Aero-engine blade adaptive measurement path planning method based on point cloud data

Abstract

The application relates to the field of digital detection of aero-engine blades, and particularly discloses an aero-engine blade self-adaptive measurement path planning method based on point cloud data, which comprises the following steps: acquiring original point cloud data of the blade and preprocessing; structurally partitioning the blade surface into sub-regions of a blade basin, a blade back, a blade edge and a blade root; extracting curvature features of each sub-region, and dividing measurement points into first-type measurement points and second-type measurement points according to the features; generating a first local path for the first-type measurement points by using a low-complexity strategy, and generating a second local path for the second-type measurement points by using a high-precision strategy; identifying a coupling relationship of boundaries of adjacent sub-regions and performing boundary collaborative correction to obtain a global measurement path; performing measurement based on the global measurement path and performing quality evaluation on the result; when a deviation exceeds an allowed range, performing root cause tracing analysis, adjusting corresponding parameters according to a tracing result, and re-executing an affected step.

Description

Technical Field

[0001] This invention belongs to the field of digital inspection of aero-engine blades, specifically an adaptive measurement path planning method for aero-engine blades based on point cloud data. Background Technology

[0002] Aero-engine blades are key components that determine an engine's aerodynamic performance and thrust-to-weight ratio. Their profiles consist of multiple regions, including the blade head, blade back, blade edge, and blade root. The blade edge region (including the leading and trailing edges) has an extremely small radius of curvature, typically only 0.1 mm to 0.5 mm, while the blade head and blade back regions exhibit relatively gentle curves, with curvature differences exceeding an order of magnitude. The geometric accuracy of the blade profile directly affects the engine's compression efficiency, fuel consumption rate, and service life. Therefore, during blade manufacturing, high-precision measurements of the profile are essential to verify whether it meets design tolerance requirements.

[0003] Point cloud data measurement technology, including non-contact measurement methods such as laser triangulation, structured light projection, and lidar, has become an important means of inspecting aero-engine blade profiles due to its advantages such as high measurement speed, high data density, and absence of contact force interference. In actual measurement processes, it is necessary to pre-plan the measurement path based on the geometric characteristics of the blade profile, i.e., determine parameters such as the scanning trajectory of the measuring equipment in space, the distribution density of sampling points, and the probe attitude. The quality of the measurement path planning directly determines the accuracy of the final measurement data and the efficiency of the measurement process.

[0004] In existing technologies, measurement path planning for blade-like freeform surface parts typically employs two methods: First, a uniform sampling strategy is used to generate equally spaced measurement paths across the entire blade profile, meaning the same sampling step size and path planning parameters are used in the blade base, blade back, blade edge, and blade root regions. Second, operators manually increase the density of measurement points in key areas such as the blade edge based on experience, and then software automatically generates continuous measurement trajectories. The former is simple to implement and highly automated, but because the curvature of the blade edge region is much greater than that of the blade base region, the uniform sampling strategy generates a large number of redundant measurement points in the blade base region, while the sampling density in the blade edge region is often insufficient to accurately reproduce the true contour of that region, making it difficult to achieve a reasonable balance between measurement accuracy and measurement efficiency. The latter, although it can increase the density of measurement points in key areas through manual intervention, heavily relies on the operator's experience level, and the manual procedure is cumbersome and has poor repeatability, making it difficult to meet the consistency and high efficiency requirements of batch blade inspection. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an adaptive measurement path planning method for aero-engine blades based on point cloud data. This method solves the problems in the prior art where the use of a uniform measurement strategy for all regions of the blade leads to redundant measurement points in flat areas and insufficient sampling density in high-curvature areas, making it difficult to adaptively balance measurement accuracy and efficiency.

[0006] This invention provides an adaptive measurement path planning method for aero-engine blades based on point cloud data, comprising the following steps:

[0007] Step 1: Obtain the raw point cloud data of the aero-engine blades, and preprocess the raw point cloud data to obtain standardized point cloud data.

[0008] Step 2: Based on the standardized point cloud data, perform structured partitioning on the leaf surface to divide the leaf surface into multiple sub-regions, which include at least the leaf basin region, leaf underside region, leaf edge region, and leaf root region.

[0009] Step 3: Extract curvature features for each sub-region. Based on the curvature features, divide the measurement points in each sub-region into a first type of measurement point and a second type of measurement point. The first type of measurement point is the measurement point with a curvature feature value lower than a preset curvature threshold, and the second type of measurement point is the measurement point with a curvature feature value greater than or equal to the preset curvature threshold.

[0010] Step 4: Use a first path planning strategy to generate a first local measurement path for the first type of measurement points, and use a second path planning strategy to generate a second local measurement path for the second type of measurement points. The computational complexity of the first path planning strategy is lower than that of the second path planning strategy.

[0011] Step 5: Identify the boundary coupling relationship between adjacent sub-regions, perform boundary collaborative correction on the local measurement paths of each sub-region to ensure the continuity and consistency of the measurement paths of adjacent sub-regions at the boundary, and merge the collaboratively corrected local measurement paths of each sub-region to obtain the global measurement path.

[0012] Step 6: Perform actual measurements based on the global measurement path, obtain measurement results, evaluate the quality of the measurement results, and calculate the deviation value between the measurement results and the preset quality standard;

[0013] Step 7: When the deviation value exceeds the preset allowable range, perform a source analysis on the root cause of the deviation, adjust the preset curvature threshold in Step 3 or the path planning parameters in Step 4 according to the source analysis results, and re-execute Steps 3 to 6 until the deviation value falls within the preset allowable range.

[0014] Preferably, the structured partitioning process in step two includes: performing principal component analysis on the standardized point cloud data, extracting the principal axis direction of the leaf, and dividing the leaf point cloud into several transverse slices along the principal axis direction; extracting the contour of the point cloud in each transverse slice, and identifying the leaf basin side contour points and the leaf back side contour points; and identifying the leaf edge region boundary and the leaf root region boundary based on the minimum distance change rate between the leaf basin side contour points and the leaf back side contour points in each transverse slice, thereby dividing the leaf surface into the leaf basin region, the leaf back region, the leaf edge region, and the leaf root region.

[0015] Preferably, the curvature feature extraction in step three includes: for each measurement point, fitting a local surface based on the neighborhood point cloud of the measurement point, calculating the principal curvatures k1 and k2 of the measurement point, and calculating the average curvature based on the principal curvatures. The absolute value of the average curvature is used as the curvature feature value of the measurement point; when the curvature feature value is less than the preset curvature threshold, the measurement point is classified as a first type of measurement point; when the curvature feature value is greater than or equal to the preset curvature threshold, the measurement point is classified as a second type of measurement point.

[0016] Preferably, the first path planning strategy in step four includes: generating a sequence of measurement pose points in the region where the first type of measurement points are located with a preset uniform sampling step size, each measurement pose point containing measurement position coordinates and measurement direction vector, and connecting adjacent measurement pose points with a straight line trajectory to form a first local measurement path.

[0017] Preferably, the second path planning strategy in step four includes: dynamically adjusting the sampling step size based on the curvature feature value of the second type of measurement points, with a smaller sampling step size for regions with larger curvature feature values; for each second type of measurement point, using the normal vector direction of the measurement point as the measurement direction, and determining the measurement position offset based on the local surface curvature radius at the location of the measurement point; and connecting the measurement pose points of each second type of measurement point with spline curve trajectories to form a second local measurement path.

[0018] Preferably, the boundary collaborative correction in step five includes: for adjacent first and second sub-regions, extracting the first boundary measurement point set on the boundary of the first sub-region and the second boundary measurement point set on the boundary of the second sub-region, respectively; calculating the positional deviation between the actual spatial position of each measurement point in the first boundary measurement point set and the actual spatial position of the corresponding measurement point in the second boundary measurement point set; when the positional deviation is greater than a preset boundary deviation threshold, performing positional interpolation correction on the measurement pose points near the boundary of at least one of the first and second sub-regions, so that the corrected boundary positional deviation is less than or equal to the preset boundary deviation threshold; and performing trajectory smoothing connection on the corrected first sub-region boundary measurement path and the second sub-region boundary measurement path.

[0019] Preferably, the quality assessment in step six includes: reconstructing the measurement results into a measured blade surface; comparing the measured blade surface with the blade design model; calculating the deviation values ​​of multiple assessment dimensions, wherein the multiple assessment dimensions include at least profile deviation, surface roughness deviation, and key feature size deviation; weighted summing of the deviation values ​​of each assessment dimension to obtain a comprehensive quality score; and determining that the deviation value exceeds the preset allowable range when the comprehensive quality score is lower than a preset quality threshold.

[0020] Preferably, the root cause tracing and adjustment in step seven includes: checking the existence of each candidate cause in the following order: the proportion of measurement point misclassification exceeds a preset misclassification tolerance threshold, the actual sampling density of the local path is lower than a preset minimum sampling density threshold, the deviation of the sub-region boundary position is greater than a preset boundary deviation threshold, and the repeatability error or accuracy error of the measurement equipment exceeds the nominal accuracy index of the equipment; for each confirmed candidate cause, assess its contribution to the current deviation, and determine the candidate cause with the highest contribution as the root cause; when the root cause is that the proportion of measurement point misclassification exceeds the preset misclassification tolerance threshold, adjust the preset curvature threshold and re-execute step three; when the root cause is that the actual sampling density of the local path is lower than the preset minimum sampling density threshold, reduce the sampling step size of the second path planning strategy and re-execute step four; when the root cause is that the deviation of the sub-region boundary position is greater than a preset boundary deviation threshold, adjust the preset boundary deviation threshold and re-execute step five.

[0021] Preferably, after merging the global measurement path in step five, a path feasibility verification step is also included: for each measurement pose point in the global measurement path, verify whether the measurement pose point satisfies the kinematic constraints of the measurement device, the kinematic constraints including the range of motion constraints of each joint, velocity constraints, and acceleration constraints; for each pair of adjacent measurement pose points in the global measurement path, verify whether the connection trajectory satisfies the reachability constraints and collision-free constraints of the blade profile; when any constraint is not satisfied, shrink and correct the current measurement pose point or connection trajectory within the constraint boundary range so that the corrected global measurement path satisfies all constraints.

[0022] Preferably, before step four is executed, a parameter adaptive adjustment step is also included: obtaining the measurement quality index of the same sub-region in the previous round of measurement; when the measurement quality index is lower than the preset quality lower limit, automatically adjusting the sampling step size of the first path planning strategy and the curvature sensitivity coefficient of the second path planning strategy; after adjustment, verifying the adjustment effect with the local point cloud data of the sub-region; if the verification is successful, step four is executed with the adjusted parameters; if the verification fails, the parameters are adjusted until the verification is successful.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] By dividing the blade profile into structured sub-regions with different geometric features, such as the leaf basin, leaf back, leaf edge, and leaf root, and adopting a two-layer progressive processing mechanism for the curvature differences of each sub-region, a low-complexity strategy is used to quickly generate paths for regions with gentle curvature, while a high-precision strategy is used to finely plan the distribution of measurement points for regions with drastic curvature changes, thus achieving the optimal synergy between measurement accuracy and computational efficiency.

[0025] By identifying the coupling relationship between adjacent sub-regions through a boundary collaborative correction mechanism and implementing location interpolation and trajectory smoothing, the boundary discontinuity problem caused by independent planning of zones is eliminated, ensuring the integrity and consistency of the global measurement path.

[0026] Furthermore, by introducing a closed-loop optimization architecture driven by measurement result feedback, when deviations are detected in the quality assessment, the system can accurately locate the source of anomalies in specific aspects such as measurement point classification, sampling density, boundary coordination, or equipment calibration through root cause analysis. After adjusting the corresponding parameters, the affected steps are re-executed, enabling the system to proactively detect errors, accurately locate root causes, and automatically correct strategies. This effectively overcomes several technical challenges of traditional open-loop measurement systems, such as the difficulty in balancing accuracy and efficiency in the inspection of complex blade surfaces, poor data stitching quality in boundary areas, and the gradual degradation of detection capabilities due to the lack of feedback correction mechanisms during long-term operation. It can be widely applied to high-precision adaptive measurement of aero-engine blades and other complex freeform surface components. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the process for acquiring and structurally partitioning blade point cloud data in Embodiment 1 of the present invention;

[0028] Figure 2 This is a schematic diagram of the curvature feature extraction and two-layer progressive path planning process in Embodiment 2 of the present invention;

[0029] Figure 3 This is a schematic diagram of the boundary collaborative correction and global path merging process in Embodiment 3 of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Example 1: This example details the specific implementation method for acquiring and structurally partitioning blade point cloud data. The specific steps are as follows: Figure 1 As shown.

[0032] 1. Acquisition and preprocessing of point cloud data

[0033] Laser scanners or structured light scanners are used to scan aero-engine blades from multiple angles to obtain raw point cloud data of the blades. Taking structured light scanning as an example, the blade is clamped on a measuring stand, and a structured light projection device projects an coded structured light pattern onto the blade surface. Two or more industrial cameras simultaneously acquire structured light images of the blade surface. The three-dimensional spatial coordinates corresponding to each pixel are calculated through phase calculation and triangulation principles, thereby obtaining dense point cloud data of the blade surface.

[0034] Let the original point cloud dataset obtained be... ,in, For the total number of point clouds, each point This represents the three-dimensional spatial coordinates of a sampling point on the blade surface.

[0035] Preprocessing the raw point cloud data includes the following sub-steps:

[0036] (1) Noise removal; outliers are removed using statistical filtering; for each point Calculate its to The average distance between the nearest neighbors :

[0037]

[0038] in, express The The nearest neighbor point, Denotes the Euclidean norm; calculates the mean of the average distances between all points. and standard deviation :

[0039]

[0040] when At that time, the point These were identified as outliers and removed. The preset standard deviation multiple threshold is usually taken as... .

[0041] (2) Filtering; A bilateral filtering algorithm is used to smooth the point cloud, removing noise while preserving the geometric details of feature areas such as leaf edges; The weighting function of the bilateral filter is:

[0042]

[0043] in, For point The normal vector, For spatial domain weighting functions, The feature-preserving weight function is usually in the form of a Gaussian function.

[0044] (3) Downsampling processing; Voxel grid filters are used to downsample the point cloud to reduce the amount of data and improve the efficiency of subsequent processing; The point cloud space is divided into sections with a side length of A cubic mesh is used, where the point closest to the center of the mesh is retained within each mesh, and all other points are discarded; voxel side length The value of is determined based on the blade size and measurement accuracy requirements, and is usually taken as . .

[0045] After the above preprocessing, a standardized point cloud dataset is obtained. ,in This represents the total number of point clouds after preprocessing.

[0046] 2. Structural partitioning treatment of blade surface

[0047] The standardized point cloud data is subjected to structured partitioning, dividing the leaf surface into leaf basin region, leaf back region, leaf edge region and leaf root region.

[0048] (1) Extraction of principal axis direction; for standardized point cloud data Perform principal component analysis (PCA); calculate the covariance matrix of the point cloud:

[0049]

[0050] in, Let the centroid of the point cloud be denoted by the covariance matrix. Perform eigenvalue decomposition:

[0051]

[0052] in, For eigenvalues, , , The corresponding eigenvectors; the largest eigenvalue. corresponding feature vector That is, the direction of the blade's main axis.

[0053] (2) Lateral slicing; along the main axis of the blade Divide the point cloud data into equally spaced sections. There are 10 transverse slices; each slice is a thin layer perpendicular to the principal axis, with a thickness of 100 mm. For the first slice Its position range along the principal axis is:

[0054]

[0055] in, .

[0056] (3) Identification of the contours of the leaf basin and the underside of the leaf; for each transverse slice Contour extraction is performed on the point cloud within the slice; the point cloud within the slice is projected onto a plane perpendicular to the principal axis to obtain a two-dimensional point set; the convex hull algorithm is used to extract the slice contour points, and the points are sorted by angle to obtain an ordered sequence of contour points. .

[0057] For an ordered sequence of contour points, the curvature of each contour point is calculated; the three-point method is used to estimate the contour points. Curvature at:

[0058]

[0059] in, For vectors and The angle between them.

[0060] Based on the distribution characteristics of the curvature of the contour points, the contour points on the leaf basin side and the leaf back side are identified; contour segments with smaller curvature and continuous changes correspond to the leaf basin region and the leaf back region; contour segments with significantly increased curvature and abrupt changes in direction correspond to the leaf margin region (leading edge and trailing edge).

[0061] (4) Leaf margin and leaf root boundary identification; for each transverse slice, calculate the minimum distance between the leaf basin side contour point and the leaf back side contour point:

[0062]

[0063] calculate Rate of change along the principal axis:

[0064]

[0065] when When the rate of change exceeds a preset threshold, this location is identified as the boundary of the leaf root region (the connection between the blade and the tenon); when When the value suddenly decreases to a local minimum and the profile curvature increases significantly, the location is identified as the leaf margin region (leading or trailing edge).

[0066] After the above partitioning process, the leaf surface is divided into four sub-regions: leaf basin region leaf underside area Leaf margin area (Including the leading and trailing margins) and the leaf root region Each sub-region satisfies:

[0067]

[0068] Example 2: This example details the specific implementation of curvature feature extraction and two-layer progressive path planning, as follows: Figure 2 As shown.

[0069] 1. Curvature Feature Extraction

[0070] For each sub-region Curvature features were extracted separately.

[0071] For sub-regions Each measurement point in The curvature characteristics of this point are estimated using the local surface fitting method.

[0072] (1) Establish a local neighborhood; adopt A k-dimensional tree (KD-tree) data structure is used for spatial indexing of point clouds; for target points... Search for its nearest neighbors in the KD-tree. Each set of 10 neighboring points constitutes a local neighborhood point set. ,in Usually taken .

[0073] (2) Local surface fitting; in the local neighborhood point set Above, the least squares method is used to fit the quadratic surface; the origin of the local coordinate system is placed at... At that point, along the z-axis If the direction of the normal vector is given, then the quadratic surface can be represented as:

[0074]

[0075] in, Let the coordinates be the coordinates of the tangent plane in the local coordinate system; the coefficients are solved by minimizing the following objective function. , , , , :

[0076]

[0077] (3) Calculate the principal curvature and the mean curvature; construct the coefficient matrix of the first and second basic forms from the coefficients of the fitted quadratic surface; the coefficients of the first basic form are:

[0078] , ,

[0079] in, , , These are the coefficients of the first fundamental form of the surface, used to describe the intrinsic metric of the surface.

[0080] The second fundamental form coefficient is:

[0081] , ,

[0082] in, , , is the second fundamental form coefficient of the surface, used to describe the curvature of the surface in space.

[0083] Gaussian curvature of the surface and mean curvature They are respectively:

[0084]

[0085]

[0086] Principal curvature and These are the two roots of the curvature equation:

[0087]

[0088] Solving for:

[0089]

[0090] In terms of the absolute value of the mean curvature This serves as the curvature characteristic value of the measurement point.

[0091] (4) Classify measurement points; set preset curvature threshold. For sub-regions Each measurement point in :

[0092] like Then Classified as the first type of measurement point ;

[0093] like Then Classified as the second type of measurement point .

[0094] Curvature threshold The typical range of values ​​is determined based on a comprehensive consideration of blade material, profile accuracy requirements, and the performance of measuring equipment. The curvature of the leaf basin area and the leaf underside area is usually smaller, so the first type of measurement points are mainly used; the curvature of the leaf margin area and the leaf root area is usually larger, so the second type of measurement points are mainly used.

[0095] 2. First Path Planning Strategy (Low Complexity Path)

[0096] For the first type of measurement points A first path planning strategy with low computational complexity is used to generate local measurement paths.

[0097] (1) Generate a sequence of measurement pose points; with a preset uniform sampling step size Generate a sequence of measurement pose points within the region where the first type of measurement points are located; for each first type of measurement point... Its measurement location coordinates are Its position, either itself or offset along the normal vector direction:

[0098]

[0099] in, The calibrated working distance for measuring equipment (for laser probes, typically...) ), For point The unit normal vector at that location.

[0100] The measurement direction vector is taken as the direction of the normal vector at that point. (For laser triangulation measurement, the measurement direction is the direction of the laser beam, which needs to be determined based on the equipment calibration results.)

[0101] (2) Trajectory connection; Connect adjacent measurement pose points with straight-line trajectories to form the first local measurement path:

[0102]

[0103] in, and For adjacent measurement pose points, and For the corresponding path parameters.

[0104] 3. Second path planning strategy (high-complexity fine-grained path)

[0105] For the second type of measurement points A second path planning strategy with high computational complexity is used to generate fine local measurement paths.

[0106] (1) Adaptive sampling step size; dynamically adjust the sampling step size based on the curvature feature value of the second type of measurement point. For curvature of Measurement points, sampling step size Determined in the following manner:

[0107]

[0108] in, This is the maximum sampling step size (corresponding to the region where the curvature is slightly higher than the threshold). This is the minimum sampling step size (corresponding to the region with maximum curvature). The curvature sensitivity coefficient controls the decay rate of the sampling step size as curvature changes; a typical value is [value missing]. , , .

[0109] From the above formula, we can see that the curvature eigenvalue The larger the sampling step size The smaller the value, the denser the measurement points.

[0110] (2) The measurement pose is precisely determined; for each type II measurement point The direction of the normal vector at that point is taken as the measurement direction:

[0111]

[0112] The measurement position offset is determined by the radius of curvature of the local surface at the point; at the point... At that point, along the principal curvature and The principal radii of curvature in the directions are respectively and For non-contact measurement, the coordinates of the measurement position are determined as follows:

[0113]

[0114] in, This is the curvature compensation amount, used to correct measurement errors caused by surface curvature. Its calculation method is as follows:

[0115]

[0116] (3) Trajectory connection; connect the measurement pose points of each type II measurement point with spline curve trajectories to form a second local measurement path; use cubic... Spline curve trajectory interpolation; [The following is a list of parameters:] ... One measurement pose point Then three times A spline curve is defined as:

[0117]

[0118] in, For three Spline basis functions are defined by the following recursive formula:

[0119]

[0120]

[0121] pass Spline curve interpolation is used to obtain a smooth and continuous second local measurement path.

[0122] 4. Parameters of Path Planning Implementation Examples

[0123] In a typical implementation example, the length of the aero-engine blade is chord width is The maximum thickness is The number of point clouds after preprocessing is approximately Points; curvature threshold The average curvature of the leaf basin area is approximately The first path planning strategy is adopted, with a uniform sampling step size. The average curvature of the leaf margin region is approximately The second path planning strategy is adopted, and the adaptive sampling step size is from Change to Curvature sensitivity coefficient The total length of the overall measurement path is approximately The total number of measurement points is approximately 3,500.

[0124] Example 3: This example details the specific implementation of boundary collaborative correction and global path merging, as follows: Figure 3 As shown.

[0125] 1. Boundary Coupling Relationship Identification

[0126] After completing the local path planning for each sub-region, it is necessary to identify the boundary coupling relationships between adjacent sub-regions; let the sub-regions be... and For adjacent sub-regions, their shared boundary is defined as:

[0127]

[0128] in, and These are the sets of boundary points for the two sub-regions, respectively.

[0129] For each pair of adjacent sub-regions Extract respectively Set of first boundary measurement points on the boundary and Set of second boundary measurement points on the boundary .

[0130] 2. Calculation of boundary position deviation

[0131] Since the local path planning for the two sub-regions is performed independently, there may be positional deviations at the boundaries; it is necessary to calculate the positional deviations between the sets of boundary measurement points.

[0132] The Iterative Closest Point (ICP) algorithm is used to... and Perform registration to establish the correspondence between the two sets of boundary points; for the registered corresponding point pairs Calculate the positional deviation:

[0133]

[0134] The modulus of the boundary position deviation is:

[0135]

[0136] Define the maximum deviation of the boundary as:

[0137]

[0138] Define the boundary mean deviation as:

[0139]

[0140] in, This represents the number of corresponding point pairs.

[0141] 3. Position interpolation correction

[0142] When the boundary maximum deviation Greater than the preset boundary deviation threshold At this time, position interpolation correction is required for the measured pose points near the boundary; The typical range of values ​​is .

[0143] The correction method is as follows: For deviations exceeding the threshold Corresponding point pairs Calculate the corrected boundary point positions:

[0144]

[0145]

[0146] The above correction moves each of the two boundary points to the other by half the deviation exceeding the threshold, so that the corrected boundary position deviation is exactly equal to... .

[0147] For pose points near the boundary that are not directly involved in the correction, an inverse distance-weighted interpolation method is used for smooth transition; let the points in the region near the boundary be... To the The distance between the boundary points is ,but The correction amount is:

[0148]

[0149] in, For the first Correction amount for each boundary point Inverse distance weights For a very small positive number (such as (This is to prevent division by zero.)

[0150] 4. Smooth connection of boundary trajectories

[0151] Perform trajectory smoothing connection on the corrected boundary measurement path; for sub-regions arrive The transition is achieved by splicing the paths on both sides of the boundary using cubic spline curves.

[0152] set up The last side near the boundary The measurement pose points are: , The beginning of the side near the boundary The measurement pose points are: ,in and This is the corrected pose point at the boundary.

[0153] Constructing connection curves To satisfy the following boundary conditions:

[0154]

[0155]

[0156] Smooth connections are achieved using cubic Hermite interpolation curves:

[0157]

[0158] in , .

[0159] 5. Global path merging

[0160] After completing the collaborative correction of the boundaries of all adjacent sub-regions, the local measurement paths of each sub-region are merged according to the spatial topological order of the blade surface to obtain the complete global measurement path:

[0161]

[0162] in, This represents the concatenation operation of paths. For the first Internal paths within each sub-region This represents the connection path between adjacent sub-regions.

[0163] Example 4: This example details the specific implementation methods for quality assessment, root cause analysis, and feedback optimization.

[0164] 1. Quality assessment of measurement results

[0165] Based on global measurement path Perform actual measurements to obtain measurement data of the blade surface; reconstruct the measured blade surface from the measurement results. , adopting based Methods for reconstructing spline surfaces:

[0166]

[0167] in, For control point grid, and They are respectively direction and Direction Spline basis functions.

[0168] The measured blade surface With blade design model Compare the results and calculate the deviation values ​​for multiple evaluation dimensions.

[0169] (1) Profile deviation; for sampling points on the measured surface Calculate its design model Minimum distance:

[0170]

[0171] Profile deviation is defined as the root mean square value of the minimum distance between all sampling points:

[0172]

[0173] (2) Surface roughness deviation; extract the surface micro-irregularity characteristics from the measured surface and calculate the surface roughness parameters. (arithmetic mean deviation) and (Ten-point height of micro-irregularities) is compared with the roughness index required by the design:

[0174]

[0175] (3) Deviation of key feature dimensions; key feature dimensions include at least the blade chord length. Maximum thickness Leaf margin radius and Twist etc.; for each key feature size Calculate the deviation between the measured value and the design nominal value:

[0176]

[0177] (4) Overall quality score: The deviation values ​​of each evaluation dimension are weighted and summed to obtain the overall quality score. :

[0178]

[0179] in, The normalized deviation value (the deviation value divided by its allowable upper limit). For the weights of each dimension, satisfying .

[0180] When the overall quality score Below the preset quality threshold If the deviation value exceeds the preset allowable range, a feedback optimization process is triggered.

[0181] 2. Determining Dynamic Weights

[0182] In a preferred embodiment, the weights in the weighted summation are dynamic weights; after each round of quality assessment, the ratio of the deviation value of each assessment dimension to the allowable upper limit of the deviation for that dimension is compared to obtain the compliance gap ratio for each dimension:

[0183]

[0184] in, For the first The current deviation value of each evaluation dimension This represents the upper limit of the allowable deviation for this dimension.

[0185] The weight of the assessment dimension with the largest compliance gap ratio is set to the highest value, and the weight of the assessment dimension with the smallest compliance gap ratio is set to the lowest value. The weights of the remaining dimensions are determined by linear interpolation based on the compliance gap ratio.

[0186]

[0187] in, , , and These are the preset highest and lowest weights (e.g., ...). , ).

[0188] By dynamically calibrating the weights, subsequent optimization efforts are prioritized for the evaluation dimensions with the largest current gap in achieving the target.

[0189] 3. Root cause analysis

[0190] When the overall quality score fails to meet the standard, a root cause analysis is performed to trace the cause of the deviation; the following candidate causes are checked in order:

[0191] Candidate Reason 1: The proportion of misclassified measurement points exceeds the preset misclassification tolerance threshold; check the curvature threshold of the sub-region boundary. Is the configuration incorrect? The specific method is to count the proportion of measurement points in each sub-region where the actual curvature does not match the classification results. ;like ( The preset misclassification tolerance threshold is usually set to [value missing]. If the candidate cause is found to exist, then the candidate cause is determined to exist.

[0192] Candidate Reason 2: The actual sampling density of the local path is lower than the preset minimum sampling density threshold; check whether the actual measurement point density in the high curvature region meets the requirements of the Nyquist sampling theorem; for curvature... The minimum required sampling density for the region is:

[0193]

[0194] If the actual sampling density If so, then the candidate cause is determined to exist.

[0195] Candidate Reason 3: The sub-region boundary position deviation is greater than the preset boundary deviation threshold; check the actual position deviation after boundary correction. Is it still greater than Or whether the trajectory curvature at the boundary undergoes a sudden change; if If the rate of change of path curvature at the boundary exceeds a preset value, then the candidate cause is determined to exist.

[0196] Candidate Reason 4: Measurement equipment calibration error; check the calibration status of the measurement equipment; calculate the measurement repeatability error by repeatedly measuring a standard sphere or standard block. and accuracy error If any error exceeds the nominal accuracy of the equipment, then the candidate cause is determined to exist.

[0197] For each confirmed candidate cause, assess its contribution to the current deviation. The method for quantifying the contribution is as follows: after excluding the cause (i.e., adjusting the parameter corresponding to the cause to the ideal value), recalculate the improvement in the overall quality score. ,by As a measure of the degree of contribution to the cause, the candidate cause with the highest degree of contribution is identified as the root cause.

[0198] 4. Targeted Adjustment and Iterative Optimization

[0199] Based on the root cause analysis results, perform targeted adjustments:

[0200] The root cause is that the proportion of misclassified measurement points exceeds the preset misclassification tolerance threshold: Adjust the preset curvature threshold. If misclassification is mainly manifested in high curvature points being incorrectly classified as Class I, then the score will be reduced. If the main manifestation is that low curvature points are incorrectly classified as the second category, then improve... Adjust the step size to get the current threshold. After adjustment, repeat step three (reclassification of measurement points).

[0201] The root cause is that the actual sampling density of the local path is lower than the preset minimum sampling density threshold: reduce the sampling step size of the second path planning strategy; increase the curvature sensitivity coefficient. Increase Or the minimum sampling step size Decrease After adjustment, repeat step four (local path replanning).

[0202] The root cause is that the deviation of the sub-region boundary position is greater than the preset boundary deviation threshold: reduce the preset boundary deviation threshold. It usually decreases to the current value. After adjustment, re-execute step five (boundary coordination correction).

[0203] After the adjustment is complete, repeat steps three through six to obtain new measurement results and quality scores; repeat the root cause tracing and targeted adjustment process above until the overall quality score is obtained. .

[0204] In a typical implementation example, after After rounds of iterative optimization, the overall quality score improved from the initial 72 points to 93 points (out of 100), meeting the quality threshold. Requirements.

[0205] Example 5: This example details the specific implementation method of path feasibility verification and confidence assessment.

[0206] 1. Route Feasibility Verification

[0207] After merging to obtain the global measurement path, the feasibility of the path is verified.

[0208] (1) Kinematic constraint test; for the global measurement path Each measurement pose point in ,in To measure the position, To determine the direction of measurement, check whether the pose point satisfies the kinematic constraints of the measuring device.

[0209] For a five-axis measurement system, pose point Corresponding to five joint variables ( , , , , Solve using inverse kinematics:

[0210]

[0211] Examine each joint variable Does it meet its range of motion constraints?

[0212]

[0213] Verify the joint velocity and acceleration constraints between adjacent pose points:

[0214]

[0215] in, and Joints The velocity and acceleration are calculated by the joint variable difference between adjacent pose points.

[0216] (2) Reachability and collision-free constraint verification; for the connection trajectory between each pair of adjacent measurement pose points in the global measurement path. To verify whether the connection trajectory satisfies the reachability constraint and collision-free constraint of the blade profile.

[0217] The reachability constraint requires that every point on the trajectory be connected. The angle between the measurement direction and the blade surface normal vector at the location does not exceed the maximum allowable tilt angle of the measuring equipment. :

[0218]

[0219] in, For the blade surface in The normal vector at that location.

[0220] The collision-free constraint requires that the minimum distance between each point on the connection trajectory and the blade surface be greater than the safety distance. :

[0221]

[0222] in, This refers to the probe radius (for contact measurements) or the laser spot radius (for non-contact measurements).

[0223] (3) Correction of constraint violation; when any constraint is not satisfied, the current measured pose point or connection trajectory is contracted and corrected within the constraint boundary; taking kinematic constraint violation as an example, if the joint Beyond its range of motion Then the corresponding pose point will be corrected as follows:

[0224]

[0225] in, Joint variables relative pose point gradient, The step size factor is determined by line search to make the corrected step size factor... ( (This is a tiny positive margin).

[0226] 2. Confidence assessment

[0227] After the measurement results are output, the confidence score of the current measurement results is calculated.

[0228] Confidence score The evaluation is based on a comprehensive assessment of the following three indicators:

[0229] (1) Completeness of input point cloud data Defined as the ratio of the actual point cloud coverage area to the theoretical surface area of ​​the blade:

[0230]

[0231] in, The leaf surface area covered by the preprocessed point cloud (estimated by point cloud triangulation). The theoretical surface area for the blade design model; typically required .

[0232] (2) Coverage of the measurement path over the blade profile Defined as the ratio of the area traversed by the measurement path to the theoretical surface area of ​​the blade:

[0233]

[0234] in, It is the sum of the areas of the Voronoi regions of all measurement pose points along the measurement path.

[0235] (3) Success rate of measurement of similar blades in history Defined as in the most recent In measurement tasks of the same type of blade, the percentage of tasks that achieved the overall quality score was:

[0236]

[0237] in, Usually taken .

[0238] The overall confidence score is a weighted average of the three indicators:

[0239]

[0240] The weights satisfy Typical value , , .

[0241] When confidence score Below the preset confidence threshold (usually taken) If the following occurs, the subsequent processing flow will be paused, an insufficient confidence warning signal will be output, and supplementary measurement data or manual confirmation will be requested.

[0242] The embodiments of the present invention are given for the purposes of illustration and description. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Any changes, modifications, substitutions and variations made by those skilled in the art to the above embodiments within the scope of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive measurement path planning method for aero-engine blades based on point cloud data, characterized in that, Includes the following steps: Step 1: Obtain the raw point cloud data of the aero-engine blades, and preprocess the raw point cloud data to obtain standardized point cloud data. Step 2: Based on the standardized point cloud data, perform structured partitioning on the leaf surface to divide the leaf surface into multiple sub-regions, which include at least the leaf basin region, leaf underside region, leaf edge region, and leaf root region. Step 3: Extract curvature features for each sub-region. Based on the curvature features, divide the measurement points in each sub-region into a first type of measurement point and a second type of measurement point. The first type of measurement point is the measurement point with a curvature feature value lower than a preset curvature threshold, and the second type of measurement point is the measurement point with a curvature feature value greater than or equal to the preset curvature threshold. Step 4: Use a first path planning strategy to generate a first local measurement path for the first type of measurement points, and use a second path planning strategy to generate a second local measurement path for the second type of measurement points. The computational complexity of the first path planning strategy is lower than that of the second path planning strategy. Step 5: Identify the boundary coupling relationship between adjacent sub-regions, perform boundary collaborative correction on the local measurement paths of each sub-region to ensure the continuity and consistency of the measurement paths of adjacent sub-regions at the boundary, and merge the collaboratively corrected local measurement paths of each sub-region to obtain the global measurement path. Step 6: Perform actual measurements based on the global measurement path, obtain measurement results, evaluate the quality of the measurement results, and calculate the deviation value between the measurement results and the preset quality standard; Step 7: When the deviation value exceeds the preset allowable range, perform a source analysis on the root cause of the deviation, adjust the preset curvature threshold in Step 3 or the path planning parameters in Step 4 according to the source analysis results, and re-execute Steps 3 to 6 until the deviation value falls within the preset allowable range.

2. The method according to claim 1, characterized in that, The structured partitioning process described in step two includes: performing principal component analysis on the standardized point cloud data to extract the principal axis direction of the leaf, and dividing the leaf point cloud into several transverse slices along the principal axis direction; extracting the contours of the point cloud in each transverse slice to identify the leaf basin side contour points and the leaf back side contour points; and identifying the leaf edge region boundary and the leaf root region boundary based on the minimum distance change rate between the leaf basin side contour points and the leaf back side contour points in each transverse slice, thereby dividing the leaf surface into the leaf basin region, the leaf back region, the leaf edge region, and the leaf root region.

3. The method according to claim 1, characterized in that, Step 3, the curvature feature extraction, includes: for each measurement point, fitting a local surface based on the neighborhood point cloud of that measurement point, calculating the principal curvatures k1 and k2 of that measurement point, and calculating the average curvature based on the principal curvatures. The absolute value of the average curvature is used as the curvature feature value of the measurement point; when the curvature feature value is less than the preset curvature threshold, the measurement point is classified as a first type of measurement point; when the curvature feature value is greater than or equal to the preset curvature threshold, the measurement point is classified as a second type of measurement point.

4. The method according to claim 1, characterized in that, The first path planning strategy in step four includes: generating a sequence of measurement pose points in the region where the first type of measurement points are located with a preset uniform sampling step size. Each measurement pose point contains measurement position coordinates and measurement direction vector. Adjacent measurement pose points are connected by a straight line trajectory to form a first local measurement path.

5. The method according to claim 1, characterized in that, The second path planning strategy in step four includes: dynamically adjusting the sampling step size based on the curvature feature value of the second type of measurement points, with a smaller sampling step size for regions with larger curvature feature values; for each second type of measurement point, using the normal vector direction of the measurement point as the measurement direction, and determining the measurement position offset based on the local surface curvature radius at the location of the measurement point; and connecting the measurement pose points of each second type of measurement point with spline curve trajectories to form a second local measurement path.

6. The method according to claim 1, characterized in that, Step 5, the boundary collaborative correction, includes: for adjacent first and second sub-regions, extracting the first boundary measurement point set on the boundary of the first sub-region and the second boundary measurement point set on the boundary of the second sub-region, respectively; calculating the positional deviation between the actual spatial position of each measurement point in the first boundary measurement point set and the actual spatial position of the corresponding measurement point in the second boundary measurement point set; when the positional deviation is greater than a preset boundary deviation threshold, performing positional interpolation correction on the measurement pose points near the boundary of at least one of the first and second sub-regions, so that the corrected boundary positional deviation is less than or equal to the preset boundary deviation threshold; and performing trajectory smoothing connection on the corrected first sub-region boundary measurement path and the second sub-region boundary measurement path.

7. The method according to claim 1, characterized in that, The quality assessment in step six includes: reconstructing the measurement results into a measured blade surface; comparing the measured blade surface with the blade design model; calculating the deviation values ​​of multiple assessment dimensions, including at least profile deviation, surface roughness deviation, and key feature size deviation; weighted summing of the deviation values ​​of each assessment dimension to obtain a comprehensive quality score; and determining that the deviation value exceeds the preset allowable range when the comprehensive quality score is lower than a preset quality threshold.

8. The method according to claim 1, characterized in that, Step seven, the root cause tracing and adjustment, includes: checking the existence of each candidate cause in the following order: the proportion of measurement point misclassification exceeds a preset misclassification tolerance threshold, the actual sampling density of the local path is lower than a preset minimum sampling density threshold, the deviation of the sub-region boundary position is greater than a preset boundary deviation threshold, and the repeatability error or accuracy error of the measurement equipment exceeds the nominal accuracy index of the equipment; for each confirmed candidate cause, assess its contribution to the current deviation, and determine the candidate cause with the highest contribution as the root cause; when the root cause is that the proportion of measurement point misclassification exceeds the preset misclassification tolerance threshold, adjust the preset curvature threshold and re-execute step three; when the root cause is that the actual sampling density of the local path is lower than the preset minimum sampling density threshold, reduce the sampling step size of the second path planning strategy and re-execute step four; when the root cause is that the deviation of the sub-region boundary position is greater than a preset boundary deviation threshold, adjust the preset boundary deviation threshold and re-execute step five.

9. The method according to claim 1, characterized in that, After merging the global measurement path in step five, a path feasibility verification step is also included: for each measurement pose point in the global measurement path, it is verified whether the measurement pose point meets the kinematic constraints of the measurement device, the kinematic constraints include the range of motion constraints of each joint, velocity constraints, and acceleration constraints; for the connection trajectory between each pair of adjacent measurement pose points in the global measurement path, it is verified whether the connection trajectory meets the reachability constraints and collision-free constraints of the blade profile; when any constraint is not met, the current measurement pose point or connection trajectory is contracted and corrected within the constraint boundary range so that the corrected global measurement path meets all constraints.

10. The method according to claim 1, characterized in that, Before step four is executed, a parameter adaptive adjustment step is also included: obtaining the measurement quality index of the same sub-region in the previous round of measurement; when the measurement quality index is lower than the preset quality lower limit, automatically adjusting the sampling step size of the first path planning strategy and the curvature sensitivity coefficient of the second path planning strategy; after adjustment, verifying the adjustment effect with the local point cloud data of the sub-region; if the verification is successful, step four is executed with the adjusted parameters; if the verification fails, the parameters are adjusted until the verification is successful.

Images