A v-direction optimal reference iteration-based blade repair region surface reconstruction method

Abstract

The application discloses a kind of v optimal reference iteration-based blade repair area surface reconstruction algorithm.For the cross section of preprocessed point cloud data, the reference cross section point cloud set is obtained, and the reference cross section curve is obtained by fitting the reference cross section point cloud set;After that, the reference cross section curve group is adopted Gaussian weighting discrete, and the v direction curve group is obtained by fitting in v direction;Based on the optimal principle in v direction, the optimal cross section of repair area is selected, the optimal cross section is intersected with the v direction curve group, the cross section curve of repair area is obtained, the cross section curve of repair area is added to the reference cross section curve group after ICP matching, then the v direction curve group is recalculated, all repair area cross sections are fitted in the form of step-by-step iteration, and the reference cross section curve group is added to the reference cross section curve group, finally, the reference cross section curve group is lofted to obtain the blade repair surface.The accuracy of the repair surface is greatly improved by the iterative method.

Description

Technical Field

[0001] This invention belongs to the field of blade repair technology, and relates to blade three-dimensional model reconstruction technology, specifically to a method for reconstructing the surface of the blade repair area based on v-direction optimal benchmark iteration. Background Technology

[0002] Blade surface reconstruction technology is widely used in blade repair work in aviation, hydropower and other fields, especially in aero-engine turbine blades. The repair of these blades requires high precision and surface profile, and their surface distortion is significant, which is one of the current research challenges in the field of blade repair.

[0003] Currently, blade reconstruction technology mainly falls into two categories: improvements based on traditional reverse engineering algorithms and reconstruction algorithms based on blade design features. The latter often offers higher accuracy and is therefore the mainstream choice for blade reconstruction algorithms. Blade surfaces possess characteristics such as NURBS surfaces and twisted surfaces. Based on this, numerous studies have focused on surface reconstruction algorithms based on blade cross-sectional curves. Yu et al. proposed a method using adjacent cross-sectional curves as a reference, the repair area point cloud as a baseline, and adjusting curve control points to control curve deformation, ultimately obtaining a repair area cross-sectional curve that meets accuracy requirements. However, this method overemphasizes positional accuracy, requiring data point density to limit the smoothness of the reconstructed surface, and neglects the surface features in the v-direction of the blade. Zhao Xusheng et al. proposed a method based on multi-section curve lofting to fit the blade surface and adjusting the repair area cross-sectional curve based on the repair area point cloud. While this method utilizes the features of the lofted blade surface, the surface in the repair area is obtained through a single lofting, resulting in insufficient utilization of v-direction features. For repair areas with significant v-direction shape changes, reconstruction accuracy decreases. Therefore, how to fully utilize the three-dimensional information of blade point cloud data to obtain a high-precision fitted surface is a crucial issue. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, the present invention aims to provide a method for reconstructing the surface of the blade repair region based on optimal benchmark iteration in the v-direction. By utilizing the optimal iteration method of the v-direction curve, the accuracy of the cross-sectional curve of the repair region after reconstruction is significantly improved, thereby maximizing the accuracy of the repair surface after reconstruction.

[0005] To solve the above-mentioned technical problems, the present invention employs the following technical means:

[0006] A method for reconstructing the surface of the blade repair region based on v-axis optimal benchmark iteration includes the following steps:

[0007] Step 1: Based on the preprocessed blade point cloud data, the blade point cloud region is divided into repair and non-repair regions. Using a cross-section parallel to the blade tip plane, the non-repair region is divided into several reference sections. Point clouds from these reference sections are then extracted to obtain several reference section point cloud sets. N is the total number of cross-sections. Indicates the first pi The points intercepted by the sectional plane are a cloud;

[0008] Step 2: Generate a cloud of points for each reference section according to the NURBS curve fitting requirements. Sorting operations are performed on the point cloud within it;

[0009] Step 3: Based on the sorted point cloud, use the NURBS surface fitting formula to fit the point cloud set of each reference section. By performing fitting, a set of reference cross-section curves consisting of N reference cross-section curves is obtained;

[0010] Step 4: Discretize the N reference cross-sectional curves to obtain the set of discrete points for the reference cross-sectional curve group. , For the first i The first discrete curve t Discrete points of each parameter, This represents the total number of discrete points on the reference cross-section curve. Let be the set of discrete points on N curves with equal parameter t, denoted as . ;

[0011] Step 5: For the discrete point set of the reference cross-section curve group The set of discrete points for each parameter t All were obtained using NURBS curve fitting. A v-curve;

[0012] Step 6: In the blade repair area, select several cross-sectional planes parallel to the blade tip plane as the repair area sections, and select the optimal repair area section according to the optimal reference principle; compare the optimal repair area section with... The intersection of the v-direction curves yields One interception point; based on this The cross-sectional curve of the repaired area is obtained by fitting the cut-off points. , k Indicates the fitting order;

[0013] Step 7: Perform ICP matching based on the point cloud and cross-sectional curve of the repaired area, and add the matched curve as the reference cross-sectional curve to the reference cross-sectional curve group in Step 3.

[0014] Step 8: Determine whether the cross-section of the repair area has been properly fitted. If it has, proceed to step 10; otherwise, return to step 5.

[0015] Step 9: Using the updated reference section curve group after iteration, perform a lofting surface fitting operation based on all reference section curves to obtain the final blade repair surface.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0017] This invention performs cross-sectional interception processing on preprocessed point cloud data. Using the cross-sectional point cloud of the non-repaired region as a benchmark, a benchmark cross-sectional point cloud set is obtained. An improved genetic algorithm is used to sort the cross-sectional point clouds to ensure the accuracy of cross-sectional curve fitting. Based on Green's theorem, a unidirectional operation is performed to fit the sorted benchmark cross-sectional point cloud set to obtain a benchmark cross-sectional curve group composed of several benchmark cross-sectional curves. Then, the benchmark cross-sectional curve group is discretized using Gaussian weighted average and fitted in the v-direction to obtain a v-direction curve group. Based on the v-direction optimality principle, the optimal cross-section of the repaired region is selected, and the optimal cross-section is intersected with the v-direction curve group to obtain the cross-sectional curve of the repaired region. After ICP matching of the cross-sectional curve of the repaired region, it is added to the benchmark cross-sectional curve group. Then, the v-direction curve group is recalculated. Through step-by-step iteration, all repaired region cross-sections are fitted with surface curves in sequence and added to the benchmark cross-sectional curve group. Finally, the benchmark cross-sectional curve group is lofted to obtain the blade repair surface. This invention iterates according to the optimal principle in the v-direction, obtaining the optimal cross-section of the repair area each time. The intersection of the optimal repair area cross-section with the v-direction curve yields a set of intercept points. The cross-sectional curve on the optimal repair area cross-section is obtained by fitting the set of intercept points. After ICP positioning and matching, the cross-sectional curve is added to the reference cross-sectional curve set. The reference cross-sectional curve set is updated, and then the v-direction curve phase is recalculated. The v-direction curve is updated step by step through iteration, so that the accuracy of the v-direction curve becomes higher and higher, more consistent with the true shape of the blade, and the accuracy of the final repair surface is also higher. Attached Figure Description

[0018] Figure 1 This is a flowchart of the blade repair region surface reconstruction method based on v-direction optimal benchmark iteration according to the present invention.

[0019] Figure 2 It is preprocessed instance point cloud data. Preprocessing mainly includes noise reduction, region detection and repair, edge extraction and other preprocessing processes.

[0020] Figure 3 It is a schematic diagram of the point cloud cross section and projection.

[0021] Figure 4 This is a projection of the point cloud on the U-section.

[0022] Figure 5This is a graph showing the curve fitting effect of the reference section.

[0023] Figure 6 In the figure, 'a' is a schematic diagram of generating the v-direction cross-section curve based on the discrete point set of the reference cross-section curve group.

[0024] Figure 6 In Figure b, the curve matching diagram is based on the point cloud of the repaired area after the cross-sectional curve of the repaired area is obtained.

[0025] Figure 7 This is a schematic diagram representing the order of cross-sectional curve fitting within the optimal repair region cross-section in the v direction.

[0026] Figure 8 for Figure 1 A schematic diagram of the genetic algorithm sorting process.

[0027] 1-Blade point cloud, 110-Repair area, 120-Non-Repair area, 121-Upper reference area SU, 122-Lower reference area SD, 2-u-section, 3-Repair area section, 4-Enlarged view of the blade to be repaired after edge extraction, 5-Reference section, 6-Edge point cloud, 7-Repair allowance point cloud, 8-Reference section curve, 9-Reference section point cloud set, 10-v-curve, 11-Truncation point, 12-Actual point cloud, 13-Section curve. Detailed Implementation

[0028] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0029] To make the technical solution and advantages of this invention clearer, the technical solution of this invention will be described more clearly below in conjunction with the accompanying drawings. Its technical roadmap is as follows: Figure 1 As shown.

[0030] The prerequisites for implementing this invention are as follows: Point cloud data of the blades is acquired using a visual scanner or coordinate measuring machine, and the original point cloud data is calibrated using hand-eye calibration to achieve high-precision data positioning. The point cloud is then subjected to smoothing and noise reduction processing, such as bilateral filtering. Furthermore, edge extraction algorithms and clustering algorithms are used to detect and separate the repaired and non-repaired areas, thereby obtaining preprocessed data, such as... Figure 2 As shown.

[0031] Based on the preprocessed blade point cloud 1 described above, this invention provides a method for reconstructing the surface of the blade repair region based on v-direction optimal benchmark iteration, comprising the following steps:

[0032] Step 1: As Figure 2As shown, based on the preprocessed blade point cloud data, the blade point cloud 1 is divided into regions: a repair region 110 and a non-repair region 120. The main idea of ​​this invention is to reconstruct the repair region based on the point cloud of the non-repair region. Therefore, the point cloud of the non-repair region is also the reference region for reconstruction. The non-repair region 120 is further divided into an upper reference region SU121 and a lower reference region SD122. A u-section 2 (parallel to the blade tip plane) is set to divide the non-repair region 120 into several reference sections 5. The point cloud of the non-repair region is intercepted using the reference sections 5 to obtain several reference section point cloud sets. N is the total number of cross-sectional planes (the sum of cross-sectional planes within the upper reference region SU and the lower reference region SD). Indicates the first pi The points intercepted by the sectional plane are a cloud;

[0033] In this embodiment, point cloud is extracted using the cross-sectional projection method, and the spacing of the cross-sectional planes and the selection range of projection points are controlled by the point cloud density. The specific steps of the cross-sectional projection method are as follows:

[0034] The formula for point cloud density is as follows:

[0035] Formula (1)

[0036] Formula (2)

[0037] In the above formula, Distances between a selected point and all its neighbors The sum of, The total number of all neighboring points of the selected point; the mean point cloud density can be calculated using the above formula. ;

[0038] The value of the number of cross-sectional planes N depends on the actual blade size and precision, and can generally be 4-20. In this embodiment, it is set to 8.

[0039] The cross-sectional interval is set to , ,in As a control factor, the smaller the value, the greater the precision. However, this requires an increase in the number of cross-sections, which increases the computational load. Generally, the value is above 100, but in this example, it is set to 150.

[0040] The set of cross-sections can be determined by the number of cross-sections N and the interval between cross-sections, and is specifically the set of upper cross-sections. and lower cut-off plane set A schematic diagram of the cross-sectional plane distribution is shown below. Figure 3 As shown.

[0041] The cross-sectional projection method refers to projecting a point cloud within a certain thickness range on both sides of a u-section to extract the point cloud; the width of the projected point cloud is... ,in, The projection truncation factor should generally not be too large; a range of 1-5 is reasonable, and it is set to 3 here. The point cloud data truncated through the width of the projected point cloud is then dimensionality-reduced based on this truncation plane. or Projection is performed, with the projection direction being the normal direction of the u-section. The final result is a set of cross-sectional point clouds. A single reference section point cluster of 9 Renderings Figure 4 As shown.

[0042] Step 2: Generate a cloud of points for each reference section according to the NURBS curve fitting requirements. Sorting operation is performed on the inner point cloud;

[0043] Since the number of leaf point clouds is huge, and the cross-sectional point clouds are generally in the hundreds, it belongs to the TSP problem. In order to avoid the problem of low computational efficiency or local optima of conventional algorithms, the embodiments of the present invention adopt an improved genetic algorithm based on the genetic algorithm.

[0044] The improvements are as follows:

[0045] Step 2.1: Set up using a genetic algorithm Total number of point clouds Calculate the Euclidean distance between each point and store it as a sorted set of point distances: , i , j Subscript This represents the Euclidean distance between any two points. The initial population size is set to... , The population size control coefficient has a value range of 1-10, and is set to 5 in this invention example.

[0046] Step 2.2: The point cloud sorting target is distance. Optimal, set the objective function: This invention uses the reciprocal of distance as fitness. Since the results in the early stages of evolution are generally worse than the final results, this invention adds a fitness transformation function and a scaling factor to the genetic algorithm for encoding: Fitness: , The proportionality coefficient (obtained randomly). The largest sorted population Value, and the greedy algorithm to find the minimum value The proportion); is included in the fitness transformation to obtain the final fitness function. : , These represent the average fitness and minimum fitness of the population, respectively.

[0047] Step 2.3: Due to the large number of point clouds, the randomly generated initial population has poor fitness evaluation, resulting in slow algorithm convergence. This invention optimizes the initial population by using a greedy algorithm to obtain... A local optimal path, and a certain proportion of the solutions. (1%-5%, this example uses 3%) randomly generated An initial population.

[0048] Step 2.4: In terms of selection, a roulette wheel method is used, and an elite strategy is added to prioritize retaining the top 1% of individuals.

[0049] Step 2.5, Regarding the crossover aspect, set the crossover probability: (In this example, the setting is 0.2), using PMX cross-connect.

[0050] Step 2.6, Regarding mutation, set the mutation probability: (In this example, the value is set to 0.9), using a multi-point mutation method, the maximum number of mutation points is set. (This invention is set to 3), multi-point mutation probability: Furthermore, mutated individuals will only be retained if their fitness is higher than that of the original individuals. The fitness values ​​of the population after crossover and mutation will then be updated.

[0051] Step 2.7: Set the maximum number of iterations (In this invention, the example is set to 2000). When the convergence condition is met, repeat steps 2.4-2.6. If the number of iterations exceeds this number, it means that the algorithm has not converged.

[0052] Convergence criteria settings: Determined by fitness gradient, with the following conditions: Set gradient threshold. (Set to 0.001 in this paper), calculate the gradient of fitness before and after population iteration. In the gradient ascent case, calculate the number of times the gradient is continuously below a threshold. When the number is greater than... ( In this example, the value is set to 0.005, which satisfies the convergence condition. The results are shown in the table below:

[0053] Table 1. Sorting results of the improved genetic algorithm

[0054]

[0055] The original algorithm could not converge for more than 200 points, and the convergence time for 186 points was about 37 seconds. However, the improved algorithm has a much faster convergence speed. After multiple sorting, the error of the convergence result is about 0.03 mm, which shows stability.

[0056] It should be noted that the above example of using a genetic algorithm for sorting is only for illustration. Other algorithms can also be used for sorting. It should also be noted that the above only explains the genetic improvement part. For the parts not explained, the original genetic algorithm can be used.

[0057] After sorting the point clouds at each cross section, the genetic algorithm's sorting direction is random, and Green's theorem ensures that the curves are in the same direction: The first sorted set of cross-sectional points is used as the directional reference.

[0058] Step 3: Based on the sorted point cloud, use the NURBS surface fitting formula to fit the point cloud set of each reference section. The fitting was performed, and the result is shown in the figure below. Figure 5 As shown, a set of reference cross-section curves consisting of N reference cross-section curves is obtained;

[0059] Step 4: Discretize the N reference cross-sectional curves to obtain the set of discrete points for the reference cross-sectional curve group. The specific method is as follows:

[0060] Step 4.1: Perform isoparametric discretization on the N reference cross-sectional curves with a discretization point count of M0 (700 in this example; selection principle: 1-3 times the average cross-sectional point cloud count);

[0061] Step 4.2: Align the N reference cross-sectional curves after discretization of the isoparameter M0 using statistical principles; the specific method is as follows:

[0062] In statistics, the correlation coefficient of curvature between cross sections can effectively represent the correlation between them. The curvature correlation coefficient is calculated as follows: ,in , Let x and y be the expected curvatures of the curve. , For curves x , y No. i Curvature at a point, with x The M0 points of the curve are taken as starting points in sequence. y With the curve location fixed at the initial value, calculate the correlation coefficient value under M0 alignments, and then select the value with the minimum correlation coefficient. x curve starting point and yThe initial starting points of the curves are aligned to ensure parameter alignment between the two curves. Based on the objective function above, using any discrete point set of a cross-section as a reference, the parameters of all reference cross-section curves are aligned.

[0063] Step 4.3: After aligning the N reference cross-sectional curves with equal parameters, the discrete step length is corrected using a curvature-based Gaussian weighted discretization algorithm. The specific method is as follows:

[0064] Each reference cross-section curve has M0 discrete parameter points, and the first... i The first reference section curve j The curvature of the discrete parameter points is Calculate the corresponding number j The average curvature under each parameter is:

[0065] Formula (3)

[0066] The total average curvature of all discrete points on N reference cross-section curves: Formula (4)

[0067] The variance is: Formula (5)

[0068] The total average curvature obtained based on the above average discretization and variance The actual discrete step size is calculated using curvature characteristics, as detailed below:

[0069] Set the first i The parameters of the reference section curve are: t The curvature at that time is Parameters such as N reference cross-section curves t The mean curvature is: Based on the Gaussian function and the data calculated in the above steps, the parameters are obtained. t The formula for calculating the corresponding step length is:

[0070] Formula (6)

[0071] In the above formula This is the maximum step size control coefficient, where the control step size is within... , The minimum step size coefficient controls the step size. , These are the control coefficients of the Gaussian kernel function, primarily controlling the influence of curvature characteristics on the step size, with significant... The smaller the value, the more concentrated the high step size is at low curvature; in this example, it is set to 2.

[0072] Step 4.4: Using the corrected step size (step size calculation formula), re-discrete the N reference cross-section curves of the reference cross-section curve group to obtain the set of discrete points of the reference cross-section curve group. , For the first i The first discrete curve t Discrete points of each parameter, This represents the total number of discrete points on the reference cross-section curve after calculation using the Gaussian weighted discretization algorithm. Let be the set of discrete points on N curves with equal parameter t, denoted as . ;

[0073] It should be noted that the above This is not a summation formula, but a two-dimensional matrix representing a discrete set of points, expressed in matrix form as follows:

[0074]

[0075] The same is The set of one-dimensional discrete points is represented as follows:

[0076]

[0077] In step 4.4, the step length is calculated using the above formula. i The reference cross-sectional curve is re-discretized to obtain the first... i Set of discrete points of the reference cross-section curve Repeat the discretization process for all reference cross-section curves to obtain N sets of discrete points for each curve:

[0078] Formula (7)

[0079] Step 5: For the discrete point set of the reference cross-section curve group The set of discrete points for each parameter t All were obtained using NURBS curve fitting. 10 v-curves, such as Figure 6 As shown.

[0080] Step 6: In the blade repair area, select several cross-sectional planes parallel to the blade tip plane as the repair area sections, and select the optimal repair area section according to the optimal reference principle; for example... Figure 6 As shown in Figure a, the optimal repair area section is compared with... The intersection of the v-direction curves yields One interception point 11; such as Figure 6 As shown in b, based on this The cross-sectional curve of the repaired area was obtained by fitting 11 intercept points. , kIndicates the fitting order;

[0081] The optimal baseline principle in the v-direction is explained as follows:

[0082] 1) In a single iteration, the number of curves selected for the repair region should be minimized, so it is set to 1;

[0083] 2) Under the same conditions, the influence of the reference curve that is closer to the repair area in the v direction on the cross-sectional curve should be greater than that of the reference curve that is farther away; (space).

[0084] 3) Because errors are inevitably introduced during iteration, the earlier the baseline cross-section curve is set, the higher its accuracy and reliability should be (time order).

[0085] The number of cross sections in the repair area is , The coefficients that affect accuracy and efficiency are as follows: the larger the coefficient, the higher the accuracy, but the greater the computing power required. Therefore, the appropriate range is 10-30.

[0086] This evaluation index goal By combining time (the influence factor of iteration order) and spatial factors, we can ensure the reliability of the baseline of the fitted curve as much as possible.

[0087] Regarding time, considering the optimal baseline principle 3): When updating the baseline curve, errors will inevitably be introduced. The earlier the iteration sequence of the baseline curve, the greater its reliability. Therefore, an influence factor is added. For the baseline cross-sectional curve of the non-repaired region, use it as the initial baseline curve and set the influence factor. The value is 1. The influence factor of subsequently added cross-sectional curves is... = , i This represents the number of iterations. It is a logarithm with base 2.

[0088] In terms of space, based on the optimal benchmark principle 2): the smaller the v-direction distance between the benchmark curve and the cross-section of the repair area, the greater the influence should be. , This is the cross-section of the repaired area to be evaluated up to the first... ptop The distance between the planes containing the reference cross-section curves.

[0089] Finally, evaluation indicators were obtained for the cross-sectional settings of each repair area. Formula (8)

[0090] Formula (8) is the summation formula.

[0091] Based on the evaluation scores, the highest score is selected for cross-sectional and curve fitting operations. The resulting illustration is shown below. Figure 7As shown in the figure, the cross sections of the repaired area from top to bottom are section 14, 15, 16, and 17, with iteration calculations of 1, 3, 4, and 2 respectively. The evaluation index of section 15 is higher than that of section 16 because section 15 is closer to section 14, which was iterated earlier, than section 16. This also reflects the baseline principle of v-axis optimality.

[0092] Step 7: Perform ICP matching based on the point cloud and cross-sectional curve of the repaired area, and add the matched curve as the reference cross-sectional curve to the reference cross-sectional curve group in Step 3.

[0093] Step 8: Determine whether the cross-section of the repaired area has been fully fitted (i.e., the number of iterations). i Is it equal to the number of cross sections in the repair area? If the fitting is complete, proceed to step 10; otherwise, return to step 5.

[0094] Step 9: Using the updated reference section curve group after iteration, perform a lofting surface fitting operation based on all reference section curves to obtain the final blade repair surface.

[0095] Blade repair can be performed using the curved surface of the blade repair area.

[0096] The above embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Although the invention has been described in detail with reference to the embodiments, those skilled in the art should understand that various combinations, modifications, or equivalent substitutions of the technical solutions of the invention do not depart from the spirit and scope of the invention and should be covered within the scope of the claims of the invention.

Claims

1. A method for reconstructing the surface of a blade repair region based on v-direction optimal benchmark iteration, characterized in that, Includes the following steps: Step 1: Based on the preprocessed blade point cloud data, the blade point cloud region is divided into repair and non-repair regions. Using a cross-section parallel to the blade tip plane, the non-repair region is divided into several reference sections. Point clouds from these reference sections are then extracted to obtain several reference section point cloud sets. N is the total number of cross-sections. Indicates the first pi The points intercepted by the sectional plane are a cloud; Step 2: Generate a cloud of points for each reference section according to the NURBS curve fitting requirements. Sorting operations are performed on the point cloud within it; Step 3: Based on the sorted point cloud, use the NURBS surface fitting formula to fit the point cloud set of each reference section. By performing fitting, a set of reference cross-section curves consisting of N reference cross-section curves is obtained; Step 4: Discretize the N reference cross-sectional curves to obtain the set of discrete points for the reference cross-sectional curve group. , For the first i The first discrete curve t Discrete points of each parameter, This represents the total number of discrete points on the reference cross-section curve. Let be the set of discrete points on N curves with equal parameter t, denoted as . ; Step 5: For the discrete point set of the reference cross-section curve group The set of discrete points for each parameter t All were obtained using NURBS curve fitting. A v-curve; Step 6: In the blade repair area, select several cross-sectional planes parallel to the blade tip plane as the repair area sections, and select the optimal repair area section according to the optimal reference principle; compare the optimal repair area section with... The intersection of the v-direction curves yields One interception point; based on this The cross-sectional curve of the repaired area is obtained by fitting the cut-off points. , k Indicates the order of fitting; Step 7: Perform ICP matching based on the point cloud and cross-sectional curve of the repaired area, and add the matched curve as the reference cross-sectional curve to the reference cross-sectional curve group in Step 3. Step 8: Determine whether the cross-section of the repair area has been properly fitted. If it has, proceed to step 10; otherwise, return to step 5. Step 9: Using the updated reference section curve group after iteration, perform a lofting surface fitting operation based on all reference section curves to obtain the final blade repair surface. In step 6, the number of cross-sections in the repair area is: , The coefficients affecting accuracy and efficiency are used to set evaluation indicators for each repair area cross-section. Each time, the cross section of the repaired area with the highest evaluation index is selected and intersected with the v-direction curve, and the cross section curve is fitted. in, , i For the number of iterations, It is the logarithm to the base 2; In the above, This is the cross-section of the repaired area to be evaluated up to the first... ptop The distance between the planes containing the reference cross-section curves; In step 1, during the point cloud cropping process of the non-repair area, a point cloud set is obtained by projecting the point cloud within a certain width range onto the cutting plane, and the interval of the cutting plane and the selection range of the projection points are controlled by the point cloud density. The formula for point cloud density is as follows: In the above formula, Distances between a selected point and all its neighbors The sum of, The total number of all neighboring points of the selected point; The interval of the cutting plane is , The control coefficient is a set constant. Width of point cloud projection on the cross-section plane The projection intercept coefficient is a set constant.

2. The method for reconstructing the curved surface of the blade repair area according to claim 1, characterized in that: In step 2, the fitness evaluation is optimized based on the improved genetic algorithm, the initial population is optimized by using the local optimum solution of the greedy algorithm to improve the convergence speed, and multi-point mutation is used to obtain the global optimal sorting of the cross-sectional point cloud. The properties of Green's theorem are used to ensure the unidirectionality of the sorted point clouds.

3. The method for reconstructing the curved surface of the blade repair area according to claim 1, characterized in that: In step 2, a cloud of points is generated for each reference section. After the point cloud within is sorted, Green's formula is used to ensure that the curves are in the same direction, and the first sorted cross-sectional point cloud set is used as the direction reference.

4. The method for reconstructing the curved surface of the blade repair area according to claim 1, characterized in that: In step 4, the Gaussian weighted discretization algorithm is used for discretization, as detailed below: Step 4.1: Perform isoparametric discretization on the N reference cross-sectional curves with a discretization point count of M0; Step 4.2: Align the N reference cross-sectional curves after discretization of the equal parameter M0 using statistical principles; Step 4.3: After the N reference cross-sectional curves are parametrically aligned, the discrete step length is corrected using a curvature-based Gaussian weighted discretization algorithm. Step 4.4: Re-discrete the reference cross-section curves within the reference cross-section curve group using the corrected step size to obtain the set of discrete points for the reference cross-section curve group. .

5. The method for reconstructing the curved surface of the blade repair area according to claim 4, characterized in that: In step 4.3, the specific method for correcting the distance step length is as follows: Each reference cross-section curve has M0 discrete parameter points, and the first... i The first reference section curve j The curvature of the discrete parameter points is Calculate the corresponding number j The average curvature under each parameter is: ； The total average curvature of all discrete points on N reference cross-section curves: ; The variance is: Based on total mean curvature and variance The actual discrete step size is calculated using curvature characteristics, as detailed below: Set the first i The parameters of the reference section curve are: t The curvature at that time is Parameters such as N reference cross-section curves t The mean curvature is: Based on the Gaussian function and the data calculated in the above steps, the parameters are obtained. t The formula for calculating the corresponding step length is: In the above formula This is the maximum step size control coefficient, where the control step size is within... , The minimum step size coefficient controls the step size. , represents the control coefficients of the Gaussian kernel function.

6. The method for reconstructing the curved surface of the blade repair area according to claim 5, characterized in that: In step 4.4, the step length is calculated using the above formula. i The reference cross-sectional curve is re-discretized to obtain the first... i Set of discrete points of the reference cross-section curve Repeat the discretization process for all reference cross-section curves to obtain N sets of discrete points for each curve: 。

Images