Method and system for accurately measuring thickness of aircraft engine blades

By dividing voxel blocks and calculating eigenvalues ​​on the air engine blade point cloud data, connecting feature voxel blocks for denoising, solving the problem of noise point removal in blade thickness measurement, and achieving more accurate thickness measurement.

CN118364659BActive Publication Date: 2024-11-05SUZHOU HANWEI MATERIAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410788456.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-19
Publication Date
2024-11-05
Estimated Expiration
2044-06-19

AI Technical Summary

Technical Problem

In the thickness measurement of aircraft engine blades, due to the complex structure and severe occlusion, the point cloud data density is uneven. It is difficult for traditional noise denoising algorithms to effectively remove noise points, resulting in inaccurate thickness measurement.

Method used

By obtaining the point cloud data of the aero engine blade and dividing it into voxel blocks, the blade structure characteristic value of each voxel block is calculated, the curve and score value are constructed, the feature voxel blocks are connected to form a connecting interval, and the point cloud data is used to use the feature model interval. Denoising is performed to calculate the blade thickness.

Benefits of technology

It improves the accuracy of aircraft engine blade thickness measurement, effectively removes noise points, and ensures the accuracy of thickness measurement.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118364659B_ABST
    Figure CN118364659B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of data processing technology, and in particular to a method and system for accurately measuring the thickness of an aircraft engine blade, comprising: obtaining point cloud data of an aircraft engine blade and dividing it into a plurality of voxel blocks, obtaining the blade structure characteristic value of each voxel block according to the correlation between each voxel block and other voxel blocks in the neighborhood, calculating the blade structure score value of each voxel block in the layer and the two adjacent layers according to the difference of the blade structure characteristic value, obtaining the blade structure possibility of each voxel block according to the fluctuation of the blade structure score value of the three-layer voxel block in the neighborhood range of each voxel block, connecting some voxel blocks to obtain a plurality of connected intervals, and then obtaining the characteristic model interval to denoise the point cloud data and calculate the thickness of the aircraft engine blade at each position. The present invention denoises the point cloud data of the blade according to the shape characteristics of the aircraft engine blade, and completes the accurate measurement of the thickness of the aircraft engine blade.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a method and system for accurately measuring the thickness of an aircraft engine blade. Background Art

[0002] Accurate thickness measurement of aircraft engine blades is an important part of ensuring blade strength and engine performance. With the rapid development of 3D scanning technology, the application of 3D point cloud data in the field of blade inspection has become more and more extensive, making the measurement of blade thickness more efficient and accurate.

[0003] In the process of monitoring aircraft engine blades, point cloud data can be used to analyze the integrity of the structure. When collecting laser point cloud data for complex aircraft engine blades, due to the complex structure and severe occlusion of aircraft engine blades, the density of the collected point cloud is extremely uneven, and there are clustered noise points, which creates a need for noise data denoising. Traditional denoising filtering algorithms mostly use a single mathematical model, which makes it difficult to obtain a complete aircraft engine blade point cloud. The random sample consensus (RANSAC) algorithm can select part of the data from the data group containing noise points, and iteratively estimate the model parameters in the source data that conform to a certain mathematical model. However, the structure of aircraft engine blades is complex, and the structural shape of the blade cannot be accurately determined in advance. Therefore, it is difficult to specify the required number of samples, resulting in over-segmentation or under-segmentation of the point cloud data, and the thickness of the aircraft engine blade cannot be accurately measured. Summary of the invention

[0004] The invention provides a method and system for accurately measuring the thickness of an aircraft engine blade to solve the existing problems.

[0005] The method and system for accurately measuring the thickness of aircraft engine blades of the present invention adopt the following technical solutions:

[0006] One embodiment of the present invention provides a method for accurately measuring the thickness of an aircraft engine blade, the method comprising the following steps:

[0007] Obtain the point cloud data of aircraft engine blades and divide it into several voxel blocks;

[0008] According to the plane angle between each voxel block and other voxel blocks in the preset neighborhood, the blade structure characteristic values ​​of other voxel blocks in the neighborhood of each voxel block are obtained; according to the blade structure characteristic values ​​of the layer where each voxel block is located, the first curve and the second curve are constructed; according to the difference in the blade structure characteristic values ​​on the first curve and the second curve, the blade structure score values ​​of each voxel block in the layer and the two adjacent layers are obtained; according to the numerical difference in the blade structure score values ​​of the three layers of voxel blocks in the neighborhood of each voxel block, the blade structure possibility of each voxel block is obtained; according to the blade structure possibility of each voxel block, some voxel blocks are connected to obtain several connected intervals; according to the number of voxel blocks and the blade structure possibility in each connected interval, the characteristic model interval score value of each connected interval is obtained;

[0009] The characteristic model interval is obtained according to the characteristic model interval score value, and the point cloud data is denoised according to the characteristic model interval to obtain denoised point cloud data; the denoised point cloud data is calculated to obtain the thickness of the aircraft engine blade at each position.

[0010] Furthermore, the step of obtaining the leaf structure characteristic values ​​of other voxel blocks in the neighborhood of each voxel block according to the plane angle between each voxel block and other voxel blocks in the preset neighborhood includes the following specific steps:

[0011] With the ath voxel block as the center, a neighborhood of size n*n*n is constructed; the calculation method for the leaf structure eigenvalue of the bth voxel block in the neighborhood of the ath voxel block is:

[0012]

[0013] In the formula, represents the leaf structure eigenvalue of the a-th voxel block and the b-th voxel block in the neighborhood; Represents the normal vector of the plane formed by the point cloud data in the ath voxel block, The normal vector of the plane formed by the point cloud data of the bth voxel block in the neighborhood of the ath voxel block; Indicates the number of point cloud data in the a-th voxel block, The number of point cloud data in the bth voxel block in the neighborhood of the ath voxel block; represents the exponential function with natural exponential as base, Represents the absolute value function.

[0014] Furthermore, the step of constructing the first curve and the second curve according to the leaf structure characteristic value of the layer where each voxel block is located includes the following specific steps:

[0015] The direction from the root of the blade to the tail of the blade is recorded as the vertical direction, and all voxel blocks except those with a blade structure characteristic value of 0 are obtained. The voxel block layer that includes the a-th voxel block and is perpendicular to the vertical direction is recorded as the layer where the a-th voxel block is located. On the layer where the a-th voxel block is located, the curves formed by the blade structure characteristic values ​​of all voxel blocks from the voxel block with the smallest blade structure characteristic value to the voxel block with the largest blade structure characteristic value are recorded as the first curve and the second curve, respectively.

[0016] Furthermore, the step of obtaining the leaf structure score of each voxel block in the layer and two adjacent layers according to the difference between the leaf structure characteristic values ​​on the first curve and the second curve includes the following specific steps:

[0017] The layer where the a-th voxel block is located is recorded as the j-th layer, and the leaf structure score of the a-th voxel block in the j-th layer is calculated as follows:

[0018]

[0019] In the formula, represents the leaf structure score of the a-th voxel block at the j-th layer; represents the slope of the line connecting the two end points of the first curve of the a-th voxel block in the j-th layer, The slope of the line connecting the two end points of the second curve representing the a-th voxel block in the j-th layer; represents the number of leaf structure eigenvalues ​​of the a-th voxel block on the first curve of the j-th layer; represents the number of leaf structure eigenvalues ​​of the a-th voxel block on the second curve of the j-th layer; represents the mean value of the leaf structure eigenvalue of the a-th voxel block on the first curve of the j-th layer; represents the mean value of the leaf structure feature value of the a-th voxel block on the second curve of the j-th layer; represents the exponential function with natural exponential as base, Represents the absolute value function.

[0020] Furthermore, the leaf structure possibility of each voxel block is obtained according to the numerical difference of the leaf structure score values ​​of the three layers of voxel blocks in the neighborhood of each voxel block, and the specific formula included is as follows:

[0021]

[0022] In the formula, represents the leaf structure possibility of the a-th voxel block; represents the variance of the leaf structure score values ​​of the three-layer voxel blocks in the neighborhood of the a-th voxel block; represents the leaf structure score of the a-th voxel block at the j-th layer; represents the normalization function; is a hyperparameter.

[0023] Furthermore, the method of connecting some voxel blocks according to the leaf structure possibility of each voxel block to obtain a plurality of connected intervals includes the following specific steps:

[0024] Among all voxel blocks, the voxel blocks whose leaf structure probability is greater than a preset feature threshold are recorded as feature voxel blocks, and the feature voxel blocks are connected with feature voxel blocks in the neighborhood to obtain a number of connected intervals.

[0025] Furthermore, the step of obtaining the feature model interval score value of each connected interval according to the number of voxel blocks and the possibility of leaf structures in each connected interval includes the following specific steps:

[0026]

[0027] In the formula, Represents the feature model interval score value of the rth connected interval; represents the number of voxel blocks in the rth connected interval; represents the possibility of the leaf structure of the t-th voxel block in the r-th connected interval; Represents the normalization function.

[0028] Furthermore, the step of obtaining the feature model interval according to the feature model interval score value and denoising the point cloud data according to the feature model interval to obtain denoised point cloud data includes the following specific steps:

[0029] The connected interval with the largest feature model interval score is recorded as the feature model interval. The product of the number of voxel blocks in the feature model interval and the minimum number of point cloud data required by the RANSAC algorithm is used as the minimum number of point cloud data selected by the RANSAC algorithm to build the model. The minimum number of point cloud data selected by the RANSAC algorithm to build the model is denoised to obtain the denoised point cloud data.

[0030] Furthermore, the denoised point cloud data is calculated to obtain the thickness of the aeroengine blade at each position, including the following specific steps:

[0031] In the denoised point cloud data, several parallel sections are selected along the vertical direction of the aircraft engine blade with the length of each voxel block as the step size. The PCL point cloud clustering algorithm is used to obtain the inner and outer contour lines in each section. The distance between the two point cloud data on the inner and outer contour lines of the section passed by a straight line perpendicular to the chord line of the engine blade is recorded as the thickness at each position on the engine blade.

[0032] The present invention also proposes a system for accurately measuring the thickness of aircraft engine blades, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of the aforementioned method for accurately measuring the thickness of aircraft engine blades.

[0033] The beneficial effects of the technical solution of the present invention are as follows: obtaining the point cloud data of the aeroengine blade and dividing it into a number of voxel blocks, so as to facilitate the analysis of the surface structure of the blade; obtaining the blade structure characteristic value of each voxel block according to the correlation between each voxel block and other voxel blocks in the neighborhood, and judging whether the point cloud data of each voxel block and the voxel blocks in the neighborhood are on the same plane; calculating the blade structure score of each voxel block in the layer and the two adjacent layers according to the difference of the blade structure characteristic value, and obtaining the blade structure possibility of each voxel block according to the fluctuation of the blade structure score of the three-layer voxel blocks in the neighborhood range of each voxel block, showing the similarity of the shapes of the two sides of the blade, and determining the possibility that the layer where each voxel block belongs to the aeroengine blade; connecting the characteristic voxel blocks to obtain a number of connected intervals, and then obtaining the characteristic model interval to denoise the point cloud data and calculate the thickness of the aeroengine blade at each position, thereby improving the accuracy of the thickness measurement of the aeroengine blade. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 The present invention is a flowchart of the steps of the method for accurately measuring the thickness of an aircraft engine blade;

[0036] Figure 2 It is a vertical schematic diagram of an aircraft engine blade;

[0037] Figure 3 It is a schematic diagram of the curved surface structure of an aircraft engine blade;

[0038] Figure 4 A schematic cross-sectional view of an aircraft engine blade. DETAILED DESCRIPTION

[0039] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method and system for accurately measuring the thickness of aircraft engine blades proposed by the present invention, its specific implementation, structure, features and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0040] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0041] The specific scheme of the method and system for accurately measuring the thickness of aircraft engine blades provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0042] See also Figure 1 , which shows a flow chart of the steps of a method for accurately measuring the thickness of an aircraft engine blade provided by an embodiment of the present invention, the method comprising the following steps:

[0043] Step S001: Acquire the point cloud data of an aircraft engine blade and divide it into a number of voxel blocks.

[0044] The purpose of this embodiment is to measure the thickness of an aircraft engine blade. First, a 3D laser scanner is used to scan the engine blade several times to obtain a number of initial point cloud data, and then the point cloud data of the aircraft engine blade is obtained by splicing. It should be noted that this embodiment uses a 3D laser scanner for scanning, so the point cloud data space is a three-dimensional space. Since the point cloud data features obtained at different positions of the aircraft engine blade are different, in this embodiment, the direction from the root of the blade to the tail of the blade is recorded as the vertical direction for analysis. The vertical direction of the blade is as follows: Figure 2 shown.

[0045] In order to improve the flow efficiency of the airflow and the overall performance of the engine, the aircraft engine blades usually adopt a curved surface design, so they present the same curved surface structure as the blades in three-dimensional space. Figure 3 As shown. Since it is relatively complicated to directly analyze large-scale point cloud data, the point cloud data is first divided into voxel blocks; specifically, all point cloud data are voxel-divided into a number of individual voxel blocks according to the octree structure. It should be noted that the octree structure is an existing method and will not be described in detail in this embodiment. At this point, a number of individual voxel blocks of all point cloud data are obtained.

[0046] Step S002: Obtain leaf structure feature values ​​of each voxel block and other voxel blocks in its neighborhood according to the correlation between each voxel block and other voxel blocks in its neighborhood.

[0047] It should be noted that since the point cloud data of aircraft engine blades in three-dimensional space appear as a curved surface structure, a single voxel block cannot reflect the characteristics of the curved surface structure. In addition to the aircraft engine blades, the voxel block also contains noise areas. Therefore, by analyzing whether the point cloud data in each voxel block and the voxel blocks in the neighborhood are on the same plane, if the adjacent voxel blocks are not in the same plane, it means that it is more likely to be on the curved surface of the blade structure, and it is preliminarily determined whether it is a voxel block on the aircraft engine blade.

[0048] Specifically, with the ath voxel block as the center, a neighborhood of size n*n*n is constructed. In this embodiment, n is preset to be 3 voxel blocks, which is described as an example. Other values ​​can be set in other implementations, and this embodiment does not limit it. The calculation method for the leaf structure feature value of the bth voxel block in the neighborhood of the ath voxel block is:

[0049]

[0050] In the formula, represents the leaf structure eigenvalue of the a-th voxel block and the b-th voxel block in the neighborhood; Represents the normal vector of the plane formed by the point cloud data in the ath voxel block, The normal vector of the plane formed by the point cloud data of the bth voxel block in the neighborhood of the ath voxel block; Indicates the number of point cloud data in the a-th voxel block, The number of point cloud data in the bth voxel block in the neighborhood of the ath voxel block; represents an exponential function with natural exponential as base, normalizing the dot product of the normal vector of the plane formed by the point cloud data in the a-th voxel block and the b-th voxel block in the neighborhood to To avoid the dot product being 0, Represents the absolute value function.

[0051] The smaller the value is, the less the planes formed by the two voxel blocks are on the same plane, the more likely it is a curved surface structure of an aircraft engine blade, the more likely it is a voxel block on an aircraft engine blade, and the larger the blade structure eigenvalue is; The smaller the value, the closer the structures of adjacent voxel blocks after three-dimensional scanning, that is, the more the number of point cloud data points is the same, the more likely it is that the voxel blocks are on the same blade with a curved surface structure, and the larger the characteristic value of the blade structure.

[0052] Similarly, according to the difference between each voxel block and other voxel blocks in its neighborhood, the leaf structure feature values ​​of all voxel blocks and each voxel block in the neighborhood are obtained.

[0053] Step S003, construct the first curve and the second curve of the leaf structure characteristic value of the layer where each voxel block is located, calculate the leaf structure score value of the layer where each voxel block is located and the two layers above and below; according to the fluctuation of the leaf structure score value of the three layers of voxel blocks in the neighborhood range of each voxel block, obtain the leaf structure possibility of each voxel block.

[0054] It should be noted that since the point cloud data of the aircraft engine blade in three-dimensional space is presented as a curved surface structure, each cross section in the vertical direction is presented as a similar teardrop shape. Figure 4 As shown, the voxel blocks are connected in sequence to form a closed curve, so for the voxel blocks on the cross section, the closer to the trailing edge of the blade, the less obvious the surface features are, and the smaller the blade structure characteristic value is, while the closer to the leading edge of the blade, the more obvious the surface features are, and the larger the blade structure characteristic value is; therefore, the similarity of the shapes of the two sides of the blade is expressed by the change of the blade structure characteristic value of the voxel blocks in the cross-sectional layer where each voxel block is located, and the possibility that the layer where each voxel block is located belongs to the aircraft engine blade is obtained.

[0055] Specifically, all voxel blocks except those with a blade structure characteristic value of 0 are obtained, and the voxel block layer including the ath voxel block and perpendicular to the vertical direction is recorded as the layer where the ath voxel block is located. On the layer where the ath voxel block is located, curves formed by the blade structure characteristic values ​​of all voxel blocks from the voxel block with the smallest blade structure characteristic value to the voxel block with the largest blade structure characteristic value are recorded as the first curve and the second curve, respectively. It should be noted that since this embodiment divides the point cloud data of the blade into voxels in the vertical direction of the blade, the layer where each voxel block is located after the segmentation is perpendicular to the vertical direction.

[0056] The layer where the a-th voxel block is located is recorded as the j-th layer, and the leaf structure score of the a-th voxel block in the j-th layer is calculated as follows:

[0057]

[0058] In the formula, represents the leaf structure score of the a-th voxel block at the j-th layer; represents the slope of the line connecting the two end points of the first curve of the a-th voxel block in the j-th layer, The slope of the line connecting the two end points of the second curve representing the a-th voxel block in the j-th layer; represents the number of leaf structure eigenvalues ​​of the a-th voxel block on the first curve of the j-th layer; represents the number of leaf structure eigenvalues ​​of the a-th voxel block on the second curve of the j-th layer; represents the mean value of the leaf structure eigenvalue of the a-th voxel block on the first curve of the j-th layer; represents the mean value of the leaf structure feature value of the a-th voxel block on the second curve of the j-th layer; represents the exponential function with natural exponential as base, Represents the absolute value function.

[0059] The smaller the value of is, the closer the slopes of the first curve and the second curve of the a-th volume pixel block in the j-th layer are, indicating that the fluctuations of the two curves are more similar, indicating that the shapes of the a-th volume pixel block on the two sides of the blade in the j-th layer are more similar, and the more likely it is to belong to a cross-section layer of the aircraft engine blade; The smaller the value is, the closer the relative blade structure eigenvalues ​​of the two sub-curve segments are. Then, the more similar the shapes of the a-th voxel block on both sides of the blade in the j-th layer are, the larger the blade structure score is, and the more likely it is a cross-sectional layer of an aircraft engine blade.

[0060] Similarly, the leaf structure scores of the previous and next layers of each voxel block are calculated. At this time, the three-layer leaf structure scores of each voxel block are obtained. According to the three-layer leaf structure scores of each voxel block, the leaf structure possibility of the voxel blocks in the entire field range of each voxel block is calculated as the leaf structure possibility of each voxel block.

[0061] Then the calculation method of the leaf structure possibility of the a-th voxel block is:

[0062]

[0063] In the formula, represents the leaf structure possibility of the a-th voxel block; represents the variance of the leaf structure score values ​​of the three-layer voxel blocks in the neighborhood of the a-th voxel block; represents the leaf structure score of the a-th voxel block at the j-th layer; represents the normalization function; is a hyperparameter. The description is made to avoid the denominator being 0, and other embodiments may set other values.

[0064] The value of is the variance of the blade structure score of the three-layer voxel block. The smaller the variance is, the closer the blade structure score of the three-layer voxel block is to the surface structure of the aircraft engine blade. The a-th voxel block is more likely to be the voxel block of the aircraft engine blade. The larger the value is, the greater the possibility that the entire neighborhood range of the a-th voxel block belongs to the structure of the blade, the closer it is to the surface structure of the blade, and the more likely it is a voxel block of an aircraft engine blade.

[0065] At this point, the leaf structure possibilities of all voxel blocks are obtained.

[0066] Step S004: merge the voxel blocks to obtain a number of connected intervals, and obtain the feature model interval score of each connected interval according to the leaf structure score and number of the voxel blocks in each connected interval.

[0067] It should be noted that since the point cloud data on the aircraft engine blades is continuous, the blade structure possibility between adjacent voxel blocks is continuous, and the voxel blocks are merged according to the continuity of the blade structure possibility between different voxel blocks. Since the idea of ​​building a model using the RANSAC method is to solve the model containing the least points, the fewer the number of voxel blocks in the connected interval, the more likely it is that the connected interval is a characteristic model interval.

[0068] Specifically, all voxel blocks whose blade structure probability is greater than a preset feature threshold are recorded as feature voxel blocks. The preset feature threshold in this embodiment is 0.7. Other values ​​can be set in other embodiments, and this embodiment does not specifically limit it. Connect the feature voxel block with the feature voxel blocks in the neighborhood to obtain several connected intervals; it should be noted that when there is only one voxel block in the connected interval, the single voxel block belongs to the surface and cannot express the surface information of the entire leaf, so the connected interval with only one voxel block is removed. Then the characteristic model interval score value of the rth connected interval is calculated as follows:

[0069]

[0070] In the formula, Represents the feature model interval score value of the rth connected interval; represents the number of voxel blocks in the rth connected interval; represents the possibility of the leaf structure of the t-th voxel block in the r-th connected interval; Represents the normalization function.

[0071] Represents the number of voxel blocks in the rth connected interval The smaller the number, the greater the possibility that the rth connected interval is the characteristic model interval. The larger the value of , the larger the score value of the rth connected interval as the characteristic model interval.

[0072] Similarly, the feature model interval score of each connected interval is obtained according to the leaf structure score and the number of voxel blocks in each connected interval.

[0073] Step S005: Obtain the thickness of the aircraft engine blade according to the characteristic model interval score value of each connected interval.

[0074] It should be noted that when the RANSAC algorithm is used to fit the surface, the minimum number of point cloud data preset in this embodiment is n. In other embodiments, other values ​​may be set, which are not specifically limited in this embodiment. Since the feature model interval is composed of multiple voxel blocks, and each voxel block constitutes a surface-like feature of an aircraft engine blade, in order to ensure the accuracy of the subsequent RANSAC algorithm fitting, each voxel block requires n point cloud data.

[0075] Specifically, the connected interval with the largest feature model interval score is recorded as the feature model interval, and the RANSAC algorithm is used to denoise the point cloud data of the aircraft engine blade. In this embodiment, the maximum number of iterations of the RANSAC algorithm is set to 20, and the minimum number of internal points is set to 10. Other values ​​can be set in other embodiments, and this embodiment does not specifically limit them. Assuming that there are Q voxel blocks in the feature model interval, the number of point cloud data required to select to build the model using RANSAC is The RANSAC algorithm described in this embodiment is a well-known technology and will not be described in detail in this embodiment.

[0076] Furthermore, in the denoised point cloud data, several parallel sections are selected along the vertical direction of the aircraft engine blade with the length of each voxel block as a step size, and the PCL point cloud clustering algorithm is used to obtain the inner and outer contour lines in each section. The distance between two point cloud data on the inner and outer contour lines of the section passed by a straight line perpendicular to the chord line of the engine blade is recorded as the thickness at each position on the engine blade. The PCL point cloud clustering algorithm described in this embodiment is an existing well-known technology and will not be described in detail in this embodiment.

[0077] On the other hand, the present invention also provides a system for accurately measuring the thickness of aircraft engine blades, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of the aforementioned method for accurately measuring the thickness of aircraft engine blades.

[0078] It should be noted that the The model only shows negative correlation and the output of the constraint model is in In the range, As the input of this model, it can be replaced by other models with the same purpose in specific implementation. The model is used as an example for description without any specific limitation.

[0079] So far, the present invention is completed.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Precise thickness measurement method for aircraft engine blades, It is characterized in that The method comprises the following steps: Obtain the point cloud data of aircraft engine blades and divide it into several voxel blocks; According to the plane angle between each voxel block and other voxel blocks in a preset neighborhood, the leaf structure characteristic values ​​of other voxel blocks in the neighborhood of each voxel block are obtained; The step of obtaining the leaf structure characteristic values ​​of other voxel blocks in the neighborhood of each voxel block according to the plane angle between each voxel block and other voxel blocks in the preset neighborhood includes the following specific steps: With the ath voxel block as the center, a neighborhood of size n*n*n is constructed; the calculation method for the leaf structure eigenvalue of the bth voxel block in the neighborhood of the ath voxel block is: In the formula, represents the leaf structure eigenvalue of the a-th voxel block and the b-th voxel block in the neighborhood; Represents the normal vector of the plane formed by the point cloud data in the ath voxel block, The normal vector of the plane formed by the point cloud data of the bth voxel block in the neighborhood of the ath voxel block; Indicates the number of point cloud data in the a-th voxel block, The number of point cloud data in the bth voxel block in the neighborhood of the ath voxel block; represents the exponential function with natural exponential as base, It represents the absolute value function; Constructing a first curve and a second curve according to the leaf structure characteristic value of the layer where each voxel block is located; obtaining the leaf structure score value of each voxel block in the layer where it is located and in the two adjacent layers according to the difference between the leaf structure characteristic values ​​on the first curve and the second curve; The step of obtaining the leaf structure score of each voxel block in the layer and two adjacent layers according to the difference between the leaf structure characteristic values ​​on the first curve and the second curve includes the following specific steps: The layer where the a-th voxel block is located is recorded as the j-th layer, and the leaf structure score of the a-th voxel block in the j-th layer is calculated as follows: In the formula, represents the leaf structure score of the a-th voxel block at the j-th layer; represents the slope of the line connecting the two end points of the first curve of the a-th voxel block in the j-th layer, The slope of the line connecting the two end points of the second curve representing the a-th voxel block in the j-th layer; represents the number of leaf structure eigenvalues ​​of the a-th voxel block on the first curve of the j-th layer; represents the number of leaf structure eigenvalues ​​of the a-th voxel block on the second curve of the j-th layer; represents the mean value of the leaf structure eigenvalue of the a-th voxel block on the first curve of the j-th layer; represents the mean value of the leaf structure feature value of the a-th voxel block on the second curve of the j-th layer; represents the exponential function with natural exponential as base, It represents the absolute value function; According to the numerical difference of the leaf structure score values ​​of the three-layer voxel blocks in the neighborhood of each voxel block, the leaf structure possibility of each voxel block is obtained; The leaf structure possibility of each voxel block is obtained according to the numerical difference of the leaf structure score values ​​of the three-layer voxel blocks in the neighborhood of each voxel block, and the specific formula included is as follows: In the formula, represents the leaf structure possibility of the a-th voxel block; represents the variance of the leaf structure score values ​​of the three-layer voxel blocks in the neighborhood of the a-th voxel block; represents the leaf structure score of the a-th voxel block at the j-th layer; represents the normalization function; is a hyperparameter; According to the leaf structure possibility of each voxel block, some voxel blocks are connected to obtain a number of connected intervals; according to the number of voxel blocks and the leaf structure possibility in each connected interval, the characteristic model interval score value of each connected interval is obtained; The specific steps of obtaining the characteristic model interval score value of each connected interval according to the number of voxel blocks and the possibility of leaf structure in each connected interval include the following: In the formula, Represents the feature model interval score value of the rth connected interval; represents the number of voxel blocks in the rth connected interval; represents the possibility of the leaf structure of the t-th voxel block in the r-th connected interval; represents the normalization function; The characteristic model interval is obtained according to the characteristic model interval score value, and the point cloud data is denoised according to the characteristic model interval to obtain the denoised point cloud data; the denoised point cloud data is calculated to obtain the thickness of the aircraft engine blade at each position.

2. The method for accurately measuring the thickness of an aircraft engine blade according to claim 1, It is characterized in that The step of constructing the first curve and the second curve according to the leaf structure characteristic value of the layer where each voxel block is located includes the following specific steps: The direction from the root of the blade to the tail of the blade is recorded as the vertical direction, and all voxel blocks except those with a blade structure characteristic value of 0 are obtained. The voxel block layer that includes the a-th voxel block and is perpendicular to the vertical direction is recorded as the layer where the a-th voxel block is located. On the layer where the a-th voxel block is located, the curves formed by the blade structure characteristic values ​​of all voxel blocks from the voxel block with the smallest blade structure characteristic value to the voxel block with the largest blade structure characteristic value are recorded as the first curve and the second curve, respectively.

3. The method for accurately measuring the thickness of an aircraft engine blade according to claim 1, It is characterized in that The method of connecting some voxel blocks according to the leaf structure possibility of each voxel block to obtain a plurality of connected intervals includes the following specific steps: Among all voxel blocks, the voxel blocks whose leaf structure probability is greater than a preset feature threshold are recorded as feature voxel blocks, and the feature voxel blocks are connected with feature voxel blocks in the neighborhood to obtain a number of connected intervals.

4. The method for accurately measuring the thickness of an aircraft engine blade according to claim 1, It is characterized in that The step of obtaining a feature model interval according to the feature model interval score value and denoising the point cloud data according to the feature model interval to obtain denoised point cloud data includes the following specific steps: The connected interval with the largest feature model interval score is recorded as the feature model interval. The product of the number of voxel blocks in the feature model interval and the minimum number of point cloud data required by the RANSAC algorithm is used as the minimum number of point cloud data selected by the RANSAC algorithm to build the model. The minimum number of point cloud data selected by the RANSAC algorithm to build the model is denoised to obtain the denoised point cloud data.

5. The method for accurately measuring the thickness of an aircraft engine blade according to claim 2, It is characterized in that The specific steps of calculating the thickness of the aeroengine blade at each position by using the denoised point cloud data are as follows: In the denoised point cloud data, several parallel sections are selected along the vertical direction of the aircraft engine blade with the length of each voxel block as the step size. The PCL point cloud clustering algorithm is used to obtain the inner and outer contour lines in each section. The distance between the two point cloud data on the inner and outer contour lines of the section passed by a straight line perpendicular to the chord line of the engine blade is recorded as the thickness at each position on the engine blade.

6. A system for accurately measuring the thickness of an aircraft engine blade, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the computer program is executed by a processor, the steps of the method for accurately measuring the thickness of an aircraft engine blade as described in any one of claims 1 to 5 are implemented.