An automatic riveting forming online quality detection method based on a 3D line scanning camera

Abstract

The application discloses an automatic riveting forming online quality detection method based on a 3D line scanning camera, and relates to the field of aviation riveting detection.The method comprises the following steps: performing a calibration operation on a three-dimensional line scanning camera, and establishing a unified measurement reference; performing multiple optical measurements on the surface of a wall plate in the vicinity of a rivet and the surface of the rivet respectively, and acquiring and recording corresponding data sets; sorting the wall plate data set and the rivet top data set according to the measurement values respectively; selecting the intermediate value of the wall plate data set and the intermediate value of the rivet data set, and calculating the algebraic difference between the two intermediate values as the rivet flushness evaluation result; and storing the evaluation results of multiple rivets into a database, and generating a structured quality detection report.The application can significantly improve the precision, efficiency and consistency of riveting flushness detection, realize full-automatic online detection and full-coverage detection, greatly reduce the missed detection rate and misjudgment risk, and has good anti-interference ability and environmental adaptability.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aviation riveting detection, and particularly relates to an automatic riveting forming online quality detection method based on a 3D line scanning camera. BACKGROUND

[0002] With the continuous development of the aerospace industry towards high quality, high efficiency and high reliability, automatic drilling and riveting technology has become a key link in the process of aircraft manufacturing. Especially in the development of large passenger aircraft and fighter aircraft, the requirement for riveting forming quality is becoming more and more strict, and the detection precision and efficiency of flushness, as one of the core indicators for evaluating riveting quality, directly affect the performance and service life of the overall structure. At present, artificial sampling inspection or traditional optical measurement methods are generally used in this field, which not only has low efficiency and is easily affected by subjective factors, but also is difficult to meet the demand for real-time full inspection of high-speed automatic production lines. In addition, the aviation components often involve various material combinations, complex surfaces and surfaces with inconsistent reflection characteristics, which further increase the difficulty of accurate measurement.

[0003] The existing flushness detection method has several significant limitations. First, the traditional manual detection method has low efficiency, limited sampling rate, and cannot achieve full coverage. Moreover, the detection results are greatly affected by the experience of the operator, and the data consistency and reliability are insufficient. Second, common optical measurement equipment is easily disturbed by stray light, environmental vibration and instantaneous deformation when dealing with strong reflection, multi-material or complex topological surfaces, resulting in increased measurement error. Third, most existing systems lack real-time noise reduction and dynamic anti-interference mechanisms for high-speed production line environments, making it difficult to meet the production line beat requirements while ensuring detection accuracy. Finally, most methods do not achieve comprehensive digitization and data traceability in the detection process, which is not conducive to deep analysis of quality data and production optimization. Therefore, based on the above problems, the present application provides an automatic riveting forming online quality detection method based on a 3D line scanning camera. SUMMARY

[0004] PURPOSE OF THE INVENTION

[0005] In order to solve the above problems, the purpose of the present application is to provide an automatic riveting forming online quality detection method based on a 3D line scanning camera, which aims to solve the outstanding problems of flushness detection in precision, efficiency and anti-interference ability in the automatic riveting process in the field of aerospace, and to ensure that high-precision, full-automatic and real-time online detection can be achieved in a high-speed automatic production line environment, in order to overcome the limitations of traditional manual sampling inspection and ordinary optical measurement methods in detection consistency, environmental adaptability and data traceability.

[0006] TECHNICAL SCHEME

[0007] In order to achieve the above object, the application provides an automatic riveting forming online quality detection method based on a 3D line scanning camera, which adopts a three-dimensional line scanning camera as a core measuring device, performs calibration, and measures multiple times in the rivet and adjacent wallboard area respectively, sorts the collected data according to numerical values and extracts the median value, calculates the height difference as the evaluation result of the rivet flushness, and integrates multiple noise reduction strategies, including outlier rejection and adaptive filtering processing, to effectively suppress mechanical vibration and surface reflection interference, finally realizes automatic recording, storage and report generation of detection data, and forms a complete and reliable online quality detection system.

[0008] In the first aspect, the application provides an automatic riveting forming online quality detection method based on a 3D line scanning camera, which includes:

[0009] Calibration operation is performed on the three-dimensional line scanning camera to establish a unified measurement reference;

[0010] Multiple non-contact optical measurements are performed on the wallboard surface adjacent to the rivet area and the top surface of the rivet, and the corresponding data sets are obtained and recorded;

[0011] The wallboard area data set and the rivet top data set are sorted in ascending order according to the measurement values respectively;

[0012] The median value of the wallboard data set and the median value of the rivet data set are selected, and the algebraic difference between the two median values is calculated as the evaluation result of the rivet flushness;

[0013] The evaluation results of multiple rivets are stored in a database, and a structured quality detection report is generated to realize full-process digital management and data traceability.

[0014] Further, the three-dimensional line scanning camera is a high-speed and high-precision optical measuring device, and the measurement process integrates multiple noise reduction processing strategies, which identify and remove abnormal values in the measurement data through statistical methods, and use edge-preserving filtering algorithm to smooth the data surface, effectively suppress environmental vibration and surface reflection interference, and improve detection consistency.

[0015] Further, the edge-preserving filtering algorithm includes adaptive bilateral filtering or Gaussian filtering, and further introduces a noise suppression mechanism based on real-time collected environmental vibration signals to dynamically adjust the filtering parameters, so as to enhance the adaptability and robustness of the system under complex working conditions.

[0016] Further, the noise suppression mechanism for dynamically adjusting the filtering parameters specifically adjusts the convolution kernel size or threshold parameter of the filtering algorithm according to the vibration frequency energy.

[0017] Further, the filtering algorithm further comprises a wavelet threshold denoising method, high-frequency noise separation and effective signal enhancement are realized by performing multi-scale decomposition and threshold reconstruction processing on the optical measurement signal.

[0018] Further, before performing data acquisition, threshold parameters of each link in the multi-stage noise suppression strategy are pre-calibrated and adaptively initialized according to the optical reflection characteristics and structural complexity of the detected surface.

[0019] In a second aspect, the application further provides an automatic rivet forming online quality detection system based on a 3D line scanning camera, which is based on the method of the first aspect and comprises:

[0020] A three-dimensional line scanning camera is used to perform optical scanning and data acquisition.

[0021] A data acquisition and processing module is configured to perform sorting of data sets, extraction of intermediate values and calculation of algebraic differences.

[0022] A data storage module is used to store detection data and associated information of the rivet.

[0023] A report generation module is used to generate a quality detection report.

[0024] A calibration module is used to perform calibration of the camera coordinate system and periodic precision rechecking.

[0025] Further, the detection path of the system is automatically calculated in real time based on the position and topological configuration of the rivet to be detected.

[0026] Further, the system is equipped with an adaptive operation control unit for sensing the state of interference sources during detection and deciding whether to start a rechecking mechanism; all sensor data and diagnostic logs are integrated into the same database to support quality traceability and production system linkage adjustment.

[0027] In a third aspect, the application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a management platform to realize the above-mentioned automatic rivet forming online quality detection method based on a 3D line scanning camera.

[0028] The application uses a high-speed and high-precision three-dimensional line scanning camera as a core measurement device, performs camera calibration, and performs multiple measurements on the rivet and adjacent wallboard regions, sorts the obtained data sets, extracts the median values, and calculates the height difference as the evaluation result of the rivet flushness. The system integrates multi-stage noise reduction and dynamic anti-interference mechanism, including outlier rejection and adaptive filtering method, effectively suppresses mechanical vibration, surface reflection and complex working condition interference, and finally realizes automatic recording, storage and report generation of detection data, forming a complete and reliable online quality detection system.

[0029] The scheme can significantly improve the accuracy, efficiency and consistency of rivet flushness detection, realize full-automatic online detection and full-coverage detection, greatly reduce the missed detection rate and misjudgment risk, and has good anti-interference ability and environmental adaptability. The system generates a full-digital quality report, supports data traceability and deep analysis, is especially suitable for high-precision requirements in the field of aerospace and other high-precision requirements, and provides reliable data support for rivet quality evaluation and process optimization.

[0030] Advantages

[0031] By implementing the above-mentioned automatic rivet forming online quality detection method based on a 3D line scanning camera, the following technical effects are achieved:

[0032] (1) By using a three-dimensional line scanning camera as the core measuring device, high-resolution capture and reconstruction of the rivet and the surface topography of the wall plate are realized, effectively overcoming the problems of low efficiency and easy damage to the surface of the workpiece in traditional contact measurement, and significantly improving the degree of automation and data acquisition accuracy of the detection.

[0033] (2) By measuring multiple times and extracting the median value of the data for height difference calculation, this method can effectively resist random errors and transient disturbances in the measurement process, improve the stability and reliability of the flushness evaluation results, and avoid the negative impact of extreme values on the detection results.

[0034] (3) The system integrates multiple noise suppression methods such as outlier rejection, bilateral filtering, Gaussian filtering and vibration response adaptive filtering, effectively dealing with complex working condition disturbances such as mechanical vibration, surface reflection and multi-material combination, ensuring that the detection system still has high robustness and accuracy in high-speed production line environment.

[0035] (4) By automatically recording the detection results in the database and generating structured reports, the rivet quality data is fully digitized, queryable and analyzable, supporting quality traceability, production statistics and process optimization, providing data basis and decision support for high-reliability manufacturing.

[0036] (5) Through the noise suppression mechanism based on real-time acquisition of environmental vibration signals to dynamically adjust the filtering parameters, the real-time linkage optimization of filtering parameters and environmental vibration is realized, effectively suppressing the periodic interference caused by mechanical vibration, and improving the adaptability and robustness of the system in complex industrial environments.

[0037] (6) Through the synergistic effect of outlier rejection and signal smoothing in the multi-level noise reduction processing strategy, the signal-to-noise ratio of the original data and the consistency of the measurement are significantly improved, laying a data foundation for high-precision quantitative evaluation, and further enhancing the detection reliability of the system in complex surface conditions such as strong reflection and multi-material combination. Attached Figure Description

[0038] To make the above-described method for online quality inspection of automatic riveting forming based on a 3D line scan camera more obvious and understandable, the accompanying drawings used in the specific embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0039] Figure 1 This is a flowchart illustrating the method described in this application;

[0040] Figure 2 This is a schematic diagram illustrating the principle of this application;

[0041] Figure 3 This describes the data processing and multi-level noise reduction strategy workflow. Detailed Implementation

[0042] Example 1:

[0043] An automatic online quality inspection method for riveting forming based on a 3D line scan camera is provided. The method flow is as follows: Figure 1 As shown, the principle is as follows Figure 2 As shown, the process includes: calibrating the 3D line scan camera to establish a unified measurement benchmark; performing multiple non-contact optical measurements on the wall panel surface and the rivet tip surface in the vicinity of the rivet, acquiring and recording the corresponding datasets; sorting the wall panel area dataset and the rivet tip dataset in ascending order according to the measured values; selecting the median value of the wall panel dataset and the median value of the rivet dataset, and calculating the algebraic difference between the two median values ​​as the rivet flatness evaluation result; storing the evaluation results of multiple rivets in a database and generating a structured quality inspection report.

[0044] The details are as follows.

[0045] First, the hardware units of the inspection system are constructed. The 3D line scan camera is fixed to the end flange of a six-axis industrial robot via a high-rigidity mechanical bracket. This configuration gives the system great mobility in three-dimensional space, enabling it to accurately reach complex or narrow spaces within the riveting area and effectively avoid blind spots. Both the camera and robot control systems are integrated with the data acquisition and processing module in the host industrial computer via industrial Ethernet.

[0046] Before the detection process starts, a strict calibration operation needs to be performed. A standard gauge block calibrated by a metrology institute is used as the reference. The robot drives the camera to scan the reference from multiple angles and positions. The internal and external parameters of the camera and the conversion relationship with the robot coordinate system are calculated by the built-in algorithm of the calibration module. A high-precision unified world coordinate system is established, and the measurement accuracy is calibrated to the micron level. This process ensures the consistency and traceability of the whole cycle measurement reference.

[0047] During formal detection, the robot positions according to the preset path, making the camera optical axis perpendicular to the rivet and its surrounding wallboard area. For single rivet measurement: the data acquisition and processing module controls the camera to trigger multiple measurements on the wallboard adjacent to the rivet, the specific number is dynamically configured according to the signal-to-noise ratio on site, such as 6 times, and then multiple measurements are also performed on the top surface of the rivet, such as 4 times. Each measurement obtains a data frame containing a three-dimensional point cloud.

[0048] Subsequently, the system performs the core median filtering and calculation process. The module sorts the multiple measurement data sets of the wallboard area and the multiple measurement data sets of the rivet top in ascending order according to the Z-axis height value. Instead of taking the average of all points, the middle values of the two sorted data sets are selected, i.e. the 3rd number of the wallboard data and the 2nd number of the rivet data. The algebraic difference between the two middle values is calculated as the flushness evaluation result of the rivet. This "median value" strategy can effectively resist the interference of outliers caused by instantaneous vibration, light intensity fluctuation or small reflective points on the surface.

[0049] To cope with the complex vibration environment of the production line, the system integrates multiple levels of noise reduction strategies, as shown in Figure 3 First, in the data preprocessing stage, a statistical outlier removal algorithm is used to calculate the average distance of each point from its nearest K neighborhood points. Points that are more than three times the standard deviation of the global average distance are removed to eliminate obvious noise. Second, an edge-preserving filtering algorithm is applied. In this embodiment, adaptive bilateral filtering is selected, with the spatial domain kernel function adjusted adaptively according to the point cloud density, and the color domain kernel function dynamically changed according to the laser line intensity information. While smoothing the surface data, it can perfectly preserve the edge features of the rivet and the wallboard, preventing the introduction of calculation errors due to edge blurring. Finally, the system collects environmental vibration signals in real time through the built-in acceleration sensor and dynamically adjusts the parameters of the filtering algorithm, such as the convolution kernel size, to achieve adaptive vibration suppression, significantly improving the robustness and measurement consistency of the system in complex working conditions.

[0050] Finally, the flushness data of each rivet and its corresponding robot position coordinates, timestamps, batch numbers, etc. are stored in the data storage module in a structured manner. The report generation module extracts data from the database on a regular basis or on demand, generates fully digital inspection reports that meet the requirements of the aviation quality system, supports data traceability and quality trend analysis by aircraft station, part number, time period, etc., forming a complete technical closed loop from detection to analysis.

[0051] The experimental object is selected as a certain type of aircraft wall plate assembly, which contains 320 rivets. The real vibration and illumination change environment of the production line is simulated. The handheld laser range finder is used for traditional manual detection, and the sampling rate is 20%. The comparison results are shown in Table 1.

[0052] Table 1, comparison results of automatic rivet forming quality detection method and traditional manual detection

[0053] Performance indicators Automatic rivet forming quality detection method Traditional manual detection Effectiveness improvement Single-point average detection time 0.90±0.05 s 3.8±0.6 s Efficiency improved by about 4.2 times Flatness measurement accuracy (standard deviation) 0.018 mm 0.095 mm Significant improvement in measurement consistency Defect detection rate 99.6% 93.75% The system realizes full detection, and manual sampling detection of 64 rivets Stability under different working conditions Error increase <0.01 mm under conventional vibration environment Error increase about 0.06 mm under vibration environment The system has excellent anti-interference ability Data recording error rate 0.1% 6.2% The system automatically records to avoid human input errors Detectable rivet type coverage 98% Limited accessibility, about 85% The system can flexibly position rivets in complex areas

[0054] The verification shows that in the case of obtaining similar average error as the above-mentioned embodiments, the system is significantly superior to the traditional manual detection method in terms of detection efficiency, measurement accuracy, defect detection capability, environmental adaptability and data reliability. The system has excellent anti-interference performance and stability, can realize full-process digital quality data management, effectively supports quality traceability and process optimization in high-precision manufacturing scenarios, and is especially suitable for application scenarios with high quality requirements and fast production rhythm in aerospace manufacturing.

[0055] Embodiment 2:

[0056] The main difference between this embodiment and embodiment 1 is the selection of specific algorithms in the multi-stage noise reduction strategy. The rest of the system composition and workflow are the same, which illustrates the flexibility of the technical solution of the present application.

[0057] In the data denoising processing stage, a wavelet transform-based noise suppression method is used instead of the adaptive bilateral filter in embodiment 1. Specifically, the data acquisition and processing module performs wavelet threshold denoising processing on the preprocessed point cloud data along the laser line direction. First, select a base wavelet suitable for representing the surface morphology of the workpiece to perform N-layer decomposition, obtaining subbands of different frequencies. Then, an adaptive threshold function is used to quantize the high-frequency detail coefficients. The determination of the threshold value is associated with the real-time acquisition of the vibration frequency energy. When the vibration energy is large, a stricter threshold value is used to suppress high-frequency noise. Finally, the processed wavelet coefficients are reconstructed to obtain the denoised smooth surface data.

[0058] This method can also effectively separate noise and real surface signals, and performs well in strong mechanical vibration interference environment, and also realizes the technical effects of edge feature preservation and dynamic anti-interference.

[0059] The same set of 320 rivet-containing aircraft panel assemblies were used as test objects on a vibration platform simulating an aircraft manufacturing site environment. The same optical measurement data were processed using a wavelet threshold denoising method and a conventional Gaussian filtering method, respectively. The comparison results are shown in Table 2.

[0060] Table 2. Comparison results of wavelet threshold denoising method and conventional Gaussian filtering method

[0061] Performance indicators Wavelet threshold denoising method Conventional Gaussian filtering method Effectiveness improvement Flatness measurement accuracy (standard deviation, mm) 0.015±0.002 0.028±0.004 Accuracy improved by 46.4% Error increment (mm) under strong vibration environment <0.008 0.022 Anti-interference ability improved by 64% Edge feature retention error (mm) 0.010 0.025 Edge retention ability improved by 60% Signal-to-noise ratio (dB) 28.5 22.3 Signal-to-noise ratio improved by 27.8% Single-point data processing time (ms) 12.5 8.2 Computing efficiency reduced by 34.4% Complex surface adaptability (pass rate%) 98.5% 89.0% Adaptability improved by 9.5%

[0062] The wavelet threshold denoising method effectively separates high-frequency noise from surface signals through multi-scale decomposition, with an error increment of only 0.008 mm in a vibration environment, which is significantly better than the 0.022 mm of the Gaussian filter. This is due to the adaptive threshold mechanism and real-time linkage with the vibration signal. The time-frequency localization property of wavelet transform enables it to accurately preserve the rivet edge features, while the global smoothing of Gaussian filtering leads to edge blurring and easy introduction of flatness calculation bias. The wavelet threshold denoising method takes 34.4% more time than the Gaussian filter, but its adaptability to complex surfaces reaches 98.5%, making it more suitable for the high-precision requirements of aviation riveting scenarios. The signal-to-noise ratio after wavelet threshold denoising is improved by 27.8% compared to Gaussian filtering, confirming that it can more effectively suppress noise and preserve effective signals. Experimental data show that in the context of online detection of aviation riveting, the wavelet threshold denoising method is significantly superior to the conventional Gaussian filter in terms of measurement accuracy, anti-vibration interference capability, edge preservation, and adaptability to complex surfaces. Although the computational efficiency is slightly lower, its overall performance better meets the needs of high-reliability manufacturing.

[0063] Those skilled in the art will appreciate that embodiments of the application can be supplied as a method, a system, or a computer program product. Thus, the application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the application can take the form of a computer program product on one or more computer-usable non-transitory storage media having computer-usable program code embodied in the medium.

[0064] The application can provide computer program instructions to the management platform of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the management platform of the computer or other programmable data processing devices generate means for implementing the system.

[0065] These computer program instructions can also be stored in a computer-readable storage medium that can guide the computer or other programmable data processing devices to work in a specific way, so that the instructions stored in the computer-readable storage medium generate a product including instruction means, which implements the functions of the system.

[0066] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable devices to generate a computer-implemented process, so that the instructions executed on the computer or other programmable devices provide steps for implementing the functions of the system.

Claims

1. A 3D line-scan camera-based automatic rivet-forming online quality detection method, characterized in that, The method comprises the following steps: Carrying out a calibration operation on a three-dimensional line scanning camera to establish a unified measurement reference; Performing multiple non-contact optical measurements on the surface of the wall plate and the surface of the rivet in the vicinity of the rivet to obtain and record corresponding data sets; the number of measurements on the surface of the wall plate and the surface of the rivet is dynamically configured according to the signal-to-noise ratio on site; Performing multi-level noise reduction processing on the obtained data sets, which comprises: identifying and removing outliers in the measurement data by a statistical outlier removal algorithm, and performing smoothing processing on the data surface by an edge-preserving filtering algorithm; the spatial domain kernel function of the edge-preserving filtering algorithm is adaptively adjusted according to the point cloud density, and the color domain kernel function is dynamically changed according to the laser line intensity information; Sorting the wall plate region data set and the rivet tip data set after noise reduction processing according to the measurement values; Selecting the median value of the wall plate data set and the median value of the rivet data set, and calculating the algebraic difference value of the two median values as the rivet flushness evaluation result; Storing the evaluation results of multiple rivets into a database and generating a structured quality detection report.

2. The method of claim 1, wherein: The three-dimensional line scanning camera is a high-speed and high-precision optical measurement device.

3. The method of claim 2, wherein: The edge-preserving filtering algorithm includes adaptive bilateral filtering or Gaussian filtering, and further introduces a noise suppression mechanism for dynamically adjusting filtering parameters based on real-time collected environmental vibration signals.

4. The method of claim 3, wherein: The noise suppression mechanism for dynamically adjusting filtering parameters specifically adjusts the convolution kernel size or threshold parameter of the filtering algorithm adaptively according to the vibration frequency energy.

5. The method of claim 3, wherein: The filtering algorithm further includes a wavelet threshold denoising method, which realizes high-frequency noise separation and effective signal enhancement by performing multi-scale decomposition and threshold reconstruction processing on the optical measurement signal.

6. The method of claim 2, wherein: Before data collection, the threshold parameters of each link in the multi-level noise suppression strategy are pre-calibrated and adaptively initialized according to the optical reflection characteristics and structural complexity of the detected surface.

7. A 3D line-scan camera-based automatic rivet-forming online quality detection system, characterized in that: The system executes the method of any one of claims 1-6, comprising: A three-dimensional line scanning camera for performing optical scanning and data collection; A data collection and processing module configured to perform data set sorting, median value extraction and algebraic difference calculation; A data storage module for storing rivet detection data and associated information; A report generation module for generating a quality detection report; A calibration module for performing camera coordinate system calibration and periodic accuracy review.

8. The system of claim 7, wherein: The detection path of the system is automatically calculated in real time based on the position and topological configuration of the rivet to be detected.

9. The system of claim 7, wherein: The system is equipped with an adaptive operation management and control unit, which is used for sensing the state of interference sources during detection and deciding whether to start a re-inspection mechanism; all sensor data and diagnosis logs are integrated into the same database to support quality traceability and production system linkage adjustment.

10. A computer readable storage medium having stored therein a computer program, characterized in that: The computer program, when run on a processor, performs the method of any one of claims 1-6.

Images