An ultrasonic phased array defect detection system and method for laser penetration of a build-up weld

Abstract

The present application belongs to the technical field of ultrasonic testing, and provides an ultrasonic phased array defect detection system and method for laser penetration overlap welding, which comprises determining the amplitude threshold value between the layers of the laser penetration overlap welding workpiece; instructing the ultrasonic phased array probe to emit sound waves to the laser penetration overlap welding workpiece to obtain echo signals; extracting the signal envelope line of the preprocessed echo signals through Hilbert transform; determining the super-threshold signal segment with an amplitude greater than the amplitude threshold value in the signal envelope line; extracting the time domain features and frequency domain features of the super-threshold signal segment, and inputting the time domain features and frequency domain features into a pre-trained double-channel CNN model to obtain the probability of incomplete fusion; through the double-channel CNN model, the time domain features and frequency domain features are mined, the time correlation law and frequency distribution mode are learned through parallel processing, and then the time domain and frequency domain features are fused, so that the subtle differences between the incomplete fusion defects and normal interfaces in the time domain and frequency domain features can be amplified.

Description

Technical Field

[0001] This invention belongs to the field of ultrasonic testing technology, and specifically relates to an ultrasonic phased array defect detection system and method for laser-penetrating lap welding. Background Technology

[0002] Laser penetration welding is a welding technology that uses a high-energy laser beam to penetrate the upper layer material, causing multi-layer workpieces to melt and form a metallurgical bond under the action of the laser. It achieves rapid fusion through the high energy density of the laser and is suitable for precision connection of dissimilar materials (such as aluminum alloy and steel) or multi-layer structures. It is widely used in automobile manufacturing, aerospace and other fields due to its high efficiency and low deformation characteristics.

[0003] Ultrasonic phased array is a non-destructive testing technology that controls the excitation time difference of each element in the array probe to achieve focusing, deflection and scanning of the sound beam. It can flexibly adjust the direction of sound wave propagation and focusing depth, and can perform all-round inspection of complex structure workpieces. It has the advantages of high resolution and fast detection speed.

[0004] During lap welding, defects such as incomplete fusion and porosity can easily occur between layers due to uneven laser energy distribution and differences in material thermal expansion. These defects can severely reduce joint strength and even cause structural failure. Therefore, precise testing is required to ensure the quality of laser penetration lap welding.

[0005] In existing technologies, traditional inspection relies on manual visual inspection and simple instrument assistance. Operators observe the weld surface with a microscope or tap the workpiece to listen to the sound to judge internal defects. This method has obvious limitations: it can only detect visible surface defects and cannot effectively identify hidden defects such as lack of fusion between layers. Moreover, the judgment results are greatly affected by experience, and the consistency of interpretation among different personnel is insufficient. The rate of missed detection and false detection remains high, making it difficult to meet the quality requirements of precision manufacturing.

[0006] With technological advancements, automated inspection based on signal capture has emerged: using X-ray inspection technology to record images of interlayer structures. However, this method still has drawbacks: while X-ray inspection can present structural images, its resolution for minute incomplete fusion defects is insufficient, it poses radiation safety hazards, it is not suitable for mass production environments, its inspection efficiency is low, and it is difficult to adapt to the high-efficiency production pace of laser-penetrating lap welding. Summary of the Invention

[0007] To address the problems in the background art, this invention proposes an ultrasonic phased array defect detection system and method for laser-penetrating lap welding.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention proposes a method for detecting defects in ultrasonic phased arrays used in laser-penetrating lap welding, comprising:

[0010] Determine the amplitude threshold between different layers of material in a clad welded workpiece;

[0011] The ultrasonic phased array probe is instructed to emit sound waves at the laser-penetrating lap-welded workpiece, acquire the echo signal, and preprocess the echo signal.

[0012] The signal envelope of the preprocessed echo signal is extracted using Hilbert transform.

[0013] Identify the over-threshold signal segments in the signal envelope whose amplitude is greater than the amplitude threshold;

[0014] Extract the time-domain and frequency-domain features of the super-threshold signal segment, and input the time-domain and frequency-domain features into a pre-trained dual-channel CNN model to obtain the non-fusion probability;

[0015] The probability of non-fusion is compared with a preset threshold to determine whether there is a defect in the laser-penetrated welded workpiece.

[0016] Preferably, determining the amplitude threshold between the layers of material in the laser-penetrated cladding workpiece specifically includes:

[0017] The camera is instructed to capture images of the laser-penetrating lap-welded workpiece to obtain the model number of the laser-penetrating lap-welded workpiece;

[0018] Based on the aforementioned model, determine the acoustic impedance of each layer of material penetrated by the laser in the lap-welded workpiece and the upper limit of the safe gap between each layer of material;

[0019] Based on the predetermined attenuation coefficient, the acoustic impedance, and the upper limit of the safety gap, the amplitude threshold for laser penetration between the layers of the laminated material is determined.

[0020] Preferably, the commanded ultrasonic phased array probe emits sound waves to the laser-penetrating lap-welded workpiece, specifically including:

[0021] The scanning path of the ultrasonic phased array probe is generated based on the predetermined three-dimensional model of the weld, and the ultrasonic phased array probe is instructed to move along the scanning path.

[0022] The excitation delay time of each array element is calculated based on the real-time position of the ultrasonic phased array probe.

[0023] The acoustic beam is emitted based on the array element excitation delay time.

[0024] Preferably, the method for preprocessing the echo signal is high-pass filtering.

[0025] Preferably, determining the over-threshold signal segment in the signal envelope with an amplitude greater than the amplitude threshold involves determining whether there is a sampling point in the signal envelope with an amplitude greater than the amplitude threshold. If so, the sampling point is taken as the starting point, and the signal is continuously tracked to a sampling point in the signal envelope with an amplitude lower than the amplitude threshold, which is taken as the end point of the over-threshold signal segment. If not, the signal segment is filtered out.

[0026] Preferably, the network architecture of the dual-channel CNN model is a parallel 1D-CNN and 2D-CNN, wherein the 1D-CNN is used to process temporal features and the 2D-CNN is used to process frequency domain features.

[0027] Preferably, the time-domain features include peak amplitude, rising edge slope, duration, peak position, and integral area; the frequency-domain features include center frequency, main frequency bandwidth, high-frequency energy ratio, and spectral peak value.

[0028] Preferably, the step of comparing the non-fusion probability with a preset threshold to determine whether there is a defect in the laser-penetrated lap welded workpiece is as follows: if the non-fusion probability is greater than or equal to the preset threshold, then there is a non-fusion defect at the position corresponding to the over-threshold signal segment.

[0029] Secondly, the present invention proposes an ultrasonic phased array defect detection system for laser-penetrating lap welding, comprising:

[0030] The amplitude threshold determination module is used to determine the amplitude threshold between different layers of material in a laser-penetrated cladding workpiece.

[0031] The signal acquisition module is used to instruct the ultrasonic phased array probe to emit sound waves to the laser-penetrating lap-welded workpiece, acquire the echo signal, and preprocess the echo signal.

[0032] The processing module is used to extract the signal envelope of the preprocessed echo signal through Hilbert transform;

[0033] An over-threshold signal segment determination module is used to determine over-threshold signal segments in the signal envelope whose amplitude is greater than the amplitude threshold.

[0034] The extraction module is used to extract the time-domain and frequency-domain features of the over-threshold signal segment, and input the time-domain and frequency-domain features into a pre-trained dual-channel CNN model to obtain the non-fusion probability;

[0035] The judgment module is used to compare the non-fusion probability with a preset threshold to determine whether there is a defect in the laser-penetrated lap-welded workpiece.

[0036] The beneficial effects of this invention are:

[0037] The method of this invention mines time-domain and frequency-domain features using a dual-channel CNN model. By learning time correlation patterns and frequency distribution patterns through parallel processing, and then fusing time-domain and frequency-domain features, it can amplify the subtle differences between the non-fusion defect and the normal interface in the dual-domain features, thereby improving the accuracy of non-fusion probability calculation. It solves the problems of blurred defect features caused by complex signals between weld layers, easy misjudgment in single-dimensional analysis, and low efficiency of manual inspection in the prior art, and achieves high accuracy and efficiency in defect detection.

[0038] The method of this invention transmits acoustic beams based on the excitation delay time of array elements, and compensates for the acoustic propagation path difference by utilizing the transmission time difference between array elements. This can significantly improve the amplitude and signal-to-noise ratio of the echo signal, increase the energy density of the focused acoustic beam, and make the reflected signals from the tiny gaps at the interlayer interface clearly distinguishable from a state that might be submerged by noise. Furthermore, it can ensure that the acoustic energy of different array elements is synchronously superimposed at the target interface, so that the acoustic beam acts on the target interface, reducing energy radiation to non-detection areas, reducing stray echo interference, and making the time-domain characteristics of the echo from the target interface more prominent, thus providing a stable and reliable foundation for subsequent signal analysis.

[0039] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description and the drawings. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A flowchart of the ultrasonic phased array defect detection method for laser-penetrating lap welding according to the present invention is shown;

[0042] Figure 2 A schematic diagram of the device framework of the present invention is shown;

[0043] Figure 3 A schematic diagram of the device structure of the present invention is shown;

[0044] Figure 4 A schematic diagram of the structure of the computer-readable medium of the present invention is shown. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Reference Figure 1 As shown, an ultrasonic phased array defect detection method for laser-penetrating cladding specifically includes the following steps:

[0047] S10. Determine the amplitude threshold between different layers of material in the laser-penetrated lap-welded workpiece;

[0048] S101, Instruct the camera to capture images of the laser-penetrating lap-welded workpiece and obtain the model of the laser-penetrating lap-welded workpiece;

[0049] S102. Based on the model, determine the acoustic impedance of laser penetration through each layer of material in the lap-welded workpiece and the upper limit of the safe gap between each layer of material;

[0050] S103. Based on the predetermined attenuation coefficient, acoustic impedance, and upper limit of the safety gap, determine the amplitude threshold for laser penetration between the layers of the cladding material. The expression for this threshold is: In the formula, For amplitude threshold, The attenuation coefficient is... , For the first Layer and First Acoustic impedance of layered materials, This represents the upper limit of the safety clearance; where acoustic impedance reflects the material's ability to impede ultrasonic waves.

[0051] In step S102, the upper limit of the safety gap between each layer of material is preset. It can be preset manually or directly use industry standards. In step S103, the attenuation coefficient is determined by the data of historical workpieces, that is, by collecting the correspondence between the "gap-reflection amplitude" of multi-layer test blocks of the same material as the actual workpiece under different gaps, and then calculating it through fitting.

[0052] In one specific embodiment, the laser penetrates the lap-welded workpiece, which is a three-layer structure of aluminum alloy-steel-aluminum alloy, with acoustic impedances of 17.01 × 10⁻⁶. 6 kg / (m²・s), 46.02×10 6 kg / (m²・s), 17.01×10 6 kg / (m²・s), upper limit of safe gap between aluminum alloy layers Upper limit of safe gap between steel layers attenuation coefficient Then the amplitude threshold between aluminum alloy and steel is determined to be... .

[0053] S20: Instruct the ultrasonic phased array probe to emit sound waves to the laser-penetrating lap-welded workpiece, acquire the echo signal, and preprocess the echo signal.

[0054] In step S20, the ultrasonic phased array probe is instructed to emit sound waves to penetrate the laser-pierced welded workpiece. Specifically, this involves generating a scanning path for the ultrasonic phased array probe based on a pre-determined three-dimensional model of the weld, instructing the ultrasonic phased array probe to move along the scanning path, calculating the excitation delay time of each array element based on the real-time position of the ultrasonic phased array probe, and emitting a sound beam based on the array element excitation delay time. The three-dimensional model of the weld has been created using laser scanning or other methods before the laser-pierced welded workpiece is detected. Generating the scanning path for the ultrasonic phased array probe based on the pre-determined three-dimensional model of the weld involves first extracting key geometric features such as the central axis, cross-sectional dimensions, and orientation of the weld from the three-dimensional model, then setting the pitch and diameter of the helix based on the effective detection range of the ultrasonic phased array probe, and then generating a continuous helical trajectory extending along the entire length of the weld according to the set parameters, using the central axis as a reference, to ensure that the probe can fully cover the weld area when moving along this trajectory. Generating a helical scanning path based on the three-dimensional model of the weld is common knowledge in the art and will not be further explained in this invention.

[0055] Specifically, the excitation delay time of each array element is calculated based on the real-time position of the ultrasonic phased array probe, and its expression is as follows: In the formula, Let n be the excitation delay time for the nth array element. For the spacing between array elements, Number the array elements. The focusing angle of the sound beam. This refers to the propagation speed of sound waves within the current layer of material. The element spacing is an inherent design parameter of the ultrasonic phased array probe, obtainable directly from the probe's specifications or the manufacturer's information. The element number is a manually assigned identifier based on the arrangement of the probe's elements, which can be preset and recorded. The acoustic beam focusing angle is a parameter determined in advance based on the testing requirements. More specifically, the acoustic beam focusing angle needs to be dynamically set in conjunction with the laser penetration of the interlayer structure of the welded workpiece, the potential depth of defects, and the differences in sound velocity between materials. For example, for a three-layer structure of aluminum alloy-steel-aluminum alloy, when the depth of the interface between aluminum alloy and steel is shallow, such as 0.5mm, The angle can be set to 30-45° to reduce sound wave attenuation in the aluminum alloy layer; when detecting deeper defects inside the steel layer, such as 1-2mm, It can be adjusted to 15-20°, and by using a gentler angle, the propagation path of sound waves in the steel layer is increased, ensuring that the energy fully reaches the target area.

[0056] It is necessary to further explain the speed of sound wave propagation in the current layer of material. It is not possible to rely solely on known data in material handbooks; calibration tests are necessary to determine the actual sound velocity. Even for the same type of material, due to differences in composition and heat treatment, the actual sound velocity may deviate from the nominal value by 5%-10% (e.g., the nominal sound velocity of aluminum alloy is 6320 m / s, but the actual velocity may fluctuate to 6100-6500 m / s due to different magnesium content). Calibration can avoid deviations in the calculation of array element excitation delay time caused by sound velocity errors, which in turn affects the focusing accuracy of the acoustic beam.

[0057] By transmitting acoustic beams based on the excitation delay time of array elements and compensating for the acoustic propagation path difference using the transmission time difference between array elements, the amplitude and signal-to-noise ratio of the echo signal can be significantly improved, the energy density of the focused acoustic beam can be increased, and the reflected signals from tiny gaps at the interlayer interface can be clearly distinguished from a state that might be submerged by noise. Furthermore, it can ensure that the acoustic energy of different array elements is synchronously superimposed at the target interface, so that the acoustic beam acts on the target interface, reducing energy radiation to non-detection areas, reducing stray echo interference, and making the time-domain characteristics of the echo from the target interface more prominent, thus providing a stable and reliable foundation for subsequent signal analysis.

[0058] In step S20 above, the echo signal is preprocessed, specifically by performing high-pass filtering on the echo signal to remove low-frequency noise.

[0059] S30. Extract the signal envelope of the preprocessed echo signal using Hilbert transform;

[0060] The Hilbert transform converts a real-valued signal into an analytic signal, thereby directly obtaining the signal's instantaneous amplitude (i.e., envelope). For real-valued signals... (The preprocessed echo signal), its Hilbert transform The expression is: In the formula, For convolution operations, Let be the kernel function of the Hilbert transform. This is the preprocessed echo signal. From a frequency domain perspective, the Hilbert transform is equivalent to applying a -90° phase shift to the signal, and its expression is: In the formula, For Fourier transform, For symbolic functions, It is the imaginary unit.

[0061] The Hilbert transform can be used to transform real-valued signals. Extended to analytic signals in complex form In the formula, The orthogonal components (90° phase-shifted signals) obtained after Hilbert transform, and the preprocessed echo signals The imaginary part constitutes the analytic signal. The preprocessed echo signal. envelope To analyze the modulus of the signal: .

[0062] S40. Identify the over-threshold signal segments in the signal envelope whose amplitudes exceed the amplitude threshold.

[0063] Specifically, it determines whether there are sampling points in the signal envelope with amplitudes greater than the amplitude threshold. If so, the sampling point is taken as the starting point and the signal envelope is continuously tracked to the sampling point with an amplitude lower than the amplitude threshold, which is taken as the end point of the signal segment exceeding the threshold. If not, it is a harmless gap reflection, and the signal segment can be filtered out.

[0064] By identifying signal segments exceeding the threshold, we can accurately pinpoint regions with significant characteristics within the signal envelope, providing a clear range for subsequent analysis. It's important to further clarify that in practice, the amplitude threshold should be set in conjunction with the envelope amplitude range of normal interface reflections in the detection scenario. For example, if the peak value of the normal interface reflection envelope is typically between 2-3V, the threshold can be set to 3.5V to filter out abnormal signal segments that may contain defects. Simultaneously, during continuous tracking, the continuity of sampling points must be considered. If isolated sampling points below the threshold appear (due to noise fluctuations), the endpoint can be determined by setting a threshold for the number of consecutive sampling points below the threshold (e.g., 3). This avoids misjudgments due to accidental fluctuations, ensuring the integrity and accuracy of the signal segments exceeding the threshold, and providing a reliable signal range basis for defect identification. In a specific embodiment, when the sampling rate is 100MHz (i.e., each sampling point is spaced 10ns apart), if the signal duration corresponding to the minimum axial dimension of the non-fused defect is 500ns (corresponding to 50 sampling points), then the threshold for the number of sampling points that are continuously below the threshold can be set to 5 (corresponding to 50ns). This can filter out the brief amplitude drop caused by instantaneous noise (such as fluctuations of 1-2 sampling points) and avoid dividing the real defect signal into multiple segments.

[0065] S50. Extract the time-domain and frequency-domain features of the signal segment exceeding the threshold, and input the time-domain and frequency-domain features into the pre-trained dual-channel CNN model to obtain the non-fusion probability.

[0066] In step S50 above, the time-domain characteristics include peak amplitude, rising edge slope, duration, peak position, and integral area. The frequency-domain characteristics include: center frequency, main frequency bandwidth, high-frequency energy proportion, and spectral peak value.

[0067] The peak amplitude is extracted as follows: the maximum value of the envelope within the signal segment exceeding the threshold, that is, the amplitude value corresponding to the peak point of the envelope. The peak amplitude is positively correlated with the area of ​​the defect reflection interface and the difference in acoustic impedance. Due to the rough interface and the presence of gaps, the peak amplitude of unfused defects is usually higher than that of normal interlayer reflection.

[0068] The expression for extracting the rising edge slope is: The time difference is the time interval during which the signal segment exceeding the threshold rises from a low-proportion peak amplitude to a high-proportion peak amplitude. The irregular interface of the unfused defect will cause the sound wave reflection to be more rapid, and the rising edge slope is steeper than that of the normal interface reflection.

[0069] Duration: The time span of the over-threshold signal segment from the start point (first time exceeding the amplitude threshold) to the end point (last time falling below the amplitude threshold), corresponding to the propagation path length of the sound wave in the defect area, which is related to the axial dimension of the defect.

[0070] Peak position can enable spatial localization of defects.

[0071] The integral area is the area enclosed by the envelope and the time axis, which comprehensively reflects the total energy reflected by the defect. Due to multi-interface scattering, the integral area of ​​unfused defects is usually larger than that of normal reflection.

[0072] Center frequency: The spectrum is obtained by performing a Fast Fourier Transform (FFT) on the signal segment exceeding the threshold. The frequency corresponding to the energy centroid of the spectrum is calculated (the sum of each frequency component × energy percentage). Due to the unevenness of the interface, the unfused defect will absorb more high-frequency energy, resulting in a lower center frequency than the normal interface reflection.

[0073] Dominant frequency bandwidth: This refers to the frequency range (upper limit frequency - lower limit frequency) within the spectrum where the energy is greater than 50% of the peak energy, reflecting the difference in sound wave reflection at different frequencies due to the defect. The irregular structure of an unfused defect scatters a wider range of frequency components, resulting in a dominant frequency bandwidth that is wider than normal reflection.

[0074] High-frequency energy ratio: The ratio of the total energy above the center frequency in the spectrum to the total energy. The high-frequency components reflected by normal interfaces are preserved intact, while the high-frequency sound waves attenuate more significantly due to the porous structure of unfused defects.

[0075] Peak frequency: The frequency corresponding to the largest amplitude value in the spectrum. This frequency is related to the characteristic size of the defect and can help distinguish the defect type.

[0076] Specifically, the dual-channel CNN model employs a parallel 1D-CNN and 2D-CNN architecture. The 1D-CNN processes temporal features, while the 2D-CNN processes frequency domain features. The 1D-CNN channel receives temporal waveform data and uses a one-dimensional convolutional kernel to capture local correlation features of the signal in the time dimension, adapting to the linear distribution characteristics of temporal signals. The 2D-CNN channel receives frequency domain spectra and uses two-dimensional convolutional kernels to extract two-dimensional distribution features of the spectrum along both the frequency and time axes, adapting to the planar distribution characteristics of frequency domain signals. After parallel processing, a feature fusion layer integrates temporal and frequency domain features, ultimately achieving accurate differentiation between unfused defects and interlayer reflection interference. It should be further noted that the basic network architecture of dual-channel CNN, as well as 1D-CNN for processing temporal data and 2D-CNN for processing image data, is common knowledge known to those skilled in the art. Furthermore, the construction, training process (such as loss function selection and optimizer parameter setting) and performance tuning methods of dual-channel CNN models have been fully described in existing public literature and technical manuals. Therefore, this invention will not provide further detailed descriptions of its specific network parameters, number of layers, and other details.

[0077] S60. Compare the probability of non-fusion with a preset threshold to determine whether there is a defect in the laser-penetrated welded workpiece;

[0078] Specifically, when the non-fusion probability output by the dual-channel CNN model for a signal segment exceeding the threshold is greater than or equal to a pre-set threshold, a non-fusion defect exists at the corresponding location of that signal segment. If the probability is less than the pre-set threshold, it is considered normal interface reflection or non-defect interference (such as micro-gap between layers). More specifically, the spatial location information of the signal envelope can be combined to mark the specific location (e.g., depth) of the defect in the 3D scanning map of the workpiece, and record the corresponding non-fusion probability value (e.g., 0.92). This provides data support for subsequent defect classification (e.g., a probability greater than or equal to 0.9 is classified as a first-level defect, and 0.8-0.9 as a second-level defect). Finally, a detection report containing the defect location, probability, and level is generated, enabling a quantitative assessment of the quality of laser-penetrated lap-welded workpieces. Furthermore, the defect classification results can be used to guide subsequent processing; for example, first-level defects require rework, second-level defects can be assessed for usage risk, and defect-free areas are directly qualified, providing a quantitative basis for production decisions.

[0079] It should be further noted that the pre-set threshold needs to be determined through statistical analysis based on a large number of labeled samples (including known non-fusion defects and normal interface signals). In a specific embodiment, if in 1000 samples, the model output probability corresponding to 99% of the real non-fusion defects is greater than 0.85, and the probability of 99% of the normal interface signals is less than or equal to 0.15, the threshold can be set to 0.8, thus balancing the false negative rate and the false positive rate.

[0080] This invention converts the preprocessed echo signal into an analytical signal using Hilbert transform. The extracted signal envelope effectively filters out high-frequency carrier interference while retaining the amplitude variation trend reflecting interface or defect characteristics, providing a foundation for the subsequent extraction of time-domain features (such as peak amplitude and duration) and frequency-domain features (such as center frequency and high-frequency energy percentage). The time-domain and frequency-domain features capture the characteristics of the over-threshold signal segment from two perspectives: morphological changes in the time dimension and energy distribution in the frequency dimension. The time-domain features reflect the physical size and reflection intensity of the defect, while the frequency-domain features reflect the differences in the defect's response to sound waves of different frequencies. The two complement each other, avoiding the one-sidedness of a single feature dimension. The dual-channel CNN model specifically mines the time-domain and frequency-domain features in depth. Through parallel processing, it learns the temporal correlation rules and frequency distribution patterns respectively, and then integrates the intrinsic correlation between the two-domain features through a feature fusion layer. This can accurately amplify the subtle differences between the unfused defect and the normal interface in the two-domain features, thereby improving the accuracy of the unfused probability calculation.

[0081] Reference Figure 2 As shown, based on the same inventive concept as the above method, this invention also proposes an ultrasonic phased array defect detection system for laser-penetrating lap welding, comprising:

[0082] The amplitude threshold determination module is used to determine the amplitude threshold between different layers of material in a laser-penetrated cladding workpiece.

[0083] The signal acquisition module is used to instruct the ultrasonic phased array probe to emit sound waves to the laser-penetrating lap-welded workpiece, acquire the echo signal, and preprocess the echo signal.

[0084] The processing module is used to extract the signal envelope of the preprocessed echo signal through Hilbert transform;

[0085] The over-threshold signal segment determination module is used to determine over-threshold signal segments in the signal envelope whose amplitude is greater than the amplitude threshold.

[0086] The extraction module is used to extract the time-domain and frequency-domain features of the signal segment exceeding the threshold, and input the time-domain and frequency-domain features into a pre-trained dual-channel CNN model to obtain the non-fusion probability;

[0087] The judgment module is used to compare the probability of non-fusion with a preset threshold to determine whether there is a defect in the laser-penetrated welded workpiece.

[0088] Reference Figure 3 As shown, based on the same inventive concept as the above method, the present invention also proposes a device, including a memory and a processor, wherein the memory stores computer instructions that can be executed on the processor, and the processor executes the above-described ultrasonic phased array defect detection method for laser penetration lap welding when executing the computer instructions.

[0089] Reference Figure 4 As shown, based on the same inventive concept as the above method, the present invention also proposes a computer-readable storage medium storing computer instructions thereon, which, when executed, can realize the above-mentioned ultrasonic phased array defect detection method for laser penetration lap welding.

[0090] Any references to memory, storage, database, or other media used in the embodiments provided in this invention may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.

[0091] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0092] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting defects in ultrasonic phased arrays used in laser-penetrating lap welding, characterized in that, include: Determining the amplitude threshold for laser penetration between different layers of material in a clad weldment workpiece specifically includes: The camera is instructed to capture images of the laser-penetrating lap-welded workpiece to obtain the model number of the laser-penetrating lap-welded workpiece; Based on the aforementioned model, determine the acoustic impedance of each layer of material penetrated by the laser in the lap-welded workpiece and the upper limit of the safe gap between each layer of material; Based on the predetermined attenuation coefficient, the acoustic impedance, and the upper limit of the safety gap, the amplitude threshold for laser penetration between the layers of the cladding material is determined, and its expression is as follows: In the formula, For amplitude threshold, The attenuation coefficient is... , For the first Layer and First Acoustic impedance of layered materials, This represents the upper limit of the safety clearance; where acoustic impedance reflects the material's ability to impede ultrasonic waves. The ultrasonic phased array probe is instructed to emit sound waves at the laser-penetrating lap-welded workpiece, acquire the echo signal, and preprocess the echo signal. The signal envelope of the preprocessed echo signal is extracted using Hilbert transform. Identify the over-threshold signal segments in the signal envelope whose amplitude is greater than the amplitude threshold; The temporal and frequency domain features of the super-threshold signal segment are extracted and input into a pre-trained dual-channel CNN model to obtain the non-fusion probability. The network architecture of the dual-channel CNN model is a parallel 1D-CNN and 2D-CNN, where 1D-CNN is used to process temporal features and 2D-CNN is used to process frequency domain features. The temporal features include peak amplitude, rising edge slope, duration, peak position, and integral area; the frequency domain features include center frequency, main frequency bandwidth, high-frequency energy proportion, and spectral peak value. The probability of non-fusion is compared with a preset threshold to determine whether there is a defect in the laser-penetrated welded workpiece. The preset threshold is determined by statistical analysis based on a large number of labeled samples.

2. The ultrasonic phased array defect detection method for laser-penetrating lap welding according to claim 1, characterized in that, The commanded ultrasonic phased array probe emits sound waves to the laser-penetrating lap-welded workpiece, specifically including: The scanning path of the ultrasonic phased array probe is generated based on the predetermined three-dimensional model of the weld, and the ultrasonic phased array probe is instructed to move along the scanning path. The excitation delay time of each array element is calculated based on the real-time position of the ultrasonic phased array probe. The acoustic beam is emitted based on the array element excitation delay time.

3. The ultrasonic phased array defect detection method for laser-penetrating lap welding according to claim 2, characterized in that, The method for preprocessing the echo signal is high-pass filtering.

4. The method for detecting defects in ultrasonic phased arrays used in laser-penetrating lap welding according to claim 2, characterized in that, The step of determining the over-threshold signal segment in the signal envelope with an amplitude greater than the amplitude threshold is to determine whether there is a sampling point in the signal envelope with an amplitude greater than the amplitude threshold. If so, the sampling point is taken as the starting point, and the signal is continuously tracked to a sampling point in the signal envelope with an amplitude lower than the amplitude threshold, which is taken as the end point of the over-threshold signal segment. If not, the signal segment is filtered out.

5. The method for detecting defects in ultrasonic phased arrays used in laser-penetrating lap welding according to claim 1, characterized in that, The step of comparing the non-fusion probability with a preset threshold to determine whether there is a defect in the laser-penetrated lap welded workpiece is as follows: if the non-fusion probability is greater than or equal to the preset threshold, then there is a non-fusion defect at the position corresponding to the over-threshold signal segment.

6. An ultrasonic phased array defect detection system for laser-penetrating lap welding, characterized in that, include: The amplitude threshold determination module is used to determine the amplitude threshold between different layers of material in a laser-penetrated cladding workpiece, specifically including: The camera is instructed to capture images of the laser-penetrating lap-welded workpiece to obtain the model number of the laser-penetrating lap-welded workpiece; Based on the aforementioned model, determine the acoustic impedance of each layer of material penetrated by the laser in the lap-welded workpiece and the upper limit of the safe gap between each layer of material; Based on the predetermined attenuation coefficient, the acoustic impedance, and the upper limit of the safety gap, the amplitude threshold for laser penetration between the layers of the cladding material is determined. The ultrasonic phased array probe is instructed to emit sound waves at the laser-penetrating lap-welded workpiece, acquire the echo signal, and preprocess the echo signal. The signal acquisition module is used to instruct the ultrasonic phased array probe to emit sound waves to the laser-penetrating lap-welded workpiece, acquire the echo signal, and preprocess the echo signal. The processing module is used to extract the signal envelope of the preprocessed echo signal through Hilbert transform; An over-threshold signal segment determination module is used to determine over-threshold signal segments in the signal envelope whose amplitude is greater than the amplitude threshold. The extraction module is used to extract the time-domain and frequency-domain features of the over-threshold signal segment, and input the time-domain and frequency-domain features into a pre-trained dual-channel CNN model to obtain the non-fusion probability; The judgment module is used to compare the non-fusion probability with a preset threshold to determine whether there is a defect in the laser-penetrated lap-welded workpiece. The preset threshold is determined based on a large number of labeled samples through statistical analysis.

Images