Steel member welding correction crack detection method and system based on ultrasonic imaging

Abstract

The invention relates to the technical field of ultrasonic detection, in particular to a steel member welding correction crack detection method and system based on ultrasonic imaging. The method comprises the following steps: acquiring an original ultrasonic signal and performing frequency domain decomposition to obtain frequency domain characteristic data; performing frequency extraction, threshold comparison and de-noising processing on the frequency domain characteristic data to obtain de-noised ultrasonic signals; extracting a waveform peak value based on the de-noised ultrasonic signal to obtain a waveform peak value sequence, and performing defect position calculation on the waveform peak value sequence to obtain a defect area coordinate; performing signal interception based on the defect area coordinates to obtain local echo data; performing gradient calculation and peak value extraction according to the local echo data to obtain a crack coordinate position, and performing interpolation based on the crack coordinate position to obtain a crack boundary point set; and removing abnormal values based on the crack boundary point set to obtain stable support nodes, and calculating a closed boundary based on the stable support nodes to obtain a final crack position.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic testing technology, and in particular to a method and system for detecting welded cracks in steel components based on ultrasonic imaging. Background Technology

[0002] In the fields of modern industrial manufacturing and construction, with the expansion of steel structure applications, how to effectively identify cracks and defects in welded parts using intelligent sensors and improve the accuracy of inspection has become a core issue that urgently needs to be addressed in the development of steel component quality inspection technology.

[0003] In a current technology, weld crack detection primarily relies on ultrasonic testing. This method monitors the amplitude, delay, or frequency characteristics of the ultrasonic echo signal; when a certain indicator exceeds a preset range, a defect is identified, thus assessing the weld quality. However, the signal interference environment in industrial settings is increasingly complex. Fixed signal analysis methods are ill-suited to the actual defect characteristics under different welding processes and material structures, leading to higher false positive rates or increased risk of missed detections. Furthermore, traditional methods are often affected by internal material inhomogeneities and the complex geometry of the weld area during signal propagation. Ultrasonic waves undergo multiple scattering and attenuation, which not only makes defect location unclear but may also lead to inaccurate crack boundary identification due to signal superposition effects.

[0004] In summary, existing technologies lack the ability to finely process ultrasonic signals in complex environments and accurately reproduce crack boundary characteristics, resulting in insufficient accuracy in locating weld cracks. Summary of the Invention

[0005] This invention provides a method and system for detecting welded cracks in steel components based on ultrasonic imaging. It can accurately locate welded cracks and restore crack boundaries, thus solving the problem of insufficient accuracy in locating welded cracks in existing technologies.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for detecting weld straightening cracks in steel components based on ultrasonic imaging, comprising: The original ultrasonic signal is acquired, and the original ultrasonic signal is decomposed in the frequency domain to obtain frequency domain feature data. Components with frequencies exceeding a preset frequency threshold are extracted from the frequency domain feature data to obtain high-frequency components. Threshold comparison and denoising processing are then performed based on the high-frequency components to obtain a denoised ultrasonic signal. Based on the denoised ultrasonic signal, waveform peaks are extracted to obtain a waveform peak sequence. The defect location is then calculated from the waveform peak sequence to obtain the coordinates of the defect area. Signal interception is performed based on the coordinates of the defect area to obtain local echo data; Gradient calculation and peak extraction are performed based on the local echo data to obtain the crack coordinate position. Interpolation is then performed based on the crack coordinate position to obtain the crack boundary point set. Outlier removal is performed on the set of crack boundary points to obtain stable support nodes. The closed boundary is then calculated based on the stable support nodes to obtain the final crack location.

[0007] In one optional implementation, acquiring the original ultrasonic signal and performing frequency domain decomposition on the original ultrasonic signal to obtain frequency domain feature data includes: The original ultrasonic signal is acquired, and wavelet transform is performed on the original ultrasonic signal to obtain the wavelet coefficient matrix. Soft thresholding is performed on the wavelet coefficient matrix to obtain a pure wavelet coefficient matrix. The pure wavelet coefficient matrix is ​​then squared to obtain frequency domain feature data.

[0008] In one optional implementation, the step of extracting components with frequencies exceeding a preset frequency threshold from the frequency domain feature data to obtain high-frequency components, and performing threshold comparison and denoising processing based on the high-frequency components to obtain a denoised ultrasonic signal includes: Based on the frequency domain feature data, components with frequencies exceeding a preset frequency threshold are extracted to obtain high-frequency components. Hilbert transform is then performed on the high-frequency components to obtain the instantaneous amplitude envelope. Based on the instantaneous amplitude envelope, regions exceeding a preset amplitude threshold are extracted to obtain an interference region set; interference suppression is performed on the interference region set to obtain denoised high-frequency components; The denoised ultrasonic signal is obtained by inverse wavelet reconstruction based on the denoised high-frequency components.

[0009] In one optional implementation, the step of extracting waveform peaks based on the denoised ultrasonic signal to obtain a waveform peak sequence, and calculating the defect location from the waveform peak sequence to obtain the defect region coordinates, includes: Based on the denoised ultrasonic signal, waveform peaks are extracted to obtain a waveform peak sequence; peak features are extracted from the waveform peak sequence to obtain a feature vector set. The defect feature vectors are classified according to the feature vector set to obtain defect feature vectors. The defect coordinates are calculated according to the defect feature vectors to obtain a defect location set. The defect location set is then spatially aggregated to obtain the defect region coordinates.

[0010] In one optional implementation, the step of extracting the signal based on the coordinates of the defect area to obtain local echo data includes: Based on the coordinates of the defect area, a local signal segment is obtained by extracting the signal from the denoised ultrasonic signal. The local signal segment is windowed to obtain local echo data.

[0011] In one optional implementation, the step of performing gradient calculation and peak extraction based on the local echo data to obtain the crack coordinate position, and then performing interpolation based on the crack coordinate position to obtain the crack boundary point set, includes: A matrix is ​​constructed based on the local echo data to obtain the echo intensity matrix. Gradient calculation is then performed based on the echo intensity matrix to obtain the intensity gradient distribution. Peak values ​​are extracted based on the intensity gradient distribution to obtain the crack coordinates. Linear interpolation is then performed based on the crack coordinates to obtain the crack boundary point set.

[0012] In one optional implementation, the step of outlier removal based on the crack boundary point set to obtain stable support nodes, and calculating the closed boundary based on the stable support nodes to obtain the final crack location, includes: The curvature is calculated based on the set of crack boundary points to obtain a curvature distribution set. Outliers are removed from the curvature distribution set to obtain stable support nodes. A polynomial function is fitted to the stable support node to obtain the final boundary point sequence. The closed boundary is then calculated based on the final boundary point sequence to obtain the final crack location.

[0013] Secondly, the present invention provides a steel component welding correction crack detection system based on ultrasonic imaging, comprising: The frequency domain decomposition module is used to acquire the original ultrasonic signal, perform frequency domain decomposition on the original ultrasonic signal, and obtain frequency domain feature data. The signal denoising module is used to extract components whose frequencies exceed a preset frequency threshold from the frequency domain feature data to obtain high-frequency components, and to perform threshold comparison and denoising processing based on the high-frequency components to obtain a denoised ultrasonic signal. The defect location module is used to extract waveform peaks based on the denoised ultrasonic signal to obtain a waveform peak sequence, and to calculate the defect location from the waveform peak sequence to obtain the coordinates of the defect area. The signal interception module is used to intercept signals based on the coordinates of the defect area to obtain local echo data. The boundary calculation module is used to perform gradient calculation and peak extraction based on the local echo data to obtain the crack coordinate position, and to perform interpolation based on the crack coordinate position to obtain the crack boundary point set; The result optimization module is used to remove outliers based on the crack boundary point set to obtain stable support nodes, and to calculate the closed boundary based on the stable support nodes to obtain the final crack location.

[0014] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention extracts high-frequency components by performing wavelet transform and frequency domain decomposition on the original ultrasonic signal, and performs Hilbert transform and interference suppression processing to obtain a denoised ultrasonic signal. This invention achieves accurate separation and suppression of high-frequency interference and noise components in ultrasonic signals under complex industrial environments, thereby effectively solving the problem that the existing technology using fixed signal analysis methods is difficult to adapt to the actual defect characteristics under different welding processes and material structures, leading to increased misjudgment rate or increased risk of missed detection.

[0015] (2) Based on the denoised ultrasonic signal, the present invention extracts the waveform peak and performs feature vector classification and spatial aggregation, thereby achieving accurate positioning of the welding defect area and deep capture of energy distribution characteristics. This effectively solves the problem that the existing technology is affected by the non-uniformity of the material and the complex geometry of the welding area during ultrasonic propagation, resulting in unclear defect location judgment.

[0016] (3) This invention calculates the gradient and extracts the peak value of the local echo data, and performs interpolation and outlier removal based on the crack coordinate position. Finally, it obtains the closed boundary by fitting a polynomial function, thereby achieving accurate restoration and complete reconstruction of the crack boundary morphology features. This effectively solves the problem of inaccurate crack boundary identification and insufficient welding crack positioning accuracy caused by signal superposition effect and multiple scattering in traditional schemes.

[0017] (4) This invention uses a density-based spatial clustering algorithm to aggregate the defect location set and combines curvature analysis to remove outliers from the crack boundary point set, thereby achieving intelligent integration of discrete defect signals and effective filtering of pseudo-boundary points, significantly improving the accuracy of defect detection and the reliability of boundary restoration. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the process for detecting welded cracks in steel components based on ultrasonic imaging, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the ultrasonic imaging-based steel component welding crack detection system provided in an embodiment of the present invention. Detailed Implementation

[0019] 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, and 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.

[0020] Reference Figure 1 This invention provides a method for detecting weld straightening cracks in steel components based on ultrasonic imaging, comprising: S11, acquire the original ultrasonic signal, and perform frequency domain decomposition on the original ultrasonic signal to obtain frequency domain feature data; S12, extract the components whose frequency exceeds a preset frequency threshold from the frequency domain feature data to obtain high-frequency components, and perform threshold comparison and denoising processing based on the high-frequency components to obtain a denoised ultrasonic signal. S13, Based on the denoised ultrasonic signal, extract the waveform peak value to obtain the waveform peak value sequence, and calculate the defect location from the waveform peak value sequence to obtain the defect area coordinates; S14, based on the coordinates of the defect area, the signal is intercepted to obtain local echo data; S15, perform gradient calculation and peak extraction based on the local echo data to obtain the crack coordinate position, and perform interpolation based on the crack coordinate position to obtain the crack boundary point set; S16, outlier removal is performed based on the crack boundary point set to obtain stable support nodes, and the closed boundary is calculated based on the stable support nodes to obtain the final crack location.

[0021] In step S11, the original ultrasonic signal is acquired, and the original ultrasonic signal is decomposed in the frequency domain to obtain frequency domain feature data, including: The original ultrasonic signal is acquired, and wavelet transform is performed on the original ultrasonic signal to obtain the wavelet coefficient matrix. Soft thresholding is performed on the wavelet coefficient matrix to obtain a pure wavelet coefficient matrix. The pure wavelet coefficient matrix is ​​then squared to obtain frequency domain feature data.

[0022] It should be noted that, firstly, an ultrasonic probe emits ultrasonic pulse signals towards the welded area of ​​the steel component to be inspected. When the pulse signals propagate inside the steel component and encounter abnormal structures such as welding defects and cracks, physical phenomena such as reflection, refraction, and scattering occur, thus generating echo signals. A multi-channel ultrasonic receiver is used to synchronously acquire the echo signals, with a sampling frequency set to 5MHz. Finally, signal processing such as pre-amplification and analog-to-digital conversion is performed to obtain the original ultrasonic signal. The original ultrasonic signal is represented as a multi-dimensional time-domain discrete sequence, denoted as x. i (n), where i represents the channel number, i=1, 2, ..., M, M is the total number of receiving channels; n represents the sampling point number, n=0, 1, 2, ..., N-1, N is the total number of sampling points for a single channel signal.

[0023] The original ultrasonic signal is decomposed using wavelet decomposition to obtain a wavelet coefficient matrix. The basis function for the wavelet decomposition is the db4 wavelet of the Daubechies wavelet family, and the decomposition order is set to 5. The original ultrasonic signal for each channel is decomposed. In the j-th decomposition of the original ultrasonic signal of the i-th channel, the signal is decomposed into two parts according to the frequency range: an approximation coefficient layer cAj_i and a detail coefficient layer cDj_i. The approximation coefficient layer represents the low-frequency trend of the signal, and the detail coefficient layer represents the high-frequency fluctuation details of the signal. The (j+1)-th decomposition further decomposes the approximation coefficient layer cAj from the i-th decomposition to obtain cAj+1_i and cDj+1_i, and so on.

[0024] The wavelet coefficient matrix includes wavelet decomposition coefficients corresponding to each channel in the original ultrasonic signal. For the i-th channel, it is denoted as {cA5_i, cD5_i, cD4_i, cD3_i, cD2_i, cD1_i}. For example, in one acquisition of the original ultrasonic signal, the sampling frequency is set to 5MHz. After five layers of db4 wavelet decomposition, the wavelet coefficient matrix obtained for the six channels shows that cD1_6 corresponds to a frequency range of 2.5MHz-5MHz; cD2_6 corresponds to a frequency range of 1.25MHz-2.5MHz; cD3_6 corresponds to a frequency range of 625kHz-1.25MHz; cD4_6 corresponds to a frequency range of 312.5kHz-625kHz; cD5_6 corresponds to a frequency range of 156.25kHz-312.5kHz; and cA5_6 corresponds to a frequency range of 0-156.25kHz.

[0025] Organize all wavelet coefficients obtained from the decomposition, and use W i Let (j, k) represent the wavelet coefficient of the i-th channel at the j-th layer and k-th position. Extract the wavelet coefficients of all M channels to construct the wavelet coefficient matrix M(j, k) = [W1(j, k), W2(j, k), ..., W...]. M (j, k)]. Further, soft thresholding is performed based on the wavelet coefficient matrix. Soft thresholding is a nonlinear denoising method. Its basic principle is to set wavelet coefficients smaller than a threshold to zero or shrink them, thereby suppressing noise components while retaining useful signal features. Specifically, for the j-th layer wavelet coefficients of the i-th channel, the soft threshold λj_i is first determined. The soft threshold can be calculated using a general threshold formula: Where cDj_i represents the detail coefficients of the j-th layer in the i-th channel, median() represents the median function, N represents the length of the wavelet coefficients in that layer, and 0.6745 is a statistical standardization constant, representing the 75th percentile of the standard normal distribution. A soft thresholding function is applied to the wavelet coefficients of each layer, defined as follows: Among them, W i (j, k) represents the wavelet coefficients, W'i(j, k) represents the clean wavelet coefficients after soft thresholding, sign() represents the sign function, and max() represents the maximum value function. For example, in a set of signal coefficients for channel 1, the soft threshold of cD3_1 is 0.3, and a certain coefficient value is 0.85. Applying the soft thresholding function to this value yields a new coefficient value of 0.85 - 0.3 = 0.55. However, for the point with a coefficient value of 0.15, since its absolute value is less than the threshold of 0.3, the new coefficient value after applying the soft thresholding function is 0.

[0026] Furthermore, for each coefficient W' in each channel of the pure wavelet coefficient matrix M(j,k) i Squaring (j, k) yields the single-channel wavelet energy coefficient E. i (j, k). To comprehensively utilize multi-channel information, cross-channel fusion processing is performed on the single-channel wavelet energy coefficients. For the same decomposition level j and time position k, the energy coefficients of all M channels are accumulated to obtain the frequency domain feature data E. sum (j, k).

[0027] In step S12, components with frequencies exceeding a preset frequency threshold are extracted from the frequency domain feature data to obtain high-frequency components. Threshold comparison and denoising processing are then performed based on these high-frequency components to obtain a denoised ultrasonic signal, including: Based on the frequency domain feature data, components with frequencies exceeding a preset frequency threshold are extracted to obtain high-frequency components. Hilbert transform is then performed on the high-frequency components to obtain the instantaneous amplitude envelope. Based on the instantaneous amplitude envelope, regions exceeding a preset amplitude threshold are extracted to obtain an interference region set; interference suppression is performed on the interference region set to obtain denoised high-frequency components; The denoised ultrasonic signal is obtained by inverse wavelet reconstruction based on the denoised high-frequency components.

[0028] First, based on the frequency domain feature data E sum(j, k) Extracts components whose frequencies exceed a preset frequency threshold to obtain high-frequency components. The preset frequency threshold, determined based on experimental data and expert experience in ultrasonic detection, distinguishes between high-frequency and low-frequency components and is set to 1.25 MHz. Specifically, for an ultrasonic signal with a sampling frequency of 5 MHz, after 5 layers of wavelet decomposition, the frequency range corresponding to each decomposition layer has been determined in step S11. Finally, the first detail layer cD1 with a frequency range of 2.5 MHz to 5 MHz and the second detail layer cD2 with a frequency range of 1.25 MHz to 2.5 MHz are extracted to obtain the high-frequency component set HF. i (j, k).

[0029] It should be noted that the Hilbert transform is a signal processing technique. In practice, the Hilbert transform is performed on the coefficient sequence of each channel and each decomposition layer in the high-frequency component set. For the high-frequency coefficient sequence HF of the i-th channel and the j-th layer... i (j, k) has a Hilbert transform result of HHF. i (j, k). Then, the original signal is combined with its Hilbert transform result to construct an analytic signal: {i Where {i} is the imaginary unit. Based on the analytic signal, the instantaneous amplitude envelope Ai(j,k) is calculated by taking the modulus of the analytic signal, i.e., A i (j, k) equals AS i (j, k) is the square root of the sum of the squares of the real and imaginary parts. The instantaneous amplitude envelope reflects the trend of energy intensity change of the signal at each moment and can effectively highlight abrupt changes and abnormal features in the signal.

[0030] Furthermore, based on the instantaneous amplitude envelope, regions exceeding a preset amplitude threshold are extracted to obtain a set of interference regions. The determination of the preset amplitude threshold requires comprehensive consideration of the signal's statistical characteristics and the noise level of the actual application scenario, and is achieved using an adaptive method. First, for each channel's instantaneous amplitude envelope A... i Calculate the statistical characteristics of (j, k), including the mean, standard deviation, and median. Calculate the arithmetic mean and standard deviation of the instantaneous amplitude envelope of the j-th layer in the i-th channel. The preset amplitude threshold for this layer is then set to the sum of the mean and 2.5 times the standard deviation. The sum of the mean and 2.5 times the standard deviation corresponds to a 99.38% confidence level, and signal points exceeding this threshold have a high probability of being abnormal interference.

[0031] Perform a point-by-point scan of the instantaneous amplitude envelope sequence, and collect all sequences that satisfy A i (j, k) > T Ai(j)The position k is marked as an interference point. After marking, all consecutive interference points are merged into the same interference region to obtain an interference region set. For example, in the instantaneous amplitude envelope A2(1,k) of the first layer of the second channel, the mean is calculated to be 0.45 and the standard deviation is 0.12. Then the preset amplitude threshold T is... A2(1) The value is 0.45 + 2.5 × 0.12 = 0.75. Scanning the instantaneous amplitude envelope sequence of this layer, it was found that between sampling point numbers 1200 and 1350, the envelope value continuously exceeded the preset amplitude threshold, and this region was identified as an interference region; between sampling point numbers 3500 and 3580, the envelope value also continuously exceeded the preset amplitude threshold and was identified as an interference region.

[0032] It should be noted that the interference suppression processing employs a combination of adaptive frequency domain filtering and time domain interpolation. For each interference region in the j-th layer of the i-th channel, the start and end positions of each interference region set are determined, and the first 50 sampling points are extracted as a pre-reference window, and the last 50 sampling points as a post-reference window. Subsequently, the interference suppression coefficient is calculated based on the statistical characteristics of the reference windows. The interference suppression coefficient reflects the degree to which the signal deviates from the normal propagation characteristics within the interference region. For the pre-reference window, the arithmetic mean of its amplitude is calculated; for the post-reference window, the arithmetic mean of its amplitude is calculated. The reference interference threshold is taken as half of the sum of the average amplitudes of the pre-reference window and the post-reference window. The suppression coefficient αi(j,k) for each sampling point k within the interference region is calculated using the following formula: Among them, T ref A is the reference interference threshold. i (j,k) represents the instantaneous amplitude envelope value at sampling point k within the interference region; β is the attenuation control parameter, set to 0.1.

[0033] To ensure the stability of the suppression effect, an upper limit constraint is set on the suppression coefficient. When the calculated suppression coefficient is greater than 1, it is forcibly set to 1 to avoid signal enhancement. The high-frequency coefficient HF' of the interference region after suppression coefficient processing is shown below. i (j, k) is equal to the original high-frequency coefficient HF. i (j, k) and the inhibition coefficient α i The product of (j, k). For high-frequency components in the non-interference region, the original coefficients are kept unchanged, i.e., HF' i (j, k) = HF i (j, k). Integrating all processed high-frequency coefficients, including the coefficients from the suppressed interference region and the coefficients from the unprocessed non-interference region, yields the denoised high-frequency component set HF'. i (j, k).

[0034] For example, in the first layer of the second channel, an interference region was identified between sampling points 1200 and 1350. The preceding reference window was extracted as sampling points 1150 to 1199, and the average amplitude of the preceding reference window was calculated to be 0.42; the following reference window was extracted as sampling points 1351 to 1400, and the average amplitude of the following reference window was calculated to be 0.48. Therefore, the reference threshold T... ref The value is (0.42+0.48) / 2=0.45. The suppression coefficient at this point is calculated as α2(1,1250)=exp[-0.1×(1.2-0.45) / 0.45]=0.85. The original high-frequency coefficient HF2(1,1250) is 0.85. After suppression, HF'2(1,1250)=0.85×0.85=0.72 is obtained, which achieves effective suppression of strong interference signals.

[0035] It should be noted that the inverse wavelet reconstruction is the inverse process of wavelet decomposition. It reconstructs the time-domain signal by inversely combining the wavelet coefficients of each layer. Specifically, for the i-th channel, the denoised high-frequency component HF' is first... i The coefficients corresponding to layers 1 and 2 in (j, k) are integrated with the detail coefficients from layers 3 to 5 and the approximation coefficients of layer 5 to form a complete wavelet coefficient set {cA5_i, cD5_i, cD4_i, cD3_i, cD'2_i, cD'1_i}, where cD'2_i and cD'1_i are denoised high-frequency coefficients after interference suppression. The inverse transform algorithm of the db4 wavelet is used to reconstruct the signal layer by layer, starting from the highest decomposition layer. First, the approximation coefficients cA5_i and the detail coefficients cD5_i of layer 5 are upsampled and filtered to obtain the reconstructed approximation coefficients cA4_i of layer 4; then, cA4_i is reconstructed with the detail coefficients cD4_i of layer 4 to obtain cA3_i; and so on, finally reconstructing cA1_i with the denoised detail coefficients cD'1_i of layer 1 to obtain the complete time-domain signal sequence x'_i(n). This sequence represents the denoised ultrasonic signal of channel i. The inverse wavelet reconstruction process is repeated for all M channels to obtain the complete set of multi-channel denoised ultrasonic signals {x'1(n), x'2(n), ..., x'}. M (n)}.

[0036] In step S13, waveform peaks are extracted based on the denoised ultrasonic signal to obtain a waveform peak sequence. The defect location is then calculated from the waveform peak sequence to obtain the defect region coordinates, including: Based on the denoised ultrasonic signal, waveform peaks are extracted to obtain a waveform peak sequence; peak features are extracted from the waveform peak sequence to obtain a feature vector set. The defect feature vectors are classified according to the feature vector set to obtain defect feature vectors. The defect coordinates are calculated according to the defect feature vectors to obtain a defect location set. The defect location set is then spatially aggregated to obtain the defect region coordinates.

[0037] It should be noted that the waveform peak extraction uses a local maximum detection algorithm. Specifically, the signal sequence is scanned point by point, and when a sampling point n meets the following condition, that point is identified as a waveform peak point: x' i (n)>x' i (n-1) and x' i (n)>x' i (n+1) means that the amplitude at this point is simultaneously greater than the amplitudes of its two adjacent sampling points. To avoid spurious peaks caused by noise, a minimum interval parameter for peak detection is set, typically 10 sampling points. A peak amplitude threshold T is also set. peak Only peak values ​​with amplitudes greater than this threshold are retained. The peak amplitude threshold T... peak The value is calculated based on the root mean square value of the denoised ultrasonic signal, using the formula T. peak =0.3×RMS i RMS i This represents the root mean square value of the denoised ultrasonic signal in the i-th channel. Each extracted peak point contains the peak position n. peak Peak amplitude A peak and peak time t peak Three basic attributes. Peak position n peak Sampling point number, peak amplitude A peak The peak value is the signal amplitude corresponding to this sampling point, and the peak time is t. peak According to the sampling frequency f s The calculation is obtained, and the formula is t. peak =n peak / f s Arrange all the peak points extracted from the i-th channel in chronological order to form the waveform peak sequence P for that channel. i (k).

[0038] For example, in the denoised ultrasonic signal x'3(n) of the third channel, the sampling frequency is 5MHz and the signal length is 10000 sampling points. The root mean square value (RMS3) of this channel signal is calculated to be 0.35, then the peak amplitude threshold T... peakThe value is set to 0.3 × 0.35 = 0.105. At sampling point number 850, the signal amplitude is 0.28. The amplitude at the preceding sampling point 849 is 0.25, and the amplitude at the following sampling point 851 is 0.26. This satisfies the local maximum condition and the amplitude exceeds the threshold of 0.105. Therefore, this point is identified as a peak point. The attributes of this peak point are: peak position n. peak =850, peak amplitude A peak =0.28, peak time t peak =850 / 5=12 microseconds.

[0039] Further, peak features are extracted from the waveform peak sequence to obtain a feature vector set. For each peak point, multi-dimensional features are extracted for subsequent defect identification and classification. The peak features include two dimensions: amplitude features and time-domain features. The amplitude features include normalized amplitude A. norm The calculation formula is A norm =A peak / max(Pi(k)), where max(Pi(k)) represents the maximum peak amplitude in the peak sequence of the channel waveform. Time-domain characteristics include the peak time t. peak Peak width W peak and peak symmetry S peak Peak width W peak Defined as the time span at which the amplitude drops to 50% of the peak amplitude at the peak point, by searching for an amplitude equal to 0.5 times A before and after the peak point. peak The location point is calculated. Peak symmetry S peak The formula S is used to measure the symmetry of the peak shape. peak =|L half -R half | / (L half +R half ), where L half R represents the width of the left half-peak at the peak point. half S represents the width of the half-peak to the right of the peak point. peak The closer the value is to 0, the more symmetrical the peak shape.

[0040] For example, the peak point identified at sampling point 850 in the third channel has a peak amplitude A. peak The normalized amplitude A is 0.28, and the maximum peak amplitude of this channel is 0.45. norm =0.28 / 0.45=0.62. By searching before and after the peak point, it was found that the amplitude drops to 0.14 at sampling point 835 and also to 0.14 at sampling point 865. Therefore, the peak width W peak =(865-835) / 5=6 microseconds. Calculate the width L of the left half-peak. halfThe half-width of the right peak is (850-835) / 5 = 3 microseconds. half The peak symmetry is (865-850) / 5=3 microseconds. peak =|3-3| / (3+3)=0, indicating that the peak shape is completely symmetrical. The eigenvector of this peak point is F. peak =[0.28, 0.62, 12, 6, 0].

[0041] Further, the feature vector set is classified to obtain defect feature vectors. The feature vectors are divided into two categories: defect feature vectors and non-defect feature vectors. Defect feature vectors correspond to ultrasonic echo signals generated by welding cracks or other defects and have specific characteristic patterns. Therefore, three classification criterion thresholds are set, including an amplitude criterion threshold T. amp The width criterion threshold T is 0.5. width The time is 8 microseconds, and the symmetry criterion threshold T sym The value is 0.3. For each eigenvector F in the eigenvector set... peak Make the following judgment: if A norm >T amp And W peak <T width And S peak <T sym If the peak points are positive, the feature vector is classified as a defect feature vector; otherwise, it is classified as a non-defect feature vector. The peak points of all feature vectors classified as defects are extracted to form the defect feature vector set F. defect .

[0042] For each defect feature vector, its corresponding peak time t peak This reflects the round-trip time of the ultrasonic wave from its emission from the probe to its reception of the defect echo. The position coordinates z of the defect in the depth direction are calculated using a depth calculation formula. defect， The formula for calculating depth is: Among them, v steel The speed of ultrasonic waves in steel is 5900 meters per second, or 5.9 millimeters per microsecond. peak Peak time, in microseconds; d probe This is the distance from the probe to the surface of the steel component, defaulted to 2 mm. Dividing by 2 is because the ultrasonic wave undergoes a round trip. This is combined with the probe's scanning position on the welded area surface (x...). probe y probe ), thus obtaining the three-dimensional spatial coordinates (x, y) of the defect. defect y defect , z defect ), where x defect =x probe y defect=y probe .

[0043] Furthermore, the spatial coordinates of defects corresponding to all defect feature vectors are summarized to obtain a set of defect locations. Spatial aggregation is then performed based on this set of defect locations to obtain the coordinates of the defect region. A density-based spatial clustering algorithm is used to aggregate spatially adjacent defect points into the same defect region. A clustering radius of 5 mm and a minimum number of cluster points of 3 are set. For each defect point in the defect location set, its Euclidean distance to other defect points is calculated. For example, if at least 3 defect points exist within a clustering radius of 5 mm, these points are aggregated into a single defect region. For each aggregated defect region, the centroid of all defect points within that region is calculated as the planar center coordinates of the defect region, and the average depth of all defect points within the region is calculated as the vertical depth of the defect region. These are then used to obtain the coordinates of the defect region.

[0044] In step S14, signal interception is performed based on the coordinates of the defect area to obtain local echo data, including: Based on the coordinates of the defect area, a local signal segment is obtained by extracting the signal from the denoised ultrasonic signal. The local signal segment is windowed to obtain local echo data.

[0045] First, calculate the theoretical time t for the ultrasonic wave to travel from the probe to the center point of the defect area and back to the receiver. center The calculation formula is as follows: Where, d probe z is the distance from the probe to the surface of the steel component. defect v represents the vertical depth coordinates of the defect area. steel The propagation speed of ultrasound in steel is set to 5900 m / s. The coefficient 2 in the formula represents the round-trip propagation process of the ultrasound from the probe to the defect and back to the probe. To ensure complete capture of the defect echo signal and its surrounding propagation characteristics, a certain time margin is added both forward and backward from the theoretical time. The time window extension parameter is set, with a default value of 10 microseconds. This determines the start and end times of signal interception, where the start time equals the theoretical time minus the time window extension parameter, and the end time equals the theoretical time plus the time window extension parameter. When the start time is negative, it is set to 0.

[0046] Based on the sampling frequency fs, the time window is converted into a range of sampling point numbers. The starting sampling point number is n. start equal to t start Multiply by the sampling frequency fs and round down to the nearest integer to terminate the sampling point number n. end equal to t endMultiply by the sampling frequency fs and round up. From the denoised ultrasonic signal x' of the i-th channel. i From (n), extract the sampling point index from n start to n end From all data points within the range, the local signal segment L corresponding to the defect area of ​​this channel is obtained. i (m), where m is the sampling point number within the local signal segment, and the value of m ranges from 0 to (n). end -n start ).

[0047] For example, suppose a defect region is identified in channel 4 with a vertical depth z defect The distance d from the probe to the surface of the steel component is 15 mm. probe The diameter is 2 mm. The speed of ultrasonic wave propagation in steel is v. steel The speed is 5900 meters per second, or 5.9 millimeters per microsecond. The one-way propagation distance is 15 + 2 = 17 millimeters, and the round-trip distance is 17 × 2 = 34 millimeters. center The time is 34 ÷ 5.9 = 5.76 microseconds. Therefore, the initial time t is... start The time is 5.76 - 10 = -4.24 microseconds. Since time cannot be negative, t... start Set to 0. Termination time t end The time interval is 5.76 + 10 = 15.76 microseconds. The sampling frequency is 5 MHz, meaning 5 sampling points per microsecond. Therefore, the starting sampling point number is n. start The value is 0, indicating the termination of sampling point number n. end =15.76×5=79. Extract the 0th to 79th sampling points from the denoised ultrasonic signal to form the local signal segment L4(m).

[0048] It should be noted that the window function used in the windowing process is the Hanning window function, denoted by w(m). The local signal segment L... i The value of each sampling point in (m) is multiplied by the corresponding window function value w(m) to obtain the windowed local echo data E. i(m). After windowing, the amplitude of the local echo data at the beginning and end is smoothly transitioned to near zero, effectively eliminating the discontinuity at the signal truncation boundary. For example, the local signal segment L4(m) of the 4th channel has a data length M=80. At m=0, w(0)=0.5×(1-cos(0))=0; at m=39, w(39)=0.5×(1-cos(2π×39 / 79))=0.998; at m=79, w(79)=0.5×(1-cos(2π))=0. The amplitude of the local signal segment at m=0 is 0.15, and after windowing, E4(0)=0.15×0=0; the amplitude at m=39 is 0.82, and after windowing, E4(39)=0.82×0.998=0.818; the amplitude at m=79 is 0.12, and after windowing, E4(79)=0.12×0=0. The amplitude at the boundaries is smoothly attenuated to zero, while the signal in the central region remains basically unchanged.

[0049] In step S15, gradient calculation and peak extraction are performed based on the local echo data to obtain the crack coordinate position. Interpolation is then performed based on the crack coordinate position to obtain the crack boundary point set, including: A matrix is ​​constructed based on the local echo data to obtain the echo intensity matrix. Gradient calculation is then performed based on the echo intensity matrix to obtain the intensity gradient distribution. Peak values ​​are extracted based on the intensity gradient distribution to obtain the crack coordinates. Linear interpolation is then performed based on the crack coordinates to obtain the crack boundary point set.

[0050] It should be noted that the matrix construction involves organizing the local echo data of all channels into a two-dimensional matrix, with the time dimension as the rows and the channel dimension as the columns, constructing an L-row, M-column echo intensity matrix I(r,c), where r corresponds to time sampling point m, c corresponds to channel i, and I(r,c) = E i (m). The gradient calculation detects edges of intensity changes, which often correspond to crack locations causing abrupt changes in echo intensity. Specifically, the gradient is calculated using a difference method. For matrix I(r,c), the gradient in the row direction Gr(r,c) = I(r+1,c) - I(r,c) and the gradient in the column direction Gc(r,c) = I(r,c+1) - I(r,c) are calculated. Finally, the intensity gradient distribution is synthesized: The total gradient value reflects the rate of intensity change at each point, with points exhibiting large gradients indicating potential boundary transition zones. For example, given the echo intensity matrix with known data I(50,2)=0.5, I(51,2)=0.8, and I(50,3)=0.6, then Gr(50,2)=0.8-0.5=0.3, Gc(50,2)=0.6-0.5=0.1, and the total gradient G(50,2)=0.32.

[0051] Further, peak values ​​are extracted based on the intensity gradient distribution to obtain the local maxima of the gradient distribution corresponding to the peak points, representing the location of the crack boundary. Specifically, thresholding and local search algorithms are employed. First, a global statistical analysis is performed on the intensity gradient distribution to calculate the arithmetic mean and standard deviation of the entire gradient distribution. The gradient peak threshold is set as the sum of the arithmetic mean of the gradient distribution and twice the standard deviation. This gradient peak threshold corresponds to approximately a 95.45% confidence level, avoiding misjudging small gradient changes within normal materials as crack boundaries. Subsequently, the intensity gradient distribution is scanned element by element, determining whether the gradient value of each element exceeds the gradient peak threshold, and simultaneously checking whether the element is a local maximum within a 3×3 window centered on it. When an element in the intensity gradient distribution is greater than the gradient peak threshold... ， When this element is the maximum value within a 3×3 window, it is marked as the peak point and recorded. Subsequently, time information is extracted based on the current peak point location, along with the corresponding channel probe's planar coordinates (x, y) at the time point specified in the time information. crack y crack Substituting the time information into the depth calculation formula in step S13, the depth value z is obtained. crack By combining the above information, the crack coordinates R are obtained. crack (x crack y crack , z crack By integrating all crack coordinates, the crack coordinate positions can be obtained.

[0052] Further, linear interpolation is performed based on the crack coordinates to obtain the crack boundary point set. This linear interpolation smooths the boundary by inserting intermediate points between adjacent peak points. Specifically, a preset interpolation threshold is first set; this threshold is an empirical coefficient used to measure whether coordinate interpolation between coordinate points is necessary. In ultrasonic testing of steel, the interpolation threshold is typically set within the range of 0.1 mm to 0.5 mm, with the specific value determined based on actual engineering constraints. For welding quality inspection of large steel structure components, due to the large component size, wide inspection range, and relatively relaxed allowable crack detection accuracy requirements, the interpolation threshold can be appropriately relaxed.

[0053] In this embodiment, considering both detection efficiency and accuracy for bridge steel structure welding defect detection, a preset interpolation threshold of 1 mm is set. The extracted crack coordinates are then arranged in ascending order of their x-coordinates to obtain a crack coordinate sequence. Next, the Euclidean distance between adjacent points is calculated. If the Euclidean distance between two points is greater than the preset interpolation threshold, an interpolation point is inserted between the two coordinates. The coordinates of each interpolation point are located at the midpoint of the shortest line segment between the two coordinate points. This interpolation process is repeated until the Euclidean distance between all coordinates in the crack coordinate sequence is less than the preset interpolation threshold, resulting in a crack boundary point set.

[0054] In step S16, outlier removal is performed based on the crack boundary point set to obtain stable support nodes. The closed boundary is then calculated based on these stable support nodes to obtain the final crack location, including: The curvature is calculated based on the set of crack boundary points to obtain a curvature distribution set. Outliers are removed from the curvature distribution set to obtain stable support nodes. A polynomial function is fitted to the stable support node to obtain the final boundary point sequence. The closed boundary is then calculated based on the final boundary point sequence to obtain the final crack location.

[0055] It should be noted that the curvature described here refers to the degree of bending of the crack boundary at each point and is an important basis for determining whether a boundary point is an outlier. For each point in the crack boundary point set, its curvature value is calculated using the three-point circular arc method. Specifically, for a given point in the crack boundary point set, the preceding and following points are selected to form a three-point combination. The curvature value at that point is calculated using a curvature approximation method based on the vector difference between adjacent points. The curvature value κ... g The calculation formula is: curvature value κ g The larger the value of S, the more pronounced the curvature of the boundary at that point; tri V1 represents the area of ​​the triangle at the three points; V2, V3 represent the three sides of the triangle, and their absolute values ​​represent their side lengths; the curvature values ​​of all points in the crack boundary point set are summarized to obtain the curvature distribution set.

[0056] For example, the crack boundary point set has three consecutive points: point P5 with coordinates (10.0, 20.0, 15.0), point P4 with coordinates (9.5, 19.8, 14.9), and point P6 with coordinates (10.5, 20.2, 15.1). The distance between points P5 and P4 is calculated to be 0.55 mm. Similarly, |P6 - P5| = 0.54 mm and |P6 - P4| = 1.08 mm are calculated. The area S of the triangle is... tri If the area is 0.15 square millimeters, then the curvature value is 1.87 millimeters.-1 .

[0057] Furthermore, stable support nodes are obtained by removing outliers from the curvature distribution set. Outliers refer to boundary points where the curvature value significantly deviates from the normal range; these points are typically pseudo-boundary points caused by signal noise, multipath effects, or boundary extraction errors. Statistical methods are used to determine the outlier threshold. First, statistical analysis is performed on the curvature distribution set to calculate the median and interquartile range of the curvature. The interquartile range is defined as the difference between the third quartile and the first quartile. Based on the interquartile range method, the upper limit threshold for curvature anomalies is set to the third quartile plus 1.5 times the interquartile range. For each curvature value, if the curvature value lies within the closed interval from 0 to the upper limit threshold for curvature anomalies... ， If the curvature value is greater than the upper limit threshold for curvature anomalies, the corresponding boundary point is considered an outlier and is removed. It should be noted that since the curvature value is always positive, all retained normal boundary points are aggregated to obtain a set of stable support nodes.

[0058] For example, assuming a crack boundary point set in a defect region contains 50 points, after calculating the curvature distribution set, statistical analysis yields a first quartile of 0.8 mm. -1 The third and fourth quartiles are 2.2 mm. -1 The interquartile range is 2.2 - 0.8 = 1.4 mm. -1 The upper limit threshold for curvature anomaly is 2.2 + 1.5 × 1.4 = 4.3 mm. -1 Since curvature values ​​are typically positive, the lower threshold is actually set to 0. Examining the curvature values ​​of all 50 boundary points, three points were found to have a curvature value of 5.2 mm. -1 4.8 mm -1 and 6.1 mm -1 All exceeded the upper limit threshold of 4.3 mm. -1 Therefore, these three points were identified as outliers and removed. The curvature values ​​of the remaining 47 points were all within the normal range, forming a stable support node set.

[0059] Furthermore, a polynomial function is fitted based on the stable support nodes to obtain the final boundary point sequence. This polynomial function fitting can smoothly connect the discrete stable support nodes, generating a continuous crack boundary curve.

[0060] A piecewise cubic polynomial fitting method is employed. Specifically, the set of stable support nodes is arranged according to their spatial position to form an ordered node sequence. For the crack boundary in three-dimensional space, fitting is performed on the xy-plane projection and the z-direction depth, respectively.

[0061] First, the xy-plane projection is processed, with the x-coordinate as the independent variable and the y-coordinate as the dependent variable, to establish a cubic polynomial fitting model. The least squares method is used to solve for the polynomial coefficients in the cubic polynomial, minimizing the sum of squared residuals between the fitted curve and the stable support nodes. The coordinates of all stable support nodes are then substituted into the cubic polynomial equation to construct an overdetermined system of equations, and the polynomial coefficients are solved using matrix operations.

[0062] For depth-direction fitting, a cubic polynomial fitting model is established using the arc length parameter in the xy-plane as the independent variable and the z-coordinate as the dependent variable. The arc length parameter represents the cumulative length along the fitted curve from the starting node, calculated starting from the first stable support node. For example, the arc length parameter for the 7th node is equal to the sum of the lengths of the first 6 straight line segments.

[0063] After solving the polynomial coefficients, dense sampling is performed on the fitted curve to generate the final boundary point sequence. The sampling interval is set to 0.2 mm. Along the fitted curve, from the starting node to the ending node, the coordinates of one boundary point are extracted every 0.2 mm. For the xy plane, the corresponding y value is calculated using the fitted polynomial based on the x value; for the depth direction, the corresponding z value is calculated using the fitted polynomial based on the arc length parameter s. All extracted boundary points are arranged in order to obtain the final boundary point sequence. For example, suppose the set of stable support nodes for a crack contains 20 nodes, with x-coordinates ranging from 8.0 mm to 18.0 mm. A cubic polynomial is fitted to these 20 nodes, and the coefficients are obtained as a3=0.002, a2=-0.05, a1=1.2, a0=15.0. Between the x-coordinates of 8.0 and 18.0, one point is sampled every 0.2 mm, generating a total of (18.0-8.0) / 0.2=50 points. For the sampling point at x = 10.0 mm, calculate y = 0.002 × 10 3 -0.05×10 2 +1.2×10+15.0=24.0 mm. Similarly, calculate the y-coordinates of all 50 sampling points, and combine them with the fitting results in the depth direction to obtain the complete three-dimensional final boundary point sequence.

[0064] Further, the closed boundary is calculated based on the final boundary point sequence to obtain the final crack location. The closed boundary refers to connecting the start and end points of the crack boundary to form a closed crack profile. First, the distance between the first and last points of the final boundary point sequence is checked. If the distance is less than the sampling interval, the crack boundary is considered to be close to closed, and the first and last points are directly connected with a straight line segment to complete the boundary closure. If the distance is greater than a preset closure threshold, linear interpolation is repeatedly performed between the first and last points, inserting several intermediate points until the distance between the first and last points does not exceed the sampling interval; wherein, the preset closure threshold can be set to a value consistent with the sampling interval, i.e., 0.2 mm.

[0065] After the boundary is closed, the geometric characteristic parameters of the crack are calculated based on the sequence of closed boundary points, including the crack length and the maximum crack width. The crack length is defined as the perimeter of the closed boundary, and is calculated by accumulating the Euclidean distances between all adjacent boundary points, including those connected end-to-end. The maximum crack width is calculated by taking the convex hull of the closed boundary point set, then solving for the Euclidean distances between all pairs of vertices on the convex hull, and taking the maximum value as the maximum crack width. The coordinates of all boundary points of the closed boundary, the crack geometric characteristic parameters, and the coordinates of the crack center are integrated to obtain the final crack location.

[0066] In summary, this invention provides a method and system for detecting welded cracks in steel components based on ultrasonic imaging. It can accurately locate weld cracks and restore crack boundaries, solving the problem of insufficient accuracy in locating weld cracks caused by the lack of fine processing capabilities of ultrasonic signals in complex environments in existing technologies.

[0067] Reference Figure 2 This invention provides a welding crack detection system for steel components based on ultrasonic imaging, characterized in that it includes: The frequency domain decomposition module is used to acquire the original ultrasonic signal, perform frequency domain decomposition on the original ultrasonic signal, and obtain frequency domain feature data. The signal denoising module is used to extract components whose frequencies exceed a preset frequency threshold from the frequency domain feature data to obtain high-frequency components, and to perform threshold comparison and denoising processing based on the high-frequency components to obtain a denoised ultrasonic signal. The defect location module is used to extract waveform peaks based on the denoised ultrasonic signal to obtain a waveform peak sequence, and to calculate the defect location from the waveform peak sequence to obtain the coordinates of the defect area. The signal interception module is used to intercept signals based on the coordinates of the defect area to obtain local echo data. The boundary calculation module is used to perform gradient calculation and peak extraction based on the local echo data to obtain the crack coordinate position, and to perform interpolation based on the crack coordinate position to obtain the crack boundary point set; The result optimization module is used to remove outliers based on the crack boundary point set to obtain stable support nodes, and to calculate the closed boundary based on the stable support nodes to obtain the final crack location.

[0068] It should be noted that the ultrasonic imaging-based steel component welding crack detection system provided in this embodiment of the invention is used to execute all the process steps of the ultrasonic imaging-based steel component welding crack detection method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0069] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a safety valve status anomaly monitoring program based on edge computing. When the processor executes the computer program, it implements the steps described in the various embodiments of the ultrasonic imaging-based steel component welded crack correction detection system, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments, such as the data preprocessing module.

[0070] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0071] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0072] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0073] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0074] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0075] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0076] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for detecting weld straightening cracks in steel components based on ultrasonic imaging, characterized in that, include: The original ultrasonic signal is acquired, and the original ultrasonic signal is decomposed in the frequency domain to obtain frequency domain feature data. Components with frequencies exceeding a preset frequency threshold are extracted from the frequency domain feature data to obtain high-frequency components. Threshold comparison and denoising processing are then performed based on the high-frequency components to obtain a denoised ultrasonic signal. Based on the denoised ultrasonic signal, waveform peaks are extracted to obtain a waveform peak sequence. The defect location is then calculated from the waveform peak sequence to obtain the coordinates of the defect area. Signal interception is performed based on the coordinates of the defect area to obtain local echo data; Gradient calculation and peak extraction are performed based on the local echo data to obtain the crack coordinate position. Interpolation is then performed based on the crack coordinate position to obtain the crack boundary point set. Outlier removal is performed on the set of crack boundary points to obtain stable support nodes. The closed boundary is then calculated based on the stable support nodes to obtain the final crack location.

2. The method for detecting welded cracks in steel components based on ultrasonic imaging according to claim 1, characterized in that, The process of acquiring the original ultrasonic signal and performing frequency domain decomposition on the original ultrasonic signal to obtain frequency domain feature data includes: The original ultrasonic signal is acquired, and wavelet transform is performed on the original ultrasonic signal to obtain the wavelet coefficient matrix. Soft thresholding is performed on the wavelet coefficient matrix to obtain a pure wavelet coefficient matrix. The pure wavelet coefficient matrix is ​​then squared to obtain frequency domain feature data.

3. The method for detecting welded straightening cracks in steel components based on ultrasonic imaging according to claim 1, characterized in that, The step of extracting components whose frequencies exceed a preset frequency threshold from the frequency domain feature data to obtain high-frequency components, and performing threshold comparison and denoising processing based on the high-frequency components to obtain a denoised ultrasonic signal includes: Based on the frequency domain feature data, components with frequencies exceeding a preset frequency threshold are extracted to obtain high-frequency components. Hilbert transform is then performed on the high-frequency components to obtain the instantaneous amplitude envelope. Based on the instantaneous amplitude envelope, regions exceeding a preset amplitude threshold are extracted to obtain an interference region set; interference suppression is performed on the interference region set to obtain denoised high-frequency components; The denoised ultrasonic signal is obtained by inverse wavelet reconstruction based on the denoised high-frequency components.

4. The method for detecting welded straightening cracks in steel components based on ultrasonic imaging according to claim 1, characterized in that, The step of extracting waveform peaks based on the denoised ultrasonic signal to obtain a waveform peak sequence, and calculating the defect location from the waveform peak sequence to obtain the defect region coordinates includes: Based on the denoised ultrasonic signal, waveform peaks are extracted to obtain a waveform peak sequence; peak features are extracted from the waveform peak sequence to obtain a feature vector set. The defect feature vectors are classified according to the feature vector set to obtain defect feature vectors. The defect coordinates are calculated according to the defect feature vectors to obtain a defect location set. The defect location set is then spatially aggregated to obtain the defect region coordinates.

5. The method for detecting welded straightening cracks in steel components based on ultrasonic imaging according to claim 4, characterized in that, The process of extracting signal data based on the coordinates of the defect area to obtain local echo data includes: Based on the coordinates of the defect area, a local signal segment is obtained by extracting the signal from the denoised ultrasonic signal. The local signal segment is windowed to obtain local echo data.

6. The method for detecting welded straightening cracks in steel components based on ultrasonic imaging according to claim 1, characterized in that, The step of performing gradient calculation and peak extraction based on the local echo data to obtain the crack coordinate position, and then performing interpolation based on the crack coordinate position to obtain the crack boundary point set, includes: A matrix is ​​constructed based on the local echo data to obtain the echo intensity matrix. Gradient calculation is then performed based on the echo intensity matrix to obtain the intensity gradient distribution. Peak values ​​are extracted based on the intensity gradient distribution to obtain the crack coordinates. Linear interpolation is then performed based on the crack coordinates to obtain the crack boundary point set.

7. The method for detecting welded straightening cracks in steel components based on ultrasonic imaging according to claim 1, characterized in that, The process of removing outliers based on the crack boundary point set to obtain stable support nodes, and calculating the closed boundary based on the stable support nodes to obtain the final crack location includes: The curvature is calculated based on the set of crack boundary points to obtain a curvature distribution set. Outliers are removed from the curvature distribution set to obtain stable support nodes. A polynomial function is fitted to the stable support node to obtain the final boundary point sequence. The closed boundary is then calculated based on the final boundary point sequence to obtain the final crack location.

8. A system for detecting weld cracks in steel components based on ultrasonic imaging, characterized in that, include: The frequency domain decomposition module is used to acquire the original ultrasonic signal, perform frequency domain decomposition on the original ultrasonic signal, and obtain frequency domain feature data. The signal denoising module is used to extract components whose frequencies exceed a preset frequency threshold from the frequency domain feature data to obtain high-frequency components, and to perform threshold comparison and denoising processing based on the high-frequency components to obtain a denoised ultrasonic signal. The defect location module is used to extract waveform peaks based on the denoised ultrasonic signal to obtain a waveform peak sequence, and to calculate the defect location from the waveform peak sequence to obtain the coordinates of the defect area. The signal interception module is used to intercept signals based on the coordinates of the defect area to obtain local echo data. The boundary calculation module is used to perform gradient calculation and peak extraction based on the local echo data to obtain the crack coordinate position, and to perform interpolation based on the crack coordinate position to obtain the crack boundary point set; The result optimization module is used to remove outliers based on the crack boundary point set to obtain stable support nodes, and to calculate the closed boundary based on the stable support nodes to obtain the final crack location.

Images