Benchmark weight driven decision-making method and device for reconstructing bending and torsional deformation blade model

Abstract

The present invention relates to a method and device for reconstructing a model of a bent and twisted deformed blade with a reference weight driven decision. The method includes acquiring point cloud data of a deformed bent and twisted blade, thereby determining the blade damage and deformation area through the point cloud data, obtaining at least one locally deformed area to be repaired, and also determining at least one non-repaired area on the blade that does not need to be repaired. Manual detection of the blade damage and deformation area is not required, thereby improving the detection efficiency and reducing the time cost; further, at least one non-repaired area can be processed to obtain a cross-sectional point cloud set for each area to be repaired, and a reference point cloud set is obtained based on all cross-sectional point cloud sets; then, at least one cross section to be repaired is reconstructed based on the reference point cloud set. When the reconstruction of all cross sections to be repaired is completed, a repaired blade surface is obtained. The repaired blade surface is a high-precision reference model, thereby allowing the blade to be put into service and reducing economic costs.

Description

Technical Field

[0001] The present invention relates to the technical field of blade model reconstruction, and in particular to a method and device for reconstructing a bending and torsional deformation blade model using a reference weight driven decision-making method. Background Art

[0002] A reliable and high-precision blade reference surface is the key to ensuring the quality of robotic grinding and polishing. However, due to the complexity of the blade profile, the control of the robotic machining process is extremely difficult, which in turn leads to the inability to ensure the surface quality of the blade. In particular, the difficulty of blade machining in the field of robotic machining has greatly limited the promotion and application of robotic machining technology.

[0003] In existing technologies, repairing damaged blades requires additive manufacturing (laser cladding) and then subtractive machining of the cladding area. Both steps require a high-precision model as a reference. However, the blades will deform and break during service, which will cause the original model to lose its reference function.

[0004] Therefore, it is urgent to propose a method and device for reconstructing a bending and torsional deformation blade model with benchmark weight-driven decision-making to solve the technical problem in the existing technology that blades will deform and break during service, causing the original model to lose its benchmark function, thereby resulting in high time and economic costs for blade repair. Summary of the Invention

[0005] In view of this, it is necessary to provide a method and device for reconstructing a bending and torsional deformation blade model with a benchmark weight-driven decision-making, so as to solve the technical problem in the existing technology that the blades will be deformed and damaged during service, causing the original CAD model to lose its benchmark function, thereby resulting in high time and economic costs for blade repair.

[0006] In order to solve the above problems, the present invention provides a method for reconstructing a bending and torsional deformation blade model using a reference weight driven decision-making method, comprising:

[0007] Acquiring point cloud data of the deformed twisted blade, and determining at least one area to be repaired and at least one non-repaired area based on the point cloud data; the area to be repaired is an area of ​​local deformation;

[0008] Processing the at least one non-repaired area to obtain a cross-sectional point cloud set of each area to be repaired, and obtaining a reference point cloud set based on all the cross-sectional point cloud sets;

[0009] The at least one section to be repaired is reconstructed according to the reference point cloud set to obtain a repaired blade surface.

[0010] In a possible implementation, processing the at least one non-repaired area to obtain a cross-sectional point cloud set for each area to be repaired includes:

[0011] Performing plane least squares fitting based on the top cross-section point cloud of the bent and twisted blade to obtain a normal vector of the top plane;

[0012] Determining, based on the normal vector and the at least one non-repaired area, a plurality of target non-repaired areas corresponding to each area to be repaired;

[0013] The point clouds in the plurality of target non-repaired areas are processed to obtain a cross-sectional point cloud set.

[0014] In a possible implementation, processing the point clouds in the plurality of target non-repaired areas to obtain a cross-sectional point cloud set includes:

[0015] Processing the point cloud data of each of the to-be-repaired areas and the corresponding target non-repaired areas to obtain an average density of the point clouds;

[0016] Calculating all point cloud data of the plurality of target non-repaired areas according to the point cloud average density to obtain a target point cloud density;

[0017] Calculating a preset width control coefficient and the target point cloud density to obtain a cross-sectional interval;

[0018] Point cloud data in the plurality of target non-repairing areas are intercepted according to the section interval to obtain a section point cloud set.

[0019] In a possible implementation, processing the point cloud data of each to-be-repaired area and the corresponding target non-repaired areas to obtain an average point cloud density includes:

[0020] Setting a bandwidth of each area to be repaired, and determining at least one cutting plane according to the bandwidth;

[0021] Triangulate the point cloud data in each target non-repaired area to obtain triangular patches;

[0022] Obtaining an initial intersection point set of each tangent plane according to the intersection points of each edge of the triangle and the at least one tangent plane;

[0023] All point cloud data in the initial intersection point set are calculated to obtain the initial point cloud density of each tangent plane, and the point cloud average density of all initial point cloud densities is calculated.

[0024] In a possible implementation, obtaining a reference point cloud set based on all cross-section point cloud sets includes:

[0025] Refining the at least one to-be-repaired region and the at least one non-repaired region according to the normal vector to obtain at least one refined to-be-repaired region and at least one refined non-repaired region;

[0026] A reference point cloud set is obtained according to the cross-sectional point cloud set of the at least one refined area to be repaired and the at least one refined non-repaired area.

[0027] In a possible implementation, thinning the at least one region to be repaired according to the normal vector to obtain the at least one refined region to be repaired includes:

[0028] Using the direction of the normal vector as the normal vector of the section plane, each area to be repaired and the previous area to be repaired are calculated to obtain a torsion angle;

[0029] According to the torsion angle, a torsion angle change rate is obtained;

[0030] updating the section interval according to the torsion angle and the torsion angle change rate to obtain a target section interval;

[0031] Each of the regions to be repaired is refined according to the target cross-sectional interval to obtain a refined region to be repaired.

[0032] In a possible implementation, reconstructing the at least one section to be repaired based on the reference point cloud set to obtain a repaired blade surface includes:

[0033] Fitting the point cloud data of each cross-section point cloud set in the reference point cloud set to obtain a reference cross-section curve set;

[0034] Reconstructing each section to be repaired according to the curves in the reference section curve set to obtain a repaired section;

[0035] All repaired sections and the at least one non-repaired section are processed based on multi-section lofting to obtain a repaired blade curved surface.

[0036] In a possible implementation, reconstructing each section to be repaired according to the curves in the reference section curve set to obtain a repaired section includes:

[0037] Set the number of iterations;

[0038] Calculating each curve in the reference section curve set according to the current section to be repaired to obtain a time weight and a spatial distance weight corresponding to each curve;

[0039] Determine a reconstructed curve according to the time weight and the spatial distance weight of all curves;

[0040] Reconstructing the damaged portion corresponding to the current section to be repaired according to the reconstruction curve to obtain a repaired section;

[0041] Determining whether the number of iterations reaches a preset number of iterations or determining whether the repaired section is completely repaired;

[0042] If not, the repaired section is reconstructed according to the curves in the reference section curve set.

[0043] In a possible implementation, determining the reconstructed curve according to the time weight and the spatial distance weight of all curves includes:

[0044] Determining a target weight corresponding to each curve according to the time weight and the spatial distance weight of each curve;

[0045] The curve corresponding to the maximum value among all target weights is determined as the initial curve;

[0046] The initial curve is optimized based on the curvature weighted VMM algorithm to obtain a reconstructed curve.

[0047] On the other hand, the present invention also provides a device for reconstructing a bending and torsional deformation blade model using a reference weight driven decision-making method, comprising:

[0048] An area determination module is configured to obtain point cloud data of the deformed, twisted blade and determine, based on the point cloud data, at least one area to be repaired and at least one non-repaired area; the area to be repaired is a locally deformed area;

[0049] a set determination module, configured to process the at least one non-repaired area to obtain a cross-sectional point cloud set of each area to be repaired, and obtain a reference point cloud set based on all cross-sectional point cloud sets;

[0050] The surface reconstruction module is used to reconstruct the at least one section to be repaired according to the reference point cloud set to obtain a repaired blade surface.

[0051] The beneficial effect of the present invention is to obtain point cloud data of deformed bent and twisted blades, so that the damage and deformation areas of the blades can be determined through the point cloud data, wherein there may be more than one deformed area, so at least one locally deformed area to be repaired can be obtained, and at the same time, the area on the blade that does not need to be repaired, that is, at least one non-repair area, can be determined, so that there is no need to manually detect the damage and deformation areas of the blades, which improves the efficiency of detection and reduces time costs; further, at least one non-repair area can be processed to obtain a cross-sectional point cloud set of each area to be repaired, and then a reference point cloud set can be obtained based on all cross-sectional point cloud sets; then at least one cross-section to be repaired can be reconstructed based on the reference point cloud set. When the reconstruction of all cross-sections to be repaired is completed, the repaired blade surface can be obtained. The repaired blade surface is a high-precision reference model, so that the blade can be put into service, reducing economic costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic flow chart of an embodiment of a method for reconstructing a bending and torsional deformation blade model using a reference weight driven decision-making method provided by the present invention;

[0053] Figure 2 For the present invention Figure 1 A schematic flow chart of an embodiment of step S102;

[0054] Figure 3 A schematic diagram of the structure of an embodiment of similar points of a point set provided by the present invention;

[0055] Figure 4 For the present invention Figure 2 A schematic flow chart of an embodiment of step S203;

[0056] Figure 5 A schematic diagram of a process flow of an embodiment of the present invention for refining the area to be repaired;

[0057] Figure 6 For the present invention Figure 1 A schematic flow chart of an embodiment of step S103;

[0058] Figure 7 For the present invention Figure 6 A schematic flow chart of an embodiment of step S602;

[0059] Figure 8 A schematic structural diagram of an embodiment of a device for reconstructing a bending and torsional deformation blade model using a reference weight driven decision-making method provided by the present invention;

[0060] Figure 9 This is a schematic structural diagram of an embodiment of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0061] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0062] Cubic Spline Interpolation (also known as Spline Interpolation) is a process of obtaining a set of curve functions by solving a set of three bending moment equations through a smooth curve with a series of shape value points.

[0063] In actual calculations, boundary conditions are required to complete the calculation. The definition of non-kinked boundaries is not explained in general calculation method books, but numerical calculation software such as Matlab uses non-kinked boundary conditions as the default boundary conditions.

[0064] like Figure 1 As shown, a specific embodiment of the present invention discloses a method for reconstructing a bending and torsional deformation blade model using a reference weight driven decision, comprising:

[0065] S101, obtaining point cloud data of a deformed, twisted blade, and determining at least one area to be repaired and at least one non-repaired area based on the point cloud data; the area to be repaired is an area with local deformation;

[0066] S102, processing at least one non-repaired area to obtain a cross-sectional point cloud set of each area to be repaired, and obtaining a reference point cloud set based on all cross-sectional point cloud sets;

[0067] S103 : Reconstruct at least one section to be repaired according to the reference point cloud set to obtain a repaired blade surface.

[0068] The embodiments of the present invention can be applied to blades of aircraft engines, blades of other robots, etc., and can be used in any field where blades need to be repaired. The deformed, twisted blade can be scanned by a laser radar to obtain point cloud data of the twisted blade. Then, the non-repaired area and the area to be repaired of the twisted blade can be quickly and accurately located using a Euclidean clustering algorithm with an adaptive threshold, thereby obtaining at least one non-repaired area and at least one area to be repaired. To improve accuracy, a large deformed area can be divided into multiple areas to be repaired, and an area without a deformed portion can be divided into multiple non-repaired areas. At least one area to be repaired can be processed using at least one non-repaired area to obtain a cross-sectional point cloud set for each area to be repaired, and then a reference point cloud set including all cross-sectional point cloud sets can be obtained. The twisted and deformed parts of at least one cross-sectional area to be repaired can be reconstructed one by one using the reference point cloud set. When all cross-sectional areas to be repaired are reconstructed, the blade surface can be obtained.

[0069] Compared with the prior art, the present embodiment provides point cloud data for obtaining deformed bent and twisted blades, so that the blade damage and deformation areas can be determined through the point cloud data, wherein there may be more than one deformed area, so at least one locally deformed area to be repaired can be obtained, and at the same time, an area on the blade that does not need to be repaired, that is, at least one non-repaired area, can be determined, so that there is no need to manually detect the blade damage and deformation areas, which improves the efficiency of detection and reduces the time cost; further, at least one non-repaired area can be processed to obtain a cross-sectional point cloud set for each area to be repaired, and then a reference point cloud set can be obtained based on all cross-sectional point cloud sets; then, at least one cross-section to be repaired can be reconstructed based on the reference point cloud set. When the reconstruction of all cross-sections to be repaired is completed, the repaired blade surface can be obtained. The repaired blade surface is a high-precision reference model, which can put the blade into service and reduce the economic cost.

[0070] In some embodiments of the present invention, Figure 2 As shown, step S102 includes:

[0071] S201, performing plane least squares fitting based on the top cross-section point cloud of the bent and twisted blade to obtain a normal vector of the top plane;

[0072] S202, determining a plurality of target non-repairing regions corresponding to each to-be-repaired region based on the normal vector and at least one non-repairing region;

[0073] S203 : Processing point clouds in multiple target non-repaired areas to obtain a cross-sectional point cloud set.

[0074] In a specific embodiment of the present invention, when determining each area to be repaired, an edge extraction algorithm based on normal information can be used to clarify the boundary of the repair area, and a plane least squares fitting is performed based on the blade tip cross-section point cloud to obtain the normal vector F of the top plane. u , the direction of this vector will be used as the normal vector of the cutting plane. At the same time, based on the edge points of the extracted repair area boundary, determine the u The position of the direction can thus intercept multiple target non-repair areas corresponding to each area to be repaired in at least one non-repair area, and use the point cloud in the area as the reference cross-section point cloud. The edge point extraction process of the repair area boundary can be as follows: (1) First set the point set { O p0}, randomly select a point p0 , add the point to the point set, and based on the established kdtree, obtain p0 of k Neighboring points.

[0075] (2) Then set the Euclidean clustering threshold l dis , and calculate p0 With each neighbor pi Euclidean distance d i po ,when d i po ≤ l dis When , the neighboring point pi and p0 As a class, add to the point set { O p0},like Figure 3 (a) shows the hollow point p1 , p2 and p0 For similar points.

[0076] (3) For { O p0 Repeat steps 1 and 2 for the newly added points in}, such as Figure 3 (b), the operation continues until { O p0}When no new points are added, exit step 3.

[0077] (4) If there are still unclassified points in the leaf point cloud, repeat steps 1, 2, and 3 until the classification is completed.

[0078] Finally, we obtain the i A collection of repair area point clouds (i.e. cladding area point clouds) { O ire}, and the non-repaired area point cloud set {O ia}.

[0079] The edge information of the repair area will determine the parameter settings of the subsequent section method and the starting point position of the final path planning. The present invention adopts an edge extraction algorithm based on normal information. Due to the irregular point cloud of the cladding area, the point cloud set of the non-repair area is { O ia} to extract edges. In order to improve efficiency, based on { O ia}Build Kdtree and use PCA (Principal Component Analysis) algorithm to calculate { O ia}The normal information of each point in the point set and the normal information of each point in the point set pf At the same time, calculate the direction vector connecting each neighbor point and pf, and project the vector to pf The maximum angle value is obtained by calculating the angle between adjacent vectors. i max , then set the threshold l θmx ,if i max < l θmx , then the point is an interior point; otherwise, the point is an edge point.

[0080] In some embodiments of the present invention, Figure 4 As shown, step S203 includes:

[0081] S401, processing the point cloud data of each area to be repaired and the corresponding multiple target non-repaired areas to obtain the average density of the point cloud;

[0082] S402, calculating all point cloud data of multiple target non-repair areas according to the average point cloud density to obtain the target point cloud density;

[0083] S403, calculating the preset width control coefficient and the target point cloud density to obtain the cross-section interval;

[0084] S404 , intercepting point cloud data in multiple target non-repairing areas according to the section interval to obtain a section point cloud set.

[0085] In some embodiments of the present invention, step S401 includes:

[0086] Setting a bandwidth for each area to be repaired, and determining at least one cutting plane based on the bandwidth;

[0087] Triangulate the point cloud data in each target non-repaired area to obtain triangular patches;

[0088] According to the intersection points of each edge of the triangle and at least one tangent plane, an initial intersection point set of each tangent plane is obtained;

[0089] Calculate all point cloud data in the initial intersection point set to obtain the initial point cloud density of each tangent plane, and calculate the point cloud average density of all initial point cloud densities.

[0090] In a specific embodiment of the present invention, the point cloud data in each target non-repaired area can be triangulated to obtain triangular patches, and then the point cloud data in each target non-repaired area can be triangulated to obtain triangular patches. The specific steps can be: setting the number of cutting plane sets to N, and the value of N needs to be based on the size of the blade and the width of the interceptable area. Since the point cloud collected by the visual equipment is messy, there will not be many points in the same plane. In order to ensure that the points on the cutting plane contain the complete edge information of the blade, it is necessary to set a certain bandwidth, including an upper bandwidth and a lower bandwidth. A section P1 can be given and two equidistant planes P2 and P3 can be generated in the normal direction of the section P1, wherein the section P1 is the section of any one of the areas to be repaired, and the sections P2 and P3 are the sections of multiple target non-repaired areas determined according to the section P1. The point clouds in the P2 and P3 areas are intercepted and projected onto the section P1. However, direct projection of the blade contour points may result in errors. Therefore, the embodiment of the present invention can intercept and triangulate the point cloud in the P2 and P3 regions based on the STL file storage principle, and then determine at least one tangent plane based on the upper and lower bandwidths. The intersection of each edge of the triangle and the tangent plane P1 is then calculated. Assuming the direction vector of the tangent plane P1 is n = (a, b, c), the equation of P1 can be expressed as shown in formula (1):

[0091] (1)

[0092] In the process of triangulating the intercepted point cloud, in order to improve the accuracy of corresponding point pairs, the embodiment of the present invention adopts a bidirectional Kdtree to search for the nearest corresponding point pairs in the "upper bandwidth" (P1 and P2 areas) point cloud and the "lower bandwidth" (P1 and P3 areas) point cloud; first, set the index of any point in the "upper bandwidth" point cloud to m, use Kdtree to search for its nearest point in the "lower bandwidth" point cloud and record its index as down[0], for the point with the index down[0] in the "lower bandwidth" point cloud, use Kdtree to search for its nearest point in the "upper bandwidth" point cloud and record its index as up[0], if formula (2) is satisfied, the corresponding point pair is determined to be correct. If formula (3) is satisfied, the corresponding point pair is eliminated. Finally, the upper and lower area point clouds are traversed to obtain the final correct correspondence relationship of the point pair. Formula (2) and formula (3) are as follows:

[0093] (2)

[0094] (3)

[0095] Where, m is the number of sections, and up[0] is the index of the nearest point to be searched.

[0096] After obtaining the corresponding point pairs in the upper and lower “bandwidths”, let the coordinates of the two points of the corresponding point pairs be M(x1, y1, z1) and N(x2, y2, z2) respectively. The intersection of the spatial line formed by MN and the tangent plane P1 is the desired section point. The standard equation of the spatial line is shown in Formula (4) and Formula (5):

[0097] (4)

[0098] (5)

[0099] The point cloud density will affect the number of intercepted point clouds and the patch error, so the section width H is adjusted adaptively based on the point cloud density. pi , which is calculated as shown in formula (6):

[0100] (6)

[0101] Where, k is the triangle projection coefficient, the appropriate k The value ensures that the complete blade cross section is obtained with high efficiency and accuracy; p The average density of point clouds in non-repaired areas of multiple targets is obtained by setting a fixed interception width. H n , obtain the point cloud set intercepted in the non-repaired area of ​​the target { O n},set up n for{ O n}Point cloud number, p i for{ O n}No. i The initial point cloud density of points is the average point cloud density of the point cloud set. p The calculation method is shown in formulas 7, 8 and 9:

[0102] (7)

[0103] (8)

[0104] (9)

[0105] Where, Dis sum The Euclidean distance between any point and each neighboring point X i of and, N m is the number of all neighboring points of the selected point, then it can be calculated by formula (9) Dis sum , and then calculate the initial point cloud density of each target non-repaired area using formula (8) p i , and then calculate the average density of the point cloud using formula (7) p .

[0106] Furthermore, after obtaining the average point cloud density, all point cloud data of multiple target non-repaired areas can be calculated using formula (7) to obtain the target point cloud density of multiple target non-repaired areas. r a , and set the preset width control coefficient W The specific preset width control coefficient can be set according to the actual situation, and the embodiment of the present invention is not limited here. The cross-sectional spacing of adjacent cutting planes is H The calculation of is shown in formula (10):

[0107] (10)

[0108] Finally, combined with the section interval, the above method is used to intercept the point cloud of the non-repaired area to obtain the section point cloud set.

[0109] In some embodiments of the present invention, step S102 includes:

[0110] Refining the at least one to-be-repaired region and the at least one non-repaired region according to the normal vector to obtain at least one refined to-be-repaired region and at least one refined non-repaired region;

[0111] A reference point cloud set is obtained according to a cross-sectional point cloud set of at least one refined area to be repaired and at least one refined non-repaired area.

[0112] In a specific embodiment of the present invention, after obtaining the cross-section point cloud set of each area to be repaired, the reference point cloud set { Q p1 ,Q p2 , Q pi , Q pN}, Q pi Indicates the p i In order to improve the accuracy of the interception, it is necessary to refine at least one area to be repaired and at least one non-repaired area.

[0113] In some embodiments of the present invention, Figure 5 As shown, at least one area to be repaired is refined according to the normal vector to obtain at least one refined area to be repaired, including:

[0114] S501, using the direction of the normal vector as the normal vector of the section plane, calculating each area to be repaired and the previous area to be repaired to obtain a torsion angle;

[0115] S502, obtaining a torsion angle change rate according to the torsion angle;

[0116] S503, updating the section interval according to the torsion angle and the torsion angle change rate to obtain a target section interval;

[0117] S504 , thinning each area to be repaired according to the target cross-section interval to obtain a thinned area to be repaired.

[0118] In a specific embodiment of the present invention, the direction of the normal vector can be used as the normal vector of the cutting plane, and the torsion angle between each section (i.e., each area to be repaired) and the previous section (i.e., the previous area to be repaired) in the direction of the normal vector can be calculated from the top section of the blade as the starting point and downward along the normal vector. i , and calculate the rate of change of the torsion angle of adjacent sections, as shown in formula (11):

[0119] (11)

[0120] Where, e i is the rate of change of the torsion angle, i i is the torsion angle of each section, i i+1 is the torsion angle of the previous section.

[0121] The torsion angle change rate threshold can be set e k and torsion angle threshold i k , if the rate of change of the torsion angle ei Greater than e k , then the corresponding area to be repaired is refined, and the section interval H is set to H / 2 to obtain the target section interval; if the torsion angle change rate yes Less than ok , then compare the torsion angle i i and torsion angle threshold i k The size of i i Greater than i k And the normal spacing between sections d i Greater than the section normal spacing threshold d max , similarly refine the section interval to obtain the target section interval, and then refine the area to be repaired based on the target section interval, and update the number of sections after refinement to m The refinement process of the non-repaired area is the same as the refinement process of the area to be repaired, and the embodiment of the present invention will not be repeated here. Then, at least one refined area to be repaired and at least one refined non-repaired area can be obtained, so that a reference point cloud set can be obtained based on the cross-sectional point cloud set of at least one refined area to be repaired and at least one refined non-repaired area.

[0122] In some embodiments of the present invention, Figure 6 As shown, step S103 includes:

[0123] S601, fitting the point cloud data of each cross-section point cloud set in the reference point cloud set to obtain a reference cross-section curve set;

[0124] S602, reconstructing each section to be repaired according to the curves in the reference section curve set to obtain a repaired section;

[0125] S603 : Process all repaired sections and at least one non-repaired section based on multi-section lofting to obtain a repaired blade surface.

[0126] In a specific embodiment of the present invention, the number of sections of the repaired area and the non-repaired area involved in the reconstruction after refinement can be determined: based on the acquired reference point cloud set, the point cloud data of each section point cloud set in the reference point cloud set is fitted using cubic spline interpolation to obtain the reference section curve set. C m ( m is the number of cutting planes).

[0127] In some embodiments of the present invention, Figure 7 As shown, step S602 includes:

[0128] S701, setting the number of iterations;

[0129] S702: Calculate each curve in the reference section curve set according to the current section to be repaired to obtain a time weight and a spatial distance weight corresponding to each curve;

[0130] S703: Determine a reconstructed curve based on the time weights and spatial distance weights of all curves;

[0131] S704, reconstructing the corresponding damaged portion of the current section to be repaired according to the reconstruction curve to obtain a repaired section;

[0132] S705: Determine whether the number of iterations reaches a preset number of iterations or whether the repaired section is completely repaired;

[0133] S706: If not, reconstruct the repaired section according to the curves in the reference section curve set.

[0134] In some embodiments of the present invention, step S703 includes:

[0135] Determine the target weight corresponding to each curve based on the time weight and spatial distance weight of each curve;

[0136] The curve corresponding to the maximum value among all target weights is determined as the initial curve;

[0137] The VMM algorithm based on curvature weight is used to optimize the initial curve to obtain the reconstructed curve.

[0138] In a specific embodiment of the present invention, during surface reconstruction using a reference cross-section curve set, as the longitudinal lofting length increases, the lofting accuracy decreases. This is defined as the spatial distance influencing factor on surface reconstruction accuracy. In the iterative method used by the present embodiment to reconstruct the repaired region's cross-section using curves from the reference cross-section curve set, the curves added to the reference curve set later after fitting and matching contain greater errors. The reference curve set is the curve reconstructed after a successful match, i.e., the set of stored reconstructed curves. This is defined as the temporal influencing factor on surface reconstruction accuracy. However, for blades with a high degree of twist, the accuracy of the surface reconstructed by cross-section lofting decreases as the twist angle between blade sections increases. However, the impact of the increasing twist angle between sections on error during blade cross-section curve acquisition has not been studied. The present embodiment proposes a new benchmark weight calculation method, specifically: setting a cross-section weight evaluation function to select an appropriate benchmark. This metric must take into account the spatial distance, time, and twist angle influencing factors on surface reconstruction accuracy. Since the benchmark curve set will inevitably introduce errors during the iterative update process, the number of iterations and the preset iteration threshold can be set. The earlier the iteration is used as the benchmark curve, the greater the confidence level. Therefore, the time weight factor is set. T i To evaluate the influence weight of the baseline curve on subsequent reconstruction, the calculation is shown in method (12):

[0139] (12)

[0140] Where, T i is the time weight factor, i For the i Matching curves.

[0141] In terms of the factors affecting spatial distance and torsion angle, the smaller the spatial distance between the reference curve and the repair area section at the top plane normal vector, the smaller the torsion angle, and the greater the influence weight on the reconstruction of the repair area section. Therefore, the spatial distance weight factor is set. d cr and i c , and the evaluation function of the repair area section weight is obtained as shown in formula (13):

[0142] (13)

[0143] Where, N d is the number of reference section curves in the current iteration order; d cr The repair area to be evaluated is the cross section to c rThe distance between the planes where the reference section curves are located; i cr The cross section of the repair area to be evaluated is c r The torsion angle of the reference section curve about the normal vector of the top plane.

[0144] The maximum value can be determined from all spatial distance weights of all curves, and the curve corresponding to the maximum value can be determined as the reconstructed curve. The reconstructed curve is then matched with the repair area curve initially fitted in the repair area, thereby realizing the reconstruction of the repair area section curve, and the reconstructed curve is added to the baseline curve set, and its iteration number is marked. Then, it can be determined whether the number of iterations reaches the preset iteration threshold or whether the repaired section is repaired. If the preset iteration threshold has been reached or the repair has been completed, then this part of the process is terminated and subsequent steps are performed. If the preset iteration threshold has not been reached or the repair has not been completed, then the above process is repeated until all sections of the repair area are reconstructed.

[0145] In a specific embodiment of the present invention, the point cloud of the blade repair area will be partially missing when it is matched with the cross-sectional curve after removing the margin, and the density of the blade point cloud obtained by scanning is also uneven. When the conventional ICP matching algorithm is applied to the situation of missing point cloud and uneven density, it is easy to cause matching tilt and local optimality. Because in the matching situation with Gaussian noise, missing point cloud and uneven density, the VMM matching algorithm has the characteristics of high accuracy and good algorithm stability. At the same time, the design feature of the blade is that the curvature of the leading and trailing edges is large, and the curvature of other parts is small and flat. After calculating all the spatial distance weights, the curve corresponding to the maximum value can be determined as the initial curve. After that, the bidirectional kdtree can be used to obtain the corresponding points, and the VMM algorithm based on curvature weight can be used to optimize the matching accuracy of the initial curve. The matching formula is shown in formula (14):

[0146] (14)

[0147] Where, d TDXwi For the i The distance from the target point to the curve is the product of the distance between the target point and the adjacent point on the curve in the normal direction of the adjacent point on the curve and the weight of the source point. TDXw For a cross section, all the actual points of the cross section (the total number is M c ) and the weighted average distance between the adjacent points on the curve, which is calculated as shown in formulas (15), (16) and (17):

[0148] (15)

[0149] (16)

[0150] (17)

[0151] Where, TDX Expressed as above TDXw , that is, the weighted average distance; w i is the weight factor representing the importance of the corresponding point to the objective function; P ssi For the i Actual points of the cross section; P ci For the i The point with the smallest distance between the actual point of the cross section and the discrete point of the curve in the normal direction of the discrete point of the curve (referred to as the curve adjacent point in this embodiment of the present invention); or ci Then it is with the i The normal direction vector of the adjacent point of the curve corresponding to the actual point of the cross section; w i The calculation of is shown in formulas (18)-(23):

[0152] (18)

[0153] (19)

[0154] (20)

[0155] (twenty one)

[0156] (twenty two)

[0157] (twenty three)

[0158] Where, CV i Represents the source point cloud point. There are two representative peaks in the curvature distribution diagram of the blade cross-section point cloud. The two peaks represent the leading and trailing edge points of the blade. Therefore d i1 The first point in the corresponding pair i From a source point cloud point to the curvature front peak point CV m1 distance, d i2 The first point in the corresponding pair i From a source point cloud point to the curvature front peak point CV m2 distance; wi1 Represents the corresponding point i From a source point cloud point to the curvature front peak point CV m1 The weight factor of w i2 Indicates the first i From a source point cloud point to the curvature front peak point CV m2 The weight factor of w i ’ is a temporary variable of weight based on curvature, s 1 and s 2 is the standard deviation parameter that controls the speed at which the weight decreases; α 1 and α 2 is the weight adjustment coefficient, which can be set according to actual conditions; m p Indicates the number of source point cloud points.

[0159] The embodiment of the present invention is based on the Gaussian formula. The absolute value of the difference between each point and the two curvature peaks (corresponding to the leading and trailing edges of the blade section) in the blade section curvature distribution diagram is used as an independent variable. The Gaussian distribution function is used to calculate the weight of each point. Finally, the weight is normalized so that the sum of the weights of all points is 1. This ensures that the leading and trailing edge points (high curvature points) of the blade section have higher weights during the matching process, and guarantees the accuracy of the reconstructed curve as much as possible. Since there is an error in the normal line of the actual cross-section point, the matching accuracy of formula (14) will be reduced. However, the normal line of the fitted curve has high overall accuracy and is smooth because it has been smoothed. The embodiment of the present invention sets the source point cloud as the discrete point of the cross-section curve and the target point cloud as the actual point of the cross-section. After matching, the curve posture conversion formula is:

[0160] (twenty four)

[0161] Where: P c 、 P c * They are the positions of a certain point on the curve before and after matching.

[0162] Based on the above matching method, the reference section with the highest weight is matched with the repair area section point cloud of the initial fitting of the repair area, so as to realize the reconstruction of the repair area section curve, and add the reconstructed repair area section curve to the reference curve set, mark its iteration number and set the iteration exit condition, and repeat the above steps until all the repair area sections are reconstructed.

[0163] The embodiment of the present invention uses an adaptive threshold Euclidean clustering algorithm and an edge extraction algorithm that considers normal information to extract the area of ​​the blade to be repaired. Then, based on the STL file storage principle, a method for adaptive width point cloud dimensionality reduction and interception that integrates point cloud density and "bidirectional Kdtree" is proposed to intercept and repair the blade section point cloud. After sorting, smoothing, and parameter alignment of the multi-section point cloud, a surface reconstruction method that considers the reference section weight is finally proposed, which includes a multi-section interpolation method that controls the section torsion angle, a reference weight calculation method that considers time and space factors, and a VMM matching algorithm that integrates curvature weight and "bidirectional Kdtree" to loft and reconstruct the repair surface. The present invention provides a new solution for reconstructing a highly reliable reference surface based on a local point cloud and post-processing of blade repair.

[0164] In order to better implement the method for reconstructing a bending and torsion deformed blade model based on the reference weight driven decision in the embodiment of the present invention, based on the method for reconstructing a bending and torsion deformed blade model based on the reference weight driven decision, the embodiment of the present invention also provides a device for reconstructing a bending and torsion deformed blade model based on the reference weight driven decision, such as Figure 8 As shown, the reference weight driven decision-making bending and torsional deformation blade model reconstruction device 800 includes:

[0165] The region determination module 801 is configured to obtain point cloud data of the deformed, twisted blade and determine, based on the point cloud data, at least one region to be repaired and at least one non-repaired region; the region to be repaired is a region with local deformation;

[0166] A set determination module 802 is configured to process at least one non-repaired area to obtain a cross-sectional point cloud set for each area to be repaired, and obtain a reference point cloud set based on all cross-sectional point cloud sets;

[0167] The surface reconstruction module 803 is used to reconstruct at least one section to be repaired according to the reference point cloud set to obtain a repaired blade surface.

[0168] The device 800 for reconstructing the model of a torsionally deformed blade based on the benchmark weight driven decision provided in the above embodiment can implement the technical solution described in the embodiment of the method for reconstructing the model of a torsionally deformed blade based on the benchmark weight driven decision. The specific implementation principles of the above modules or units can be found in the corresponding contents in the embodiment of the method for reconstructing the model of a torsionally deformed blade based on the benchmark weight driven decision, and will not be repeated here.

[0169] like Figure 9 As shown, the present invention also provides an electronic device 900. The electronic device 900 includes a processor 901, a memory 902 and a display 903. Figure 9Only some of the components of the electronic device 900 are shown, but it should be understood that implementation of all of the shown components is not required, and more or fewer components may be implemented instead.

[0170] In some embodiments, the memory 902 may be an internal storage unit of the electronic device 900, such as a hard disk or memory of the electronic device 900. In other embodiments, the memory 902 may also be an external storage device of the electronic device 900, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 900.

[0171] Furthermore, the memory 902 may include both an internal storage unit of the electronic device 900 and an external storage device. The memory 902 is used to store application software installed in the electronic device 900 and various data.

[0172] In some embodiments, the processor 901 can be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 902, such as the benchmark weight-driven decision-making bending and torsional deformation blade model reconstruction method in the present invention.

[0173] In some embodiments, the display 903 can be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 903 is used to display information about the electronic device 900 and to display a visual user interface. Components 901-903 of the electronic device 900 communicate with each other via a system bus.

[0174] In some embodiments of the present invention, when the processor 901 executes the benchmark weight driven decision-making bending and torsional deformation blade model reconstruction program in the memory 902, the following steps may be implemented:

[0175] Acquire point cloud data of the deformed twisted blade, and determine at least one area to be repaired and at least one non-repaired area based on the point cloud data; the area to be repaired is the area with local deformation;

[0176] Processing at least one non-repaired area to obtain a cross-sectional point cloud set of each area to be repaired, and obtaining a reference point cloud set based on all cross-sectional point cloud sets;

[0177] At least one section to be repaired is reconstructed according to the reference point cloud set to obtain a repaired blade surface.

[0178] It should be understood that, when the processor 901 executes the benchmark weight driven decision-making bending and torsional deformation blade model reconstruction program in the memory 902, in addition to the above functions, it can also implement other functions. For details, please refer to the description of the corresponding method embodiment above.

[0179] Furthermore, the embodiments of the present invention do not specifically limit the type of electronic device 900 mentioned. The electronic device 900 may be a portable electronic device such as a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, or laptop computer. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The portable electronic devices mentioned above may also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 900 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0180] Accordingly, an embodiment of the present application also provides a computer-readable storage medium, which is used to store computer-readable programs or instructions. When the program or instructions are executed by a processor, it can implement the steps or functions of the method for reconstructing the bending and torsional deformation blade model with the benchmark weight-driven decision-making provided in the above-mentioned method embodiments.

[0181] Those skilled in the art will appreciate that all or part of the process flow of the above-described method embodiment can be implemented by instructing related hardware (such as a processor, controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0182] The above is a detailed introduction to the method and device for reconstructing the bending and torsional deformation blade model based on the benchmark weight-driven decision-making provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for technical personnel in this field, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for reconstructing a bending and torsional deformation blade model based on a benchmark weight driven decision, characterized in that: include: Acquiring point cloud data of the deformed, twisted blade, and determining at least one area to be repaired and at least one non-repaired area based on the point cloud data; The area to be repaired is a locally deformed area; Processing the at least one non-repaired area to obtain a cross-sectional point cloud set of each area to be repaired, and obtaining a reference point cloud set based on all the cross-sectional point cloud sets; Reconstructing the at least one area to be repaired according to the reference point cloud set to obtain a repaired blade surface; Obtaining a normal vector of a top plane based on a point cloud of a top cross section of the twisted blade; Obtaining a cross-sectional interval based on point clouds in the at least one area to be repaired and the plurality of target non-repaired areas; The base point cloud set is obtained based on all cross-section point cloud sets, including: Refining the at least one to-be-repaired region and the at least one non-repaired region according to the normal vector to obtain at least one refined to-be-repaired region and at least one refined non-repaired region; Obtaining a reference point cloud set according to the cross-sectional point cloud set of the at least one refined area to be repaired and the at least one refined non-repaired area; Refining the at least one area to be repaired according to the normal vector to obtain at least one refined area to be repaired, comprising: Using the direction of the normal vector as the normal vector of the section plane, each area to be repaired and the previous area to be repaired are calculated to obtain a torsion angle; According to the torsion angle, a torsion angle change rate is obtained; updating the section interval according to the torsion angle and the torsion angle change rate to obtain a target section interval; Each of the regions to be repaired is refined according to the target cross-sectional interval to obtain a refined region to be repaired.

2. The method for reconstructing a bending and torsional deformation blade model based on a reference weight driven decision according to claim 1, characterized in that: The processing of the at least one non-repaired area to obtain a cross-sectional point cloud set of each area to be repaired includes: Performing plane least squares fitting based on the top cross-section point cloud of the bent and twisted blade to obtain a normal vector of the top plane; Determining, based on the normal vector and the at least one non-repaired area, a plurality of target non-repaired areas corresponding to each area to be repaired; The point clouds in the plurality of target non-repaired areas are processed to obtain a cross-sectional point cloud set.

3. The method for reconstructing a bending and torsional deformation blade model based on a reference weight driven decision according to claim 2, characterized in that: The step of processing the point clouds in the plurality of target non-repaired areas to obtain a cross-sectional point cloud set includes: Processing the point cloud data of each of the to-be-repaired areas and the corresponding target non-repaired areas to obtain an average density of the point clouds; Calculating all point cloud data of the plurality of target non-repaired areas according to the point cloud average density to obtain a target point cloud density; Calculating a preset width control coefficient and the target point cloud density to obtain a cross-sectional interval; Point cloud data in the plurality of target non-repairing areas are intercepted according to the section interval to obtain a section point cloud set.

4. The method for reconstructing a bending and torsional deformation blade model based on a reference weight driven decision according to claim 3 is characterized in that: The step of processing the point cloud data of each of the to-be-repaired areas and the corresponding target non-repaired areas to obtain an average density of the point clouds includes: Setting a bandwidth of each area to be repaired, and determining at least one cutting plane according to the bandwidth; Triangulate the point cloud data in each target non-repaired area to obtain triangular patches; Obtaining an initial intersection point set of each tangent plane according to the intersection points of each edge of the triangle and the at least one tangent plane; All point cloud data in the initial intersection point set are calculated to obtain the initial point cloud density of each tangent plane, and the point cloud average density of all initial point cloud densities is calculated.

5. The method for reconstructing a bending and torsional deformation blade model based on reference weight driven decision making according to claim 1, characterized in that: The step of reconstructing the at least one section to be repaired according to the reference point cloud set to obtain a repaired blade surface includes: Fitting the point cloud data of each cross-section point cloud set in the reference point cloud set to obtain a reference cross-section curve set; Reconstructing each section to be repaired according to the curves in the reference section curve set to obtain a repaired section; All repaired sections and the at least one non-repaired section are processed based on multi-section lofting to obtain a repaired blade curved surface.

6. The method for reconstructing a bending and torsional deformation blade model based on reference weight driven decision making according to claim 5, characterized in that: The step of reconstructing each section to be repaired according to the curves in the reference section curve set to obtain a repaired section includes: Set the number of iterations; Calculating each curve in the reference section curve set according to the current section to be repaired to obtain a time weight and a spatial distance weight corresponding to each curve; Determine a reconstructed curve according to the time weight and the spatial distance weight of all curves; Reconstructing the damaged portion corresponding to the current section to be repaired according to the reconstruction curve to obtain a repaired section; Determining whether the number of iterations reaches a preset number of iterations or determining whether the repaired section is completely repaired; If not, the repaired section is reconstructed according to the curves in the reference section curve set.

7. The method for reconstructing a bending and torsional deformation blade model based on reference weight driven decision making according to claim 6, characterized in that: The determining of the reconstructed curve according to the time weight and the spatial distance weight of all curves includes: Determining a target weight corresponding to each curve according to the time weight and the spatial distance weight of each curve; The curve corresponding to the maximum value among all target weights is determined as the initial curve; The initial curve is optimized based on the curvature weighted VMM algorithm to obtain a reconstructed curve.

8. A device for reconstructing a bending and torsional deformation blade model based on a benchmark weight driven decision, characterized in that: include: An area determination module is configured to obtain point cloud data of the deformed, twisted blade and determine, based on the point cloud data, at least one area to be repaired and at least one non-repaired area; the area to be repaired is a locally deformed area; a set determination module, configured to process the at least one non-repaired area to obtain a cross-sectional point cloud set of each area to be repaired, and obtain a reference point cloud set based on all cross-sectional point cloud sets; a surface reconstruction module, configured to reconstruct the at least one area to be repaired according to the reference point cloud set to obtain a repaired blade surface; Obtaining a normal vector of a top plane based on a point cloud of a top cross section of the twisted blade; Obtaining a cross-sectional interval based on point clouds in the at least one area to be repaired and the plurality of target non-repaired areas; The base point cloud set is obtained based on all cross-section point cloud sets, including: Refining the at least one to-be-repaired region and the at least one non-repaired region according to the normal vector to obtain at least one refined to-be-repaired region and at least one refined non-repaired region; Obtaining a reference point cloud set according to the cross-sectional point cloud set of the at least one refined area to be repaired and the at least one refined non-repaired area; Refining the at least one area to be repaired according to the normal vector to obtain at least one refined area to be repaired, comprising: Using the direction of the normal vector as the normal vector of the section plane, each area to be repaired and the previous area to be repaired are calculated to obtain a torsion angle; According to the torsion angle, a torsion angle change rate is obtained; updating the section interval according to the torsion angle and the torsion angle change rate to obtain a target section interval; Each of the regions to be repaired is refined according to the target cross-sectional interval to obtain a refined region to be repaired.

Images