Method for reducing test ice shape error

By performing multiple data overlap and processing in the freezing wind tunnel test, the ice-shaped error was successfully reduced, the accuracy and stability of the experiment were improved, and the quantification of ice-shaped errors was provided by using the BLSOM neural network and the Kriging regression model. Description: Solve the problem of large repeatability error in icy wind tunnel tests.

CN119043635AActive Publication Date: 2024-11-29AVIC GENERAL HUANAN AIRCRAFT IND CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411145237.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-29
Estimated Expiration
2044-08-20

AI Technical Summary

Technical Problem

There is a large repeatability error in the freezing wind tunnel test, resulting in significant differences in ice shape in single tests and multiple repeat tests, and there is a lack of effective error evaluation and reduction methods.

Method used

By conducting multiple repetitive tests under the same experimental model and the same icing conditions, point cloud data is obtained using a three-dimensional digital scanner, data overlap and denoising are performed, and clustering algorithms based on BLSOM neural network and Kriging regression model are used. Calculate the two-dimensional average ice shape and its tolerance band to quantitatively describe the target ice shape and reduce errors.

Benefits of technology

It significantly reduces the random error of a single trial, improves the stability and accuracy of the test results, provides a quantitative description of ice-shaped errors, simplifies data processing, and improves the objectivity and scientificity of the analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119043635A_ABST
    Figure CN119043635A_ABST
Patent Text Reader

Abstract

The invention provides a method for reducing test ice shape errors. The method comprises the steps that n times of repeated icing wind tunnel tests are carried out under the same test model and the same icing condition; under the same position reference condition, the obtained n groups of point cloud data of the three-dimensional test ice shape are placed in the same data space, and the n groups of point cloud data are overlapped; projecting the overlapped point cloud data to a two-dimensional plane along the height direction of the model to obtain two-dimensional overlapped point cloud data; clustering the data to obtain a two-dimensional average ice shape; calculating a tolerance zone of the two-dimensional average ice shape; quantitatively describing a target ice shape by using a two-dimensional average ice shape and a tolerance zone combination mode thereof; and carrying out regression interpolation on the feature points on the ice-shaped curve to obtain a smooth parametric ice-shaped curve of which noise points on the numerical ice-shaped curve are filtered out. The three-dimensional test ice shape error can be reduced, the accuracy of the icing wind tunnel test is improved, the cost is lower, the operation is simpler, the theory is clearer, the practicability is higher, and the effect is more remarkable.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of ice shape data processing in an icing wind tunnel test, and in particular to a method for reducing test ice shape errors. Background Art

[0002] At present, the research methods for aircraft icing include flight tests, icing wind tunnel tests and numerical simulations. Under the current industrial technology level, for icing wind tunnel tests of straight segment models, even if the quality of the icing wind tunnel fully meets the requirements of industry standards, the spatial inhomogeneity and temporal instability of its flow field, cloud field and temperature field can still lead to not only obvious differences in ice shape along the model span in a single test, but also obvious differences between ice shapes in multiple repetitive tests, that is, the repeatability error of the icing wind tunnel test is large.

[0003] There are two ways to reduce the error of icing wind tunnel tests: one is to improve the control accuracy of the icing wind tunnel equipment on the flow field, cloud field and temperature field, and improve the accuracy of a single test. This approach involves a wide range of technologies and progresses slowly. The other way is to conduct multiple repetitive tests and then process the repetitive test data to reduce or eliminate the error of the ice shape in the post-processing stage to obtain a test ice shape that is closest to the target ice shape.

[0004] Due to the particularity of icing wind tunnel tests, there is little research in the currently available information on the evaluation methods for the repeatability errors of the tests and the means to reduce the repeatability errors of the test ice shapes. Summary of the invention

[0005] The purpose of the present invention is to provide a method for reducing the error of test ice shape, which can reduce the error of three-dimensional test ice shape and improve the accuracy of icing wind tunnel test, with lower cost, simpler operation, clearer theory, stronger practicality and more significant effect.

[0006] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0007] A method for reducing test ice shape error, the method comprising the following steps:

[0008] Carry out n repetitive icing wind tunnel tests on the same test model and under the same icing conditions;

[0009] At the end of each test, a three-dimensional digital scanner is used to scan the surface of the test ice shape to obtain point cloud data of the three-dimensional test ice shape;

[0010] Under the same position reference condition, the obtained n groups of three-dimensional test ice shape point cloud data are placed in the same data space to achieve the overlapping of the n groups of point cloud data;

[0011] Projecting the overlapped point cloud data onto a two-dimensional plane along the height direction of the model to obtain two-dimensional overlapping point cloud data;

[0012] The two-dimensional overlapping point cloud data are clustered using a clustering algorithm based on the BLSOM neural network to obtain the two-dimensional average ice shape;

[0013] The tolerance band of the two-dimensional average ice shape is calculated using probability statistics methods to provide a quantitative description of the error range of the ice shape;

[0014] The target ice shape is quantitatively described by combining the two-dimensional average ice shape and its tolerance band;

[0015] The Kriging regression model is used to perform regression interpolation on the characteristic points on the ice shape curve, and a smooth parameterized ice shape curve is obtained by filtering out the noise points on the numerical ice shape curve.

[0016] According to a method for reducing test ice shape error provided by the present invention, obtaining point cloud data of a three-dimensional test ice shape includes:

[0017] According to the characteristics of the test ice shape and scanning requirements, set the scanning parameters to ensure that the scanning parameters can capture all important features of the test ice shape;

[0018] Aim the three-dimensional digital scanner at the test ice shape and scan it according to the preset scanning path;

[0019] De-noising the scanned point cloud data. If multiple scans are performed, the point cloud data at different angles and positions are spliced ​​to form a complete three-dimensional test ice shape model.

[0020] The point cloud data are further analyzed and processed to obtain point cloud data of the three-dimensional test ice shape.

[0021] According to a method for reducing test ice shape error provided by the present invention, the overlapping of n groups of point cloud data comprises:

[0022] Perform quality checks on each set of point cloud data to ensure that all point cloud data are in the same coordinate system;

[0023] Rough registration: Select the feature points in the two sets of point cloud data, and obtain a preliminary rotation and translation matrix by calculating the correspondence between these feature points; then, use this matrix to transform one set of point cloud data into the coordinate system of the other set of point cloud to achieve preliminary coincidence;

[0024] Fine registration: Based on the rough registration, the ICP algorithm iteratively finds the closest point pair and optimizes the rotation and translation matrix to minimize the error between the two sets of point clouds.

[0025] Merge the registered point cloud data into the same dataset;

[0026] De-duplicate and filter the merged point cloud data to improve data quality;

[0027] A three-dimensional model of overlapping point cloud data is formed.

[0028] According to a method for reducing test ice shape error provided by the present invention, obtaining two-dimensional overlapping point cloud data includes:

[0029] Confirm the projection direction, which is along the height direction of the model;

[0030] Load the 3D model of the overlapped and processed point cloud data;

[0031] Determine a two-dimensional plane as the projection target, traverse each point in the point cloud data, and convert its coordinates from three-dimensional (X, Y, Z) to two-dimensional (X, Y);

[0032] The projected two-dimensional overlapping point cloud data is filtered and denoised to remove noise or abnormal points generated during the projection process.

[0033] According to a method for reducing the error of experimental ice shape provided by the present invention, when calculating the two-dimensional average ice shape, the method comprises the following steps:

[0034] Define the input dataset X of two-dimensional overlapping point cloud data;

[0035] When using the clustering algorithm based on the BLSOM neural network, a one-dimensional linear array is selected as the topological structure of the neural network, and the number of neurons is determined according to the specific ice shape, and the number of neurons M is determined, and the number of learning times is K;

[0036] The principal component analysis method PCA is used to determine the initial weight vector of the neuron, which is expressed as the following formula:

[0037] b i =X av +5×σ1×T1×(iM / 2) / M

[0038] Among them, b i is the weight vector of the i-th neuron, X av is the mean vector of the data set X, σ1 is the standard deviation of the first principal component determined by PCA, and T1 is the eigenvector of the first principal component determined by PCA.

[0039] According to a method for reducing the experimental ice shape error provided by the present invention, data points in a data set X are distributed to adjacent neurons to form M point clusters, and the neuron weight vector is updated according to the following formula:

[0040] bi new =b i +α(r)×(X i av -b i )

[0041] Among them, X i av represents the average position vector of all point cloud data assigned to the i-th neuron, and α(r) represents the r-th learning efficiency, which is determined by the following formula:

[0042] α(r)=max{0.01,0.06×(1-r / 1000)}

[0043] After K times of learning, the final weight vector of M neurons is obtained, and each neuron is located at the center of its point cluster;

[0044] The neurons are connected by straight line segments according to the topological structure of the neurons themselves to form a manifold β, which is called the two-dimensional average ice shape of the overlapping point cloud data along the height direction.

[0045] According to a method for reducing the error of experimental ice shape provided by the present invention, when calculating the tolerance band of the two-dimensional average ice shape, the point cloud data is clustered using a clustering algorithm based on the BLSOM neural network to form a series of point clusters, each cluster is represented by a winning neuron at the center, so the standard deviation of the data points in the cluster relative to the winning neuron represents the discreteness of the data points, which is used to represent the uncertainty of the winning neuron.

[0046] According to a method for reducing the experimental ice shape error provided by the present invention, it is assumed that the manifold β is a first-order manifold in a two-dimensional space, which is characterized in that at each neuron b n At , the local slope of the manifold β is equal to the slope of the two nearest neurons b n-1 and b n+1 The slope of the determined straight line; assuming that all deviations of the point cloud data from the manifold β are perpendicular to the manifold β, that is, any point cloud data point x j The deviation from the manifold β is equal to x j In its winning neuron b n The projected height of the normal.

[0047] According to a method for reducing test ice shape error provided by the present invention, a single point cloud data point x is defined. j 、x j The winning neuron b n , and b on the manifold β n Two adjacent neurons b n-1 and b n+1 .

[0048] Data point x j Relative manifold β in neuron b n Position deviation dN x / j The calculation formula is as follows:

[0049] dN x / j =h×cos(γ x / j -x b / n )

[0050] Where h = [(x x / j -x b / n ) 2 +(y x / j -y b / n ) 2 ] 1 / 2 Represents data point x j (x x / j ,y x / j ) and neuron b n (x b / n ,y b / n ) in a straight line.

[0051] According to a method for reducing test ice shape error provided by the present invention, the following steps are further performed:

[0052] When calculating the two-dimensional average ice shape, the curvature of all data points on the ice shape curve and the curvature change between adjacent data points are calculated, and m feature points are selected on the numerical ice shape curve according to the magnitude relationship of the curvature change to obtain the coordinate point set of all feature points;

[0053] Select a feature points at all locations where the curvature change is less than a given threshold;

[0054] Select b feature points at all locations where the curvature change is greater than a given threshold;

[0055] The regression model is used to perform regression interpolation on the characteristic points on the ice shape curve, and the ice shape curve is used as the input object. The regression Kriging model is used to perform regression interpolation to obtain the interpolation model;

[0056] The interpolation results at all feature point positions are output to obtain a smooth parameterized ice shape curve by filtering out the noise points on the numerical ice shape curve.

[0057] It can be seen that compared with the prior art, the present invention has the following beneficial effects:

[0058] 1. The present invention can significantly reduce the random error of a single test and improve the stability and accuracy of the test results by conducting n repeated tests under the same test model and the same icing conditions. The average result of multiple tests is more representative of the actual situation than that of a single test.

[0059] 2. The present invention uses a three-dimensional digital scanner to perform high-precision scanning of the test ice shape. The acquired point cloud data can fully reflect the three-dimensional characteristics of the ice shape, including shape, size and surface details, providing a rich data basis for subsequent analysis.

[0060] 3. The present invention overlaps n groups of point cloud data under the same position reference condition and simplifies the complexity of data processing by projecting them onto a two-dimensional plane, while retaining key height information, making data analysis and processing more efficient and intuitive.

[0061] 4. The present invention uses a clustering algorithm based on the BLSOM neural network to process two-dimensional overlapping point cloud data, which can automatically identify and extract the most representative ice shape features, avoid the subjectivity of human judgment, and improve the objectivity of the analysis.

[0062] 5. The present invention calculates the tolerance band of the two-dimensional average ice shape through a probabilistic statistical method, providing a quantitative description of the error range of the ice shape. This quantitative error assessment method helps to more accurately understand the variability and uncertainty of the ice shape, and provides a scientific basis for subsequent design, analysis and optimization.

[0063] 6. The present invention uses a combination of a two-dimensional average ice shape and its tolerance band to quantitatively describe the target ice shape, thereby achieving standardization and normalization of the ice shape description.

[0064] The present invention also provides an electronic device, comprising:

[0065] a memory storing computer executable instructions;

[0066] a processor configured to execute the computer executable instructions,

[0067] Wherein, when the computer executable instructions are executed by the processor, the steps of any one of the above methods for reducing the test ice shape error are implemented.

[0068] The present invention also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is used to implement the steps of any one of the above methods for reducing the test ice shape error.

[0069] It can be seen that the present invention also provides an electronic device and a storage medium for reducing the test ice shape error, which includes: one or more memories and one or more processors. The memory is used to store program codes and intermediate data generated during program operation, storage of model output results, and storage of models and model parameters; the processor is used for processor resources occupied by code operation and multiple processor resources occupied when training models.

[0070] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a two-dimensional ice shape map of 4 positions obtained by using the "ice drawing method" in the prior art.

[0072] Figure 2 The present invention is a flow chart of a method for reducing test ice shape error.

[0073] Figure 3 It is a principle diagram for defining the point cloud projection distance in an embodiment of a method for reducing test ice shape error of the present invention.

[0074] Figure 4 In the embodiment of the present invention, a method for reducing the error of the test ice shape is used to define the angle Schematic diagram.

[0075] Figure 5 It is a schematic diagram of a two-dimensional average ice shape determined by BLSOM in an embodiment of a method for reducing experimental ice shape errors of the present invention.

[0076] Figure 6 It is a schematic diagram of the average ice shape and its tolerance band provided in an embodiment of a method for reducing test ice shape error of the present invention.

[0077] Figure 7 This is a diagram showing the comparison of the number of neurons in an embodiment of a method for reducing experimental ice shape errors of the present invention.

[0078] Figure 8 It is a schematic diagram of a three-dimensional overlapping point cloud and a two-dimensional overlapping point cloud projected onto an XY plane provided in an embodiment of a method for reducing test ice shape errors of the present invention.

[0079] Fig. 9 It is a schematic diagram of an example of the average ice shape and its tolerance zone of a two-dimensional overlapping point cloud provided in an embodiment of a method for reducing test ice shape error of the present invention. DETAILED DESCRIPTION

[0080] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0081] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0082] In this embodiment, the three-dimensional test ice shape suitable for the present invention has a geometric size that is significantly larger than sandpaper ice and has obvious single ice angle or double ice angle characteristics. The corresponding test model is an equal straight segment model, which is installed horizontally or vertically in the icing wind tunnel test section, and the height direction of the model is perpendicular to the incoming flow velocity direction.

[0083] See also Figures 2 to 9 The present invention provides a method for reducing the test ice shape error, the method comprising the following steps:

[0084] Step S1, performing n (n≥2, the larger n is, the more significant the effect is) repetitive icing wind tunnel tests on the same test model and under the same icing conditions.

[0085] Step S2: at the end of each single test, a three-dimensional digital scanner is used to scan the surface of the test ice shape to obtain point cloud data of the three-dimensional test ice shape.

[0086] Step S3, under the same position reference condition, the obtained n groups of point cloud data of three-dimensional test ice shapes are placed in the same data space to achieve overlapping of the n groups of point cloud data.

[0087] Step S4, projecting the overlapped point cloud data onto a two-dimensional plane along the model height direction to obtain two-dimensional overlapping point cloud data.

[0088] Step S5, clustering the two-dimensional overlapping point cloud data using a clustering algorithm based on a BLSOM neural network to obtain a two-dimensional average ice shape.

[0089] Step S6, using a probability statistics method to calculate the tolerance band of the two-dimensional average ice shape; where it is assumed that the position deviation dN of the point cloud in the cluster relative to the winning neuron in the cluster x / j It follows a normal distribution, and the 95% confidence interval of the winning neuron position is [μ-1.96σ,μ+1.96σ], where μ is the point cloud position deviation dN in the cluster. x / j The average value of σ is the point cloud position deviation dN in the cluster. x / j The standard deviation of .

[0090] Step S7, using a combination of a two-dimensional average ice shape and a tolerance band to quantitatively describe the target ice shape;

[0091] Step S8, using the Kriging regression model to perform regression interpolation on the characteristic points on the ice shape curve, to obtain a smooth parameterized ice shape curve by filtering out the noise points on the numerical ice shape curve.

[0092] It can be seen that the method provided in this embodiment is to reduce the test error by taking the average value through multiple repetitive tests. In terms of specific operation, the overlapping ice shapes of the repetitive test ice shapes are first obtained through appropriate equipment and software, and then the two-dimensional average ice shape of the overlapping ice shapes is obtained by using a clustering algorithm based on the BLSOM neural network. Then, the tolerance band of the two-dimensional average ice shape is calculated using a probability statistics method. Using the combination of the two-dimensional average ice shape and its tolerance band to represent the target ice shape can effectively reduce the error of the test ice shape.

[0093] In the above step S2, the point cloud data of the three-dimensional test ice shape is obtained, including:

[0094] According to the characteristics of the test ice shape and scanning requirements, set the scanning parameters to ensure that the scanning parameters can capture all important features of the test ice shape;

[0095] Aim the three-dimensional digital scanner at the test ice shape and scan it according to the preset scanning path;

[0096] De-noising the scanned point cloud data. If multiple scans are performed, the point cloud data at different angles and positions are spliced ​​to form a complete three-dimensional test ice shape model.

[0097] The point cloud data are further analyzed and processed to obtain point cloud data of the three-dimensional test ice shape.

[0098] In the above step S3, overlapping of n groups of point cloud data is achieved, including:

[0099] Perform quality checks on each set of point cloud data to ensure that all point cloud data are in the same coordinate system;

[0100] Rough registration: Select the feature points in the two sets of point cloud data, and obtain a preliminary rotation and translation matrix by calculating the correspondence between these feature points; then, use this matrix to transform one set of point cloud data into the coordinate system of the other set of point cloud to achieve preliminary coincidence;

[0101] Fine registration: Based on the rough registration, the ICP algorithm iteratively finds the closest point pair and optimizes the rotation and translation matrix to minimize the error between the two sets of point clouds.

[0102] Merge the registered point cloud data into the same dataset;

[0103] De-duplicate and filter the merged point cloud data to improve data quality;

[0104] A three-dimensional model of overlapping point cloud data is formed.

[0105] In the above step S4, obtaining two-dimensional overlapping point cloud data includes:

[0106] Confirm the projection direction, which is along the height direction of the model;

[0107] Load the 3D model of the overlapped and processed point cloud data;

[0108] Determine a two-dimensional plane as the projection target, traverse each point in the point cloud data, and convert its coordinates from three-dimensional (X, Y, Z) to two-dimensional (X, Y);

[0109] The projected two-dimensional overlapping point cloud data is filtered and denoised to remove noise or abnormal points generated during the projection process.

[0110] In this embodiment, the above-mentioned BLSOM neural network technology is an important method in unsupervised learning methods, which can be used for clustering, high-dimensional visualization, data compression, feature extraction and other purposes. In the BLSOM algorithm, the initial value of the neural network array is determined by principal component analysis (PCA), and the mapping result does not depend on the order of input data in the learning process. Figure 3 As shown, b n Represents a neuron.

[0111] When the BLSOM neural network technology is applied to ice shape description, a one-dimensional linear array is selected as the topological structure of the neural network, and the number of neurons needs to be determined according to the specific ice shape: too few neurons will lead to the loss of ice shape features, and too many neurons will lead to messy ice shape curves.

[0112] In this embodiment, when calculating the two-dimensional average ice shape, the following steps are included:

[0113] Define the input dataset X of two-dimensional overlapping point cloud data;

[0114] When using the BLSOM neural network, a one-dimensional linear array is selected as the topological structure of the neural network, the number of neurons is determined according to the specific ice shape, the number of neurons M is determined, and the number of learning times is K. The K value can be defined according to user needs. In this embodiment, K is 1000 but is not limited to 1000.

[0115] The principal component analysis method PCA is used to determine the initial weight vector of the neuron, which is expressed as the following formula:

[0116] b i =X av +5×σ1×T1×(iM / 2) / M

[0117] Among them, b i is the weight vector of the i-th neuron, X av is the mean vector of the data set X, σ1 is the standard deviation of the first principal component determined by PCA, and T1 is the eigenvector of the first principal component determined by PCA.

[0118] Assign the data points in the data set X to adjacent neurons to form M point clusters, and update the neuron weight vector according to the following formula:

[0119] b i new =b i +α(r)×(X i av -b i )

[0120] Among them, X i av represents the average position vector of all point cloud data assigned to the i-th neuron, and α(r) represents the r-th learning efficiency, which is determined by the following formula:

[0121] α(r)=max{0.01,0.06×(1-r / 1000)}

[0122] After K times of learning, the final weight vector of M neurons is obtained, and each neuron is located at the center of its point cluster;

[0123] like Figure 5 As shown, neurons are connected by straight line segments according to their own topological structure to form a manifold β, which is called the two-dimensional average ice shape of three-dimensional sandpaper ice along the height direction.

[0124] In this embodiment, when calculating the tolerance band of the two-dimensional average ice shape, a clustering algorithm based on the BLSOM neural network is used to cluster the point cloud data to form a series of point clusters, each cluster is represented by a winning neuron at the center. Therefore, the standard deviation of the data points in the cluster relative to the winning neuron represents the discreteness of the data points, which is used to represent the uncertainty of the winning neuron.

[0125] Assume that the manifold β is a first-order manifold in two-dimensional space, characterized by the fact that at each neuron b n At , the local slope of the manifold β is equal to the slope of the two nearest neurons b n-1 and b n+1 The slope of the determined straight line; assuming that all deviations of the point cloud data from the manifold β are perpendicular to the manifold β, that is, any point cloud data point x j The deviation from the manifold β is equal to x j In its winning neuron b n The projected height of the normal.

[0126] like Figure 3 and Figure 4 As shown, Figure 3 Shows a single point cloud data point x j 、x j The winning neuron b n , and b on the manifold β n Two adjacent neurons b n-1 and b n+1 In this paper, it is assumed that the manifold β is a first-order manifold in two-dimensional space, characterized by the fact that at each neuron b n At , the local slope of the manifold β is equal to the slope of the two nearest neurons b n-1 and b n+1 The slope of the line determined by the point cloud data is assumed to be perpendicular to the manifold β. In other words, any point cloud data point x j The deviation from the manifold β is equal to x j In its winning neuron b n The projected height of the normal.

[0127] exist Figure 3 middle, Represents vector b n-1 b n+1 The angle with the X axis, that is, the manifold β at neuron b n The tangent angle at Figure 4 As shown, due to When the vector b n-1 b n+1 When it is located in the first and fourth quadrants of the XY coordinate plane, it can be directly calculated by the inverse tangent function When the vector b n-1 b n+1 When located in the second and third quadrants of the XY coordinate plane, At this time, you need to add π to the value of the inverse tangent function to get the correct value. therefore, The complete expression of is the formula:

[0128]

[0129] Depend on Figure 3 It can be seen that the manifold β is n Normal angle at Data point x j Relative neuron b n The direction angle is Based on The same considerations, The complete expression of is the formula:

[0130]

[0131] So the data point x j Relative manifold β in neuron b n Normal projection height at The calculation formulas are:

[0132] dN x / j =h×cos(γ x / j -x b / n )

[0133] Where h = [(x x / j -x b / n ) 2 +(y x / j -y b / n ) 2 ] 1 / 2 , represents the data point x j With neuron b n The straight-line distance.

[0134] In this embodiment, it is assumed that the position deviation dN of the point cloud in the “cluster” relative to the winning neuron in the “cluster” is x / j The 95% confidence interval (95% CI) of the winning neuron position is [μ-1.96σ, μ+1.96σ], where μ is the point cloud position deviation dN in the “cluster”. x / j The average value of , σ is the position deviation dN of the point cloud in the “cluster” x / j The standard deviation of .

[0135] like Figure 6 As shown, in Figure 6 2 shows a schematic diagram of the two-dimensional average ice shape calculated by the present invention and its 95% probability tolerance band. The two-dimensional average ice shape reflects the overall trend of the three-dimensional test ice shape, and the tolerance band reflects the uncertainty of the two-dimensional average ice shape.

[0136] When calculating the two-dimensional average ice shape, the curvature of all data points on the ice shape curve and the curvature change between adjacent data points are calculated, and m feature points are selected on the numerical ice shape curve according to the magnitude relationship of the curvature change to obtain the coordinate point set of all feature points;

[0137] Select a feature points at all locations where the curvature change is less than a given threshold;

[0138] Select b feature points at all locations where the curvature change is greater than a given threshold.

[0139] The regression model is used to perform regression interpolation on the characteristic points on the ice shape curve, and the ice shape curve is used as the input object. The regression Kriging model is used to perform regression interpolation to obtain the interpolation model;

[0140] The interpolation results at all feature point positions are output to obtain a smooth parameterized ice shape curve by filtering out the noise points on the numerical ice shape curve.

[0141] In this embodiment, the interpolation model specifically includes: converting the ice shape curve data point set into a polar coordinate data point set; and performing a Kriging model regression using the polar coordinate data point set to obtain an interpolation model.

[0142] In this embodiment, filtering out noise points on the ice shape curve to form a smooth parameterized ice shape curve specifically includes:

[0143] The coordinate point set of m feature points is converted into the corresponding polar coordinate point set; each polar coordinate is substituted into the interpolation model to obtain a new polar coordinate point set; it is converted into a coordinate point set in a Cartesian coordinate system to obtain a parameterized ice shape curve with noise filtered out.

[0144] In summary, this embodiment uses the Romer Absolute scanner to scan the three-dimensional test ice shape, obtains the point cloud data of the three-dimensional test ice shape, and then projects the three-dimensional point cloud data onto the XY plane along the model height direction to form two-dimensional point cloud data. The clustering algorithm based on the BLSOM neural network is used to "cluster" the two-dimensional point cloud data, and the "one-dimensional linear array" is selected as the topological structure of the neuron. Figure 7 As shown, Figure 7 In (a), the ice shape lines are clear, but the appearance of the upper and lower ice corners is not well captured; Figure 7 In (b), the ice shape lines are clearer, and the appearance of the upper and lower ice corners is better captured; Figure 7 The ice shape in (c) has the best capture effect on the upper and lower ice angles, but due to the large number of neurons, the ice shape lines are messy. Therefore, comprehensively speaking, a neural network with 30 neurons is selected to describe the ice shape. Then the two-dimensional average ice shape is calculated. Finally, the tolerance band of the two-dimensional average ice shape is calculated. Among them, Figure 8 and Fig. 9 As shown, Figure 8 is a schematic diagram of a three-dimensional overlapping point cloud provided in this embodiment and a two-dimensional overlapping point cloud projected onto an XY plane, Fig. 9 An example graph of the 2D average ice shape and its tolerance band is shown.

[0145] In addition, there are many burrs on the ice shape contour curve of the icing test, that is, there is "noise" in the curve data. After regression Kriging interpolation, the output parameterized ice shape curve well retains the main characteristics of the ice shape contour curve of the icing test, successfully filters out the noise of the curve data, and obtains a parameterized ice shape curve with smooth curvature.

[0146] In summary, this embodiment can significantly reduce the random error of a single test and improve the stability and accuracy of the test results by conducting n repetitive tests under the same test model and the same icing conditions. The average results of multiple tests are more representative of the actual situation than a single test. This embodiment uses a three-dimensional digital scanner to perform high-precision scanning of the test ice shape, and the acquired point cloud data can reflect the three-dimensional characteristics of the ice shape in detail, including shape, size and surface details, providing a rich data basis for subsequent analysis. This embodiment overlaps n groups of point cloud data under the same position reference conditions, and by projecting them onto a two-dimensional plane, it simplifies the complexity of data processing, while retaining key height information, making data analysis and processing more efficient and intuitive. This embodiment uses a clustering algorithm based on a BLSOM neural network to process two-dimensional overlapping point cloud data, which can automatically identify and extract the most representative ice shape features, avoid the subjectivity of human judgment, and improve the objectivity of the analysis. This embodiment calculates the tolerance band of the two-dimensional average ice shape by a probability statistical method, and provides a quantitative description of the error range of the ice shape. This quantitative error evaluation method helps to more accurately understand the variability and uncertainty of the ice shape, and provides a scientific basis for subsequent design, analysis and optimization. This embodiment uses a combination of a two-dimensional average ice shape and its tolerance band to quantitatively describe the target ice shape, and realizes the standardization and normalization of the ice shape description.

[0147] In one embodiment, an electronic device is provided, which may be a server. The computer device includes a processor, a memory, and a network interface connected via a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the electronic device is used to store data. The network interface of the electronic device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for reducing the test ice shape error is implemented.

[0148] Those skilled in the art will understand that the electronic device structure shown in this embodiment is only a partial structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device may include more or fewer components than those shown in this embodiment, or combine certain components, or have a different arrangement of components.

[0149] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0150] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0151] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0152] It can be seen that this embodiment also provides an electronic device and a storage medium for reducing the test ice shape error, which includes: one or more memories and one or more processors. The memory is used to store program codes and intermediate data generated during program operation, storage of model output results, and storage of models and model parameters; the processor is used for processor resources occupied by code operation and multiple processor resources occupied when training models.

[0153] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0154] The above-mentioned embodiments are only preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantial changes and substitutions made by technicians in this field on the basis of the present invention shall fall within the scope of protection required by the present invention.

Claims

1. A method for reducing test ice shape error, characterized in that: The Method The following steps are involved: Carry out n repetitive icing wind tunnel tests on the same test model and under the same icing conditions; At the end of each test, a three-dimensional digital scanner is used to scan the surface of the test ice shape to obtain point cloud data of the three-dimensional test ice shape; Under the same position reference condition, the obtained n groups of three-dimensional test ice shape point cloud data are placed in the same data space to achieve the overlapping of the n groups of point cloud data; Projecting the overlapped point cloud data onto a two-dimensional plane along the height direction of the model to obtain two-dimensional overlapping point cloud data; The two-dimensional overlapping point cloud data are clustered using a clustering algorithm based on the BLSOM neural network to obtain the two-dimensional average ice shape; The tolerance band of the two-dimensional average ice shape is calculated using probability statistics methods to provide a quantitative description of the error range of the ice shape; The target ice shape is quantitatively described by combining the two-dimensional average ice shape and its tolerance band; The Kriging regression model is used to perform regression interpolation on the characteristic points on the ice shape curve, and a smooth parameterized ice shape curve is obtained by filtering out the noise points on the numerical ice shape curve.

2. The method according to claim 1, characterized in that The step of obtaining point cloud data of a three-dimensional test ice shape comprises: According to the characteristics of the test ice shape and scanning requirements, set the scanning parameters to ensure that the scanning parameters can capture all important features of the test ice shape; Aim the three-dimensional digital scanner at the test ice shape and scan it according to the preset scanning path; De-noising the scanned point cloud data. If multiple scans are performed, the point cloud data at different angles and positions are spliced ​​to form a complete three-dimensional test ice shape model. The point cloud data are further analyzed and processed to obtain point cloud data of the three-dimensional test ice shape.

3. The method according to claim 1, characterized in that The implementation of overlapping n groups of point cloud data includes: Perform quality checks on each set of point cloud data to ensure that all point cloud data are in the same coordinate system; Rough registration: Select the feature points in the two sets of point cloud data, and obtain a preliminary rotation and translation matrix by calculating the correspondence between these feature points; then, use this matrix to transform one set of point cloud data into the coordinate system of the other set of point cloud to achieve preliminary coincidence; Fine registration: Based on the rough registration, the ICP algorithm iteratively finds the closest point pair and optimizes the rotation and translation matrix to minimize the error between the two sets of point clouds. Merge the registered point cloud data into the same dataset; De-duplicate and filter the merged point cloud data to improve data quality; A three-dimensional model of overlapping point cloud data is formed.

4. The method according to claim 3, characterized in that The obtaining of two-dimensional overlapping point cloud data comprises: Confirm the projection direction, which is along the height direction of the model; Load the 3D model of the overlapped and processed point cloud data; Determine a two-dimensional plane as the projection target, traverse each point in the point cloud data, and convert its coordinates from three-dimensional (X, Y, Z) to two-dimensional (X, Y); The projected two-dimensional overlapping point cloud data is filtered and denoised to remove noise or abnormal points generated during the projection process.

5. The method according to claim 1, characterized in that: When calculating the two-dimensional average ice shape, the following steps are included: Define the input dataset X of two-dimensional overlapping point cloud data; When using the clustering algorithm based on the BLSOM neural network, a one-dimensional linear array is selected as the topological structure of the neural network, and the number of neurons is determined according to the specific ice shape, and the number of neurons M is determined, and the number of learning times is K; The principal component analysis method PCA is used to determine the initial weight vector of the neuron, which is expressed as the following formula: b i =X av +5×σ1×T1×(i-M / 2) / M Among them, b i is the weight vector of the i-th neuron, X av is the mean vector of the data set X, σ1 is the standard deviation of the first principal component determined by PCA, and T1 is the eigenvector of the first principal component determined by PCA.

6. The method according to claim 5, characterized in that: Assign the data points in the data set X to adjacent neurons to form M point clusters, and update the neuron weight vector according to the following formula: b i new =b i +α(r)×(X i av -b i ) Among them, X i av represents the average position vector of all point cloud data assigned to the i-th neuron, and α(r) represents the r-th learning efficiency, which is determined by the following formula: α(r)=max{0.01,0.06×(1-r / 1000)} After K times of learning, the final weight vector of M neurons is obtained, and each neuron is located at the center of its point cluster; The neurons are connected by straight line segments according to the topological structure of the neurons themselves to form a manifold β, which is called the two-dimensional average ice shape of the overlapping point cloud data along the height direction.

7. The method according to claim 6, characterized in that: When calculating the tolerance band of the two-dimensional average ice shape, the point cloud data is clustered using a clustering algorithm based on the BLSOM neural network to form a series of point clusters. Each cluster is represented by a winning neuron at the center. Therefore, the standard deviation of the data points in the cluster relative to the winning neuron represents the discreteness of the data points, which is used to represent the uncertainty of the winning neuron.

8. The method according to claim 6, characterized in that: Assume that the manifold β is a first-order manifold in two-dimensional space, characterized by the fact that at each neuron b n At , the local slope of the manifold β is equal to the slope of the two nearest neurons b n-1 and b n+1 The slope of the determined straight line; assuming that all deviations of the point cloud data from the manifold β are perpendicular to the manifold β, that is, any point cloud data point x j The deviation from the manifold β is equal to x j In its winning neuron b n The projected height of the normal.

9. The method according to claim 6, characterized in that: Define a single point cloud data point x j 、x j The winning neuron b n , and b on the manifold β n Two adjacent neurons b n-1 and b n +1 . Data point x j Relative manifold β in neuron b n Position deviation dN x / j The calculation formula is formula (4): dN x / j =h×cos(γ x / j -x b / n (4) Where h = [(x x / j -x b / n ) 2 +(y x / j -y b / n ) 2 ] 1 / 2 Represents data point x j (x x / j ,y x / j ) and neuron b n (x b / n ,y b / n ) in a straight line.

10. The method according to claim 9, characterized in that Also execute: When calculating the two-dimensional average ice shape, the curvature of all data points on the ice shape curve and the curvature change between adjacent data points are calculated, and m feature points are selected on the numerical ice shape curve according to the magnitude relationship of the curvature change to obtain the coordinate point set of all feature points; Select a feature points at all locations where the curvature change is less than a given threshold; Select b feature points at all locations where the curvature change is greater than a given threshold; The regression model is used to perform regression interpolation on the characteristic points on the ice shape curve, and the ice shape curve is used as the input object. The regression Kriging model is used to perform regression interpolation to obtain the interpolation model; The interpolation results at all feature point positions are output to obtain a smooth parameterized ice shape curve by filtering out the noise points on the numerical ice shape curve.

Citation Information

Patent Citations

  • Wind tunnel model 3D ice shape on-line measuring method based on three-dimensional ultrasonic imaging technology

    CN106248340A

  • Parameterization method of ice shape of ice wind tunnel icing test

    CN111368367A

  • Ice shape parameterization method for wind tunnel test

    CN112985753A

  • Complex ice shape measurement error evaluation method

    CN117191334A