: If the current iteration number of the iterative equation reaches the upper limit of iteration, or if the point-to-point error term and the point-to-surface error term simultaneously satisfy the convergence threshold condition, then output the transformation matrix.

[0113] In the specific implementation, refer to Figure 3 The measuring device constructs an iterative equation based on a linearized approximation of the adaptive weighted error function. Assuming the rigid body transformation parameters are the rotation vector w and the translation vector t, a first-order approximation of the error function is performed using Taylor expansion to obtain the residual vector r and the corresponding Jacobian matrix J. This leads to the normal equation:

[0114]

[0115] in This is the combined update amount for rotation and translation.

[0116] Then, update the rigid body transformation, find the optimal rigid body transformation matrix, and precisely align the target blade point cloud to the coordinate system of the standard blade point cloud. Increment the obtained rotation vector. Transformed into rotation matrix increments via exponential mapping:

[0117]

[0118] And update the current rotation and translation matrices:

[0119]

[0120] in To ensure that the updated rotation matrix remains orthogonal, it is an antisymmetric matrix.

[0121] Finally, an iterative decision is made. If the number of iterations k reaches the upper limit, or the two types of errors meet the convergence threshold condition, that is:

[0122] k ,or

[0123] If the registration process is considered converged, the optimal transformation matrix is ​​output, achieving accurate registration between the source and target point clouds. If the iteration termination condition is not met, the weights and iteration count k are updated based on the current error after each iteration, and then the iteration is repeated.

[0124] Step S60: Perform pose transformation on the point cloud of the target blade model based on the transformation matrix, and calculate the three-dimensional measurement accuracy of the blade of the power device under test based on the point cloud of the target blade model after pose transformation and the point cloud of the standard blade model.

[0125] It should be noted that three-dimensional measurement accuracy can be an indicator that evaluates the accuracy of measurement results by quantifying geometric deviations, and is used to assess the geometric error between the blade of the power unit under test and the blade of a standard power unit.

[0126] In some embodiments, the measuring device completes pose transformation by applying a transformation matrix to the target blade point cloud, calculates the root mean square error (RMSE) between the transformed point cloud and the standard point cloud, generates a blade surface deviation heat map based on the error distribution, and calculates the three-dimensional measurement accuracy of the power unit blade under test. This eliminates scanning and installation errors, realizes an objective quantitative assessment of manufacturing quality, and provides a reliable basis for quality control.

[0127] Furthermore, in order to accurately calculate the measurement accuracy, step S60 above may include:

[0128] Step S601: Perform pose transformation on the point cloud of the target blade model based on the transformation matrix;

[0129] Step S602: Determine the matching points in the standard blade model point cloud corresponding to each point in the target blade model point cloud after pose transformation using the nearest neighbor search algorithm, and obtain the matching point pairs between the target blade model point cloud after pose transformation and the standard blade model point cloud;

[0130] Step S603: Calculate the root mean square error between the point cloud of the target blade model after pose transformation and the point cloud of the standard blade model based on the Euclidean distance of each matching point pair, and calculate the three-dimensional measurement accuracy of the blade of the power device under test based on the root mean square error.

[0131] In the specific implementation, the pose of the target blade point cloud is adjusted by applying the optimal transformation matrix, and the nearest neighbor search method is used to find the nearest point in the standard blade point cloud corresponding to each point in the target blade point cloud after the pose adjustment.

[0132] Step S52: Calculate the root mean square error between the target point cloud after pose adjustment and the standard point cloud to obtain the measurement accuracy. For the registered point cloud, For standard point clouds, Let be the Euclidean distance between the i-th pair of points. Then the root mean square error is defined as:

[0133]

[0134] in, The root mean square error is a core indicator used to quantify the geometric deviation between the blade under test and the standard blade. The smaller the value, the higher the measurement accuracy. This represents the first point in the target leaf point cloud after pose transformation. One point; Indicating standard point cloud and Corresponding points; This represents the total number of valid point pairs in the point cloud.

[0135] This embodiment uses a line structured light scanner to scan and model the blade of the power unit under test, obtaining a target blade model point cloud. The blade cross-section point clouds of both the target and standard blade model point clouds are segmented into blade base point clouds, blade back point clouds, and leading and trailing edge point clouds. The standard blade model point cloud is obtained by pre-scanning and modeling a standard power unit blade. Based on the matching point pairs between the blade base and blade back point clouds, the mid-arc point clouds of both the target and standard blade model point clouds are extracted. An adaptive weighted error function is constructed based on the leading and trailing edge point clouds and mid-arc point clouds of both the target and standard blade model point clouds. The adaptive weight error function is solved using the Gauss-Newton optimization algorithm to obtain the transformation matrix. Based on the transformation matrix, the target blade model point cloud is pose-transformed, and the 3D measurement accuracy of the power unit blade is calculated based on the pose-transformed target blade model point cloud and the standard blade model point cloud. Since this embodiment uses the mid-arc line and leading and trailing edge parts of different heights to replace the overall blade profile, and uses the mid-arc line to offset the error generated by the overall blade point cloud in the registration, it effectively solves the problem of inaccurate segmentation and registration failure caused by feature loss in traditional measurement of weak textured metal blades, and significantly improves the high-precision registration of 3D measurement of power unit blades.

[0136] refer to Figure 4 , Figure 4 This is a flowchart illustrating the second embodiment of the three-dimensional measurement method for power unit blades based on mid-arc line constraints according to the present invention.

[0137] Based on the first embodiment described above, in this embodiment, step S20 further includes:

[0138] Step S201: Extract the key geometric features of each layer of the point cloud of the target blade model and the standard blade model.

[0139] It should be noted that the key geometric features include the total number of points, the span in the X direction, and the span in the Y direction.

[0140] In practical implementation, the measuring equipment can utilize the flat bottom of the power unit blades to find a reference plane at the bottom. A fixed layer thickness, set along the normal vector direction of the reference plane, is used to create equally spaced layers of point cloud. Key geometric features are extracted and statistically analyzed for each layer, and the total number of points is calculated. The X-axis span and Y-axis span are used to transform disordered point clouds into structured hierarchical data, providing a quantitative geometric basis for subsequent analysis and effectively avoiding the difficulty of region division caused by feature loss in weakly textured leaves.

[0141] Step S202: Based on the key geometric features, analyze the variation trend characteristics between adjacent layers in the point cloud of the target blade model and the point cloud of the standard blade model, and identify the structural abrupt changes between the tenon and the blade body.

[0142] Understandably, the measuring equipment can calculate the changes in the number of points between adjacent layers and the changes in the X / Y span, analyze the changing trends between adjacent layers, and identify the structural abrupt changes in the tenon and the blade, thereby achieving the purpose of separating the tenon and the blade. This enables precise separation of the tenon and the blade, avoids interference from irrelevant areas on key measurements, and lays the foundation for subsequent analysis focusing on the blade section.

[0143] In practice, the measuring equipment extracts the number of points between layers and the rate of change of the lateral range based on the layer-by-layer scanning point cloud data along the normal vector direction of the reference plane to identify the structural boundary location. The specific steps include the following:

[0144] Step a1: Scan layer by layer along the normal vector direction of the reference plane:

[0145] The blade point cloud data is read from top to bottom along the normal vector direction of the reference plane, and the coordinate information of each point is obtained. The scanning layer thickness is set to... The system is divided into K layers. Key parameters of the leaf point cloud data in each layer are calculated, including the number of points. X-coordinate range Y-coordinate range The coordinate range uses the difference between the maximum and minimum values, calculated using the following formula:

[0146]

[0147]

[0148] Step a2: Calculate the rate of change:

[0149] A sequence consisting of point cloud quantity and lateral size Calculate the rate of change between adjacent layers:

[0150]

[0151] Step a3: Determine the boundary:

[0152] Set a threshold for the rate of change; if the condition is met at a certain level... The following conditions must be met simultaneously:

[0153] Points change:

[0154] Horizontal width variation:

[0155] Then this height This is the dividing point between the blade and the tenon.

[0156] Step S203: Extract the blade cross-sectional profile point cloud based on the reference plane and search accuracy, and project the blade cross-sectional profile point cloud onto a two-dimensional coordinate system to obtain a two-dimensional point cloud.

[0157] It should be noted that the two-dimensional coordinate system can be the xOy plane, which can be the horizontal reference plane with Z=0 in the three-dimensional rectangular coordinate system, used to simplify contour analysis.

[0158] Understandably, the measuring device can define the reference plane as the bottom surface of the point cloud and its parallel surface when the blade is placed horizontally, define the search precision p as the distance value in the direction of the normal vector of the specified height surface, extract the point cloud of the blade section, and project the extracted blade section contour point cloud onto the xOy plane, thereby reducing the complex three-dimensional problem to two-dimensional analysis, greatly reducing the computational complexity, while retaining the core geometric features of the blade section, creating conditions for accurate contour extraction.

[0159] Step S204: Extract the blade boundary contour from the two-dimensional point cloud based on the α-shape algorithm.

[0160] It should be noted that the α-shape algorithm is a point cloud boundary extraction method based on Delaunay triangulation, and the parameter α controls the degree of detail preservation of the contour.

[0161] In the specific implementation, the measuring device dynamically sets the α parameter (such as 5% of the blade width) according to the blade size; constructs the Delaunay triangulation of the point cloud and filters triangles covered by empty spheres with a radius ≤ α; extracts the boundary edges to form a closed contour; filters isolated edges through angle threshold to ensure the continuity of the contour, so that the blade boundary can still be accurately extracted under weak texture conditions, effectively distinguishing between sharp leading edges and gentle trailing edges, and solving the boundary blurring problem caused by feature loss in traditional methods.

[0162] Step S205: Determine the leading edge endpoint and trailing edge endpoint based on the change in the directional angle between adjacent line segments on the blade boundary profile.

[0163] It should be noted that the leading edge endpoint is the sharpest point at the very tip of the blade, the first point the airflow contacts. The trailing edge endpoint is the sharpest point at the very back of the blade, the last point the airflow leaves. The change in directional angle is the difference in angle between the directional vectors of adjacent line segments on the boundary profile, reflecting changes in local curvature.

[0164] Understandably, the measuring equipment can use the α-shape algorithm to extract the contour, calculate the change in the angle between adjacent line segments on the boundary, and, based on the characteristic that the curvature of the leading edge of the blade is greater than that of the trailing edge, select the intersection of the two adjacent line segments with the largest angle change as the leading edge endpoint, and select the intersection of the two adjacent line segments with the second largest angle change as the trailing edge endpoint.

[0165] In practical implementation, the measuring device reads the point set of the two-dimensional cross-sectional point cloud data. The coordinate sequence of the outer contour boundary points is extracted using the α-shape algorithm as follows:

[0166]

[0167] Calculate the vector increment of adjacent points on the boundary:

[0168]

[0169] Then calculate the direction angle of adjacent line segments:

[0170]

[0171] Since the boundary is a closed curve, the change in direction (angle change) between adjacent line segments is calculated as follows:

[0172]

[0173] The `unwrap` function is used to eliminate abrupt changes in the azimuth angle, making the angle change continuous. The changes are sorted in descending order of magnitude, and the two groups of line segments with the largest changes are extracted. The intersection points of these segments are the leading and trailing edge endpoints of the blade (the largest value is the leading edge endpoint, and the second largest is the trailing edge endpoint).

[0174] Step S206: Determine the leading edge boundary point and the trailing edge boundary point based on the chord connecting the leading edge endpoint and the trailing edge endpoint.

[0175] It should be noted that the chord can be a straight line connecting the two ends of the leading or trailing edge, serving as a boundary reference. Boundary points are used to define the feature points of the leading and trailing edge regions, and are used to divide functional areas.

[0176] It should be understood that the measuring device can use the chord formed by connecting the extracted front and rear edge endpoints to cut the boundary point A at the front edge according to its length ratio. Then, it can search for candidate points B in the neighborhood of point A and select the point B that has the closest angle to the direction of the front edge endpoint to 90°. The same method is used to find the boundary point for the rear edge, thereby accurately defining the boundary of the front and rear edge areas and avoiding misclassifying the leaf base / leaf back into the front and rear edge areas, thus ensuring the accuracy of functional area segmentation.

[0177] In its implementation, the measuring device connects the leading and trailing edge endpoints LE and TE, and its chord length is defined as:

[0178]

[0179] On the chord, according to the set proportional coefficient and To determine the boundary points near the leading edge and the trailing edge, respectively, denoted as:

[0180]

[0181] The typical range of the proportionality coefficient is within At a given chord position point A ( and At point A, with point A as the center and a radius equal to the chord length... ( The range is 0.10–0.25, in the point cloud set. Search candidate point set in the middle:

[0182]

[0183] For each candidate point B, construct the thickness direction vector. And calculate the midpoint of this thickness segment:

[0184]

[0185] Then define the reference vector using the reference edge point R:

[0186]

[0187] The reference edge point R is either the leading edge point LE or the trailing edge point TE, and the specific choice depends on whether the chord position point A is located in the leading edge portion or the trailing edge portion.

[0188] The angle between the candidate thickness vector w and the reference vector u can be obtained by using the dot product of the vectors:

[0189]

[0190] The absolute value operation is used to ensure directional symmetry. Based on this, the angular deviation metric is defined as follows:

[0191]

[0192] The system calculates the aforementioned angular deviation for each candidate point and selects the one that makes the deviation acceptable. The point with the smallest value is taken as the optimal matching point:

[0193]

[0194] Step S207: Construct the point clouds of the leading and trailing edges of the target blade model point cloud and the standard blade model point cloud based on the leading edge endpoint, the trailing edge endpoint, the leading edge boundary point, and the trailing edge boundary point.

[0195] It should be noted that the point clouds at the leading and trailing edges include both the leading edge point cloud and the trailing edge point cloud. The leading edge point cloud can be the set of point clouds in the foremost region of the blade, which has a significant impact on aerodynamic performance. The trailing edge point cloud can be the set of point clouds in the aftmost region of the blade, which affects flow separation characteristics.

[0196] It is understandable that the extracted boundary points A and B are connected by a straight line, and a straight line with the same slope passing through the endpoint of the leading edge is constructed. All point clouds between the two straight lines are called the leading edge part. Similarly, the trailing edge part can be obtained.

[0197] Step S208: Segment the target blade model point cloud and the standard blade model point cloud based on the point clouds of the leading and trailing edges, and extract the remaining regions to be segmented.

[0198] Understandably, the measuring device decomposes the complex leaf point cloud into functionally defined sub-regions by removing the segmented leading and trailing edge point clouds from the original point cloud; using a region growing algorithm to identify the remaining regions; ensuring region connectivity through curvature consistency checks; and outputting a point cloud to be segmented that contains only the leaf base and leaf back. This process improves the efficiency and accuracy of subsequent segmentation.

[0199] Step S209: Perform high-order parabolic fitting on the remaining region to be segmented, and segment the point cloud of the target leaf model and the point cloud of the standard leaf model according to the opening orientation of the parabolic fitting, to obtain the leaf basin point cloud and the leaf back point cloud.

[0200] Understandably, after the measuring equipment extracts the point cloud from the front and rear edges, it performs high-order parabolic fitting on the remaining middle area point cloud, and determines the segmentation of the leaf basin and leaf back point cloud based on the direction of the parabola's opening.

[0201] It should be noted that during the initial exploration process, we found that simply fitting a parabola did not yield ideal results. The commonly used parabola equation is... The parabola is fitted using a straight line connecting the two endpoints as a reference. If the slope of this line is not zero, the fitted parabola cannot accurately segment the leaf base and leaf back point clouds. In this case, the parabola equation should contain miscellaneous terms such as xy and y, but since it is impossible to determine which miscellaneous terms exist, and adding miscellaneous terms will also increase the fitting complexity. When the line is rotated to a slope of 0, the same rotation matrix is ​​used for the cross-sectional point cloud, with the front and rear endpoints as the beginning and end parts of the parabola. The parabola obtained by fitting the parabola equation above can successfully segment the leaf base and leaf back point clouds.

[0202] It should be understood that the point cloud of a certain blade cross section is a set:

[0203]

[0204] Calculation from the leading edge and trailing edge Vectors formed by point clouds Its rotation angle is:

[0205]

[0206] Construct a two-dimensional rotation matrix to rotate the line to the horizontal:

[0207]

[0208] Rotate and transform all cross-sectional points to obtain the point set in normalized coordinates:

[0209]

[0210] In the rotated coordinate system, we conducted multiple experiments and found that the sixth-order polynomial curve could meet our segmentation requirements. Therefore, we selected all cross-sectional points after rotation. A sixth-order polynomial curve fitting is performed, and its functional form is:

[0211]

[0212] To determine the coefficient vector The least squares method is introduced to optimize the fitting error. Let the dataset have N points, where the fitting objective is to minimize the sum of squared prediction errors for all points:

[0213]

[0214] Construct the fitting matrix equation:

[0215]

[0216] The least squares method simplifies to matrix form:

[0217]

[0218] Its analytical solution is:

[0219]

[0220] After fitting, the parabolic expression can be obtained. However, since the orientation of the point cloud in the rotated section is unknown, the orientation of the fitted curve is also unknown. Accurately segmenting the point clouds of the leaf base and leaf underside becomes very difficult in this situation, but the orientation of the fitted curve can be determined based on parameters. To judge, when When >0, the opening faces upwards. When the value is less than 0, the opening faces downwards. Then, based on the relative position of each point cloud on the fitted parabola, the point cloud can be divided into leaf basin and leaf underside point clouds. The following is the segmentation logic:

[0221] When the opening faces upwards:

[0222]

[0223] When the opening faces downwards:

[0224]

[0225] This embodiment extracts key geometric features from each layer of the point cloud of the target blade model and the standard blade model. These key geometric features include the total number of points, the span in the X direction, and the span in the Y direction. Based on these key geometric features, the variation trend between adjacent layers in the target blade model and the standard blade model point cloud is analyzed to identify the structural abrupt changes between the tenon and the blade body. The blade body cross-sectional contour point cloud is extracted based on the reference plane and search accuracy, and then projected onto a two-dimensional coordinate system to obtain a two-dimensional point cloud. The blade boundary contour is extracted from the two-dimensional point cloud using the α-shape algorithm. Based on the change in the directional angle between adjacent line segments on the blade boundary contour, the leading edge endpoint and the trailing edge endpoint are determined. The leading edge dividing point and the trailing edge endpoint are determined based on the chord connecting the leading edge endpoint and the trailing edge endpoint. The trailing edge boundary point is used to construct the leading and trailing edge point clouds of the target blade model and the standard blade model, respectively, based on the leading edge endpoint, the trailing edge endpoint, the leading edge boundary point, and the trailing edge boundary point. The leading and trailing edge point clouds include the leading edge point cloud and the trailing edge point cloud. The target blade model point cloud and the standard blade model point cloud are segmented based on the leading and trailing edge point clouds, and the remaining regions to be segmented are extracted. The remaining regions to be segmented are fitted with a high-order parabola, and the target blade model point cloud and the standard blade model point cloud are segmented according to the opening orientation of the parabola fitting, respectively, to obtain the blade basin point cloud and the blade back point cloud. In this way, by focusing on geometric features rather than texture information, the measurement bottleneck of weakly textured metal blades is overcome, and the measurement of complex curved blades is transformed from experience-dependent to a standardized process, ensuring the consistency and repeatability of the measurement results.

[0226] refer to Figure 5 , Figure 5This is a flowchart illustrating the third embodiment of the three-dimensional measurement method for power unit blades based on mid-arc line constraints according to the present invention.

[0227] Based on the above embodiments, in this embodiment, step S30 further includes:

[0228] Step S301: Sort the leaf basin point cloud and the leaf back point cloud to obtain the leaf basin point set and the leaf back point set.

[0229] It should be noted that sorting can be used to rearrange unordered point clouds according to specific geometric rules to ensure that the point set follows the topological structure of the blade cross section.

[0230] It should be noted that the leaf basin point set can be a sorted and sequentially arranged set of point clouds of the leaf basin region, preserving the continuity of the cross-sectional geometric features. The leaf back point set can be a sorted and sequentially arranged set of point clouds of the leaf back region, forming a one-to-one correspondence with the leaf basin point set.

[0231] Understandably, this embodiment first projects the point clouds of the leaf base and leaf back onto a reference plane (such as the xOy plane) based on the geometric characteristics of the leaf cross-section; then, it uses a polar coordinate-based sorting method to calculate the polar angle of each point with the leaf center as the origin; the angle sorting ensures that the point set is arranged in the natural order of the cross-sectional contour; finally, it applies a local smoothing algorithm (such as moving average) to eliminate abnormal jumps in the sorting process, obtaining a continuous and ordered point set of the leaf base and leaf back, realizing the topological structuring of the point cloud, making the subsequent matching operation have a clear geometric order, effectively avoiding erroneous matching caused by disordered point clouds, and solving the problem that weakly textured leaves are difficult to establish corresponding relationships due to the lack of obvious features.

[0232] Step S302: Search for the Euclidean distance between each point in the leaf basin point set and each point in the leaf underside point set to obtain the search results.

[0233] In some embodiments, the measuring device can use a KD-Tree spatial index structure to accelerate distance calculation; for each point in the leaf basin point set, a k-nearest neighbor search (k=1) is performed in the leaf back point set; a pruning strategy is used to exclude point pairs that significantly exceed the leaf thickness range; abnormal matches are removed by filtering with a distance threshold (e.g., the maximum distance does not exceed 1.5 times the average leaf thickness); finally, a search result set containing point pair coordinates and distance values ​​is generated, thereby ensuring that potential matching point pairs can still be found quickly even when the point cloud of weakly textured leaves is unevenly distributed, solving the real-time problem caused by the large amount of computation in traditional methods.

[0234] Step S303: Determine the matching point pairs between the leaf basin point set and the leaf back point set based on the search results.

[0235] It should be noted that a matching point pair can be a correspondence between a point in the leaf basin point set and a point in the leaf back point set, satisfying the conditions of geometric proximity and topological consistency.

[0236] In some embodiments, the measuring device employs a bidirectional nearest neighbor matching strategy: first, it finds the nearest point in the set of leaf back points starting from the set of leaf base points, and then verifies whether the point considers the leaf base point as the nearest point; combining prior knowledge of leaf thickness, it excludes matches where the distance exceeds a reasonable range; it applies local consistency constraints, requiring that the angle between the normal vectors of the matching point pairs be less than a threshold; and finally, it optimizes the global matching order through a dynamic programming algorithm to ensure the continuity and rationality of the matching results.

[0237] Step S304: Construct the intersection of the perpendicular bisectors of each matching point pair, and extract the mid-arc point cloud of the target blade model point cloud and the standard blade model point cloud based on the intersection of the perpendicular bisectors.

[0238] It should be noted that the intersection point of the perpendicular bisectors is the point formed by the intersection of the perpendicular planes of the lines connecting the matching points, and it can be located on the geometric center line of the blade section. The mid-curve point cloud can be a point cloud composed of the set of intersection points of the matching points and the perpendicular bisectors, representing the geometric center line of the blade section.

[0239] In practical implementation, the measuring device segments the cross-sectional point cloud, resulting in:

[0240] Leaf Back Point Collection Leaf pot collection .

[0241] Each point The coordinates of the mid-arc point are represented in a two-dimensional plane. The steps for extracting the mid-arc point are as follows:

[0242] For each point in the leaf pot point set On the leaf back point set Searching for the one with the smallest Euclidean distance It satisfies the following formula:

[0243]

[0244] For each pair of matching points Construct the intersection of the perpendicular bisectors of the pair of points (i.e., the mid-arc point):

[0245]

[0246] The set of all midpoints constitutes the mid-arc point set:

[0247]

[0248] The points formed by all the mid-arc points are the mid-arc point set required in this embodiment. The line connecting all the mid-arc points with the front and rear edges as the beginning and end is the mid-arc line, which geometrically represents the equidistant curve between the leaf back and the leaf base.

[0249] This embodiment sorts the leaf base point cloud and leaf back point cloud to obtain the leaf base point set and leaf back point set. It then searches for the Euclidean distance between each point in the leaf base point set and each point in the leaf back point set to obtain search results. Based on these search results, it determines matching point pairs between the leaf base point set and the leaf back point set, constructs the perpendicular bisector intersection of each matching point pair, and extracts the mid-arc line point cloud of the target leaf model point cloud and the standard leaf model point cloud based on the intersection of the perpendicular bisectors. Because this embodiment uses a geometrically based matching strategy, it effectively avoids the feature recognition difficulties caused by the smooth surface of weakly textured leaves, ensuring stable and reliable mid-arc line features under various manufacturing conditions, and making the measurement results unaffected by fluctuations in the leaf surface quality.

[0250] Furthermore, this embodiment of the invention also proposes a computer-readable storage medium storing a three-dimensional measurement program for power plant blades based on mid-arc line constraints. When the three-dimensional measurement program for power plant blades based on mid-arc line constraints is executed by a processor, it implements the steps of the three-dimensional measurement method for power plant blades based on mid-arc line constraints as described above.

[0251] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0252] The aforementioned computer-readable storage medium may be included in a three-dimensional measurement device for power plant blades based on mid-arc line constraints; or it may exist independently and not assembled into a three-dimensional measurement device for power plant blades based on mid-arc line constraints.

[0253] Furthermore, this invention also proposes a computer program product, including a three-dimensional measurement program for power plant blades based on mid-arc line constraints. When the three-dimensional measurement program for power plant blades based on mid-arc line constraints is executed by a processor, it implements the steps of the three-dimensional measurement method for power plant blades based on mid-arc line constraints as described above.

[0254] The specific implementation of the computer program product of the present invention is basically the same as the embodiments of the above-mentioned three-dimensional measurement method for power unit blades based on mid-arc line constraints, and will not be repeated here.

[0255] Reference Figure 6 , Figure 6 This is a structural block diagram of the first embodiment of the three-dimensional measurement device for power unit blades based on mid-arc line constraints of the present invention.

[0256] like Figure 6 As shown, the three-dimensional measurement device for power unit blades based on mid-arc line constraints proposed in this embodiment of the invention includes:

[0257] The point cloud modeling module 10 is used to scan and model the blade of the power device under test using a line structured light scanner to obtain the point cloud of the target blade model.

[0258] The blade segmentation module 20 is used to segment the blade cross-section point cloud of the target blade model point cloud and the standard blade model point cloud into the blade basin point cloud, the blade back point cloud and the leading and trailing edge point cloud, respectively. The standard blade model point cloud is obtained by scanning and modeling a standard power unit blade in advance.

[0259] The mid-arc extraction module 30 is used to extract the mid-arc point clouds of the target leaf model point cloud and the standard leaf model point cloud based on the matching point pairs between the leaf basin point cloud and the leaf back point cloud.

[0260] Error function construction module 40 is used to construct an adaptive weighted error function based on the point clouds of the leading and trailing edges and the mid-arc line of the point cloud of the target blade model and the point cloud of the standard blade model.

[0261] The transformation matrix construction module 50 is used to solve the adaptive weight error function using the Gauss-Newton optimization algorithm to obtain the transformation matrix;

[0262] The blade 3D registration module 60 is used to perform pose transformation on the target blade model point cloud based on the transformation matrix, and to calculate the 3D measurement accuracy of the power unit blade under test based on the pose-transformed target blade model point cloud and the standard blade model point cloud.

[0263] This embodiment uses a line structured light scanner to scan and model the blade of the power unit under test, obtaining a target blade model point cloud. The blade cross-section point clouds of both the target and standard blade model point clouds are segmented into blade base point clouds, blade back point clouds, and leading and trailing edge point clouds. The standard blade model point cloud is obtained by pre-scanning and modeling a standard power unit blade. Based on the matching point pairs between the blade base and blade back point clouds, the mid-arc point clouds of both the target and standard blade model point clouds are extracted. An adaptive weighted error function is constructed based on the leading and trailing edge point clouds and mid-arc point clouds of both the target and standard blade model point clouds. The adaptive weight error function is solved using the Gauss-Newton optimization algorithm to obtain the transformation matrix. Based on the transformation matrix, the target blade model point cloud is pose-transformed, and the 3D measurement accuracy of the power unit blade is calculated based on the pose-transformed target blade model point cloud and the standard blade model point cloud. Since this embodiment uses the mid-arc line and leading and trailing edge parts of different heights to replace the overall blade profile, and uses the mid-arc line to offset the error generated by the overall blade point cloud in the registration, it effectively solves the problem of inaccurate segmentation and registration failure caused by feature loss in traditional measurement of weak textured metal blades, and significantly improves the high-precision registration of 3D measurement of power unit blades.

[0264] The three-dimensional measurement device for power plant blades based on mid-arc line constraints provided in this application employs the three-dimensional measurement method for power plant blades based on mid-arc line constraints described in the above embodiments, and can solve the technical problem of three-dimensional measurement of power plant blades based on mid-arc line constraints. Compared with the prior art, the beneficial effects of the three-dimensional measurement device for power plant blades based on mid-arc line constraints provided in this application are the same as the beneficial effects of the three-dimensional measurement method for power plant blades based on mid-arc line constraints provided in the above embodiments, and other technical features in the three-dimensional measurement device for power plant blades based on mid-arc line constraints are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0265] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0266] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0267] In addition, for technical details not described in detail in this embodiment, please refer to the three-dimensional measurement method of power unit blades based on mid-arc line constraints provided in any embodiment of the present invention, which will not be repeated here.

[0268] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system 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 system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0269] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0270] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0271] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A three-dimensional measurement method for power unit blades based on mid-arc line constraints, characterized in that, The method includes: A line structured light scanner was used to scan and model the blade of the power unit under test, and the point cloud of the target blade model was obtained. The point cloud of the target blade model and the point cloud of the standard blade model are respectively divided into the point cloud of the blade basin, the point cloud of the blade back, and the point cloud of the leading and trailing edges. The point cloud of the standard blade model is obtained by scanning and modeling a standard power unit blade in advance. Based on the matching point pairs between the leaf basin point cloud and the leaf back point cloud, the mid-arc point cloud of the target leaf model point cloud and the standard leaf model point cloud are extracted respectively. An adaptive weighted error function is constructed based on the point clouds of the leading and trailing edges and the mid-arc line of the target blade model point cloud and the standard blade model point cloud. The adaptive weight error function is solved using the Gauss-Newton optimization algorithm to obtain the transformation matrix; The target blade model point cloud is pose-transformed based on the transformation matrix, and the three-dimensional measurement accuracy of the power unit blade under test is calculated based on the pose-transformed target blade model point cloud and the standard blade model point cloud. The adaptive weighted error function is constructed based on the point clouds of the leading and trailing edges and the mid-arc line of the target blade model point cloud and the standard blade model point cloud, including: A point-to-point error term is constructed based on the point clouds of the leading and trailing edges of the target blade model point cloud and the point clouds of the leading and trailing edges of the standard blade model point cloud. A point-to-surface error term is constructed based on the mid-arc point cloud of the target blade model point cloud and the mid-arc point cloud of the standard blade model point cloud. The point-to-point error term and the point-to-surface error term are fused based on the error complementary weight to obtain an adaptive weight error function; The step of solving the adaptive weight error function using the Gauss-Newton optimization algorithm to obtain the transformation matrix includes: The adaptive weight error function is expanded using the Gauss-Newton optimization algorithm, transforming it into the product of the residual vector and the corresponding Jacobian matrix. An iterative equation is then constructed based on the product of the residual vector and the Jacobian matrix, where the rigid body transformation parameters of the iterative equation are rotation vectors and translation vectors. The rotation and translation matrices are solved iteratively for the iterative equation, and the rigid body transformation parameters are updated accordingly. If the current iteration number of the iterative equation reaches the upper limit of the iteration number, or if the point-to-point error term and the point-to-surface error term simultaneously satisfy the convergence threshold condition, then the transformation matrix is ​​output. The step of performing pose transformation on the target blade model point cloud based on the transformation matrix, and calculating the three-dimensional measurement accuracy of the power unit blade under test based on the pose-transformed target blade model point cloud and the standard blade model point cloud, includes: The pose transformation of the point cloud of the target blade model is performed based on the transformation matrix; The matching points in the standard blade model point cloud corresponding to each point in the target blade model point cloud after pose transformation are determined by the nearest neighbor search algorithm, thereby obtaining the matching point pairs between the target blade model point cloud after pose transformation and the standard blade model point cloud. The root mean square error between the point cloud of the target blade model after pose transformation and the point cloud of the standard blade model is calculated based on the Euclidean distance of each matching point pair, and the three-dimensional measurement accuracy of the blade of the power device under test is calculated based on the root mean square error.

2. The three-dimensional measurement method for power unit blades based on mid-arc line constraints as described in claim 1, characterized in that, The process involves using a line structured light scanner to scan and model the blades of the power unit under test, obtaining a point cloud of the target blade model, including: The line structured light scanner is automatically calibrated using a calibration plate. The position and angle of the blades of the power unit under test, as well as the brightness of the line structured light scanner, are adjusted. In response to meeting the preset scanning conditions, the turntable stepping angle is set to the preset angle, and the line structured light scanner is started to perform multi-view scanning on the blade of the power device under test. The preset adjustment conditions include: the laser emitted by the line structured light scanner overlaps on the blade, and there is a preset number of identifiable marker points in the image preview field of view. The point cloud images of the blade of the power unit under test obtained from multiple perspective scans are stitched together to obtain the point cloud of the target blade model.

3. The three-dimensional measurement method for power unit blades based on mid-arc line constraints as described in claim 2, characterized in that, The step of dividing the point cloud of the target leaf model and the point cloud of the standard leaf model into the point cloud of the leaf basin, the point cloud of the leaf back, and the point cloud of the leading and trailing edges includes: Extract the key geometric features of each layer of the point cloud of the target blade model and the point cloud of the standard blade model. The key geometric features include the total number of points, the span in the X direction, and the span in the Y direction. Based on the key geometric features, the variation trend characteristics between adjacent layers in the point cloud of the target blade model and the point cloud of the standard blade model are analyzed to identify the structural abrupt changes between the tenon and the blade body. Based on the reference plane and search accuracy, the blade cross-section contour point cloud is extracted, and the blade cross-section contour point cloud is projected onto a two-dimensional coordinate system to obtain a two-dimensional point cloud. The blade boundary contour is extracted from the two-dimensional point cloud based on the α-shape algorithm; The leading edge endpoint and trailing edge endpoint are determined based on the change in the directional angle between adjacent line segments on the blade boundary profile. The leading edge boundary point and the trailing edge boundary point are determined based on the chord connecting the leading edge endpoint and the trailing edge endpoint; Based on the leading edge endpoint, the trailing edge endpoint, the leading edge boundary point, and the trailing edge boundary point, the leading and trailing edge point clouds of the target blade model point cloud and the standard blade model point cloud are constructed respectively, and the leading and trailing edge point clouds include the leading edge point cloud and the trailing edge point cloud. Based on the point clouds of the leading and trailing edges, the point clouds of the target blade model and the standard blade model are segmented, and the remaining regions to be segmented are extracted. The remaining regions to be segmented are fitted with a high-order parabola, and the point clouds of the target leaf model and the standard leaf model are segmented according to the opening orientation of the parabola fitting, to obtain the leaf basin point cloud and the leaf back point cloud.

4. The three-dimensional measurement method for power unit blades based on mid-arc line constraints as described in claim 3, characterized in that, The step of extracting the mid-arc point cloud of the target leaf model point cloud and the standard leaf model point cloud based on the matching point pairs between the leaf basin point cloud and the leaf underside point cloud includes: The leaf basin point cloud and the leaf underside point cloud are sorted to obtain the leaf basin point set and the leaf underside point set. Search for the Euclidean distance between each point in the leaf basin point set and each point in the leaf underside point set to obtain the search results; Based on the search results, a matching point pair between the leaf basin point set and the leaf underside point set is determined; Construct the intersection of the perpendicular bisectors of each matching point pair, and extract the mid-arc point cloud of the target blade model point cloud and the standard blade model point cloud based on the intersection of the perpendicular bisectors.

5. A three-dimensional measurement device for power unit blades based on mid-arc line constraints, applying the three-dimensional measurement method for power unit blades based on mid-arc line constraints according to any one of claims 1 to 4, characterized in that, The three-dimensional measurement device for power unit blades based on mid-arc line constraints includes: The point cloud modeling module is used to scan and model the blades of the power unit under test using a line structured light scanner to obtain the point cloud of the target blade model. The blade segmentation module is used to segment the blade cross-section point cloud of the target blade model point cloud and the standard blade model point cloud into the blade basin point cloud, the blade back point cloud, and the leading and trailing edge point cloud, respectively. The standard blade model point cloud is obtained by scanning and modeling a standard power unit blade in advance. The mid-arc line extraction module is used to extract the mid-arc line point clouds of the target leaf model point cloud and the standard leaf model point cloud based on the matching point pairs between the leaf basin point cloud and the leaf back point cloud. The error function construction module is used to construct an adaptive weighted error function based on the point clouds of the leading and trailing edges and the mid-arc line of the target blade model point cloud and the point cloud of the standard blade model point cloud. The transformation matrix construction module is used to solve the adaptive weight error function using the Gauss-Newton optimization algorithm to obtain the transformation matrix; The blade 3D registration module is used to perform pose transformation on the point cloud of the target blade model based on the transformation matrix, and to calculate the 3D measurement accuracy of the blade of the power device under test based on the point cloud of the target blade model after pose transformation and the point cloud of the standard blade model.

6. A three-dimensional measurement device for power unit blades based on mid-arc line constraints, characterized in that, The three-dimensional measurement device for power plant blades based on mid-arc line constraints includes: a memory, a processor, and a three-dimensional measurement program for power plant blades based on mid-arc line constraints stored in the memory. The processor is used to run the three-dimensional measurement program for power plant blades based on mid-arc line constraints, and the three-dimensional measurement program for power plant blades based on mid-arc line constraints is configured to implement the three-dimensional measurement method for power plant blades based on mid-arc line constraints as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a three-dimensional measurement program for power plant blades based on mid-arc line constraints. When the three-dimensional measurement program for power plant blades based on mid-arc line constraints is executed by a processor, it implements the three-dimensional measurement method for power plant blades based on mid-arc line constraints as described in any one of claims 1 to 4.

Images