Self-adaptive cleaning method and system for damaged blade and aircraft engine blade repairing method

Abstract

The invention belongs to the technical field of aero-engine blade repairing, and discloses a damaged blade self-adaptive cleaning method and system and an aero-engine blade repairing method.The method comprises the steps that scanning point cloud of a damaged blade and sampling point cloud of a blade design model are compared, the distance between the scanning point cloud and the sampling point cloud to the nearest point of the opposite side is calculated, and a threshold value is set; determining a damaged area and a to-be-cladded area; feature segmentation is carried out on the to-be-cladded region based on a region growing method, and the to-be-cladded region is segmented into a suction surface, a pressure surface and a blade edge; designing the shape of a bounding box based on the damaged area and the feature surface, obtaining a cleaning model of the target area, and filling a cavity of a cleaning section; according to the method, a rectangular or trapezoidal bounding box is constructed based on a feature segmentation result of a to-be-cladded area of a damaged blade, points in the bounding box are removed, a blade mechanical cleaning model with a regular cleaning section shape and a reasonable direction is adaptively obtained according to different damage conditions, convenience is provided for follow-up laser cladding path planning, and the working efficiency is improved. And the subsequent cladding quality is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of aero-engine blade repair, and particularly relates to an adaptive cleaning method and system for damaged blades and an aero-engine blade repair method. Background Art

[0002] The repair process of in-service damaged blades generally includes damaged area identification, mechanical cleaning, and additive and subtractive composite manufacturing. In the additive and subtractive composite manufacturing means during the blade repair process, generally, the laser cladding technology is first used to fill the damaged area of the blade with the same metal, and then the excess material after additive manufacturing is removed by means of adaptive milling to ensure that the surface quality of the repaired blade meets the requirements for re-service.

[0003] During the blade repair process, if the material mechanical properties of the damaged area have changed significantly, it is usually necessary to remove the material near the damaged area to ensure that the repaired blade has consistent performance and strength. At present, mechanical cleaning is generally carried out by manually operating a machine tool for milling, and the cross-sectional shape and cross-sectional direction of the cleaning are relatively arbitrary. This method has a low degree of automation, poor adaptability to diverse damage situations, and is likely to affect the subsequent cladding quality. Summary of the Invention

[0004] In view of the above defects or improvement requirements of the prior art, the present invention provides an adaptive cleaning method and system for damaged blades and an aero-engine blade repair method, which improve the adaptability of the mechanical cleaning process of damaged blades to different damage situations, provide convenience for the subsequent laser cladding path planning, and help improve the subsequent cladding quality.

[0005] To achieve the above object, according to one aspect of the present invention, an adaptive cleaning method for damaged blades is provided, including the following steps:

[0006] S1: By comparing the scanned point cloud of the damaged blade with the sampled point cloud of the blade design model, calculate the distances from each other to the nearest points respectively and set a threshold to determine the damaged area and the area to be cladded;

[0007] S2: Based on the region growing method, perform feature segmentation on the area to be cladded obtained in step S1, and divide the area to be cladded into a suction surface, a pressure surface, and a blade edge;

[0008] S3: Design the shape of the bounding box based on the damaged area identified in step S1 and the feature surfaces segmented in step S2, obtain the cleaning model of the target area, and fill the holes in the cleaning cross-section.

[0009] Preferably, step S1 specifically includes the following steps:

[0010] S11: using a three-dimensional scanning method to obtain an actual scanning point set P of the damaged blade and performing point cloud sampling on the blade design model to obtain a design point set Q, aligning the actual scanning point set P with the design point set Q, and transforming them into the same coordinate system;

[0011] S12: Search the coordinate system of step S11 for the point p in the actual scanning point set P. i The closest point q to the design point set Q j and the midpoint q of the design point set Q i The closest point p to the actual scan point set P j , and calculate the Euclidean distance between the two and record it as D pi , D qi :

[0012] D pi =min(dist(p i ,q j ))

[0013] D qi =min(dist(q i ,p j ))

[0014] S13: Based on step S12, a screening threshold is set to obtain point clouds of the damaged area and the area to be clad.

[0015] As a preference, in step S13, twice the resolution 2*re is used as the screening threshold. pi >2*re, then p i Store the damaged area point set E; if D qi >2*re, then q i Store the point set W of the area to be clad.

[0016] Preferably, step S2 specifically includes the following steps:

[0017] S21: assigning an initial value l=0 to all point area labels in the point set W of the area to be clad obtained in step S13, and setting a normal vector difference threshold θ th and the curvature threshold c th ;

[0018] S22: construct the curvature ascending sequence {C} of the area label l=0 point in step S21, and take the point with the smallest curvature as the initial seed point S[0];

[0019] S23: Pop point S[0] from the seed point queue {S} in step S21, and traverse the neighborhood I of S[0]. k [S[0]], k is the number of neighboring points, and the normal vector angle θ between the neighboring point and the seed point is calculated. c and the curvature c of the neighborhood point c;

[0020] S24: If θ c ≤ θ th , then add the current point to the growth region, set the corresponding region label for this neighborhood point, and then continue the comparison. If c c < c th , then add the current point to the seed point set {S}; if θ c > ht θ, then skip this point;

[0021] S25: Repeat steps S23 and S24. Finally, when the seed point set {S} is emptied, it represents the completion of one region growth;

[0022] S26: For the points with the remaining region label l = 0, continue to construct a curvature ascending sequence, and repeat steps S22, S23, and S24 until the region labels of all points l ≠ 0, then the region growth is completed. The three regions respectively correspond to the suction surface, the blade edge, and the pressure surface of the area to be cladded.

[0023] Preferably, step S3 specifically includes the following steps:

[0024] S31: Perform principal component analysis on the suction surface or pressure surface point cloud after segmentation in step S2, where is the first principal direction and is the second principal direction, and use as the projection direction of the area to be cladded point set;

[0025] S32: Identify the blade edge points based on the K-Means algorithm to obtain the blade edge direction;

[0026] S33: Adjust the position of the area to be cladded point cloud according to the projection direction obtained in step S31 and the blade edge direction obtained in step S32 to obtain the point cloud E of the damaged area after rotation and projection, rp the point cloud D of the damaged blade scan point set D after rotation and projection rp ;

[0027] S34: Based on the damaged area identified in step S1, construct a cleaning bounding box for local repair and generate a cleaning model for the target area accordingly;

[0028] S35: Fill the cleaned hole cross-section according to the cleaning bounding box parameters in step S34.

[0029] Preferably, step S34 specifically includes the following steps:

[0030] S341: Calculate the centroid G rp , D rp of the point clouds E e (x e , z e ), Gd (x d , z d ), calculate E rp The maximum and minimum values of x and z on the X and Z axes, x min , x max , z min , z max and E rp The resolution re of the point cloud;

[0031] Specifically,

[0032] S342: If x e ≤ x d , let l1: x + z - x max - z max - a = 0, l2: x - z - x max + z min + b = 0, l3: x = x min , if x e > x d , let l1: x - z - x min + z max - a = 0, l2: x + z - x min - z min + b = 0, l3: x = x max ;

[0033] S343: Calculate the set {S1} of the distances from each point of the point cloud E rp to l1 and the set {S2} of the distances from each point of the point cloud E rp to l2, and sort the sets {S1} and {S2} in ascending order;

[0034] S344: Compare the first item S1[0] in the set {S1} with δ2. If S1[0] ≤ δ2, update a = a + re and return to step S343;

[0035] Compare the first item S2[0] in the set {S2} with δ2. If S2[0] ≤ δ2, update b = b + re and return to step S343;

[0036] S345: Expand the final l1, l2, and l3 outward by a cleaning depth d. The straight lines l1, l2, and l3 enclose a trapezoidal closed area. Calculate the points of the point cloud D rp located within this closed area, record their indices as I1, D rp the points located outside this closed area have indices I2; Extract the points in I1 and I2 from the damaged blade scan point set D, and denote the removed point cloud as R and the cleaned point cloud as C.

[0037] Preferably, for the damage of the blade body and blade angle, step S34 specifically includes the following steps:

[0038] S341': The first principal direction of point cloud E through principal component analysis rp of Calculate the centroid G rp of point cloud E e (x e , z e ), and determine the straight line l through the principal direction and the centroid G e : Ax + By + Cz + D = 0; g

[0039] S342': The straight line l g divides the points in point cloud E rp into two sides, retains the point set on the same side as the root point and records it as H. Calculate the Euclidean distance dist(h i to l g ), and obtain the maximum Euclidean distance dist i (h g , l max ). The straight line l i is offset by a distance of dist g (h g , l max ) towards the root side; i g

[0040] S343': The straight line l g divides the two-dimensional point cloud D rp into a cleaning area and a post-cleaning area. The removed point cloud is recorded as R, and the post-cleaned point cloud is recorded as C'.

[0041] Preferably, step S35 specifically includes the following steps:

[0042] S351: Take R as the search point set, and calculate whether there is a point in point cloud C i within the range of the radius that can be searched in point set R. If found, record the index in index set I3. The points in R that are indexed in I3 are the boundary point set of the damaged blade cleaning area, denoted as B;

[0043] S352: Perform inverse transformation on the three sides l1, l2, l3 of the trapezoid according to the rotation matrix M to obtain the planes l 1r , l 2r , l 3r . Construct filling points at certain intervals on the three planes. After construction, extract the points within the boundary B as the final cross-section filling point set A, and combine A with C as the final post-cleaning point cloud model.

[0044] ​​​To achieve the above object, according to one aspect of the present invention, an adaptive cleaning system for damaged blades is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it executes the above-mentioned adaptive cleaning method for damaged blades.

[0045] To achieve the above object, according to one aspect of the present invention, a method for repairing aeroengine blades is provided, including the following steps: (1) obtaining a cleaning model of the mechanical target area of the aeroengine blade by using the above-mentioned adaptive cleaning method for the point cloud of damaged blades, and (2) performing laser cladding and material removal based on the cleaning model of the mechanical target area.

[0046] Generally speaking, compared with the prior art through the above technical solutions conceived by the present invention, the adaptive cleaning method, device and application for the point cloud of damaged blades provided by the present invention mainly have the following beneficial effects:

[0047] 1. The present invention adaptively generates corresponding mechanical cleaning models according to different damage modes and deformation degrees of the blades, with a wide range of applications and high automation.

[0048] 2. By designing the cleaning section direction, the present invention makes the stacking direction of the cladding material consistent with the main direction of the blade, overcoming the problem that the thin-walled blade body has a small bearing area for the molten stacking material and is prone to molten pool dripping. Compared with the section direction generated by the convex hull in the traditional method, it is more conducive to subsequent cladding.

[0049] 3. By designing the cleaning section as a trapezoid, the present invention can provide more contact points at the bevel angle and promote faster heat conduction, which helps to improve the cladding strength and reduce the risk of thermal deformation in the subsequent cladding process. Setting the trapezoid angle to 135° is convenient for the cladding head to penetrate for cladding and is not easy to interfere with the blade. Compared with the quadratic surface section, the trapezoid section fully considers the engineering reality and provides convenience and quality guarantee for the subsequent cladding process. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a flowchart of an adaptive cleaning method for damaged blades provided by the present invention;

[0051] Figure 2 is a point cloud diagram of the area to be clad calculated in step S1 of the embodiment of the present invention;

[0052] Figure 3 is an algorithm flowchart for feature segmentation of the point set W of the area to be clad in step S2 of the embodiment of the present invention;

[0053] Figure 4 is the feature segmentation result of blades with different damage modes in step S2 of the embodiment of the present invention;

[0054] Figure 5 It is the flowchart of the cleaning algorithm in step S3 of the embodiment of the present invention under the condition of leaf edge breakage;

[0055] Figure 6 It is the overall cleaning process diagram of the embodiment of the present invention under the conditions of leaf edge breakage and leaf corner breakage. Specific embodiments

[0056] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0057] Please refer to Figure 1 and Figure 2 , the present invention proposes an adaptive cleaning method for damaged blades. The method can adaptively obtain a mechanical cleaning model for damaged blades according to different damage conditions, providing convenience for subsequent laser cladding path planning and helping to improve the subsequent cladding quality.

[0058] The method mainly includes the following steps:

[0059] S1: By comparing the scanned point cloud of the damaged blade with the sampled point cloud of the blade design model, calculate the distances from both to the nearest points of each other respectively and set a threshold to determine the damaged area and the area to be clad;

[0060] Further, step S1 includes:

[0061] S11: Obtain the actual scanned point set P of the damaged blade by three-dimensional scanning and obtain the designed point set Q by point cloud sampling of the blade design model. Register the actual scanned point set P and the designed point set Q and transform them into the same coordinate system;

[0062] That is, obtain the scanned data of the damaged blade by a three-dimensional scanner, filter and denoise the scanned point set to improve the quality of the scanned point cloud. Perform uniform downsampling on the blade design model and the scanned data of the damaged blade to obtain scanned data P and designed data Q with a fixed number of points. Register the point set P and the point set Q and transform them into the same coordinate system; Obtain the actual scanned point set P of the damaged blade by three-dimensional scanning and the corresponding normal vector set NP(p i ) of each point p i in it, and calculate the curvature CP(p i ) corresponding to each point p i according to the normal vector; Obtain the designed point set Q by point cloud sampling of the blade design model, find the patch in the STL model closest to the sampled point, and use the normal vector of the patch as q in the designed point set Qi The normal vector NQ(q i ), and calculate each point q according to the normal vector i The corresponding curvature CQ(q i ); The actual scan point set P is registered with the design point set Q using the ICP algorithm and transformed into the same coordinate system. The uniform downsampling method is used to make the two point sets have the same preservation resolution, denoted as re.

[0063] S12: Search the coordinate system of step S11 for the point p in the actual scanning point set P. i The closest point q to the design point set Q j and the midpoint q of the design point set Q i The closest point p to the actual scan point set P j , and calculate the Euclidean distance between the two and record it as D pi , D qi :

[0064] D pi =min(dist(p i ,q j ))

[0065] D qi =min(dist(q i ,p j ))

[0066] S13: Based on step S12, a screening threshold is set to obtain point clouds of the damaged area and the area to be clad. The twice resolution 2*re is used as the screening threshold. If D pi >2*re, then p i Store the damaged area point set E, if D qi >2*re, then q i The point set W of the area to be clad is stored. The areas to be clad extracted under different damage modes are as follows: Figure 2 As shown, the error band is set according to the curvature of each point, and the number in the error band is the curvature value of the point.

[0067] S2: performing feature segmentation on the area to be clad obtained in step S1 based on the region growing method, and dividing the area to be clad into a suction surface, a pressure surface, and a blade edge;

[0068] like Figure 3 As shown, specifically including:

[0069] S21: assigning an initial value l=0 to all point area labels in the point set W of the area to be clad obtained in step S13, and setting a normal vector difference threshold θ th and the curvature threshold c th ;

[0070] S22: Ascendingly sort the curvature sequence {C} of the point with the constructed region label l = 0 in step S21, and use the point with the minimum curvature as the initial seed point S[0].

[0071] S23: Pop the point S[0] from the seed point queue {S} in step S21, and traverse the neighborhood I k [S[0]], where k is the number of neighborhood points, and calculate the included angle θ between the normal vector of the neighborhood point and the seed point c and the curvature c of this neighborhood point c ;

[0072] S24: If θ c ≤θ th , then add the current point to the growing region, set the corresponding region label for this neighborhood point, and then continue the comparison. If c c <c th , then add the current point to the seed point set {S}; if θ c > ht θ, then skip this point;

[0073] S25: Repeat steps S23 and S24. Finally, when the seed point set {S} is emptied, it represents the completion of one region growth;

[0074] S26: Continue to construct the ascending curvature sequence for the remaining points with region label l = 0, and repeat steps S22, S23, and S24 until the region labels of all points l≠0, then the region growth is completed. The three regions respectively correspond to the suction surface, the blade edge, and the pressure surface of the area to be cladded.

[0075] S26. Continue to construct the ascending curvature sequence for the remaining points with region label l = 0, and repeat steps S22, S23, and S24 until the region labels of all points l≠0, then the region growth is completed. The three regions respectively correspond to the suction surface, the blade edge, and the pressure surface of the area to be cladded.

[0076] S3: Design the shape of the bounding box based on the damaged area identified in step S1 and the feature surfaces segmented in step S2, obtain the cleaning model of the target area, and fill the holes in the cleaning section;

[0077] Step S3 specifically includes the following steps:

[0078] S31: Perform principal component analysis on the point cloud of the suction surface or pressure surface segmented in step S2, where is the first principal direction and is the second principal direction, and use as the projection direction of the point set of the area to be cladded;

[0079] Specifically, write the point cloud coordinates of the suction surface (or pressure surface) into the matrix and calculate the average coordinates of the point cloud on the three coordinate axes is:

[0080]

[0081] wherein, n is the number of points in the suction surface (or pressure surface) point cloud.

[0082] Perform de-centralization processing on the point cloud coordinates to obtain the de-centralized matrix X q , and calculate the covariance matrix X c is

[0083]

[0084] Solve for the eigenvalues v1, v2, v3 of X c and the corresponding eigenvectors Arrange them in descending order according to the eigenvalue magnitude. After the descending order arrangement, the eigenvectors corresponding to each eigenvalue are the first, second, and third principal components. Taking the case of v1 > v2 > v3 as an example, the first principal component direction and the second principal component direction retain the most point cloud information. Therefore and the plane spanned by is the main projection plane of the suction surface (or pressure surface) point cloud, and the normal vector of this plane is used as the projection direction of the point set W

[0085] S32: Identify the leaf edge points based on the K-Means algorithm to obtain the leaf edge direction;

[0086] Cluster the curvatures of each point in the one-dimensional point set W of the area to be clad, obtain the leaf edge point recognition threshold and screen the leaf edge points; Calculate the distances between the leaf edge points, and connect the two leaf edge points with the maximum Euclidean distance to obtain the leaf edge direction

[0087] The K-Means algorithm specifically includes the following steps:

[0088] S321: Set the number of clusters to 2, and use the minimum and maximum values of the suction surface point cloud curvature as the initial centers θ = θ1, θ2;

[0089] S322: For each point W i in the point set W of the area to be clad, calculate its distances to the two cluster centers and assign it to the class corresponding to the cluster center with the minimum distance;

[0090] S323: Assume that the kth cluster contains n k data points, denoted as the set For each class, recalculate its cluster center:

[0091]

[0092] Among them is the cluster center of the k-th updated cluster.

[0093] S324: Repeat the two operations of S322 and S323 until the change in the iteration error is less than 0.01. After clustering, the cluster center with a large curvature value is used as the leaf margin point recognition threshold to identify leaf margin points.

[0094] S33: Adjust the position of the point cloud in the area to be clad according to the projection direction obtained in step S31 and the leaf margin direction obtained in step S32;

[0095] That is, according to the projection direction and the leaf margin direction adjust the position of the point cloud so that the normal vector points to the positive direction of the Y axis, and the leaf margin direction is consistent with the positive direction of the Z axis, and calculate the rotation matrix M. For the damage of the blade body and blade angle, the leaf margin direction does not need to be calculated. Multiply the damaged area point set E and the damaged blade scan point set D by the rotation matrix M and project them onto the XOZ plane to obtain the point cloud E rp after rotation and projection of the damaged area point set E, and the point cloud D rp after rotation and projection of the damaged blade scan point set D.

[0096] S34. Construct a cleaning bounding box for local repair based on the damaged area obtained in step S33, and generate a cleaning model for the target area accordingly;

[0097] As Figure 5 shown, it specifically includes:

[0098] S341. Calculate the centroids G rp , D rp of the point clouds E e (x e , z e ), G d (x d , z d ), calculate the maximum and minimum values x rp , x min , z max , z min of E max on the X and Z axes, and the resolution re of the point cloud of E rp . Set the initial values a = 0, b = 0.

[0099] The calculation process of the point cloud centroid and resolution is as follows:

[0100]

[0101] Traverse the point cloud E rpEach point ep in i , search for the point eb rp nearest to ep in the point cloud E i , calculate the Euclidean distance between the two points, accumulate the sum of the Euclidean distances between the two points, and divide by the number of points in the point cloud to calculate the resolution re of E i : rp The resolution re of E:

[0102]

[0103] S342: If x e ≤x d , let l1: x + z - x max -z max -a = 0, l2: x - z - x max +z min +b = 0, l3: x = x min , if x e >x d , let l1: x - z - x min +z max -a = 0, l2: x + z - x min -z min +b = 0, l3: x = x max ;

[0104] S343: Calculate the set {S1} of the distances of each point in the point cloud E rp from l1 and the set {S2} of the distances of each point in the point cloud E rp from l2, and sort the sets {S1} and {S2} in ascending order;

[0105] S344: Compare the first item S1[0] in the set {S1} with δ2. If S1[0] ≤ δ2, update a = a + re and go back to step S343;

[0106] Compare the first item S2[0] in the set {S2} with δ2. If S2[0] ≤ δ2, update b = b + re and go back to step S343.

[0107] S345: Expand the final l1, l2, and l3 outward by the cleaning depth d. The straight lines l1, l2, and l3 enclose a trapezoidal closed area. Calculate the points in the point cloud D rp located within this closed area, record their indices as I1, and the indices of the points in the point cloud D rp located outside this closed area as I2; Extract the points in I1 and I2 from the damaged blade scan point set D, record the removed point cloud as R, and the cleaned point cloud as C.

[0108] For damage to the blade body and blade angle, the steps of S34 are different, specifically as follows:

[0109] S341’: Determine the first principal direction of the point cloud E through principal component analysis rp of the point cloud E Calculate the centroid G rp of the point cloud E e (x e , z e ), and determine the straight line l through the principal direction e and the centroid G g : Ax + By + Cz + D = 0;

[0110] S342’: The straight line l g divides the points in the point cloud E rp into two sides, and the point set on the same side as the blade root point is retained and denoted as H. Calculate the Euclidean distance dist(h i to l g ) of each point h in the point set H. Obtain the maximum Euclidean distance dist i (h g , l max ). The straight line l i is offset by a distance of dist g (h g , l max ) towards the blade root side; i g g

[0111] S343’: The straight line l g divides the two-dimensional point cloud D rp into a cleaning area and a post-cleaning area. The cut-off point cloud is denoted as R, and the post-cleaned point cloud is denoted as C’.

[0112] S35: Clean the bounding box parameters in step S34 and fill the holes in the cross-section after cleaning.

[0113] Step S35 specifically includes the following steps:

[0114] S351: Use R as the search point set, and calculate whether there is a point in the point set R within the range of radius δ3 (usually taken as twice the point cloud resolution) for each point c i in the point cloud C. If found, record the index in the index set I3. The points in R that are indexed by I3 are the boundary point set of the damaged blade cleaning area, denoted as B;

[0115] S352: Perform inverse transformation on the three sides l1, l2, l3 of the trapezoid according to the rotation matrix M to obtain the planes l 1r , l 2r , l 3r . Construct filling points at a certain interval on the three planes. After construction, extract the points within the boundary B as the final cross-section filling point set A, and combine A and C as the final post-cleaned point cloud model.

[0116] The entire cleaning process is as follows Figure 6 shown. On the left side in the figure is the damaged scanned point cloud data, on the right side is the target model after cleaning, and the removed area during cleaning is shown in the middle.

[0117] The present invention also provides a self - adaptive cleaning system for damaged blades. The system includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it executes the self - adaptive cleaning method for damaged blades as described above.

[0118] The present invention also provides a repair method for aero - engine blades. The repair method first obtains a cleaned model of the mechanical target area of the aero - engine blade by using the self - adaptive cleaning method for the damaged blade point cloud as described above, and then performs laser cladding and material removal based on the cleaned model.

[0119] The present invention also provides a computer - readable storage medium. The computer - readable storage medium stores machine - executable instructions. When the machine - executable instructions are called and executed by a processor, the machine - executable instructions cause the processor to implement the self - adaptive cleaning method for damaged blades as described above.

[0120] Those skilled in the art can easily understand that the above - mentioned are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive cleaning method for damaged blades, characterized in that, It includes the following steps: S1: By comparing the scanned point cloud of the damaged blade with the sampled point cloud of the blade design model, calculate the distances from each to the nearest point of the other respectively and set thresholds to determine the damaged area and the area to be cladded; S2: Based on the region growing method, perform feature segmentation on the area to be cladded obtained in step S1, and divide the area to be cladded into the suction surface, pressure surface, and blade edge; S3: Design the shape of the bounding box based on the damaged area identified in step S1 and the feature surfaces segmented in step S2, obtain the cleaning model of the target area, and fill the holes in the cleaning section.

2. The adaptive cleaning method for damaged blades according to claim 1, wherein Step S1 specifically includes the following steps: S11: Use a three-dimensional scanning method to obtain the actual scanned point set P of the damaged blade and the designed point set Q obtained by point cloud sampling of the blade design model, register the actual scanned point set P and the designed point set Q, and transform them into the same coordinate system; S12: Search for the nearest point q from the actual scanned point set P to the designed point set Q and the nearest point p from the designed point set Q to the actual scanned point set P in the coordinate system of step S11, and calculate the Euclidean distances between them, denoted as D and D respectively: i j i j pi qi ​​​​​​ D pi = min(dist(p i , q j )) D qi = min(dist(q i , p j )) S13: Set a screening threshold based on step S12 to obtain the point clouds of the damaged area and the area to be cladded.

3. The self - adaptive cleaning method for damaged blades according to claim 2, wherein, In step S13, twice the resolution 2*re is used as the screening threshold. If D pi > 2*re, then p i is stored in the damaged area point set E; if D qi > 2*re, then q i is stored in the cladding area point set W to be processed.

4. The self - adaptive cleaning method for damaged blades according to claim 3, wherein Step S2 specifically includes the following steps: S21: Assign an initial value of l = 0 to the point region labels of all points in the point set W of the to-be-clad region obtained in step S13, and set the normal vector difference threshold θ th and the curvature threshold c th ; S22: Arrange the curvature ascending sequence {C} of the points with the constructed region label l = 0 in step S21, and use the point with the minimum curvature as the initial seed point S[0]; S23: Pop the point S[0] from the seed point queue {S} in step S21, and traverse the neighborhood I k [S[0]], where k is the number of neighborhood points, and calculate the angle θ between the normal vectors of the neighborhood points and the seed point c and the curvature c of this neighborhood point c ; S24: If θ c ≤ θ th , add the current point to the growth region, set the corresponding region label for this neighborhood point, and then continue the comparison. If c c < c th , add the current point to the seed point set {S}; if θ c > ht θ, skip this point; S25: Repeat steps S23 and S24. When the final seed point set {S} is emptied, it represents the completion of one region growth; S26: Continue to construct the curvature ascending sequence for the remaining points with region label l = 0, and repeat steps S22, S23, and S24 until the region labels l ≠ 0 for all points, then the region growth is completed, and the three regions respectively correspond to the suction surface, blade edge, and pressure surface of the area to be cladded.

5. The self - adaptive cleaning method for damaged blades according to claim 1, wherein Step S3 specifically includes the following steps: S31: Perform principal component analysis on the suction surface or pressure surface point cloud after segmentation in step S2, Take it as the first principal direction and Take it as the second principal direction, and use As the projection direction of the point set of the area to be cladded; S32: Identify the blade edge points based on the K-Means algorithm to obtain the blade edge direction; S33: Adjust the point cloud position of the area to be cladded according to the projection direction obtained in step S31 and the leaf edge direction obtained in step S32 to obtain the point cloud E after rotation and projection of the damaged area point set E rp , the point cloud D after rotation and projection of the damaged blade scan point set D rp ; S34: Based on the damaged area obtained in step S33, construct a cleaning bounding box for local repair, and generate the cleaning model of the target area accordingly; S35: For the parameters of the cleaning bounding box in step S34, fill the cross-section holes after cleaning.

6. The adaptive cleaning method for damaged blades according to claim 5, characterized in that, Step S34 specifically includes the following steps: S341: Calculate point cloud E rp , D rp centroid G e (x e , z e ), G d (x d , z d ), calculate E rp maximum and minimum values of x of E on the X and Z axes min , x max , z min , z max and rp resolution re of the point cloud E; S342: If x e ≤x d , let l1: x + z - x max -z max -a = 0, l2: x - z - x max +z min +b = 0, l3: x = x min If x e > x d Let l1: x - z - x min + z max - a = 0, l2: x + z - x min - z min + b = 0, l3: x = x max ; S343: Calculate point cloud E rp The set {S1} of the distances of each point from l1, point cloud E rp The set {S2} of the distances of each point from l2, sort the sets {S1} and {S2} in ascending order; S344: Compare the first item S1[0] in the set {S1} with δ2. If S1[0] ≤ δ2, update a = a + re, and return to step S343; Compare the first item S2[0] in the set {S2} with δ2. If S2[0] ≤ δ2, update b = b + re, and return to step S343; S345: Expand the final l1, l2, and l3 outward by the cleaning depth d. The straight lines l1, l2, and l3 enclose a trapezoidal closed area, and calculate the point cloud D rp For the points located within this closed area, record their indices as I1, D rp For the points located outside this closed area, the indices are I2; Extract the points in I1 and I2 from the scanned point set D of the damaged blade, record the removed point cloud as R, and the cleaned point cloud as C 7. The self-adaptive cleaning method for damaged blades according to claim 5, wherein For damage to the blade body and blade corner, step S34 specifically includes the following steps: S341’: Point cloud E by principal component analysis rp The first principal direction Calculate point cloud E rp Centroid G e (x e , z e ), through the principal direction And centroid G e Determine line l g : Ax + By + Cz + D = 0; S342’: Straight line l g Divide the points in the point cloud E rp into two sides, and retain the point set on the same side as the leaf root point, denoted as H. Calculate the Euclidean distance dist(h i to l g ), and obtain the maximum Euclidean distance dist i (h g , l max (h i , l g ). Offset l g towards the leaf root side by a distance of dist max (h i , l g ); S343’: Straight line l g Divide the two-dimensional point cloud D rp into a cleaning area and a post-cleaning area, denote the excised point cloud as R, and the post-cleaned point cloud as C'.

8. An adaptive cleaning device for the point cloud of a damaged blade according to claim 6 or 7, characterized in that Step S35 specifically includes the following steps: S351: Take R as the search point set, and calculate each point c in the point cloud C i to check if there is a point in the point set R within the range of the radius. If found, record the index in the index set I3. Extract the points in the index set I3 from R, which are the boundary point set of the damaged blade cleaning area, denoted as B; S352: The three sides l1, l2, and l3 of the trapezoid are inversely transformed according to the rotation matrix M to obtain the plane l 1r , l 2r , l 3r . Fill-in points are constructed at certain intervals on the three planes. After construction, the points within the boundary B are extracted as the final cross-section fill-in point set A. A and C are combined to form the final cleaned point cloud model.

9. An adaptive cleaning system for damaged blades, characterized in that, It includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it executes the damaged blade adaptive cleaning method according to any one of claims 1-8.

10. A method for repairing aeroengine blades, characterized in that: It includes the following steps: (1) Use the adaptive cleaning method for the damaged blade point cloud according to any one of claims 1-8 to obtain the cleaning model of the mechanical target area of the aeroengine blade, (2) Perform laser cladding and material removal based on the cleaning model of the mechanical target area.

Images