An aircraft fastener head population quality evaluation method based on three-dimensional topography clustering statistics

Abstract

The application discloses an aircraft fastener head group quality evaluation method based on three-dimensional topography clustering statistics, comprising: acquiring point cloud data of all fastener heads in a target area; extracting a multi-dimensional topographic feature vector for each fastener head to construct a group feature vector set; processing the feature vector set by using a clustering analysis algorithm to identify a main cluster, an abnormal cluster, an outlier, and corresponding group quality distribution characteristics; based on the group quality distribution characteristics, calculating group quality evaluation indexes including cluster offset degree, cluster dispersion degree, cluster separation degree, and abnormal sample proportion; and based on the group quality evaluation indexes, performing multi-level threshold quality evaluation to realize mode anomaly identification and process deviation early warning. The application realizes fastener head group quality evaluation, can find process deviation trends when single parameters have not yet exceeded the tolerance, and is beneficial to improving the process quality control capability of aircraft fastener installation.

Description

Technical Field

[0001] This invention relates to the field of quality control technology in aircraft manufacturing processes, and in particular to a method for evaluating the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics. Background Technology

[0002] In aircraft manufacturing and maintenance, riveting and fastener installation are critical processes. A modern civil airliner contains hundreds of thousands to millions of rivets and fasteners, and their installation quality directly affects the aircraft's structural strength, aerodynamic performance, and service life. Currently, the inspection of fastener head morphology in aircraft manufacturing mainly relies on manual visual inspection, contact gauge inspection, or automated visual inspection of individual fasteners. These methods all use individual fasteners as inspection units, comparing their head convexity, height, and other geometric parameters with preset tolerances to provide a binary judgment of pass or fail.

[0003] However, traditional techniques have the following shortcomings: First, the inspection perspective is limited to single-piece judgment, failing to reveal quality patterns at the group level. Within a workstation or a joint, even if the individual parameters of all fasteners are within tolerance, their overall distribution may exhibit significant systematic deviations, which is a crucial early warning signal of process out of control. Second, they lack the ability to control the overall process through statistical analysis. Quality problems in batch riveting are often not isolated; factors such as riveting tool wear, riveting force drift, or material batch differences can cause the morphology of fastener heads in a region to exhibit overall deviations or dispersion trends, which are difficult to detect using only individual acceptance judgment. Therefore, there is an urgent need for a new method for fastener head morphology quality assessment that shifts from single-point inspection to group evaluation and from acceptance judgment to process control. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention proposes a group quality assessment method for aircraft fastener heads based on three-dimensional morphology clustering statistics, aiming to overcome the shortcomings of existing technologies that only focus on the morphology detection of individual fasteners and lack the ability to perform group statistical analysis and pattern anomaly identification.

[0005] To achieve the above-mentioned technical objectives, the present invention provides the following technical solution:

[0006] A method for assessing the quality of aircraft fastener head groups based on three-dimensional topographic clustering statistics, specifically including the following steps:

[0007] Acquire the 3D point cloud data of all fastener heads within the target area, preprocess and segment the 3D point cloud data to obtain the independent point cloud of each fastener head after preprocessing.

[0008] Multidimensional topographic features are extracted based on the independent point cloud of each fastener head and the surrounding reference surface, and topographic feature vectors are constructed. The topographic feature vectors of all fastener heads in the target area constitute a set of group feature vectors.

[0009] A baseline model is constructed based on the morphological feature vectors of the heads of historical qualified fasteners;

[0010] After normalizing or standardizing the set of population feature vectors, cluster analysis is performed to obtain the main cluster, abnormal clusters, outliers, and the corresponding population quality distribution characteristics; the abnormal clusters are those whose sample size reaches a preset lower limit and whose cluster center is more than a preset threshold away from the main cluster or the benchmark model.

[0011] Based on the population quality distribution characteristics of the main cluster, anomalous clusters, and outliers, and combined with the benchmark model, population quality assessment indicators are calculated; the population quality assessment indicators include cluster offset (CSI), cluster dispersion (CDI), inter-cluster separation (ISI), and outlier percentage (ASR).

[0012] The group quality assessment index is compared with the benchmark model and the preset group quality threshold to conduct a group quality assessment of all fastener heads in the target area.

[0013] Furthermore, the extraction of multi-dimensional topographic features based on the independent point cloud of each fastener head and the surrounding reference surface, and the construction of the topographic feature vector, specifically involves:

[0014] Calculate the height difference between the fastener head and the surrounding reference surface, and use it as the head concavity / convexity.

[0015] Calculate the diameter of the circumcircle or least-squares fitted circle of the fastener head profile, and use it as the equivalent diameter of the head.

[0016] Calculate the degree of deviation between the fastener head profile and the standard circle, and use this as the head roundness.

[0017] Calculate the offset between the geometric center of the fastener head profile and the center of its least squares fitted circle, and use this offset as the profile eccentricity.

[0018] Calculate the slope distribution characteristics of the transition region at the edge of the fastener head, and use it as the edge slope.

[0019] Calculate the maximum or average value of the difference between the height of the sampling points in the transition area of ​​the fastener head edge and the height of the surrounding reference surface, and use it as the transition height difference between the head and the surrounding reference surface.

[0020] The topographic feature vector of each fastener head consists of at least two topographic features from the following: head concavity / convexity, head equivalent diameter, head roundness, profile eccentricity, edge slope, and transition height difference between the head and the surrounding reference surface.

[0021] The surrounding reference surface is a reference surface obtained by fitting the local skin surface of the target area or the theoretical design surface.

[0022] Furthermore, the specific steps for constructing the benchmark model based on the morphological feature vectors of historical qualified fastener heads are as follows:

[0023] Perform group quality assessment on fastener heads by batch; obtain three-dimensional morphological data of fastener heads from historical batches that have been confirmed to be of qualified quality, extract morphological feature vectors and construct a historical group feature vector set; cluster and clean the historical group feature vector set, remove outliers, and obtain historical main clusters;

[0024] The mean vector and covariance matrix of the historical principal clusters are calculated and used as the baseline model parameters to obtain the baseline model; the formula for the baseline model parameters is expressed as follows:

[0025] ;

[0026] ;

[0027] in, As the main cluster of history, Each sample corresponds to a shape feature vector of a fastener head; for The number of samples in the sample; The mean vector of the baseline model. The covariance matrix of the baseline model;

[0028] The benchmark model is classified and stored according to installation area, fastener specifications, and installation process conditions, and a dynamic update mechanism is adopted: when the preset conditions are met, the benchmark model is updated using the morphological feature vector of the latest qualified batch.

[0029] Furthermore, the cluster offset CSI is obtained by calculating the Mahalanobis distance between the current main cluster center and the reference model center, specifically:

[0030] ;

[0031] ;

[0032] in, As the current primary cluster, Each sample corresponds to a shape feature vector of a fastener head; This is the current mean vector of the main cluster. The mean vector of the baseline model. The covariance matrix of the baseline model; for The number of samples in the cluster; each element of the mean vector is actually obtained independently by averaging the eigenvalues ​​of the corresponding dimension in the morphological feature vectors of all fastener heads within that cluster.

[0033] When the covariance matrix of the benchmark model is singular or non-invertible, for Regularization is performed first, and then it is used in the calculation of cluster offset (CSI).

[0034] Furthermore, the cluster dispersion (CDI) is obtained by calculating the ratio of the trace of the current principal cluster covariance matrix to the trace of the baseline model covariance matrix, specifically:

[0035] ;

[0036] in, The current principal cluster covariance matrix, The covariance matrix of the baseline model, Represents the trace of a matrix.

[0037] Furthermore, the inter-cluster separation index (ISI) is obtained by calculating the Mahalanobis distance between the centers of anomalous clusters and the centers of the baseline model, specifically as follows:

[0038] ;

[0039] in, This is the mean vector of the outlier cluster. This is the current mean vector of the main cluster. The covariance matrix of the baseline model;

[0040] When there are multiple anomalous clusters, calculate the inter-cluster separation degree for each anomalous cluster separately.

[0041] Furthermore, the anomalous sample ratio (ASR) is obtained by calculating the proportion of fasteners not belonging to the main cluster to the total number of fasteners in the target area, specifically as follows:

[0042] ;

[0043] in, This represents the number of fasteners belonging to the abnormal cluster. This represents the number of fasteners that belong to outliers. This represents the total number of fasteners within the target area.

[0044] Furthermore, the step of comparing the group quality assessment index with the benchmark model and the preset group quality threshold to perform a group quality assessment on all fastener heads within the target area specifically involves:

[0045] Preset CSI first threshold CSI second threshold CDI first threshold CDI second threshold ISI threshold and ASR threshold Based on the comparison results between the group quality assessment indicators and the multi-level preset thresholds, the corresponding group quality level is output:

[0046] First determine if ASR exceeds When ASR does not exceed hour:

[0047] If CSI does not exceed And CDI does not exceed Output "Group quality level 1";

[0048] When ASR does not exceed hour:

[0049] If CSI only exceeds Or CDI only exceeds Output "Population Quality Level 2" if CSI exceeds Or CDI exceeds It outputs "Group Quality Level 3" and simultaneously provides process deviation warnings based on the CSI / CDI threshold comparison results.

[0050] When ASR exceeds hour:

[0051] If CSI does not exceed And CDI did not exceed Output "Population Quality Level 3"; if CSI exceeds... And CDI exceeds It outputs "Group Quality Level 4"; at the same time, it performs process deviation warning based on the CSI / CDI threshold comparison results, triggers ISI threshold comparison, and performs pattern anomaly identification based on the ISI threshold comparison results.

[0052] The final output includes population quality assessment results, process deviation warning results, and pattern anomaly identification results.

[0053] More specifically, the process offset early warning based on the CSI / CDI threshold comparison results specifically refers to:

[0054] When CSI only exceeds Or CDI only exceeds The output is "Slight process offset";

[0055] when When this happens, output "process offset";

[0056] when When this happens, output "process offset";

[0057] When CSI exceeds And CDI exceeds The output is "Severe process offset".

[0058] More specifically, the process of identifying pattern anomalies based on ISI threshold comparison results includes:

[0059] If the ASR exceeds the threshold and there is only one anomalous cluster with an ISI exceeding the threshold, output "Single-mode defect exists";

[0060] If the ASR exceeds the threshold and there are at least two anomalous clusters with ISI exceeding the threshold, output "Multi-mode defect exists";

[0061] If the ASR exceeds the threshold and there are no anomalous clusters with ISI exceeding the threshold, then the output is "Occasional Individual Defect".

[0062] Based on the above technical solution, the present invention has at least the following beneficial effects:

[0063] This invention shifts the detection perspective from judging the pass / fail status of a single fastener to a holistic assessment of the consistency and stability of the morphology of a group of fasteners within an entire area, a whole joint, or an entire workstation. This better meets the actual needs of process quality control. It can detect process deviation trends and pattern anomalies in advance by clustering distribution changes before individual parameters exceed tolerances. Furthermore, by establishing a benchmark model and calculating indicators such as cluster offset and cluster dispersion, it quantifies the overall quality status, thereby improving the scientific nature and traceability of the assessment. Attached Figure Description

[0064] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0065] Figure 1 This is a flowchart of a method for assessing the quality of aircraft fastener head groups based on three-dimensional morphology clustering statistics in an embodiment of the present invention;

[0066] Figure 2 This is a schematic diagram of the distribution of multi-dimensional group morphological features in an embodiment of the present invention;

[0067] Figure 3 This is a schematic diagram of calculating the cluster offset CSI in an embodiment of the present invention;

[0068] Figure 4 This is a schematic diagram illustrating the calculation of inter-cluster separation index (ISI) in an embodiment of the present invention;

[0069] Figure 5 This is a flowchart of the population quality assessment in an embodiment of the present invention. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0071] Although the steps in this invention are arranged by reference numerals, this is not intended to limit the order of the steps. Unless the order of the steps is explicitly stated or the execution of a step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" as used herein refers to and covers any and all possible combinations of one or more of the associated listed items.

[0072] This embodiment provides a method for assessing the quality of aircraft fastener head groups based on three-dimensional topographic clustering statistics. The method provided in this embodiment can be executed on a server, computer, mobile phone, or similar computing device. Figure 1 As shown, the method includes the following steps:

[0073] S1. Obtain the three-dimensional point cloud data of all fastener heads in the target area, preprocess and segment the three-dimensional point cloud data to obtain the independent point cloud of each fastener head after preprocessing.

[0074] In a preferred embodiment, step S1 specifically includes:

[0075] Set up structured light equipment and camera equipment to perform three-dimensional shape scanning on all fasteners in the target area (such as a seam, a work station or a skin area) to obtain original three-dimensional point cloud data covering multiple fastener heads.

[0076] To ensure that the structured light equipment accurately captures the surface features of the measured part, it needs to undergo precise calibration and adjustment to eliminate errors such as optical distortion. This includes: synchronizing the structured laser and the camera equipment, adjusting the position and angle to ensure that the light stripe is clearly and consistently imaged on the fastener surface; calibrating the camera equipment to obtain an image that truly reflects the geometric features of the part; and after setting appropriate exposure time and scanning parameters, the structured laser projects the light stripe, the camera equipment captures the image and converts it into point cloud data to complete the high-precision measurement.

[0077] The original point cloud data is then denoised, filtered, and corrected for reference planes. By setting reasonable thresholds, point cloud data within a specific dimensional range is allowed to pass through, while data exceeding the preset range is removed. A point cloud segmentation algorithm is then used to separate each fastener head region from the overall point cloud, resulting in independent point clouds for each fastener head region. This provides accurate and cleaned point cloud data for subsequent processing.

[0078] S2. Extract multi-dimensional topographic features based on the independent point cloud and surrounding reference plane of each fastener head, construct topographic feature vectors, and form a set of group feature vectors composed of the topographic feature vectors of all fastener heads in the target area.

[0079] In a preferred embodiment, step S2 specifically comprises:

[0080] The reference surface obtained by fitting the local skin surface or theoretical design surface of the target area is taken as the surrounding reference surface;

[0081] Calculate the height difference between the fastener head and the surrounding reference surface, as the head concavity / convexity. In this embodiment, the average height difference between the sampling point on the top surface of the head and the corresponding local reference surface is taken as the head concavity / convexity amount. ;

[0082] Calculate the diameter of the circumcircle or least-squares fitted circle of the fastener head profile, and use it as the equivalent diameter of the head. ;

[0083] Calculate the degree of deviation between the fastener head profile and the standard circle, as the head roundness. , ,in and These are the maximum and minimum distances from the contour points to the center of the fitted circle, respectively.

[0084] The offset between the geometric center of the fastener head profile and the center of its least-squares fitted circle is calculated and used as the profile eccentricity. ;

[0085] Calculate the slope distribution characteristics of the transition region at the edge of the fastener head (the annular area from the top surface of the head to the surrounding skin), such as the average slope angle, as the edge slope. ;

[0086] Calculate the maximum or average difference between the height of the sampling points in the transition area of ​​the fastener head edge and the height of the surrounding reference surface, and use this as the transition height difference between the head and the surrounding reference surface. ;

[0087] The topographic feature vector of each fastener head It is composed of at least two morphological features from the following: head concavity / convexity, head equivalent diameter, head roundness, profile eccentricity, edge slope, and the transition height difference between the head and the surrounding reference surface; it can be expressed as: Or other combinations thereof;

[0088] Assuming there are N fasteners within the target area, the set of group feature vectors is as follows: .

[0089] S3. Construct a benchmark model based on the morphological feature vectors of historical qualified fastener heads; provide prior knowledge of normal fluctuation range and a unified comparison reference;

[0090] In a preferred embodiment, step S3 specifically comprises:

[0091] Perform group quality assessment on fastener heads by batch; obtain three-dimensional morphological data of fastener heads from historical batches that have been confirmed to be of qualified quality, extract morphological feature vectors and construct a historical group feature vector set; cluster and clean the historical group feature vector set, remove outliers, and obtain historical main clusters;

[0092] The mean vector and covariance matrix of the historical principal clusters are calculated and used as the baseline model parameters to obtain the baseline model; the formula for the baseline model parameters is expressed as follows:

[0093] ;

[0094] ;

[0095] in, As the main cluster of history, Each sample corresponds to a shape feature vector of a fastener head; for The number of samples in the sample; The mean vector of the baseline model. The covariance matrix of the baseline model;

[0096] The benchmark model is categorized and stored according to installation area, fastener specifications, and installation process conditions, and a dynamic update mechanism is adopted: when preset conditions are met, the benchmark model is updated using the morphological feature vector of the latest qualified batch (e.g., updating the mean vector using an exponentially weighted moving average, and then updating the covariance matrix). In this embodiment, the preset conditions are: the group quality assessment indicators of multiple consecutive batches (e.g., 10 batches) are all within the normal range, with no abnormal warnings.

[0097] S4. After normalizing or standardizing the set of group feature vectors, cluster analysis is performed. In this embodiment, the DBSCAN method is preferred. Density connectivity is defined by two parameters: the neighborhood radius ε and the minimum number of core point samples MinPts. Figure 2 As shown, the data is divided into main clusters, outlier clusters, and outliers, and the corresponding population quality distribution characteristics are obtained. The population quality distribution characteristics include the mean vector, covariance matrix, and number of samples within a single cluster. For outliers, this refers to the number of outlier samples.

[0098] The main cluster can be defined as the cluster with the highest proportion of samples and that meets the density connectivity condition; the abnormal cluster is the cluster whose sample number reaches a preset lower limit and whose cluster center is more than a preset threshold away from the main cluster or the benchmark model; the outlier can be defined as a sample that does not meet the core point or boundary point condition; for normal riveting batches, the morphological features are highly consistent and the clustering results are usually a single compact main cluster; when there is a process deviation or local anomaly, abnormal clusters and / or outliers that deviate from the main cluster will appear.

[0099] Besides the DBSCAN method, cluster analysis can also employ other algorithms capable of recognizing clusters of arbitrary shapes without requiring a pre-defined number of clusters, such as hierarchical clustering, spectral clustering, or a structure-aware multi-subspace granular-spherical clustering framework. After clustering is complete, the principal clusters are... It contains most normal fasteners; if there are deviations from the sample, it includes at least one anomalous cluster. and / or set of outliers .

[0100] In this embodiment, the concept of "abnormal clusters" is introduced to identify local batch anomalies: an abnormal cluster is a dense group of multiple abnormal individuals with similar characteristics but deviating from the main cluster as a whole. Unlike outliers (isolated abnormal individuals), abnormal clusters reflect a batch of fasteners with abnormal installation quality caused by systematic deviations in a specific time period, specific workstation, or specific tool. This "clustered anomaly" is a key signal of process quality problems. By distinguishing between "main cluster + outlier" (occasional individual defects) and "main cluster + abnormal cluster" (local pattern defects), different directions of troubleshooting suggestions can be provided for the field - the former points to accidental factors (such as local foreign objects), and the latter points to systematic factors.

[0101] S5. Based on the population quality distribution characteristics of the main cluster, anomalous clusters, and outliers, and in conjunction with the benchmark model, calculate the population quality assessment index; the population quality assessment index includes cluster offset (CSI), cluster dispersion (CDI), inter-cluster separation (ISI), and outlier percentage (ASR).

[0102] In a preferred embodiment, the cluster offset CSI is obtained by calculating the Mahalanobis distance between the current master cluster center and the reference model center, specifically:

[0103] ;

[0104] ;

[0105] in, As the current primary cluster, Each sample corresponds to a shape feature vector of a fastener head; This is the current mean vector of the main cluster. The mean vector of the baseline model. The covariance matrix of the baseline model; for The number of samples in the mean vector; each element of the mean vector is actually obtained independently by taking the mean of the eigenvalues ​​of the corresponding dimension in the morphological feature vectors of all fastener heads in the cluster.

[0106] When the covariance matrix of the benchmark model is singular or non-invertible, for Regularization is performed first, and then the result is used in the calculation of the cluster offset (CSI). In this embodiment, the regularization process is represented as follows: ,in For regularization parameters, for The identity matrix, is the dimension of the shape feature vector.

[0107] Reference Figure 3 , Figure 3 This is a schematic diagram of calculating the cluster offset CSI in an embodiment of the present invention;

[0108] In a two-dimensional projection space, the reference covariance matrix Define a vector of benchmark means The confidence ellipse is centered on the center, and the Mahalanobis distance from any point on the ellipse to the center is 1, which is a standard deviation isosurface. The Mahalanobis distance from the current main cluster center to the reference center corresponds to the relative proportion of the line connecting the center of the confidence ellipse to that point in the radial direction of the confidence ellipse, thus visually displaying the magnitude of the cluster offset CSI.

[0109] In a preferred embodiment, the cluster dispersion (CDI) is obtained by calculating the ratio of the trace of the current principal cluster covariance matrix to the trace of the baseline model covariance matrix, specifically:

[0110] ;

[0111] in, The current principal cluster covariance matrix, The covariance matrix of the baseline model, Represents the trace of a matrix.

[0112] In a preferred embodiment, the inter-cluster separation index (ISI) is obtained by calculating the Mahalanobis distance between the centers of anomalous clusters and the center of the baseline model, specifically as follows:

[0113] ;

[0114] in, This is the mean vector of the outlier cluster. This is the current mean vector of the main cluster. The covariance matrix of the baseline model;

[0115] When there are multiple anomalous clusters, calculate the inter-cluster separation degree for each anomalous cluster separately.

[0116] like Figure 4 As shown, a confidence ellipse centered on the mean vector of the main cluster is defined by the covariance of the benchmark model; the Mahalanobis distances from the centers of different anomalous clusters to the center of the main cluster correspond to the relative proportions of the line connecting the center of the confidence ellipse to that point in the radial direction of the confidence ellipse, thus visually demonstrating the magnitude of the inter-cluster separation (ISI).

[0117] In a preferred embodiment, the abnormal sample ratio (ASR) is obtained by calculating the proportion of fasteners not belonging to the main cluster to the total number of fasteners in the target area, specifically as follows:

[0118] ;

[0119] in, This represents the number of fasteners belonging to the abnormal cluster. This represents the number of fasteners that belong to outliers. This represents the total number of fasteners within the target area.

[0120] The above describes the entire process of calculating the population quality assessment index in this embodiment. It should be noted that, in addition to Mahalanobis distance, Euclidean distance, weighted distance, or standardized distance can also be selected as alternatives according to actual needs. For example, cluster offset can be calculated using standardized Euclidean distance, or weighted distance can be used to highlight key features. However, in the multidimensional morphology evaluation of this application, Mahalanobis distance is more suitable due to the consideration of the correlation between features.

[0121] S6. Compare the group quality assessment index with the benchmark model and the preset group quality threshold, and conduct a group quality assessment on all fastener heads in the target area.

[0122] As a preferred embodiment, in this embodiment, such as Figure 5 As shown, step S6 specifically involves:

[0123] Preset CSI first threshold CSI second threshold CDI first threshold CDI second threshold ISI threshold and ASR threshold Based on the comparison results between the group quality assessment indicators and the multi-level preset thresholds, the corresponding group quality level is output:

[0124] First determine if ASR exceeds When ASR does not exceed hour:

[0125] If CSI does not exceed And CDI does not exceed The output "Group Quality Level 1" indicates that the fastener head group quality of the current evaluation batch is excellent, and the group consistency and stability are normal.

[0126] When ASR does not exceed hour:

[0127] If CSI only exceeds Or CDI only exceeds Output "Population Quality Level 2" if CSI exceeds Or CDI exceeds It outputs "Group Quality Level 3" and simultaneously provides process deviation warnings based on the CSI / CDI threshold comparison results.

[0128] In this embodiment, the offset warning is specifically as follows:

[0129] When CSI only exceeds Or CDI only exceeds The output is "Slight process offset";

[0130] when When this happens, output "process offset";

[0131] when When this happens, output "process offset";

[0132] When CSI exceeds And CDI exceeds The output is "Severe process offset".

[0133] When ASR exceeds hour:

[0134] If CSI does not exceed And CDI did not exceed Output "Population Quality Level 3"; if CSI exceeds... And CDI exceeds The system outputs "Population Quality Level 4"; simultaneously, it issues a process offset warning based on the CSI / CDI threshold comparison result and triggers an ISI threshold comparison, then performs pattern anomaly identification based on the ISI threshold comparison result; in this embodiment, pattern anomaly identification specifically involves:

[0135] If the ASR exceeds the threshold and there is only one anomalous cluster with an ISI exceeding the threshold, the output "Single-mode defect exists" indicates that there is only one anomalous cause.

[0136] If the ASR exceeds the threshold and there are at least two anomalous clusters with ISI exceeding the threshold, output "Multi-mode defect exists", indicating that there are multiple different anomalous causes.

[0137] If the ASR exceeds the threshold and there are no anomalous clusters with ISI exceeding the threshold, then the output is "Occasional Individual Defect".

[0138] The final output includes population quality assessment results, process deviation warning results, and pattern anomaly identification results.

[0139] In summary, the method proposed in this invention shifts the detection perspective from judging the pass / fail status of a single fastener to a holistic assessment of the consistency and stability of the fastener group morphology across an entire area, a whole joint, or an entire workstation. This approach better aligns with the actual needs of process quality control. It can detect process deviation trends and batch anomalies in advance through cluster distribution changes, even before individual parameters exceed tolerances. By establishing a benchmark model and calculating indicators such as cluster offset and cluster dispersion, the overall quality status is quantified, improving the scientific rigor and traceability of the assessment. This invention achieves group quality assessment of fastener heads, enabling the detection of process deviation trends before individual parameters exceed tolerances, which is beneficial for improving the process quality control capabilities of aerospace fastener installation.

[0140] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0141] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0142] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics, characterized in that, Includes the following steps: Acquire the 3D point cloud data of all fastener heads within the target area, preprocess and segment the 3D point cloud data to obtain the independent point cloud of each fastener head after preprocessing. Multidimensional topographic features are extracted based on the independent point cloud of each fastener head and the surrounding reference surface, and topographic feature vectors are constructed. The topographic feature vectors of all fastener heads in the target area constitute a set of group feature vectors. A baseline model is constructed based on the morphological feature vectors of the heads of historical qualified fasteners; After normalizing or standardizing the set of population feature vectors, cluster analysis is performed to obtain the main cluster, abnormal clusters, outliers, and the corresponding population quality distribution characteristics; the abnormal clusters are those whose sample size reaches a preset lower limit and whose cluster center is more than a preset threshold away from the main cluster or the benchmark model. Based on the population quality distribution characteristics of the main cluster, anomalous clusters, and outliers, and combined with the benchmark model, population quality assessment indicators are calculated; the population quality assessment indicators include cluster offset (CSI), cluster dispersion (CDI), inter-cluster separation (ISI), and outlier percentage (ASR). The group quality assessment index is compared with the preset group quality threshold to perform a group quality assessment on all fastener heads in the target area.

2. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 1, characterized in that, The extraction of multi-dimensional topographic features based on the independent point cloud of each fastener head and the surrounding reference surface, and the construction of the topographic feature vector, are specifically as follows: Calculate the height difference between the fastener head and the surrounding reference surface, and use it as the head concavity / convexity. Calculate the diameter of the circumcircle or least-squares fitted circle of the fastener head profile, and use it as the equivalent diameter of the head. Calculate the degree of deviation between the fastener head profile and the standard circle, and use this as the head roundness. Calculate the offset between the geometric center of the fastener head profile and the center of its least squares fitted circle, and use this offset as the profile eccentricity. Calculate the slope distribution characteristics of the transition region at the edge of the fastener head, and use it as the edge slope. Calculate the maximum or average value of the difference between the height of the sampling points in the transition area of ​​the fastener head edge and the height of the surrounding reference surface, and use it as the transition height difference between the head and the surrounding reference surface. The topographic feature vector of each fastener head consists of at least two topographic features from the following: head concavity / convexity, head equivalent diameter, head roundness, profile eccentricity, edge slope, and transition height difference between the head and the surrounding reference surface. The surrounding reference surface is a reference surface obtained by fitting the local skin surface of the target area or the theoretical design surface.

3. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 1, characterized in that, The specific steps for constructing the benchmark model based on the morphological feature vectors of historical qualified fastener heads are as follows: Perform group quality assessment on fastener heads by batch; obtain three-dimensional morphological data of fastener heads from historical batches that have been confirmed to be of qualified quality, extract morphological feature vectors and construct a historical group feature vector set; Cluster and clean the historical group feature vector set, remove outliers, and obtain the historical main clusters; The mean vector and covariance matrix of the historical principal clusters are calculated and used as the baseline model parameters to obtain the baseline model; the formula for the baseline model parameters is expressed as follows: ； ； in, As the main cluster of history, Each sample corresponds to a shape feature vector of a fastener head; for The number of samples in the sample; The mean vector of the baseline model. The covariance matrix of the baseline model; The benchmark model is classified and stored according to installation area, fastener specifications, and installation process conditions, and a dynamic update mechanism is adopted: when the preset conditions are met, the benchmark model is updated using the morphological feature vector of the latest qualified batch.

4. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 1, characterized in that, The cluster offset (CSI) is obtained by calculating the Mahalanobis distance between the current master cluster center and the baseline model center, specifically: ； ； in, As the current primary cluster, Each sample corresponds to a shape feature vector of a fastener head; This is the current mean vector of the main cluster. The mean vector of the baseline model. The covariance matrix of the baseline model; for The number of samples in the sample; When the covariance matrix of the benchmark model is singular or non-invertible, for Regularization is performed first, and then it is used in the calculation of cluster offset (CSI).

5. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 1, characterized in that, The cluster dispersion (CDI) is obtained by calculating the ratio of the trace of the current principal cluster covariance matrix to the trace of the baseline model covariance matrix, specifically: ； in, The current principal cluster covariance matrix, The covariance matrix of the baseline model, Represents the trace of a matrix.

6. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 1, characterized in that, The inter-cluster separation index (ISI) is obtained by calculating the Mahalanobis distance between the centers of anomalous clusters and the centers of the baseline model, specifically: ； in, This is the mean vector of the outlier cluster. This is the current mean vector of the main cluster. The covariance matrix of the baseline model; When there are multiple anomalous clusters, calculate the inter-cluster separation degree for each anomalous cluster separately.

7. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 1, characterized in that, The outlier sample ratio (ASR) is obtained by calculating the proportion of fasteners not belonging to the main cluster to the total number of fasteners in the target area, specifically: ； in, This represents the number of fasteners belonging to the abnormal cluster. This represents the number of fasteners that belong to outliers. This represents the total number of fasteners within the target area.

8. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 1, characterized in that, The step of comparing the group quality assessment index with a preset group quality threshold to perform a group quality assessment on all fastener heads within the target area specifically involves: Preset CSI first threshold CSI second threshold CDI first threshold CDI second threshold ISI threshold and ASR threshold Based on the comparison results between the group quality assessment indicators and the multi-level preset thresholds, the corresponding group quality level is output: First determine if ASR exceeds When ASR does not exceed hour: If CSI does not exceed And CDI does not exceed Output "Population quality level 1"; When ASR does not exceed hour: If CSI only exceeds Or CDI only exceeds Output "Population Quality Level 2" if CSI exceeds Or CDI exceeds Output "Group Quality Level 3"; simultaneously, perform process deviation warning based on CSI / CDI threshold comparison results; When ASR exceeds hour: If CSI does not exceed And CDI did not exceed Output "Population Quality Level 3" if CSI exceeds And CDI exceeds Output "Group Quality Level 4"; simultaneously, perform process offset warning based on CSI / CDI threshold comparison results, trigger ISI threshold comparison, and perform pattern anomaly identification based on ISI threshold comparison results; The final output includes population quality assessment results, process deviation warning results, and pattern anomaly identification results.

9. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 8, characterized in that, The process offset early warning based on the CSI / CDI threshold comparison results specifically refers to: When CSI only exceeds Or CDI only exceeds The output is "Slight process offset"; when When this happens, output "process offset"; when When this happens, output "process offset"; When CSI exceeds And CDI exceeds The output is "Severe process offset".

10. The method for assessing the quality of aircraft fastener head groups based on three-dimensional morphological clustering statistics according to claim 8, characterized in that, The specific steps for identifying pattern anomalies based on ISI threshold comparison results are as follows: If the ASR exceeds the threshold and there is only one anomalous cluster with an ISI exceeding the threshold, output "Single-mode defect exists"; If the ASR exceeds the threshold and there are at least two anomalous clusters with ISI exceeding the threshold, output "Multi-mode defect exists"; If the ASR exceeds the threshold and there are no anomalous clusters with ISI exceeding the threshold, then output "Occasional Individual Defect".

Images