Precisely-forged blade intake and exhaust edge double-notch blade profile tolerance detection method

Abstract

The invention discloses a precision forging blade intake and exhaust edge double-gap blade profile tolerance detection method, which comprises the following steps: a) carrying out independent coordinate system division on an intake / exhaust edge region based on a double-domain decoupling strategy, and inhibiting multi-gap coupling interference; b) performing multi-level noise suppression processing on the noisy point-containing cloud data, wherein the multi-level noise suppression processing comprises dynamic clustering analysis and a deformation compensation mechanism; c) adopting a bidirectional collaborative iterative optimization algorithm to synchronously execute air inlet side forward fitting and exhaust side reverse fitting, and realizing convergence control through cross-domain residual verification; d) dynamically adjusting a data fitting interval based on a self-adaptive segmentation strategy of the feature sensitive area; and e) constructing an incremental data management module, and realizing versioning storage and compression optimization of an iteration process. A double-domain decoupling strategy is adopted to overcome multi-gap coupling interference; the double-gap synchronous detection is realized, and compared with the traditional method, the detection time is obviously shortened.

Description

Technical Field

[0001] The present invention relates to the technical field of blade profile detection, and specifically to a method for detecting the profile of the intake and exhaust edges of a precision forging blade with double notches. Background Technique

[0002] In the detection of the profile of precision forging blades, the measurement of the profile in the double-notch area of the intake / exhaust edges faces a series of technical bottlenecks, which seriously affect the accuracy and efficiency of detection: First, multi-notch coupling interference. When detecting the profile of the double-notch area of the intake / exhaust edges, existing methods (such as discrete sampling by a coordinate measuring machine) will generate coupling errors when processing the point cloud data in this area. This coupling error is caused by the relatively complex double-notch structure of the intake / exhaust edges and the mutual influence between each notch. When detecting the intake and exhaust edges simultaneously, this coupling error will be further amplified, and the maximum deviation can reach 0.1 mm. This means that in actual detection, the measurement result may deviate from the true value by 0.1 mm. For precision forging blades of aeroengines, which have extremely high precision requirements, such a deviation is unacceptable. Second, the lack of data correlation. Currently, commercial software (such as PolyWorks 2023) lacks topological correlation analysis when processing data in the double-notch area. Topological correlation analysis is crucial for understanding the relationship between the data of the intake and exhaust edges. Due to the lack of this analysis, a relatively high error rate will occur when matching the data of the intake and exhaust edges, and its error rate is greater than 8%. This means that in the data processing process, the data of the intake and exhaust edges cannot be accurately corresponded and integrated, thus affecting the accuracy of the entire profile detection. Third, the lack of algorithm adaptability. When the existing RANSAC algorithm processes the double-curvature mutation area (when the curvature difference is greater than 0.1 mm⁻¹), there is a problem of excessive fitting residuals. The fitting residual is an important indicator to measure the fitting effect of the algorithm. In this case, the fitting residual of this algorithm exceeds 0.1 mm and cannot meet the requirements of the ASME B89.3.4-2020 standard. This indicates that the existing algorithm cannot effectively perform data fitting when facing the special curvature change situation in the double-notch area of the intake and exhaust edges of precision forging blades, thus affecting the accuracy of profile detection. Summary of the Invention

[0003] To solve the problems of the existing technology, the present invention provides a method for detecting the profile of the intake and exhaust edges of a precision forging blade with double notches, including the following steps: a) Based on the double-domain decoupling strategy, independently divide the coordinate system for the intake / exhaust edge area to suppress multi-notch coupling interference; b) Implement multi-level noise suppression processing on the noisy point cloud data, including dynamic clustering analysis and deformation compensation mechanism; c) Adopt a two-way collaborative iterative optimization algorithm to synchronously perform forward fitting of the intake edge and reverse fitting of the exhaust edge, and achieve convergence control through cross-domain residual verification; d) Based on the adaptive segmentation strategy of the feature-sensitive region, dynamically adjust the data fitting interval; e) Construct an incremental data management module to achieve versioned storage and compression optimization during the iterative process.

[0004] Furthermore, the dual-domain decoupling strategy includes: establishing independent coordinate systems according to the geometric features and functional weights of the intake / exhaust edges; dynamically allocating the regional weights of the intake / exhaust edges through a curvature weighting model, where the weight range of the intake edge is 0.5 - 0.6, and the weight range of the exhaust edge is 0.4 - 0.5.

[0005] Furthermore, the multi-level noise suppression processing specifically includes: The first layer: a clustering algorithm that dynamically adjusts the density based on material properties to suppress the noise of the hardened layer; The second layer: a compensation function that combines temperature and deformation parameters to correct the plastic deformation error; The neighborhood radius parameter of the clustering algorithm is dynamically calculated according to the hardness and surface roughness of the forging, and the expression is: Eps = k1(1 + k2), where k1 is the basic neighborhood radius and k2 is the material property correction coefficient; The noise suppression processing further includes: a noise classification module that identifies the noise type based on surface roughness and material hardness; Gaussian filtering is used for high-frequency noise, and median filtering is used for low-frequency noise; The input parameters of the noise classification module include: the surface hardness value range of the forging is 30 - 60 HRC; the roughness value range is Ra = 0.8 - 2.5 μm.

[0006] Furthermore, the two-way collaborative iterative optimization algorithm meets the following convergence conditions: the spatial residual threshold ≤ 0.03 mm; the curvature gradient continuity conforms to the C² continuity standard; The triggering conditions for the cross-domain residual verification include: performing a residual comparison every N iterations, where N is a preset positive integer; the verification threshold is dynamically adjusted based on the intake / exhaust edge weights, and the range ≤ 0.01 mm.

[0007] Furthermore, the adaptive segmentation strategy includes: a dynamic interruption mechanism based on curvature mutation, and the curvature change threshold is set to 0.05 - 0.15 mm -1 ; the cumulative error threshold triggers segment reconstruction, and the error range ≤ 0.03 mm; The dynamic interruption mechanism includes: using morphological filtering to eliminate isolated noise points, and the filter kernel size is dynamically selected according to the feature-sensitive region; retaining high-curvature feature points through two-way geometric constraints.

[0008] Furthermore, the incremental data management module includes: a differential encoding model that only records the data differences between iterations; a compression rate evaluation model that calculates the storage optimization efficiency in real time; The implementation method of the differential encoding model is: losslessly compress and store the initial data set; in subsequent iterations, only store the difference matrix between the current data set and the previous version; The calculation formula of the compression rate evaluation model is: compression rate = 1 - (incremental data size / original data size), and the compression rate threshold is set to ≥ 50%.

[0009] Furthermore, the method further includes: realizing the initial alignment of the measured point cloud and the theoretical model based on the improved ICP algorithm, and the feature weight distribution is leading edge > exhaust edge > blade back / blade basin; The improved ICP algorithm is optimized in the following way: using KD-Tree to accelerate the nearest neighbor search, and the node size is 0.05 - 0.15 mm.

[0010] Furthermore, a non-rigid deformation field is constructed based on the thin plate spline function to compensate for the non-linear displacement during the forging process; The construction of the non-rigid deformation field includes: defining a four-dimensional deformation tensor based on the material constitutive relationship; using the rigid translation amount and the non-linear displacement component as the initial constraints of the thin plate spline function; The deformation tensor model includes: Linear component: strain tensor (ε x , ε y , γ xy ) and rigid translation amount (δ x , δ y ); Non-linear component: displacement field interpolation based on the thin plate spline function; The tensile or compressive strain of the material in the x-axis direction is denoted as ε x , the tensile or compressive strain of the material in the y-axis direction is denoted as ε y , the shear slip deformation of the material in the x-y plane is denoted as γxy, the rigid translation amount of the material in the x-axis direction is denoted as δx, and the rigid translation amount of the material in the y-axis direction is denoted as δy.

[0011] Furthermore, in the method, the curvature calculation adopts: mapping the local curvature to the feature protection module, disabling compensation for high curvature regions (κ > 0.1 mm -1 ); enabling curvature adaptive filtering for medium-high curvature regions (0.05 mm -1 < κ < 0.1 mm -1 ).

[0012] Furthermore, based on the damping factor adaptive update mechanism of the LM algorithm, the update formula is: ; Among them, is a regulation factor, is the th iteration residual; The LM algorithm is optimized as: parameter update under the constraint of the Jacobian matrix condition number; historical Jacobian matrix reuse mechanism to reduce the amount of repeated calculation.

[0013] Furthermore, the weight distribution model satisfies: the weight of the intake edge is positively correlated with the local curvature; the weight of the exhaust edge is negatively correlated with the deviation from the theoretical model.

[0014] Furthermore, search for the minimum entropy compensation scheme based on the simulated annealing algorithm to optimize the residual distribution of the deformation field; The search process of the minimum entropy compensation scheme includes: using the entropy value of the residual field as the objective function; controlling the convergence speed of the simulated annealing through the temperature decay coefficient.

[0015] Furthermore, data preprocessing includes: implementing radius compensation for the laser scan point cloud to eliminate the inherent measurement error of the device; converting the theoretical CAD model into a NURBS subdivision surface with a subdivision level ≥ Lv4.

[0016] Furthermore, the detection performance indicators include: the time-consuming for double-notch synchronous detection ≤ 1.5 times the time-consuming for single-sided detection; the matching error rate of the intake / exhaust edge ≤ 1.2%; the residual RMS value ≤ 0.02 mm.

[0017] Furthermore, the real-time data quality monitoring module detects abnormal point cloud density and triggers data re-acquisition; The triggering conditions of the data quality monitoring module include: the local point cloud density is lower than the preset threshold (≤ 150 points / mm²); the coverage rate of the curvature mutation area < 90%.

[0018] Furthermore, the dynamic parameter adjustment includes: the minimum number of neighborhood points (MinPts) of the clustering algorithm is dynamically set based on the data density; the number of fitting iterations is adaptively allocated according to the curvature complexity; The dynamic setting formula of the minimum number of neighborhood points is: .

[0019] Furthermore, the multi-sensor data fusion module integrates the laser scan and coordinate measuring data; eliminates the systematic error between sensors through weighted averaging; The weight distribution of the sensor data fusion is based on: the confidence of the laser scan data (positively correlated with the surface reflectivity); the spatial resolution of the coordinate measuring data. Brief Description of the Drawings

[0020] Figure 1 is the schematic diagram of profile comparison provided by the present invention; Figure 2 The method flow chart is provided for the present invention. Specific embodiments

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0022] The present invention provides a method for detecting the profile of the double-notch air inlet and outlet edges of precision forging blades, including the following steps: a) Based on the double-domain decoupling strategy, an independent coordinate system is divided for the air inlet / outlet edge area to suppress the coupling interference of multiple notches; b) Implement multi-level noise suppression processing on the noisy point cloud data, including dynamic clustering analysis and deformation compensation mechanism; c) Adopt a two-way collaborative iterative optimization algorithm to synchronously perform forward fitting of the air inlet edge and reverse fitting of the air outlet edge, and achieve convergence control through cross-domain residual verification; d) Based on the adaptive segmentation strategy of the feature-sensitive area, dynamically adjust the data fitting interval; e) Construct an incremental data management module to achieve versioned storage and compression optimization of the iterative process.

[0023] Furthermore, the double-domain decoupling strategy includes: establishing an independent coordinate system according to the geometric features and functional weights of the air inlet / outlet edges; dynamically allocating the weights of the air inlet / outlet edge areas through a curvature weighting model, where the weight range of the air inlet edge is 0.5 - 0.6, and the weight range of the air outlet edge is 0.4 - 0.5.

[0024] Furthermore, the multi-level noise suppression processing specifically includes: The first layer: a clustering algorithm that dynamically adjusts the density based on material properties to suppress the noise of the hardened layer; The second layer: a compensation function that combines temperature and deformation parameters to correct the plastic deformation error; The neighborhood radius parameter of the clustering algorithm is dynamically calculated according to the hardness and surface roughness of the forging, and the expression is: Eps = k1(1 + k2), where k1 is the basic neighborhood radius and k2 is the material property correction coefficient; The noise suppression processing further includes: a noise classification module that identifies the noise type based on surface roughness and material hardness; Gaussian filtering is used for high-frequency noise, and median filtering is used for low-frequency noise; The input parameters of the noise classification module include: the surface hardness value range of the forging is 30 - 60 HRC; the roughness value range is Ra = 0.8 - 2.5 μm.

[0025] Further, the bidirectional collaborative iterative optimization algorithm satisfies the following convergence conditions: the spatial residual threshold ≤ 0.03 mm; the curvature gradient continuity conforms to the C² continuity criterion; The triggering conditions for cross-domain residual verification include: performing residual comparison every N iterations, where N is a preset positive integer; the verification threshold is dynamically adjusted based on the intake / exhaust edge weights, with a range ≤ 0.01 mm.

[0026] Further, the adaptive segmentation strategy includes: a dynamic interruption mechanism based on curvature mutation, with the curvature change threshold set to 0.05 - 0.15 mm -1 ; cumulative error threshold triggers segment reconstruction, with the error range ≤ 0.03 mm; The dynamic interruption mechanism includes: using morphological filtering to eliminate isolated noise points, and the filter kernel size is dynamically selected according to the feature sensitive area; retaining high-curvature feature points through bidirectional geometric constraints.

[0027] Further, the incremental data management module includes: a differential coding model that only records the data differences between iterations; a compression rate evaluation model that calculates the storage optimization efficiency in real time; The implementation method of the differential coding model is: performing lossless compression storage on the initial data set; only storing the difference matrix between the current data set and the previous version in subsequent iterations; The calculation formula of the compression rate evaluation model is: compression rate = 1 - (incremental data size / original data size), and the compression rate threshold is set to ≥ 50%.

[0028] Further, the method further includes: achieving the initial alignment of the measured point cloud and the theoretical model based on the improved ICP algorithm, with the feature weight distribution being leading edge > exhaust edge > blade back / shell; The improved ICP algorithm is optimized in the following way: using a KD-Tree to accelerate the nearest neighbor search, with the node size being 0.05 - 0.15 mm.

[0029] Further, a non-rigid deformation field is constructed based on thin plate spline functions to compensate for the non-linear displacement during the forging process; The construction of the non-rigid deformation field includes: defining a four-dimensional deformation tensor based on the material constitutive relationship; using the rigid translation amount and the non-linear displacement component as the initial constraints of the thin plate spline function; The deformation tensor model includes: a linear component: the strain tensor (εx, εy, γxy) and the rigid translation amount (δx, δy); a non-linear component: displacement field interpolation based on thin plate spline functions; The tensile or compressive strain of the material in the x-axis direction is denoted as ε x , and the tensile or compressive strain in the y-axis direction is denoted as ε y, the shear slip deformation of the material in the x-y plane is denoted as γxy, the rigid translation amount of the material in the x-axis direction is denoted as δx, and the rigid translation amount of the material in the y-axis direction is denoted as δy.

[0030] Further, in the method, curvature calculation is performed as follows: local curvature is mapped to the feature protection module, and compensation is disabled for high-curvature regions (κ > 0.1 mm -1 ); curvature adaptive filtering is enabled for medium-high curvature regions (0.05 mm -1 < κ < 0.1 mm -1 ).

[0031] Further, based on the damping factor adaptive update mechanism of the LM algorithm, the update formula is: ; where, is the adjustment factor, is the -th iteration residual; The LM algorithm is optimized as: parameter update under the constraint of the condition number of the Jacobian matrix; historical Jacobian matrix reuse mechanism to reduce the amount of repeated calculation.

[0032] Further, the weight distribution model satisfies: the weight of the intake edge is positively correlated with the local curvature; the weight of the exhaust edge is negatively correlated with the deviation from the theoretical model.

[0033] Further, based on the simulated annealing algorithm, search for the minimum entropy compensation scheme to optimize the residual distribution of the deformation field; The search process of the minimum entropy compensation scheme includes: using the entropy value of the residual field as the objective function; controlling the convergence speed of the simulated annealing through the temperature decay coefficient.

[0034] Further, data preprocessing includes: performing radius compensation on the laser scan point cloud to eliminate the inherent measurement error of the device; converting the theoretical CAD model into a NURBS subdivision surface, and the subdivision level ≥ Lv4.

[0035] Further, the detection performance indicators include: the time-consuming for double-notch synchronous detection ≤ 1.5 times the time-consuming for single-sided detection; the matching error rate of the intake / exhaust edge ≤ 1.2%; the residual RMS value ≤ 0.02 mm.

[0036] Further, the real-time data quality monitoring module detects abnormal point cloud density and triggers data re-acquisition; The triggering conditions of the data quality monitoring module include: the local point cloud density is lower than the preset threshold (≤ 150 points / mm²); the coverage rate of the curvature mutation region < 90%.

[0037] Furthermore, the dynamic parameter adjustment includes: the minimum number of neighboring points (MinPts) of the clustering algorithm is dynamically set based on data density; the number of fitting iterations is adaptively allocated according to the curvature complexity. The formula for dynamically setting the minimum number of neighboring points is: .

[0038] Furthermore, the multi-sensor data fusion module integrates laser scanning and coordinate measuring data; and eliminates the systematic error between sensors through weighted averaging. The weight allocation for the sensor data fusion is based on: the confidence level of the laser scanning data (positively correlated with the surface reflectivity); the spatial resolution of the coordinate measuring data.

[0039] In some embodiments, adopting a dual-domain decoupling strategy to overcome the multi-notch coupling interference includes: achieving regional decoupling by establishing independent coordinate systems on the intake side and the exhaust side. The weight of the intake side is set to 0.55, and the weight of the exhaust side is set to 0.45. The weight allocation is obtained by comprehensively considering the importance of the intake side and the exhaust side in the entire blade structure and their own geometric characteristic factors. Specifically, in order to overcome the problem of multi-notch coupling interference, the present invention adopts a dual-domain decoupling strategy. Specifically, regional decoupling is achieved by establishing independent coordinate systems for the intake / exhaust sides. When determining the weight allocation of the independent coordinate systems, according to the characteristics of the intake / exhaust sides, the weight of the intake side is set to 0.55, and the weight of the exhaust side is set to 0.45. This weight allocation is obtained by comprehensively considering various factors such as the importance of the intake / exhaust sides in the entire blade structure and their own geometric characteristics.

[0040] The weight is obtained based on a mathematical model, and the mathematical model is expressed as: ; ; where represents the intake weight, represents the exhaust weight, and represent the local curvature of the intake and the local curvature of the exhaust respectively. The final weight allocation result is = 0.55, = 0.45.

[0041] Through such a mathematical model, the relationship between the intake / exhaust sides in the independent coordinate system can be more accurately described, thereby effectively achieving regional decoupling and reducing the influence of coupling errors on the profile detection.

[0042] Dual-domain decoupled initialization includes: First, it is necessary to load the forging point cloud data. In this embodiment, the.asc format data is used, and its data density is 200 points / mm². This data format and density are determined according to actual detection requirements and equipment characteristics, and can meet the subsequent processing requirements. Second, when performing region segmentation, it is necessary to set the recognition thresholds for the intake edge and the exhaust edge. For the intake edge, the recognition threshold is set to curvature ≥ 0.08mm -1 ; for the exhaust edge, the recognition threshold is set to curvature ≥ 0.07mm -1 . These thresholds are determined through the analysis of a large number of precision forging blade samples and considering the geometric characteristics of the intake / exhaust edges, and can accurately distinguish the intake / exhaust edges from the overall data, laying a foundation for the subsequent dual-domain decoupled strategy.

[0043] Processing the cloud data with noise points using an improved clustering algorithm includes: Preprocessing the noise-resistant point cloud by improving the DBSCAN clustering algorithm, and setting the parameters Eps = 0.05mm and MinPts = 6 to process the point cloud data with noise points; when dealing with the noise problem caused by the surface characteristics of the forging, the present invention adopts an improved DBSCAN clustering algorithm. Through specific parameter settings (Eps = 0.05mm, MinPts = 6), this algorithm can effectively process the point cloud data with noise points. After being processed by this algorithm, the noise filtering rate can reach ≥ 95%, greatly improving the quality of the point cloud data.

[0044] Noise-resistant point cloud processing includes: The input data is the original data with 15% noise points. This high noise content data is a common situation in the actual detection process, reflecting the influence of the forging surface characteristics on the scanned data. Set the parameters of the improved DBSCAN clustering algorithm to Eps = 0.05mm and MinPts = 6. These parameter settings are obtained through a large number of experiments and optimizations, and can make the algorithm achieve the best effect when processing this kind of data with noise points. After the above processing, the output is effective point cloud data, and the noise content ≤ 5%. This indicates that after the noise-resistant point cloud preprocessing, the quality of the point cloud data has been significantly improved and can meet the requirements of subsequent profile detection for data quality.

[0045] Use the method of bidirectional progressive fitting to process the data of the intake edge and the exhaust edge, synchronously perform the forward iteration of the intake edge and the reverse iteration of the exhaust edge, and perform cross-domain residual verification every 5 iterations; the convergence condition of the bidirectional progressive fitting is that the spatial residual satisfies ≤ 0.02mm and the curvature gradient is continuous.

[0046] Among them, the present invention uses a two-way progressive fitting method to process the data of the intake / exhaust edges. The core lies in simultaneously performing iterative fitting in different directions for the intake edge and the exhaust edge. The intake edge adopts forward iteration, that is, the fitting operation is carried out in sequence according to the natural order of the data; the exhaust edge adopts reverse iteration, that is, starting from the end of the data and performing fitting in reverse. This two-way operation mode can fully consider the respective characteristics of the intake / exhaust edges. For example, there are differences in aspects such as airflow influence and geometric shape between the intake edge and the exhaust edge. Two-way fitting can capture these differences more accurately and improve the accuracy of fitting.

[0047] During the iterative process, cross-domain residual verification is carried out every 5 iterations. The residual refers to the difference between the fitting result and the actual data. Cross-domain residual verification is to check the difference between the fitting results of the intake edge and the exhaust edge. By regularly performing cross-domain residual verification, possible deviations in the fitting process can be detected in a timely manner. For example, if after several iterations, there is a large inconsistency between the fitting results of the intake edge and the exhaust edge, it indicates that there may be a fitting deviation. At this time, the fitting process can be adjusted and corrected to ensure that the final fitting result is more in line with the actual situation. Its convergence conditions include two aspects: one is that the spatial residual needs to satisfy ≤0.02mm), this standard ensures the accuracy of the fitting result in terms of spatial position; the other is to ensure the continuity of the curvature gradient (C² continuity), this condition ensures the smoothness of the fitting curve from a mathematical perspective and avoids situations such as sudden curvature changes that do not conform to the actual blade shape.

[0048] The convergence conditions include: Spatial residual condition: The spatial residual needs to satisfy ≤0.02mm. The spatial residual reflects the degree of deviation between the fitting result and the actual data in terms of spatial position. Setting this standard is to ensure a high degree of accuracy of the fitting result in terms of spatial position. For example, in the design and manufacture of blades, the spatial position accuracy of the intake / exhaust edges directly affects the aerodynamic performance of the blades. If the spatial residual is too large, it may cause problems such as airflow disorder when the blades are working. Curvature gradient continuity condition: It is necessary to ensure the continuity of the curvature gradient (C² continuity). Curvature describes the degree of bending of a curve. Continuity of the curvature gradient means that the bending change of the curve is smooth and there will be no sudden curvature change. In the actual blade shape, the curves of the intake / exhaust edges are usually smoothly transitioned, and sudden curvature changes do not conform to the actual situation. From a mathematical perspective, C² continuity ensures the continuity of the second derivative of the curve, making the fitting curve smoother mathematically and more in line with the actual shape requirements of the blades.

[0049] Bidirectional progressive fitting includes: for the intake edge, a forward iteration method is adopted, and the LM algorithm is used for fitting operations, where μ = 0.005. This parameter value is determined according to the specific data characteristics and fitting requirements of the intake edge, and can ensure effectively approaching the true contour curve during the forward iteration process.

[0050] For the exhaust edge, a reverse iteration method is adopted, and the LM algorithm is also used. However, here μ = 0.008. The setting of this parameter takes into account the differences in geometric shape and data distribution between the exhaust edge and the intake edge. By adjusting the μ value to adapt to the characteristics of the exhaust edge, the accuracy of the reverse iteration is ensured.

[0051] During the bidirectional progressive fitting process, cross-domain residual verification is performed every 5 iterations. The verification threshold is set as ΔRMS ≤ 0.005 mm. Through this regular verification mechanism, possible deviations in the fitting process of the intake edge and the exhaust edge can be detected and corrected in a timely manner, ensuring that the entire fitting process proceeds in the accurate direction.

[0052] The process of processing the data of the intake edge and the exhaust edge by the method of bidirectional progressive fitting includes: 1. Data preprocessing: Load the measured point cloud data (in the.las1.4 format) after radius compensation. The point cloud data is a set of a large number of discrete points obtained by a three-dimensional measurement device (such as a laser scanner). These points record the three-dimensional coordinate information of the object surface. During the measurement process, due to reasons such as the characteristics of the measurement device, there may be radius errors. Radius compensation is to correct these errors so that the point cloud data can more accurately reflect the actual shape of the intake / exhaust edge..las1.4 is a common point cloud data storage format, which has the characteristics of high efficiency and flexibility, and can store a large amount of point cloud data and related attribute information, such as the color and intensity of the points. Loading this format of data is the basis for subsequent processing.

[0053] Import the theoretical CAD model and convert it into a subdivided NURBS surface (subdivision level: Lv5). The theoretical CAD model is a three-dimensional model of the blade created according to the design requirements, and it represents the ideal shape of the blade. Import it into the processing system as a reference for comparing and fitting with the measured point cloud data.

[0054] The NURBS (Non-Uniform Rational B-Splines) surface is a mathematical model that can accurately represent complex geometric shapes, such as the intake / exhaust edges of the blade. The subdivision operation is to further refine the NURBS surface to improve the accuracy and details of the surface. The subdivision level Lv5 indicates that the surface has been subdivided five levels. The higher the subdivision level, the better the accuracy and details of the surface, and it can be more accurately matched and fitted with the measured point cloud data.

[0055] 2. Initial coordinate system alignment: Execute the improved ICP algorithm, and perform feature weight assignment: the leading edge weight is 0.6, the exhaust edge weight is 0.3, and the back / front weight is 0.1. The ICP (Iterative Closest Point) algorithm is a commonly used point cloud registration algorithm for aligning point cloud data in different coordinate systems to the same coordinate system. The improved ICP algorithm is optimized based on the traditional ICP algorithm and can complete the registration task more efficiently and accurately.

[0056] The feature weight assignment is determined according to the importance of the intake / exhaust edges, leading edge, and back / front in the blade. The leading edge plays a key role in guiding the airflow, and its shape and position directly affect the intake efficiency and aerodynamic performance of the blade, so a higher weight of 0.6 is assigned. The exhaust edge has an important impact on the discharge of the airflow, with a weight of 0.3. The back / front has relatively less direct impact on the airflow, with a weight of 0.1. Through this weight assignment, more attention can be paid to the alignment of important features during the registration process, improving the accuracy of registration.

[0057] Use KD-Tree to accelerate the nearest point search (node size: 0.1mm): In the ICP algorithm, it is necessary to continuously find the nearest point pairs in the point cloud. The KD-Tree (K-Dimensional Tree) is an efficient spatial indexing structure that can partition and organize points in a high-dimensional space, thus quickly finding the point closest to a given reference point.

[0058] The node size is set to 0.1mm, and this parameter affects the partitioning accuracy of the KD-Tree. A smaller node size can improve the search accuracy but increase the tree construction and search time; a larger node size has the opposite effect. Selecting a node size of 0.1mm is a balance between search accuracy and efficiency, which can accelerate the nearest point search speed while ensuring search accuracy, thus improving the efficiency of the entire registration process.

[0059] 3. Iterative optimization process: Dynamic interruption strategy: Perform adaptive segmentation according to curvature and error.

[0060] Based on the curvature change and error accumulation, perform adaptive segmentation on the intake edge and exhaust edge data, and use the dynamic interruption strategy to optimize the iterative process for optimization; The formula for detecting sudden curvature change is: ; In the formula, represents the curvature change amount between the i-th point and the (i + 1)-th point; represents the curvature of the i-th point; represents a preset curvature change threshold; The formula for error accumulation judgment is: ; represents the coordinates of the j-th actual measurement point, which represents the real physical position data in the point cloud data; is the fitting function value with respect to j, that is, the theoretical coordinate value at the corresponding position obtained by the curve fitting algorithm; represents a preset error threshold.

[0061] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for detecting the profile of a double-notch air inlet and outlet edge of a precision forging blade, characterized in that It includes the following steps: a) Independently divide the coordinate system for the inlet / exhaust edge region based on the dual-domain decoupling strategy to suppress the coupling interference of multiple notches; b) Implement multi-level noise suppression processing on the noisy point cloud data, including dynamic clustering analysis and deformation compensation mechanism; c) Adopt a two-way collaborative iterative optimization algorithm to synchronously perform forward fitting of the inlet edge and reverse fitting of the exhaust edge, and achieve convergence control through cross-domain residual verification; d) Based on the adaptive segmentation strategy of the feature-sensitive region, dynamically adjust the data fitting interval; e) Construct an incremental data management module to realize versioned storage and compression optimization during the iterative process.

2. The method according to claim 1, wherein The dual-domain decoupling strategy includes: establishing an independent coordinate system according to the geometric characteristics and functional weights of the inlet / exhaust edge; dynamically allocating the weights of the inlet / exhaust edge regions through a curvature weighting model, where the weight range of the inlet edge is 0.5 - 0.6, and the weight range of the exhaust edge is 0.4 - 0.

5.

3. The method according to claim 1, wherein The multi-level noise suppression processing specifically includes: The first layer: a clustering algorithm that dynamically adjusts the density based on material properties to suppress the noise of the hardened layer; The second layer: a compensation function that combines temperature and deformation parameters to correct the plastic deformation error; The neighborhood radius parameter of the clustering algorithm is dynamically calculated according to the hardness and surface roughness of the forging, and the expression is: Eps = k1(1 + k2), where k1 is the basic neighborhood radius and k2 is the material property correction coefficient; The noise suppression processing further includes: a noise classification module that identifies the noise type based on surface roughness and material hardness; using Gaussian filtering for high-frequency noise and median filtering for low-frequency noise; The input parameters of the noise classification module include: the forging surface hardness value range is 30 - 60 HRC; the roughness value range is Ra = 0.8 - 2.5 μm.

4. The method according to claim 1, characterized in that The two-way collaborative iterative optimization algorithm meets the following convergence conditions: the spatial residual threshold ≤ 0.03 mm; the curvature gradient continuity conforms to the C² continuity standard; The triggering conditions for the cross-domain residual verification include: performing a residual comparison every N iterations, where N is a preset positive integer; the verification threshold is dynamically adjusted based on the weights of the inlet / exhaust edge, and the range ≤ 0.01 mm.

5. The method according to claim 1, characterized in that The adaptive segmentation strategy includes: a dynamic interruption mechanism based on curvature mutation, and the curvature change threshold is set to 0.05 - 0.15 mm -1 ; the cumulative error threshold triggers segmentation reconstruction, and the error range ≤ 0.03 mm; The dynamic interruption mechanism includes: using morphological filtering to eliminate isolated noise points, and the size of the filtering kernel is dynamically selected according to the feature-sensitive region; retaining high-curvature feature points through two-way geometric constraints.

6. The method according to claim 1, characterized in that, The incremental data management module includes: a differential coding model that only records the data differences between iterations; a compression rate evaluation model that calculates the storage optimization efficiency in real time; The implementation method of the differential coding model is: perform lossless compression storage on the initial data set; only store the difference matrix between the current data set and the previous version in subsequent iterations; The calculation formula of the compression rate evaluation model is: compression rate = 1 - (incremental data size / original data size), and the compression rate threshold is set to ≥ 50%; 7. The method according to claim 1, characterized in that The method also includes: realizing the initial alignment of the measured point cloud and the theoretical model based on the improved ICP algorithm, and the feature weight distribution is leading edge > exhaust edge > blade back / shell; The improved ICP algorithm is optimized in the following way: using a KD-Tree to accelerate the nearest neighbor search, and the node size is 0.05 - 0.15 mm.

8. The method according to claim 1, wherein Construct a non-rigid deformation field based on thin plate spline function to compensate for the non-linear displacement during forging process; The construction of the non-rigid deformation field includes: defining a four-dimensional deformation tensor based on the material constitutive relation; using the rigid translation amount and the non-linear displacement component as the initial constraints of the thin plate spline function; The deformation tensor model includes: Linear component: strain tensor (ε x , ε y , γ xy ), and rigid translation (δ x , δ y ); Nonlinear component: displacement field interpolation based on thin plate spline function The tensile or compressive strain of the material in the x-axis direction is denoted as ε x , the tensile or compressive strain of the material in the y-axis direction is denoted as ε y , the shear slip deformation of the material in the x-y plane is denoted as γ xy , the rigid translation amount of the material in the x-axis direction is denoted as δ x , the rigid translation amount of the material in the y-axis direction is denoted as δ y。 9. The method according to claim 1, wherein In the method, the curvature calculation adopts: locally mapping the curvature to the feature protection module, and for the high-curvature region where κ > 0.1 mm -1 Disable compensation; for the medium-high curvature region of 0.05 mm -1 < κ < 0.1 mm -1 Enable curvature adaptive filtering.

10. The method according to claim 1, wherein The method further includes: a damping factor adaptive update mechanism based on the LM algorithm, and the update formula is: ； Among them, is a regulation factor, is the residual of the -th iteration. The LM algorithm is optimized as: parameter update under the condition number constraint of the Jacobian matrix; a historical Jacobian matrix reuse mechanism to reduce the amount of repeated calculation.

Images