Nondestructive discrimination method for self-piercing riveted joint, electronic device, readable medium

Abstract

The disclosure provides a nondestructive discrimination method, an electronic device and a readable medium for a self-piercing riveted joint, the nondestructive discrimination method comprising: acquiring a pressure signal in a self-piercing riveting process; preprocessing the pressure signal to obtain a homogenized pressure signal; performing detrending processing on the homogenized pressure signal to obtain a detrended pressure signal; performing wavelet analysis on the detrended pressure signal to identify a target characteristic frequency band near a device inherent frequency; performing narrowband wavelet reconstruction in the target characteristic frequency band to obtain a bimodal characteristic curve; extracting first and second peak amplitudes A and B and a peak ratio R=B / A from the bimodal characteristic curve; inputting the first and second peak amplitudes into a macroscopic deformation discrimination model established based on offline calibration data to discriminate a joint macroscopic deformation mode and obtain a macroscopic discrimination result; inputting the peak ratio into a microstructure discrimination model established based on offline calibration data to discriminate a joint microstructure evolution state and obtain a microstructure discrimination result; and outputting a macro-micro morphology discrimination result of the joint according to the macroscopic and microstructure discrimination results.

Description

Technical Field

[0001] This disclosure relates to the fields of lightweight material joining technology and non-destructive quality monitoring technology, and in particular to a non-destructive identification method for self-piercing friction riveting joints, an electronic device, and a computer-readable medium. Background Technology

[0002] In the assembly and connection of aluminum alloy components, traditional fusion welding processes (such as laser welding and resistance spot welding) are prone to defects such as coarsening of strengthening phases, hot cracking and porosity. Conventional mechanical connection processes (such as traditional self-piercing riveting) may have problems such as joint cracking, poor rivet forming and unreliable mechanical interlocking when connecting low-plasticity high-strength aluminum alloys and cast aluminum alloys, which are difficult to meet the connection performance requirements of high-end equipment.

[0003] Self-piercing friction riveting (F-SPR) is a joining process that combines the characteristics of friction stir welding and self-piercing riveting, achieving a composite connection method that combines mechanical interlocking and solid-state bonding. The flat-die F-SPR process eliminates the need for a guiding die, resulting in a relatively smooth joint structure. However, in the application of F-SPR, joint quality is difficult to inspect and accurately assess online. Existing inspection methods mainly rely on metallographic sectioning and destructive tensile testing, which have low efficiency and cannot meet the continuous inspection requirements of production lines, thus limiting the engineering application of this process to some extent. Summary of the Invention

[0004] This disclosure provides a non-destructive evaluation method for self-piercing friction riveting joints, an electronic device, and a computer-readable medium. In existing self-piercing friction riveting (F-SPR) processes, the joint forming process is transiently closed, making material flow and microstructure evolution unobservable, and hindering accurate online non-destructive evaluation of joint quality. Traditional offline testing methods such as metallographic sectioning and tensile failure are inefficient and costly, failing to meet the real-time monitoring requirements of mass production lines. Existing process signal-based monitoring methods are mostly designed for processes like friction stir welding, only capable of single-factor or strength evaluation, and cannot identify the unique macroscopic deformation patterns and microstructure types of F-SPR joints. Therefore, there is an urgent need for an online non-destructive macro- and microscopic morphology evaluation method suitable for F-SPR joints.

[0005] In a first aspect, embodiments of this disclosure provide a non-destructive discrimination method for self-piercing friction riveting joints, comprising: acquiring a pressure signal during the self-piercing friction riveting process; preprocessing the pressure signal to obtain a homogenized pressure signal; performing detrending processing on the homogenized pressure signal to obtain a detrended pressure signal; performing wavelet analysis on the detrended pressure signal to identify a target characteristic frequency band located near the inherent frequency of the device based on the wavelet energy distribution; performing narrowband wavelet reconstruction within the target characteristic frequency band to obtain a bimodal characteristic curve containing a first peak amplitude A and a second peak amplitude B; and extracting the first peak amplitude A, the second peak amplitude B, and the peak value from the bimodal characteristic curve. The peak ratio R = B / A, where the first peak amplitude A and the second peak amplitude B are used to characterize the macroscopic deformation state of the joint, and the peak ratio R is used to characterize the microscopic evolution state of the joint. The first peak amplitude A and the second peak amplitude B are input into a macroscopic deformation discrimination model established based on offline calibration data to discriminate the macroscopic deformation mode of the joint and obtain a macroscopic discrimination result. The peak ratio R is input into a microscopic organization discrimination model established based on offline calibration data to discriminate the microscopic evolution state of the joint and obtain a microscopic discrimination result. Based on the macroscopic discrimination result and the microscopic discrimination result, the macroscopic and microscopic morphology determination result of the joint is output.

[0006] In some embodiments, the pressure signal is preprocessed to obtain a homogenized pressure signal, including: extracting an effective signal segment from the pressure signal from the point where the rivet contacts the plate to the completion of the riveting; removing zero-point drift and environmental interference noise from the effective signal segment to output a denoised effective signal segment; and resampling the denoised effective signal segment at equal intervals to obtain a homogenized pressure signal.

[0007] In some embodiments, detrending the homogenized pressure signal to obtain a detrended pressure signal includes: dividing the homogenized pressure signal into multiple continuous segments, wherein adjacent segments of the multiple continuous segments have overlapping regions; fitting local trend curves to each continuous segment; performing weighted smoothing on the local trend curves of the adjacent segments in the overlapping regions to reconstruct a complete trend curve; and subtracting the homogenized pressure signal from the complete trend curve point by point to obtain the detrended pressure signal.

[0008] In some embodiments, the first peak amplitude A and the second peak amplitude B together constitute a two-dimensional feature space of the macroscopic deformation state of the joint, and different macroscopic deformation modes correspond to different distribution regions in the two-dimensional feature space; the peak ratio R=B / A is used to characterize the microstructure evolution state of the joint, and different microstructure types correspond to different peak ratio ranges.

[0009] In some embodiments, the first peak amplitude is the first excitation response peak amplitude corresponding to the material insertion and initial flow stage in the early riveting phase of the bimodal characteristic curve, and the second peak amplitude is the second excitation response peak amplitude corresponding to the rivet deformation, material filling and mechanical interlocking formation stage in the later riveting phase of the bimodal characteristic curve.

[0010] In some embodiments, the macroscopic deformation discrimination model and the microstructure discrimination model are implemented using a support vector machine classifier, and the discrimination boundary is determined in advance by training with offline calibration data; wherein, the offline calibration data includes the pressure signal features of joint samples prepared under different process parameters, as well as the corresponding macroscopic deformation mode labels and microstructure evolution state labels.

[0011] In some embodiments, the macro-micro morphology determination result includes the macro deformation mode and microstructure type of the rivet; a joint determination is made based on the macro deformation mode and the microstructure type of the rivet; when the macro deformation mode of the rivet is a preset optimal macro morphology and the microstructure type is a preset optimal structure type, the joint is determined to be a high-performance qualified joint.

[0012] In some embodiments, the non-destructive discrimination method further includes: acquiring the pressure signal of the newly added joint sample and extracting the corresponding first peak amplitude, second peak amplitude, and peak ratio; performing macroscopic cross-sectional measurement and / or metallographic sectioning and annotation on the newly added joint sample to obtain the corresponding macroscopic deformation mode label and microstructure type label; adding the first peak amplitude, the second peak amplitude, the peak ratio, and the corresponding macroscopic deformation mode label and microstructure type label to the training sample library, and retraining or calibrating the discrimination boundary of the macroscopic deformation discrimination model and the microstructure discrimination model based on the updated training sample library.

[0013] In a second aspect, embodiments of this disclosure provide an electronic device, including: one or more processors; and a memory storing one or more programs thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the lossless discrimination method according to the first aspect.

[0014] Thirdly, embodiments of this disclosure provide a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the lossless discrimination method according to the first aspect.

[0015] The non-destructive testing method provided in this disclosure extracts the bimodal characteristic curve of the pressure signal during the F-SPR riveting process and combines it with a dual-discrimination model architecture to achieve simultaneous online non-destructive testing of the macro- and micro-morphology of the joint. Compared with traditional offline testing methods that rely on metallographic sectioning and tensile failure, this disclosure can complete the joint quality assessment on the production line without damaging the workpiece, significantly improving testing efficiency and reducing testing costs. Experimental verification shows that this method exhibits high discrimination accuracy under different process parameters and aluminum alloy material conditions. This method can be used for quality monitoring of the connection process of aluminum alloy components in automobile bodies, aerospace structures, and rail transit equipment, and can be integrated into F-SPR intelligent riveting equipment to realize online quality inspection functions. Attached Figure Description

[0016] Figure 1 This is a flowchart of a non-destructive identification method for self-piercing friction riveting joints according to an embodiment of this disclosure.

[0017] Figure 2 This is a schematic diagram of the module structure and data flow of a non-destructive testing system for self-piercing friction riveting joints according to an embodiment of this disclosure.

[0018] Figure 3 This is a bimodal characteristic curve of the riveting pressure signal in a specific embodiment of this disclosure.

[0019] Figure 4 This is a schematic diagram of the spatial distribution of macro- and micro-morphological discrimination features of the connector based on peak features, as described in the embodiments of this disclosure.

[0020] Figure 5 This is a typical macro- and micro-morphological reference diagram of the F-SPR connector involved in the embodiments of this disclosure.

[0021] Figure 6 This is a block diagram of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions of this disclosure will be described in detail below with reference to the accompanying drawings.

[0023] Exemplary embodiments will be described more fully below with reference to the accompanying drawings; however, these exemplary embodiments may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will enable those skilled in the art to fully understand the scope of this disclosure.

[0024] Where there is no conflict, the various embodiments of this disclosure and the features thereof in the embodiments may be combined with each other.

[0025] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0026] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. As used herein, the singular forms “a” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded.

[0027] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this disclosure, and will not be interpreted as having an idealized or overly formal meaning, unless expressly so defined herein.

[0028] Unless otherwise specified in this disclosure, the following technical terms shall be interpreted as follows: Friction self-piercing (F-SPR) is a solid-state composite joining process that combines friction stir welding and self-piercing riveting. It achieves a dual connection of mechanical interlocking and solid-state bonding of aluminum alloy sheets through the heat generated by the rotational friction of the rivet and the plastic deformation by stamping.

[0029] The macro- and micro-morphological features of a joint (or the macro- and micro-morphological features of a joint) are a collective term for two core morphological characteristics of a joint: macro-morphology refers to the external deformation morphology of the rivet; micro-morphology refers to the grain structure morphology of the metal inside the joint, which together determine the mechanical properties of the joint.

[0030] Wavelet transform feature extraction (or wavelet analysis feature extraction) is a signal analysis method that can accurately extract hidden subtle features from the original pressure signal and pinpoint the effective signal corresponding to the inherent frequency of the equipment.

[0031] The SVM classifier (Support Vector Machine) is a commonly used intelligent classification model that can automatically classify data and determine the results based on data features. This disclosure is used for the automatic identification of joint morphology.

[0032] Support vectors are key sample points that determine the location of the decision boundary in an SVM classification model. They are the core data for building the classification model. Only a small number of support vectors are needed to determine the discrimination rules. In this disclosure, they are used to accurately delineate the optimal discrimination interval for the macro- and micro-morphological features of the joint.

[0033] The natural frequency (or the inherent vibration frequency of the equipment or the natural frequency of the system) is the inherent vibration frequency of the riveting equipment itself. Signals at this frequency can most sensitively reflect subtle changes in the riveting process.

[0034] Online non-destructive identification (or online non-destructive testing, online non-destructive judgment) refers to a method that completes joint quality inspection and judgment in real time during the production process without cutting or damaging the workpiece.

[0035] Segmented overlap detrending (or segmented smoothing detrending, segmented overlap trend elimination) is a custom signal preprocessing method of this disclosure. It processes the pressure signal in segments, eliminates slow fluctuation trends, retains overlapping areas between segments and smooths the transition, avoids signal distortion, and highlights effective features.

[0036] The characteristic curve (or bimodal characteristic curve) is the core signal feature defined in this disclosure. The pressure characteristic curve, which contains two obvious peaks and is extracted by wavelet transform, is the main basis for judging the macro- and micro-morphology of the joint.

[0037] This embodiment uses the quality control of aluminum alloy component assembly and connection as a typical application scenario to illustrate a method for online non-destructive identification of the macro- and micro-morphology of self-piercing friction riveting (F-SPR) joints based on the pressure signal of the riveting process.

[0038] This disclosed method is primarily applicable to aluminum alloy components, especially to the online assessment and intelligent monitoring of connection quality for low-ductility aluminum alloy components such as 7-series aluminum alloys and cast aluminum alloys. After recalibrating samples and training the model for different material systems, this method can also be extended to other lightweight metallic materials suitable for the F-SPR process.

[0039] This disclosure can be applied to the quality control of the connection process of lightweight aluminum alloy components such as automobile bodies, aerospace structures, and rail transit equipment, and realizes online non-destructive testing and intelligent identification of the macroscopic deformation morphology and microstructure morphology of the joint.

[0040] This disclosure addresses the need for online non-destructive testing of the macro- and micro-morphology of aluminum alloy flat die self-piercing friction riveting (F-SPR) joints by providing a morphology recognition method based on a combination of frequency domain features of process pressure signals and machine learning models.

[0041] This embodiment uses double-layer aluminum alloy sheets as the connected components, employs suitable steel semi-hollow rivets, and utilizes a flat die self-piercing friction riveting device equipped with a pressure acquisition module to complete the connection process. The pressure acquisition module is located at the device base and is used to collect pressure signals in real time throughout the riveting process. The entire solution relies on conventional signal processing methods and intelligent classification models to achieve automatic identification of the macro- and micro-morphological features of the joint.

[0042] The core idea of ​​this non-destructive discrimination method is as follows: Pressure signals are acquired in real-time during the F-SPR riveting process. After preprocessing and segmented overlap detrending, a bimodal feature curve corresponding to the characteristic excitation of the riveting process is extracted using wavelet analysis, and the amplitudes of the first and second peaks, A and B, are obtained. Based on the extracted feature parameters, a two-dimensional feature space is constructed using A and B to characterize the macroscopic deformation state of the joint, and the peak ratio R = B / A is used to characterize the microstructure state of the joint. These are then input into pre-trained macroscopic and microscopic discrimination models respectively to achieve simultaneous identification of the macroscopic deformation mode and microstructure type of the joint.

[0043] The discrimination boundary of the classification model can be obtained through the annotation of macroscopic cross-sectional morphology and microstructure of offline samples. In practical applications, only the pressure signal of the riveting process needs to be collected to complete the online non-destructive identification of the macroscopic and microscopic morphology of the joint, without the need for cutting or other destructive testing. In addition, the classification model can be retrained or the discrimination boundary can be calibrated based on newly added labeled samples to adapt to the application requirements of different aluminum alloy materials and different process conditions.

[0044] The implementation process of this disclosure will be described in detail below with reference to specific embodiments.

[0045] Figure 1 This is a flowchart of a non-destructive identification method for self-piercing friction riveting joints according to an embodiment of this disclosure.

[0046] Firstly, referring to Figure 1 This disclosure provides a non-destructive testing method for self-piercing friction riveting joints. The method is based on the correspondence between the pressure signal during the riveting process and the macro- and micro-morphological features of the joint, enabling online non-destructive testing of the joint's macroscopic deformation mode and microstructure type. Specifically, the non-destructive testing method includes: S11. Obtain the pressure signal during the self-piercing friction riveting process; During the F-SPR riveting process, the pressure signal of the entire riveting process is acquired in real time by a pressure acquisition module installed on the riveting equipment. The pressure acquisition module originally acquires the pressure-time signal during the riveting process; since the feed speed in the F-SPR process is a set value, the time can be converted into feed depth / pressing depth, so the synchronized curve can also be represented as a pressure-feed depth signal.

[0047] S12. Preprocess the pressure signal to obtain a homogenized pressure signal; In some embodiments, step S12 includes: S121. Extract the effective signal segment from the pressure signal from the point where the rivet contacts the plate to the completion of the riveting; Extract the valid signal segment from the point where the rivet contacts the plate to the completion of the riveting from the collected complete pressure signal, and remove invalid time periods before and after the riveting begins.

[0048] S122. Remove zero-point drift and environmental interference noise from the effective signal segment and output the denoised effective signal segment; The effective signal segment is subjected to noise processing to remove zero drift, environmental vibration interference, and random equipment noise, resulting in a denoised effective signal segment.

[0049] S123. The effective signal segment of the noise reduction is resampled at equal intervals to obtain a homogenized pressure signal.

[0050] The effective denoised signal segment is resampled at equal intervals, so that the sampling points are evenly distributed on the time axis or feed depth axis to obtain a homogenized pressure signal. Specifically, the number of sampling points can be set according to the subsequent feature extraction requirements, preferably one less than the original number of points. Through resampling, redundant data interference is eliminated while fully preserving key process information, so that the signal samples are evenly distributed on the target axis.

[0051] Optionally, the signal can be differentially divided, normalized, or otherwise preprocessed before resampling to enhance the subsequent feature extraction effect.

[0052] In other implementations, equal-interval resampling can be achieved by the following alternative methods: resampling at a uniform time step, resampling at a uniform feed depth step, normalizing signals of different lengths and then interpolating to a fixed number of points, or resampling using linear interpolation, cubic spline interpolation, or shape-preserving interpolation.

[0053] S13. Perform detrending processing on the homogenized pressure signal to obtain a detrended pressure signal; This step processes the low-frequency trend term and baseline drift component in the homogenized pressure signal. The aim is to reduce the influence of low-frequency background components such as equipment stiffness changes, thermal accumulation effects, and overall downward pressure trends on subsequent feature extraction, while retaining the local dynamic response features related to rivet insertion, material flow, rivet expansion, and mechanical interlocking.

[0054] In some embodiments, step S13 includes: S131. The homogenized pressure signal is divided into multiple continuous segments, wherein adjacent segments of the multiple continuous segments are provided with overlapping regions; The homogenized pressure signal is divided into multiple continuous segments, with overlapping regions between adjacent segments. Specifically, the number of segments (i.e., the number of sub-segments) is determined based on the effective signal length, sampling frequency, and the rate of change of the low-frequency trend in the signal. Preferably, the number of segments can be 5 to 20, or the length of each segment can be set to cover the time range corresponding to the target low-frequency trend change characteristics, thereby ensuring that the trend fitting within each segment is statistically significant. Meanwhile, the overlap ratio between adjacent segments is preferably 10% to 50%, and more preferably 20%. This overlapping region provides a buffer zone for smooth transitions when subsequently splicing the trend curves of each segment, effectively avoiding signal jumps or distortions at the boundary connections caused by segmented fitting.

[0055] S132. Fit local trend curves to each continuous segment respectively; For each segment, local trend fitting is performed. Specifically, the least squares polynomial fitting method can be used to fit the low-frequency trend term in each continuous segment to obtain the local trend curve of the corresponding segment.

[0056] S133. In the overlapping area, the local trend curves of the adjacent segments are weighted and smoothed to reconstruct the complete trend curve; Within the overlapping areas of adjacent segments, the corresponding local trend curves are weighted and fused, so that the trend values ​​within the overlapping area are continuously transitioned from the two local trend curves according to preset weights, thereby forming a continuous and smooth trend curve connection result. Specifically, the weighted fusion can be implemented using Tukey window, Hanning window, Hamming window, or other weighted window functions.

[0057] S134. Subtract the homogenized pressure signal from the complete trend curve point by point to obtain the detrended pressure signal.

[0058] The homogenized pressure signal is subtracted point-by-point from the reconstructed complete trend curve at the same sampling point to obtain the detrended pressure signal. This detrended pressure signal primarily retains the local fluctuation components and process excitation information in the pressure changes, while the low-frequency trend term is significantly weakened. The output is the detrended pressure fluctuation signal, not a simply smoothed signal.

[0059] The above processing can effectively reduce the influence of low-frequency background components such as overall downward pressure trend, equipment stiffness changes and heat accumulation, while retaining the local dynamic response characteristics related to material flow, rivet deformation and mechanical interlocking, thereby improving the stability and repeatability of subsequent bimodal feature extraction.

[0060] S14. Perform wavelet analysis on the detrended pressure signal and identify the target characteristic frequency band located near the device's natural frequency based on the wavelet energy distribution; In the signal feature extraction stage, wavelet analysis is performed on the detrended pressure signal. By performing broadband continuous wavelet analysis on the detrended pressure signal, a full-band time-frequency scan is conducted. Based on the wavelet energy distribution at different frequency scales, response frequency bands with concentrated energy are identified, and target feature frequency bands located near the inherent frequency of the equipment system are determined. These target feature frequency bands correspond to the main dynamic responses generated by processes such as rivet insertion, material flow, and rivet deformation, providing an analysis frequency band for subsequent feature extraction.

[0061] In some embodiments, the target characteristic frequency band is located near the natural frequency of the equipment system. Because the self-piercing friction riveting process generates dynamic excitations on the equipment system during stages such as rivet insertion, material plastic flow, rivet deformation, and the formation of mechanical interlocks, the pressure signal experiences energy amplification near the equipment's natural frequency. Therefore, wavelet energy distribution can be used to identify the target characteristic frequency band containing joint forming feature information.

[0062] S15. Perform narrowband wavelet reconstruction within the target characteristic frequency band to obtain a bimodal characteristic curve containing the first peak amplitude A and the second peak amplitude B. After determining the target characteristic frequency band, narrowband continuous wavelet analysis is performed around the target characteristic frequency band to improve the temporal resolution of the dynamic response within the target frequency band, and wavelet coefficients within the corresponding scale range are extracted. Signal reconstruction is then performed based on these wavelet coefficients to obtain a bimodal characteristic curve. The bimodal characteristic curve includes a first peak amplitude A and a second peak amplitude B. This bimodal characteristic curve can characterize the dynamic response features at different stages of the riveting process, providing characteristic parameters for subsequent discrimination of the macroscopic deformation state and microstructure evolution state of the joint.

[0063] In this embodiment, the Morse wavelet is preferably used as the mother wavelet for continuous wavelet transform. The Morse wavelet possesses good time-frequency localization capabilities and adjustable parameter characteristics, enabling stable time-frequency localization near the target characteristic frequency band and ensuring that the extracted first peak amplitude A and second peak amplitude B have good peak distinguishability, repeatability, and class separability. It should be understood that other wavelet functions such as the Morlet wavelet, complex Morlet wavelet, Mexican Hat wavelet, Daubechies wavelet, and Symlet wavelet can also be used. After appropriate parameter adaptation, target characteristic frequency band identification and bimodal feature extraction can also be achieved.

[0064] S16. Extract the first peak amplitude A, the second peak amplitude B, and the peak ratio R=B / A from the bimodal characteristic curve, wherein the first peak amplitude A and the second peak amplitude B are used to characterize the macroscopic deformation state of the joint, and the peak ratio R is used to characterize the microstructure evolution state of the joint. To illustrate the correspondence between the bimodal characteristic parameters and the macro- and micro-morphological features of the joint, the physical meaning of the bimodal characteristic parameters will be explained below.

[0065] In this embodiment, through macroscopic cross-sectional observation and microstructure analysis of multiple joint samples, it was found that there is a correspondence between the two-dimensional feature space formed by the first peak amplitude A and the second peak amplitude B and the macroscopic deformation mode of the rivet; and there is a correspondence between the peak ratio R=B / A and the microstructure evolution state of the joint. Therefore, online non-destructive discrimination of the macro- and micro-morphology of the joint can be achieved based on A, B, and R.

[0066] Specifically, the first peak amplitude A is the first response peak amplitude in the bimodal characteristic curve corresponding to the early stage of rivet insertion and the initial material flow stage. The first peak amplitude A is not the direct pressure peak in the original pressure signal, but a characterization parameter of the dynamic response characteristics corresponding to the target characteristic frequency band. It mainly reflects the dynamic excitation response generated by the early interaction between the rivet and the material, interfacial friction, and the local deformation process of the material.

[0067] The second peak amplitude B is the second response peak amplitude in the bimodal characteristic curve corresponding to the later stages of riveting: rivet deformation, material filling of the rivet cavity, rivet expansion, and mechanical interlock formation. The second peak amplitude B is also a dynamic response characteristic parameter, rather than the direct pressure peak in the original pressure signal.

[0068] Therefore, the first peak amplitude A and the second peak amplitude B can jointly reflect the macroscopic forming state of the joint, while the peak ratio R=B / A can reduce the influence of the overall amplitude change on the discrimination result and more effectively characterize the relative relationship between the early riveting response and the later forming response, thus being used to characterize the microstructure evolution state of the joint.

[0069] The bimodal formation mechanism can be summarized as follows: Significant differences exist in the material flow state and rivet stress state at different stages of the F-SPR riveting process, resulting in two main excitation enhancement stages near the target characteristic frequency of the equipment system. Specifically, the first peak amplitude A mainly reflects the friction, insertion, and material torsion states in the early stages of riveting; the second peak amplitude B mainly reflects the rivet expansion, material filling, mechanical interlocking formation, and local instability states in the later stages of riveting.

[0070] It should be noted that, in addition to the first main peak A and the second main peak B, the bimodal characteristic curve may also contain several local fluctuation peaks with lower amplitudes. These local fluctuation peaks are not used as discriminative features; only the first main peak A and the second main peak B are used for subsequent macroscopic and microscopic morphology discrimination.

[0071] Based on the correspondence between the above-mentioned characteristic parameters and the macro- and micro-morphology of the joint, this embodiment further sets up two pre-trained discrimination models, which are used for automatic discrimination of the macro-deformation mode and micro-structure type of the rivet, respectively.

[0072] S17. Input the first peak amplitude A and the second peak amplitude B into a macroscopic deformation discrimination model established based on offline calibration data to discriminate the macroscopic deformation mode of the joint and obtain a macroscopic discrimination result; input the peak ratio R into a microstructure discrimination model established based on offline calibration data to discriminate the microstructure evolution state of the joint and obtain a microstructure discrimination result; In the intelligent discrimination stage, pre-trained macroscopic deformation discrimination models and microstructure discrimination models are used to automatically discriminate the macroscopic deformation mode and microstructure evolution state of the joint.

[0073] The macroscopic deformation discrimination model uses the first peak amplitude A and the second peak amplitude B as input features, that is, it uses the two-dimensional feature vector (A, B) as the discrimination criterion to output the macroscopic deformation mode of the rivet corresponding to the joint. Analysis of offline calibration samples reveals that different macroscopic deformation modes have different distribution areas in the (A, B) two-dimensional feature space. Therefore, the corresponding macroscopic deformation mode can be determined based on the first peak amplitude A and the second peak amplitude B. Preferably, the macroscopic deformation discrimination model can classify the joint into three typical macroscopic deformation modes, including: no obvious rivet deformation (macroscopic mode 1), rivet outward deformation (macroscopic mode 2), and rivet bending deformation (macroscopic mode 3); wherein, rivet outward deformation is a preset optimal macroscopic shape.

[0074] The microstructure discrimination model uses the peak ratio R = B / A as input feature and outputs the microstructure type corresponding to the joint. Analysis of offline calibrated samples revealed that different microstructure types correspond to different peak ratio ranges. Therefore, the corresponding microstructure evolution state (or microstructure type) can be determined based on the peak ratio R. Preferably, the microstructure discrimination model can identify different microstructure types such as layered fine-grained regions (microstructure mode I), continuous fine-grained regions (microstructure mode II), and flat fine-grained regions (microstructure mode III); among which, the flat fine-grained region is a preset optimal microstructure type.

[0075] The peak ratio R=B / A can effectively characterize the microstructure because this ratio can reduce the influence of the degree of material softening and the overall pressure amplitude change on the absolute value of a single peak under different process parameters. It can better reflect the relative relationship between the material torsion state in the early stage of riveting and the forming excitation state in the later stage, and this relative relationship is closely related to the microstructure evolution process of the joint.

[0076] Both the macroscopic deformation discrimination model and the microstructure discrimination model are pre-trained discrimination models. During the offline training phase, joint samples are prepared using process parameters such as different rotational speeds and feed rates. Pressure signals are collected, and the first peak amplitude A, the second peak amplitude B, and the peak-to-peak ratio R are extracted. Simultaneously, corresponding macroscopic deformation mode labels and microstructure type labels are obtained through macroscopic cross-sectional observation, metallographic section analysis, and / or EBSD analysis. The discrimination boundary is determined based on the aforementioned feature parameters and corresponding labels. In the online application phase, only the pressure signal of the new joint needs to be collected and the corresponding feature parameters extracted to complete the non-destructive automatic discrimination of the joint's macroscopic and microscopic morphology, without the need for sectioning or other destructive testing.

[0077] Preferably, the macroscopic deformation discrimination model and the microscopic organization discrimination model are implemented using a support vector machine (SVM) classifier. Those skilled in the art will understand that, when trained using the same feature parameters, machine learning models such as neural networks, decision trees, random forests, logistic regression, and K-nearest neighbors can also be used to discriminate the macroscopic and microscopic morphology of the joint.

[0078] In a specific example, the macroscopic deformation discrimination model is implemented using a Gaussian kernel soft-margin support vector machine. Its model parameters can be obtained through offline training, such as the kernel function parameter σ, the penalty coefficient C, and the bias term b (e.g., σ=74.55, C=386.95, b=8.61). These parameters are merely a set of example parameters obtained from training in this embodiment and do not constitute a limitation on the scope of this disclosure. Those skilled in the art can redetermine the corresponding model parameters based on the training samples, or use other support vector machine models or other machine learning models to achieve the same function.

[0079] When a joint simultaneously meets the preset optimal macroscopic morphology and preset optimal microstructure type, it can be further determined that the joint is a high-performance qualified joint.

[0080] S18. Based on the macroscopic discrimination result and the microscopic discrimination result, output the macroscopic and microscopic morphology determination result of the connector.

[0081] In some embodiments, the macro-micro morphology determination results include the macroscopic deformation mode and microstructure type of the rivet; a joint determination is made based on the macroscopic deformation mode and the microstructure type of the rivet to obtain the final quality evaluation result of the joint.

[0082] Preferably, when the macroscopic deformation mode of the rivet is a preset optimal macroscopic shape (i.e., rivet outward deformation, macroscopic mode 2) and the microstructure type is a preset optimal structure type (i.e., flat fine grain region, microstructure mode III), the joint is determined to be a high-performance qualified joint.

[0083] In this embodiment of the disclosure, the quality level of the joint can be further output based on the macroscopic discrimination result and the microscopic discrimination result, such as qualified joint, joint to be optimized and unqualified joint, for use in quality classification management and process parameter optimization during the production process.

[0084] The judgment results can be used for online quality monitoring of the production line, adjustment of process parameters, and traceability of joint quality.

[0085] Through the above steps, the simultaneous non-destructive online identification of the macro- and micro-morphology of F-SPR joints is achieved. Without cutting or damaging the workpiece, the macro-deformation mode of the rivet and the micro-structure type inside the joint can be identified at the same time. It has the advantages of high identification accuracy, strong adaptability and high detection efficiency, and can meet the application requirements of real-time quality monitoring of automated production lines.

[0086] In some embodiments, the non-destructive discrimination method further includes online updating and optimization of the discrimination model, specifically including: acquiring the pressure signal of the newly added joint sample and extracting the corresponding first peak amplitude, second peak amplitude, and peak ratio; performing macroscopic cross-sectional measurement and / or metallographic sectioning and annotation on the newly added joint sample to obtain the corresponding macroscopic deformation mode label and microstructure type label; adding the first peak amplitude, the second peak amplitude, the peak ratio, and the corresponding macroscopic deformation mode label and microstructure type label to the training sample library, and retraining or calibrating the discrimination boundary of the macroscopic deformation discrimination model and the microstructure discrimination model based on the updated training sample library.

[0087] In this embodiment, the new samples can be obtained from joint samples acquired during subsequent production. As the training sample library expands, the discrimination model can continuously optimize its discrimination performance using the new samples to adapt to the joint morphology discrimination requirements under different aluminum alloy materials, different rivet specifications, and different process parameters.

[0088] Among them, "retraining" refers to rebuilding the classification model parameters using the expanded training sample library; "discrimination boundary calibration" refers to adjusting and optimizing the discrimination boundary using new samples while keeping the original model structure unchanged, so as to improve the model's adaptability to current production conditions.

[0089] Optionally, the retraining or discrimination boundary calibration process can be triggered when the number of new samples reaches a preset sample number threshold, the discrimination accuracy of the discrimination model falls below a preset accuracy threshold, or the joint material system or process parameter range changes. It should be understood that the above triggering methods are merely examples, and this disclosure does not limit specific update triggering conditions.

[0090] Through the above update mechanism, the discrimination model can continuously learn from new sample information, achieve adaptive optimization of the discrimination model, and thus improve the model's generalization ability and discrimination accuracy under different material systems and process conditions.

[0091] Figure 2 This is a schematic diagram of the module structure and data flow of a non-destructive testing system for self-piercing friction riveting joints according to an embodiment of this disclosure.

[0092] The system includes a pressure signal acquisition module, a signal preprocessing module, a segmented overlap detrending unit, a feature curve extraction unit, a dual discriminator unit, and a result output unit. The modules are connected sequentially according to the data flow direction. Specifically, the pressure signal acquisition module acquires the pressure signal during the riveting process; the signal preprocessing module performs effective segment truncation, noise reduction, and homogenization on the pressure signal; the segmented overlap detrending unit extracts and removes low-frequency trend terms from the pressure signal; the feature curve extraction unit obtains a bimodal feature curve containing a first peak amplitude A and a second peak amplitude B through wavelet analysis; the dual discriminator unit discriminates the macroscopic deformation mode and microstructure type of the joint based on the first peak amplitude A, the second peak amplitude B, and the peak ratio R=B / A; and the result output unit outputs the macroscopic and microscopic morphology determination results of the joint based on the discrimination results.

[0093] Figure 2 Each module of the system shown is used to implement the corresponding steps in the aforementioned non-destructive discrimination method. The specific processing procedures and implementation methods of each module can be found in the aforementioned method embodiments; the same content will not be repeated here. The following combines... Figure 2 The composition and data flow of each module are explained. Those skilled in the art will understand that... Figure 2 Each module shown can be implemented in software, hardware circuits, programmable logic devices, or a combination of software and hardware; this disclosure does not limit the implementation in this regard.

[0094] (a) Pressure signal acquisition module The pressure acquisition device for the corresponding equipment can be installed on the base of the flat die self-piercing friction riveting equipment to acquire riveting pressure signals in real time during the riveting process. The pressure acquisition module originally acquires a pressure-time signal; since the feed rate in the F-SPR process is usually known or preset, the time axis can be converted into a feed depth axis, and correspondingly, the pressure signal can also be expressed as a pressure-feed depth signal.

[0095] (ii) Signal preprocessing module The signal preprocessing module is used to preprocess the acquired pressure signal, including effective signal segment truncation, noise reduction, zero drift removal, and homogenization. Homogenization can be achieved through methods such as equal-interval sampling, resampling, and interpolation.

[0096] (iii) Segmented Overlapping Detrending Units The segmented overlapping detrending unit is used to extract and remove low-frequency trend terms from the pressure signal to obtain a detrended pressure signal. This unit can construct a complete trend curve using methods such as segmented trend fitting, overlapping weighting, and smooth splicing, and obtain the detrended pressure signal by subtracting the trend curve from the homogenized pressure signal point by point.

[0097] The output of this unit preserves local pressure fluctuations and process excitation characteristics, reduces slow baseline drift, and provides input for subsequent wavelet frequency domain extraction of bimodal features.

[0098] (iv) Feature curve extraction unit (wavelet transform feature extraction unit) The feature curve extraction unit performs wavelet analysis on the detrended pressure signal to identify the target characteristic frequency band and extract a bimodal feature curve containing a first peak amplitude A and a second peak amplitude B. The feature curve extraction process can be implemented using a combination of broadband and narrowband wavelet analysis. The bimodal feature curve serves as the input feature for subsequent macroscopic deformation classifiers and microscopic tissue classifiers.

[0099] (v) Dual discriminator unit The dual discriminator unit includes a pre-trained macroscopic deformation discrimination model and a microstructure discrimination model. The macroscopic deformation discrimination model takes a first peak amplitude A and a second peak amplitude B as input and outputs the macroscopic deformation pattern of the rivet; the microstructure discrimination model takes the peak ratio R = B / A as input and outputs the joint microstructure type. The macroscopic deformation discrimination model and the microstructure discrimination model can be implemented using support vector machines or other machine learning models and are pre-trained based on labeled samples.

[0100] (vi) Result Output Unit The result output unit generates macro- and micro-morphological judgment results for the joint based on the macro- and micro-discrimination results, and can further output the joint quality grade. The macro- and micro-morphological judgment results can be displayed on the human-machine interface, or sent to the production line control system, quality monitoring system, or data management platform.

[0101] (vii) Model update unit (optional) In some embodiments, the system further includes a model update unit, used to update the training sample library based on newly added labeled samples, and to retrain or calibrate the discrimination boundaries of the macroscopic deformation discrimination model and the microstructure discrimination model, so as to improve the model's adaptability to different materials and process conditions. The model update unit is an optional module and does not affect the basic implementation of the nondestructive discrimination system of this disclosure.

[0102] Figure 3 This is a bimodal characteristic curve of the riveting pressure signal in a specific embodiment of this disclosure.

[0103] like Figure 3 As shown, the curve is Figure 2 The bimodal feature curve output by the feature curve extraction unit is obtained by wavelet transform of the detrended pressure signal. Figure 3 The horizontal axis represents the riveting feed depth (mm), and the vertical axis represents the characteristic signal amplitude.

[0104] The bimodal characteristic curve exhibits two distinct main peaks, corresponding to the amplitude of the first peak A and the amplitude of the second peak B, respectively.

[0105] Among them, the first peak amplitude A is located in the early stage of the riveting process, corresponding to the first excitation response peak generated when the rivet penetrates the upper plate and the early flow stage of the material. It reflects the excitation response intensity generated by the initial contact between the rivet and the plate, the interface friction, and the local plastic deformation process of the material on the system.

[0106] The second peak amplitude B is located in the later stage of the riveting process. It corresponds to the second excitation response peak generated during the rivet deformation, material filling of the rivet cavity, and mechanical interlocking stages. It reflects the excitation response intensity generated by the system's natural frequency during the later forming stage.

[0107] It should be noted that, in addition to the two main peaks corresponding to the first peak amplitude A and the second peak amplitude B, there may also be several local fluctuation peaks with smaller amplitudes on the bimodal characteristic curve. These local fluctuation peaks are mainly caused by transient disturbances during the riveting process, local material flow fluctuations, or equipment vibrations, and are not used as the main characteristic parameters for subsequent morphology identification.

[0108] Figure 3 The results of the bimodal characteristic curve formation and the extraction methods of the first peak amplitude A and the second peak amplitude B are intuitively demonstrated, providing feature inputs for subsequent macroscopic deformation discrimination models and microscopic tissue discrimination models to distinguish the macroscopic and microscopic morphology of joints.

[0109] Figure 4 This is a schematic diagram of the feature space distribution of the connector macro-micro morphology discrimination feature based on peak features involved in the embodiments of this disclosure, used to show the distribution law of F-SPR connector samples prepared under different process parameters in the feature space and the corresponding discrimination boundary.

[0110] Figure 4 The horizontal axis represents the first peak amplitude A, and the vertical axis represents the second peak amplitude B. Each data point represents an F-SPR connector sample. The coordinates of the point are determined by the (A, B) extracted from the connector, and are labeled with the corresponding process parameter labels (in the format of "rotation speed-feed speed", for example, "60-2" means rotation speed of 60 r / s and feed speed of 2 mm / s).

[0111] Figure 4 Divided into two sub-images, left and right: Figure 4 Figure 'a' shows the distribution of macroscopic deformation patterns in the feature space. Three typical macroscopic deformation patterns are marked with different shapes: no obvious deformation of the rivet (macroscopic pattern 1), outward deformation of the rivet (macroscopic pattern 2, optimal), and bending deformation of the rivet (macroscopic pattern 3). Different patterns exhibit clear clustering distributions in the (A, B) feature space. The dashed line in the figure represents the classification decision boundary determined by the macroscopic deformation discrimination model, used to distinguish between the optimal and non-optimal macroscopic deformation patterns.

[0112] Figure 4 b represents the distribution of microstructure types in the feature space. The figure uses different shapes to label three typical microstructure types: layered fine-grained regions (microstructure mode I), continuous fine-grained regions (microstructure mode II), and flat fine-grained regions (microstructure mode III, optimal). Different microstructure types exhibit a regular distribution along the peak ratio direction. The dashed line in the figure represents the discrimination threshold line corresponding to the peak ratio R=B / A, which is represented as a straight line passing through the origin in the AB plane, used to distinguish between optimal and non-optimal microstructure types.

[0113] Figure 4 The correspondence between the feature parameters established in this disclosure and the macro- and micro-morphology of the joint is illustrated, verifying the feasibility of using the first peak amplitude A, the second peak amplitude B, and the peak ratio R as features for discriminating the macro- and micro-morphology of the joint. Specifically, a stable clustering mapping relationship exists between the two-dimensional feature space formed by A and B and the macro-deformation mode of the rivet, and a stable threshold correspondence exists between R=B / A and the microstructure type of the joint. Based on the above correspondence, after extracting the first peak amplitude A and the second peak amplitude B from a new joint, the macro-deformation mode can be discriminated based on their positions in the two-dimensional feature space, and the microstructure type can be discriminated based on the positional relationship between the peak ratio R=B / A and the threshold line, thereby achieving online non-destructive discrimination of the macro- and micro-morphology of the joint.

[0114] Figure 5 This is a typical macro- and micro-morphological reference diagram of the F-SPR joint involved in the embodiments of this disclosure. It is used to show the standard reference morphology of the joint’s typical macro-deformation mode and typical micro-structure type, and to provide a reference for offline sample labeling, discrimination model training and online discrimination result verification.

[0115] (I) Macroscopic Deformation Model Diagram The top row displays the joint cross-sectional morphology of three typical macroscopic deformation modes, each corresponding to a different rivet deformation state.

[0116] Macroscopic Pattern 1 (No obvious deformation of the rivet): After the rivet penetrates the sheet metal, it basically maintains its original shape, the rivet leg does not expand significantly, and the mechanical interlock with the underlying material is weak. This pattern usually corresponds to insufficient material flow.

[0117] Macroscopic mode 2 (rivet outward deformation, optimal): The rivet leg expands outward uniformly and forms a stable mechanical interlocking structure. The material is fully filled and the overall forming quality of the joint is high, which corresponds to the optimal macroscopic deformation mode defined in this disclosure.

[0118] Macroscopic Mode 3 (Rivet Bending Deformation): The rivet leg undergoes significant deflection or bending, leading to asymmetry or localized failure of the interlocking structure and a decline in joint forming quality. This mode typically corresponds to abnormal material flow or rivet stress imbalance.

[0119] (II) Microstructure Type Diagram The bottom row displays metallographic images of three typical microstructure types side by side, with the dotted lines in the images indicating the distribution range of the fine-grained region.

[0120] Microscopic pattern I (layered fine-grained region): The fine-grained structure is distributed in layers and has poor continuity.

[0121] Microscopic Model II (Continuous Fine-Grained Region): The fine-grained structure is continuously distributed along the interface, and the range of the fine-grained region is larger than that of Model I, with improved microstructure continuity.

[0122] Microstructure Mode III (Flat Fine-grained Region, Optimal): The fine-grained structure forms a wide and continuous flat fine-grained band in the connecting region. The fine-grained region is relatively large and has good uniformity, which corresponds to the optimal microstructure type defined in this disclosure.

[0123] Figure 5 The macroscopic models 1-3 and microscopic models I-III shown are Figure 4 The standard labels corresponding to each training sample. Specifically, Figure 4 In the left figure, macroscopic patterns 1, 2, and 3 correspond to respectively Figure 5 The top row shows three macroscopic deformation forms; Figure 4 The microscopic models I, II, and III in the right figure correspond to respectively Figure 5 The bottom row shows three microstructure types. These labels were obtained through joint cross-sectional observation, metallographic analysis, and / or EBSD analysis, and serve as standard labels during the training and validation of the discrimination model.

[0124] During the offline training phase, Figure 5 The standard reference morphology shown serves as the basis for sample annotation; during the online discrimination process, the discrimination model outputs the corresponding macroscopic pattern and microstructure type based on the feature parameters (first peak amplitude A, second peak amplitude B, and peak ratio R=B / A), and compares them with... Figure 5The standard reference morphology is compared and verified. In a specific example, the macroscopic optimal mode 2 can be implemented using a Gaussian kernel soft-margin support vector machine classifier with the first peak amplitude A and the second peak amplitude B as inputs (e.g., σ=74.55, C=386.95, b=8.61), and the microscopic optimal tissue type III can be distinguished by comparing the peak ratio R=B / A with a preset threshold (e.g., T0=5.543). The above parameters are only one example implementation of this disclosure, and this disclosure is not limited thereto.

[0125] therefore, Figure 5 The standard reference morphology shown can achieve a one-to-one correspondence between feature parameters, classification results and actual macro- and micro-morphology of the joint, and verify the accuracy of the discrimination model output results.

[0126] Based on the non-destructive discrimination method disclosed herein, this disclosure has the following technical effects: 1) Achieve simultaneous non-destructive online identification of the macro- and micro-morphological characteristics of F-SPR connectors. To address the technical problem that existing technologies cannot simultaneously identify the macroscopic and microscopic morphology of F-SPR joints and typically rely on offline destructive testing, this disclosure achieves online non-destructive testing of F-SPR joint quality by using a complete processing chain including pressure signal preprocessing, segmented overlap detrending, wavelet transform bimodal feature extraction, and joint discrimination using dual discriminant models. This enables simultaneous discrimination of the macroscopic deformation mode of the rivet and the microscopic structure type inside the joint without cutting or damaging the workpiece.

[0127] 2) The joint morphology is highly accurate and has good process adaptability. To address the limitations of existing methods in adapting to the F-SPR process and their insufficient discrimination stability, this disclosure employs a dual-classifier structure that combines a macroscopic deformation discrimination model and a microstructure discrimination model. Using the first peak amplitude A, the second peak amplitude B, and the peak ratio R = B / A as feature inputs, stable classification of macroscopic and microscopic morphologies is achieved. Experimental results show that, under different combinations of process parameters and different aluminum alloy materials, the method of this disclosure maintains high and stable discrimination performance, demonstrating good generalization ability and engineering adaptability.

[0128] 3) Reduce testing costs and improve production line testing efficiency This disclosure replaces traditional offline testing methods such as metallographic sectioning and tensile failure, enabling rapid identification of individual joints on the production line without the need for sample preparation and destructive testing. This significantly reduces testing costs and the burden of manual testing, while improving the overall production cycle time. It is suitable for the online quality monitoring needs of mass automated production lines.

[0129] 4) Support process optimization and closed-loop quality control This disclosure establishes a correspondence between characteristic parameters A, B, and R=B / A and macro- and micro-morphological features, enabling quantifiable characterization of the joint morphology. This can be used to locate signal feature intervals corresponding to different macro- and micro-morphological features, thus providing a basis for optimizing F-SPR process parameters and upgrading the quality control mode from "post-inspection" to "process feedback control." Simultaneously, the discrimination model can be updated or its parameters calibrated based on new sample data to continuously improve discrimination stability and process adaptability.

[0130] Figure 6 This is a block diagram of an electronic device according to an embodiment of the present disclosure.

[0131] Secondly, referring to Figure 6 This disclosure provides an electronic device comprising: One or more processors 601; The memory 602 stores one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the non-destructive identification method for self-piercing friction riveting joints described in any of the above-mentioned methods. One or more I / O interfaces 603 are connected between the processor and the memory and configured to enable information exchange between the processor and the memory.

[0132] The processor 601 is a device with data processing capabilities, including but not limited to a central processing unit (CPU); the memory 602 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), and flash memory (FLASH); the I / O interface (read / write interface) 603 is connected between the processor 601 and the memory 602, enabling information exchange between the processor 601 and the memory 602, including but not limited to a data bus (Bus).

[0133] In some embodiments, the processor 601, memory 602, and I / O interface 603 are interconnected via bus 604, and thus connected to other components of the computing device.

[0134] The electronic device may be an industrial computer, an embedded system, an edge computing gateway, or a control unit integrated into the riveting equipment.

[0135] Thirdly, embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the non-destructive identification method for self-piercing friction riveting joints described in any of the preceding claims.

[0136] The non-destructive discrimination method disclosed herein can be widely applied in the following fields, including but not limited to: 1) Automotive manufacturing sector: Applied to the body assembly line of new energy vehicles or traditional fuel vehicles, covering the connection process of integrated die-cast aluminum structural parts, deformed aluminum alloy profiles and cover parts, realizing real-time monitoring of joint quality and automatic optimization of process parameters, adapting to the mass production needs of lightweight automotive bodies.

[0137] 2) Aerospace field: Used for F-SPR connection quality inspection of aluminum alloy thin-walled structures and frame components of aerospace vehicles to meet the high reliability and high precision quality control requirements of aerospace equipment.

[0138] 3) Rail transit sector: Applied to the connection process of aluminum alloy car body structure of high-speed rail and subway cars, enabling online detection of car body joint quality and ensuring the structural strength and operational safety of rail transit equipment.

[0139] 4) Intelligent riveting equipment integration: It can be directly embedded into F-SPR automated riveting equipment as a standard intelligent quality monitoring module, realizing the intelligent upgrade of riveting equipment and adapting to the unmanned quality control scenario of intelligent manufacturing factory.

[0140] 5) Light Alloy Joining Technology Development: This involves the research and optimization of self-piercing friction riveting (F-SPR) technology for aluminum alloys. This includes joint performance analysis under different aluminum alloy material systems, research on the influence of process parameters on the macro- and micro-morphology of the joints, and evaluation of connection quality. By extracting and analyzing the characteristic parameters A, B, and R=B / A of the pressure signal, it can guide the optimization of F-SPR process parameters and the evaluation of joint performance, thereby shortening the process development and verification cycle.

[0141] To enable those skilled in the art to more clearly understand the technical solutions provided by the embodiments of this disclosure, the technical solutions provided by the embodiments of this disclosure will be described in detail below through specific embodiments: Example 1 In this embodiment, two layers of JDA1b cast aluminum alloy plates are used as the connecting parts, each with a thickness of 3.0 mm. 35CrMo steel semi-hollow rivets are used, with an outer diameter of 6.0 mm, a shank length of 5.75 mm, a wall thickness of 1.0 mm, and an inner cavity depth of 6.5 mm. A flat die F-SPR device is used for the riveting operation, and the pressure sensor sampling frequency is 200 Hz.

[0142] 1. Sample preparation and data acquisition A 4×4 full-factor process parameter design was adopted, with rotational speeds of 900 rpm, 1800 rpm, 2700 rpm, and 3600 rpm, feed rates of 1 mm / s, 2 mm / s, 4 mm / s, and 6 mm / s, and a fixed compression depth of 5.8 mm, resulting in a total of 16 process parameter combinations.

[0143] For each set of process parameters, at least one joint sample was prepared for cross-sectional morphology observation and microstructure analysis, and three additional samples were prepared for tensile and shear performance testing.

[0144] 2. Feature extraction and labeling Pressure signals during the riveting process of each group of samples were collected, and preprocessed, segmented overlapping detrending and wavelet transform were performed according to the preset signal processing method. The bimodal feature curve was extracted to obtain the amplitude of the first peak A, the amplitude of the second peak B and the peak ratio R=B / A.

[0145] Macroscopic cross-sectional observation and microscopic organization analysis were performed on each group of samples to construct corresponding macroscopic deformation mode labels and microscopic organization type labels. The macroscopic labels include Mode 1, Mode 2, and Mode 3, and the microscopic labels include Type I, Type II, and Type III.

[0146] 3. Model Training The first peak amplitude A and the second peak amplitude B are used as input features for the macroscopic deformation discrimination model, and the peak ratio R=B / A is used as input features for the microscopic organization discrimination model. Classification models are constructed respectively.

[0147] The macroscopic deformation discrimination model uses a Gaussian kernel soft-margin support vector machine (SVM), while the microscopic organization discrimination model uses an R-based threshold classification method or a one-dimensional SVM model.

[0148] After training and validation, the model demonstrated good classification performance on experimental samples.

[0149] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0150] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in connection with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in connection with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of this disclosure as set forth by the appended claims.

Claims

1. A non-destructive identification method for self-piercing friction riveting joints, characterized in that, include: Acquire pressure signals during the self-piercing friction riveting process; The pressure signal is preprocessed to obtain a homogenized pressure signal; The homogenized pressure signal is detrended to obtain a detrended pressure signal; Wavelet analysis is performed on the detrended pressure signal to identify the target characteristic frequency band located near the device's natural frequency based on the wavelet energy distribution; Narrowband wavelet reconstruction is performed within the target characteristic frequency band to obtain a bimodal characteristic curve containing a first peak amplitude A and a second peak amplitude B; The first peak amplitude A, the second peak amplitude B, and the peak ratio R = B / A are extracted from the bimodal characteristic curve, wherein the first peak amplitude A and the second peak amplitude B are used to characterize the macroscopic deformation state of the joint, and the peak ratio R is used to characterize the microstructure evolution state of the joint. The first peak amplitude A and the second peak amplitude B are input into a macroscopic deformation discrimination model established based on offline calibration data to discriminate the macroscopic deformation mode of the joint and obtain macroscopic discrimination results; the peak ratio R is input into a microstructure discrimination model established based on offline calibration data to discriminate the microstructure evolution state of the joint and obtain microstructure discrimination results. Based on the macroscopic discrimination results and the microscopic discrimination results, the macroscopic and microscopic morphology determination results of the connector are output.

2. The non-destructive discrimination method according to claim 1, characterized in that, The preprocessing of the pressure signal to obtain a homogenized pressure signal includes: Extract the effective signal segment from the time the rivet contacts the plate to the completion of the riveting from the pressure signal; Zero-point drift and environmental interference noise are removed from the effective signal segment to output a denoised effective signal segment; The effective denoised signal segment is resampled at equal intervals to obtain a homogenized pressure signal.

3. The non-destructive discrimination method according to claim 1, characterized in that, The process of detrending the homogenized pressure signal to obtain a detrended pressure signal includes: The homogenized pressure signal is divided into multiple continuous segments, wherein adjacent segments of the multiple continuous segments are provided with overlapping regions; Local trend curves were fitted to each continuous segment; In the overlapping region, the local trend curves of the adjacent segments are weighted and smoothed to reconstruct the complete trend curve; The detrended pressure signal is obtained by subtracting the homogenized pressure signal from the complete trend curve point by point.

4. The non-destructive discrimination method according to claim 1, characterized in that, The first peak amplitude A and the second peak amplitude B together constitute a two-dimensional feature space of the macroscopic deformation state of the joint. Different macroscopic deformation modes correspond to different distribution regions in the two-dimensional feature space. The peak ratio R=B / A is used to characterize the microstructure evolution state of the joint. Different microstructure types correspond to different peak ratio ranges.

5. The non-destructive discrimination method according to claim 1, characterized in that, The first peak amplitude A is the first excitation response peak amplitude corresponding to the material insertion and initial flow stage in the early stage of riveting in the bimodal characteristic curve, and the second peak amplitude B is the second excitation response peak amplitude corresponding to the rivet deformation, material filling and mechanical interlocking formation stage in the later stage of riveting in the bimodal characteristic curve.

6. The non-destructive discrimination method according to claim 1, characterized in that, The macroscopic deformation discrimination model and the microstructure discrimination model are implemented using a support vector machine classifier, and the discrimination boundary is determined in advance by training with offline calibration data; wherein, the offline calibration data includes the pressure signal features of joint samples prepared under different process parameters, as well as the corresponding macroscopic deformation mode labels and microstructure evolution state labels.

7. The non-destructive discrimination method according to claim 1, characterized in that, The macro-micro morphology determination results include the macro deformation mode and microstructure type of the rivet; a joint determination is made based on the macro deformation mode and the microstructure type of the rivet; when the macro deformation mode of the rivet is a preset optimal macro morphology and the microstructure type is a preset optimal structure type, the joint is determined to be a high-performance qualified joint.

8. The non-destructive discrimination method according to claim 1, characterized in that, Also includes: Acquire the pressure signal of the newly added connector sample and extract the corresponding first peak amplitude, second peak amplitude, and peak ratio; Macroscopic cross-sectional measurements and / or metallographic sectioning and annotations are performed on the newly added joint samples to obtain the corresponding macroscopic deformation mode labels and microstructure type labels; The first peak amplitude, the second peak amplitude, the peak ratio, and the corresponding macroscopic deformation mode label and microscopic tissue type label are added to the training sample library, and the macroscopic deformation discrimination model and the microscopic tissue discrimination model are retrained or the discrimination boundary is calibrated based on the updated training sample library.

9. An electronic device, comprising: One or more processors; A memory having stored one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the non-destructive discrimination method according to any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program, when executed by a processor, implements claim 1. The non-destructive discrimination method described in any one of the 8.

Images