Interference signal identification method for weld defects under different lifts-off conditions

Abstract

The invention discloses a method for identifying interference signals of weld defects under different lifts-off conditions, which comprises the following steps of: S1, de-trending processing: carrying out de-trending processing on original detection signals; s2, Gaussian wavelet transform: carrying out Gaussian wavelet transform on the detrended signal; s3, optimal wavelet basis selection: by calculating correlation coefficients or energy ratios of different wavelet basis and defect signals, selecting the wavelet basis with the highest matching degree for reconstruction; s4, envelope processing: extracting a signal envelope based on Hilbert transform; s5, mean filtering: applying sliding window mean filtering to the envelope signal; and S6, threshold processing: setting a self-adaptive threshold screening signal, and retaining the feature points of which the amplitudes exceed the threshold. According to the method, de-trending and Gaussian wavelet transform are used for de-noising enhancement. According to the method, wavelet functions of different orders are constructed and matched with defects, secondary signal enhancement is carried out by selecting a filtering method, finally, threshold stripping interference is calculated, reliable defect identification is carried out, and technical support is provided for uneven welding seam quality monitoring.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic nondestructive testing technology, and in particular to a method for identifying interference signals of weld defects under different lift-off conditions. Background Technology

[0002] Alternating current electromagnetic field (ACFM) testing is a relatively new electromagnetic non-destructive testing (NDT) technology that can accurately measure cracks. However, due to the uneven surface of crane circumferential welds, it is prone to missed or false detections. Currently, there is a lack of signal analysis methods to address the lift-off effect caused by weld surface unevenness, and no signal processing methods exist for identifying weld defects when the lift-off effect occurs. Therefore, the presence of the lift-off problem in actual working conditions makes accurate defect identification difficult.

[0003] Therefore, when using ACFM testing instruments to detect weld defects in existing specimens, changes in probe lift-off height lead to baseline drift and increased random noise. Traditional filtering methods (such as mean filtering and wavelet denoising) struggle to separate defect features in complex noise environments. Envelope analysis is susceptible to low-frequency interference, resulting in misjudgments. CN112730405A uses a single wavelet basis for denoising without optimizing wavelet basis matching for defect features; US20210055121A1's threshold processing lacks pre-noise suppression, leaving residual high-frequency interference. Therefore, a method for identifying interference signals from weld defects under different lift-off conditions is urgently needed to address these issues. Summary of the Invention

[0004] In view of this, the present invention provides a method for identifying interference signals of weld defects under different lift-off conditions, utilizing detrending methods and Gaussian wavelet transform for noise reduction and enhancement. Wavelet functions of different orders are constructed and matched with defects. A secondary signal enhancement method is performed using a selected filtering method. Finally, a threshold is calculated to remove interference, enabling reliable defect identification. This also provides technical support for monitoring the quality of uneven welds.

[0005] This invention provides a method for identifying interference signals of weld defects under different lift-off conditions, including:

[0006] S1. Detrending Processing

[0007] The original detection signal is detrended by using the moving average method or polynomial fitting method to eliminate the signal baseline drift. The main purpose is to adjust the signal baseline drift and eliminate the low-frequency offset caused by instrument or environmental factors, so that the signal is more stable and lays the foundation for subsequent processing.

[0008] S2. Gaussian wavelet transform

[0009] Gaussian wavelet transform is performed on the detrended signal, and multi-scale decomposition is performed using Gaussian wavelets to suppress random noise and initially highlight defect features. It is mainly used to suppress random noise in the signal and reduce interference caused by baseline drift, and initially highlight the defect signal, but noise interference still exists at this time.

[0010] S3. Optimal Wavelet Basis Selection

[0011] By calculating the correlation coefficient or energy ratio between different wavelet bases and the defect signal, the wavelet base with the highest matching degree is selected for reconstruction; the wavelet base that is most similar to the defect signal is identified. The defect signal is significantly enhanced at this stage, while the noise signal is weakened, further improving the signal's recognizability.

[0012] S4. Envelope Processing

[0013] The signal envelope is extracted based on Hilbert transform, high-frequency noise is suppressed and defect amplitude information is enhanced; when the defect signal has already emerged, the interference of weak noise on the defect signal is effectively suppressed, thus improving signal quality.

[0014] S5. Mean Filtering

[0015] Applying sliding window mean filtering to the envelope signal smooths out local noise fluctuations; however, minor noise interference still exists after the envelope signal is applied, so noise in some channels is suppressed to further highlight defect characteristics.

[0016] S6. Threshold processing

[0017] An adaptive threshold is set to filter signals, retaining feature points whose amplitude exceeds the threshold. Threshold processing is applied to filter the signals, removing noise below the set threshold and retaining defect signals with significant characteristics.

[0018] After the above processing, the final signal not only highlights the defect features in the weld but also effectively removes noise interference. This process achieves defect signal extraction and noise suppression, greatly improving the accuracy and reliability of weld defect detection.

[0019] Preferably, in S1, the original detection signal x(t) is fitted using a binomial fitting method to obtain the baseline correction signal x. d (t), where the window length L = N / 10, and N is the number of sampling points.

[0020] Preferably, in S2, the first-order Gaussian derivative wavelet is selected. As the mother wavelet, a 5-8 scale discrete wavelet transform is performed, with the scale set s = {21, 22, ..., 2k}, and the output detail coefficients D are generated. s (t) and approximation coefficient A s (t).

[0021] Preferably, in S3, the selection of the optimal wavelet basis includes the following steps:

[0022] (1) Construct a wavelet base library, which includes, but is not limited to, the Daubechies (dbN), Symlet (symN), and Coiflet (coifN) series;

[0023] (2) Calculate the correlation coefficients between each wavelet basis and the defect template signal.

[0024] (3) Select the optimal wavelet basis reconstruction signal x with ρ > 0.85. w (t).

[0025] Preferably, in S4, the optimal wavelet basis reconstructed signal is subjected to Hilbert transform, as shown in the formula:

[0026] Generate envelope signal:

[0027] Preferably, in S5, the mean filtering includes the following steps:

[0028] (1) Design a variable window length filter, where the window width is... f s f is the sampling rate. c These are the defect characteristic frequencies obtained through spectrum analysis;

[0029] (2) Perform a sliding window average, the formula is:

[0030] Preferably, in step S6, the threshold processing includes the following steps:

[0031] (1) Calculate the noise standard deviation, σ n The median of the envelope signal is estimated using the following formula: σ n =median(|e f (t)|) / 0.6745;

[0032] (2) Set dynamic threshold: T h =k1σ n ,k1∈[3.0,4.0] is the high threshold,:T l =k2σ n k2∈[1.5,2.5] represents the low threshold;

[0033] (3) Output the set of defect locations: {t i |e f (t i )>T h}∪{t j |ef (t j )>T l And the domain has T h point}.

[0034] Compared with the prior art, the advantages of this invention are:

[0035] This invention processes signals under different lift-off conditions ranging from 1 to 5 mm. Testing with signals from a 1 mm lift-off condition accurately identifies defect signals. Verification analysis is performed using original defect signals from other lift-off conditions to validate the method's feasibility. Finally, appropriate evaluation indicators are selected for evaluation, using quantitative data to demonstrate that the lift-off effect can be addressed, while providing technical support for monitoring uneven weld quality. Timely handling of the lift-off effect enables defect identification, facilitates rework, ensures weld quality, and reduces safety hazards.

[0036] This invention has the following characteristics: high resistance to lift-off interference, suppressing >90% of low-frequency drift through a combination of detrending processing, Gaussian wavelet transform, and optimal wavelet basis selection; high defect enhancement, with the defect signal-to-noise ratio improved by ≥15dB after optimal wavelet basis selection, envelope processing, and mean filtering; strong generalization ability, with the optimal wavelet basis adaptively matching different defect types (porosity / cracks / lack fusion); and strong real-time performance, with relatively low algorithm complexity, supporting real-time processing in embedded systems. Attached Figure Description

[0037] Figure 1 The waveform and image of weld defect signal under 1-3mm lift-off.

[0038] Figure 2 The waveform and image of weld defect signal under 4-5mm lift-off.

[0039] Figure 3 The original signals and images were obtained from the 1mm lift-off test of the weld defects in this invention.

[0040] Figure 4 This invention provides signal and image processing for the back-and-forth detection trend of weld defects under 1mm lift-off.

[0041] Figure 5 Gaussian wavelet signals and images were used to detect weld defects under 1mm lift-off conditions in this invention.

[0042] Figure 6 This invention provides the signal and image processing for the back-and-forth detection envelope of weld defects under 1mm lift-off conditions.

[0043] Figure 7 This invention provides back-and-forth detection, filtering, and noise reduction signals and images of weld defects under 1mm lift-off conditions.

[0044] Figure 8 This invention provides the threshold processing signals and images for the back-and-forth detection of weld defects at a 1mm lift-off point.

[0045] Figure 9 This is a comparison of the original signal and the result of weld defect signal processing under 1-3mm lift-off in this invention.

[0046] Figure 10 This is a comparison of the original signal and the result of weld defect signal processing under 4-5mm lift in this invention.

[0047] Figure 11 This is a comparison of the original images and results of weld defect signal processing under 1-3mm lift-off in this invention.

[0048] Figure 12 This is a comparison of the original images and results of weld defect signal processing under 4-5mm lift-off in this invention. Detailed Implementation

[0049] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0050] To clearly illustrate the technical features of this patent, the following detailed description is provided through specific embodiments and in conjunction with the accompanying drawings.

[0051] like Figure 1-2 The image shows the waveform and imaging of weld defect signals after 1-5mm lift-off. Significant complex interference exists in the original signal, making direct identification of the defect signal difficult. Detailed analysis of the defect signal revealed the following key issues:

[0052] (1) Lift-off effect: The crack signal is significantly affected by the lift-off effect, especially as the lift-off distance gradually increases, resulting in uneven signal intensity attenuation. This manifests as signal baseline drift and weaker signals at the defect location, making it impossible to accurately and intuitively distinguish the defect location. This phenomenon makes it difficult to distinguish the defect signal from background noise, thus affecting the accurate location of the defect.

[0053] (2) Signal interference and baseline drift: The original signal baseline drifted, which may be caused by ambient noise, equipment performance fluctuations, or uneven surface conditions of the specimen. The baseline drift of the signal further increases the difficulty of defect identification, especially when the defect signal is weak.

[0054] (3) Problems in detecting minute defects: Due to the complexity of signal interference, it is difficult to clearly extract the signal characteristics of minute cracks from the original signal. Although the image shows signs of defects in some areas, the defect signal is still relatively blurry because noise is not effectively suppressed during signal processing.

[0055] Therefore, in order to ensure that the defect signals detected under different working conditions can be effectively identified, it is necessary to perform signal processing on the detected weld defect signals, propose reasonable signal processing methods, and accurately and efficiently complete the identification of defect signals.

[0056] This invention proposes a method for identifying interference signals of weld defects under different lift-off conditions, including:

[0057] S1. Detrending Processing

[0058] The original detection signal is detrended by using the moving average method or polynomial fitting method to eliminate the signal baseline drift. The main purpose is to adjust the signal baseline drift and eliminate the low-frequency offset caused by instrument or environmental factors, so that the signal is more stable and lays the foundation for subsequent processing.

[0059] S2. Gaussian wavelet transform

[0060] Gaussian wavelet transform is performed on the detrended signal, and multi-scale decomposition is performed using Gaussian wavelets to suppress random noise and initially highlight defect features. It is mainly used to suppress random noise in the signal and reduce interference caused by baseline drift, and initially highlight the defect signal, but noise interference still exists at this time.

[0061] S3. Optimal Wavelet Basis Selection

[0062] By calculating the correlation coefficient or energy ratio between different wavelet bases and the defect signal, the wavelet base with the highest matching degree is selected for reconstruction; the wavelet base that is most similar to the defect signal is identified. The defect signal is significantly enhanced at this stage, while the noise signal is weakened, further improving the signal's recognizability.

[0063] S4. Envelope Processing

[0064] The signal envelope is extracted based on Hilbert transform, high-frequency noise is suppressed and defect amplitude information is enhanced; when the defect signal has already emerged, the interference of weak noise on the defect signal is effectively suppressed, thus improving signal quality.

[0065] S5. Mean Filtering

[0066] Applying sliding window mean filtering to the envelope signal smooths out local noise fluctuations; however, minor noise interference still exists after the envelope signal is applied, so noise in some channels is suppressed to further highlight defect characteristics.

[0067] S6. Threshold processing

[0068] An adaptive threshold is set to filter signals, retaining feature points whose amplitude exceeds the threshold. Threshold processing is applied to filter the signals, removing noise below the set threshold and retaining defect signals with significant characteristics.

[0069] After the above processing, the final signal not only highlights the defect features in the weld but also effectively removes noise interference. This process achieves defect signal extraction and noise suppression, greatly improving the accuracy and reliability of weld defect detection.

[0070] Three surface crack defects were repeatedly scanned using an ACEM testing instrument, and the resulting signals were processed. The test workpiece was a 10mm thick steel plate with a 2mm weld reinforcement. The defects were surface cracks, totaling three. The weld reinforcement was measured to be approximately 2mm. Acrylic plates of different thicknesses were placed on both sides of the weld to simulate the lift-off effect caused by uneven surfaces under actual working conditions. The test simulated a lift-off height of 1-5mm; a height of 5mm was sufficient to address the problems that would arise under actual working conditions.

[0071] Example 1

[0072] Processing of defect signals under 1mm lift-off conditions, and methods for identifying interference signals of weld defects under 1mm lift-off conditions, including:

[0073] S1. Detrending Processing

[0074] The original detection signal x(t) is fitted using a binomial fitting method to obtain the baseline correction signal x. d (t), where the window length L = N / 10, and N is the number of sampling points;

[0075] S2. Gaussian wavelet transform

[0076] Gaussian first derivative wavelet As the mother wavelet, a 5-8 scale discrete wavelet transform is performed, with the scale set s = {21, 22, ..., 2k}, and the output detail coefficients D are generated. s (t) and approximation coefficient A s (t);

[0077] S3. Optimal wavelet basis selection includes the following steps:

[0078] (1) Construct a wavelet base library, which includes, but is not limited to, the Daubechies (dbN), Symlet (symN), and Coiflet (coifN) series;

[0079] (2) Calculate the correlation coefficients between each wavelet basis and the defect template signal.

[0080] (3) Select the optimal wavelet basis reconstruction signal x with ρ > 0.85. w (t);

[0081] S4. Envelope Processing

[0082] The Hilbert transform of the reconstructed signal based on the optimal wavelet basis is given by the following formula:

[0083] Generate envelope signal:

[0084] S5. Mean filtering, including the following steps:

[0085] (1) Design a variable window length filter, where the window width is... f s f is the sampling rate. c These are the defect characteristic frequencies obtained through spectrum analysis;

[0086] (2) Perform a sliding window average, the formula is:

[0087] S6. Threshold processing, including the following steps:

[0088] (1) Calculate the noise standard deviation, σ n The median of the envelope signal is estimated using the following formula: σ n =median(|e f (t)|) / 0.6745;

[0089] (2) Set dynamic threshold: T h =k1σ n ,k1∈[3.0,4.0] is the high threshold,:T l =k2σ n k2∈[1.5,2.5] represents the low threshold;

[0090] (3) Output the set of defect locations: {t i |e f (t i )>T h}∪{t j |e f (t j )>T l And the domain has T h point}.

[0091] like Figures 3-8 As shown, the original detection signal under 1mm lift undergoes detrending processing, Gaussian wavelet transform, optimal wavelet basis selection, envelope processing, mean filtering, and thresholding processing to process the signal and image.

[0092] Example 2

[0093] like Figures 9-12 As shown, the defect signals under 2-5mm lifting were processed respectively, and the signals and images before processing were compared with the results after processing.

[0094] In the analysis of signal processing results for weld defects, signal-to-noise ratio (SNR) and noise reduction ratio (DNR) are selected as evaluation indicators because these two indicators can reflect the effect of signal processing and the ability to detect weld defects.

[0095] Signal-to-noise ratio (SNR): SNR is an important parameter for measuring signal quality; it represents the ratio of the intensity of the effective signal to that of noise. A higher SNR means a clearer detection signal, less noise interference, and improved accuracy in defect identification. In weld defect inspection, a higher SNR helps to accurately identify minute defects and reduce false positives.

[0096] Noise Reduction Ratio (DNR): The noise reduction ratio measures the effectiveness of noise suppression during signal processing. A high noise reduction ratio means that noise has been effectively removed or suppressed after signal processing, preserving more useful signal information. In weld inspection, noise reduction technology can enhance the identifiability of defect signals. Especially in complex welding environments, good noise reduction capabilities help improve the stability and reliability of inspection.

[0097] By comprehensively analyzing these two indicators, the performance of weld defect signal processing technology can be fully evaluated, detection methods can be optimized, and efficient identification and accurate assessment of defects can be ensured.

[0098]

[0099] Table 1 Signal Quality Evaluation

[0100] Signal processing was performed at different lift-off points ranging from 1-5 mm. Testing with signals from a 1 mm lift-off point accurately identified defect signals. Verification analysis was conducted using original defect signals from other lift-off points to validate the method's feasibility. Finally, appropriate evaluation indicators were selected for evaluation, using quantitative data to demonstrate that the lift-off effect can be addressed, providing technical support for monitoring uneven weld quality. Timely handling of the lift-off effect enables defect identification, facilitates rework, ensures weld quality, and reduces safety hazards.

[0101] There are many ways to implement this invention. The above description is only a preferred embodiment of this invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of this invention, and these improvements should also be considered within the scope of protection of this invention.

Claims

1. A method of identifying interference signals from different lift-off underweld defects, characterized by, Comprise: S1. Detrending Detrend the original detection signal, adopt moving average method or polynomial fitting method to eliminate signal baseline drift; S2. Gaussian wavelet transform Perform Gaussian wavelet transform on the signal after detrending, use Gaussian wavelet to carry out multi-scale decomposition, suppress random noise and highlight the defect characteristics preliminarily; S3. Optimal wavelet basis selection Select the wavelet basis with the highest matching degree by calculating the correlation coefficient or energy ratio of different wavelet bases and defect signals for reconstruction; S4. Envelope processing Extract signal envelope based on Hilbert transform, suppress high-frequency noise and enhance defect amplitude information; S5. Mean filtering Apply sliding window mean filtering to the envelope signal to smooth local noise fluctuations; S6. Threshold processing Set adaptive threshold to filter the signal, and retain the characteristic points with amplitude exceeding the threshold.

2. The method of claim 1, wherein, In S1, the original detection signal x(t) is fitted by binomial fitting method to obtain the baseline correction signal x d (t), wherein the window length L=N / 10, and N is the number of sampling points.

3. The method of claim 1, wherein, In S2, Gaussian first derivative wavelet is selected As mother wavelet, 5-8 scale discrete wavelet transform is performed, scale set s = {2l, 2 2, ··· 2 k}, and output detail coefficient D s (t) and approximation coefficient A s (t).

4. The method of claim 1, wherein, In S3, the optimal wavelet basis selection includes the following steps: (1) Construct a wavelet basis library, which includes but is not limited to Daubechies (dbN), Symlet (symN) and Coiflet (coifN) series; (2) calculating the correlation coefficient of each wavelet base and the defect template signal (3) The optimal wavelet base with p>0.85 is selected to reconstruct the signal x w (t).

5. The method of claim 1, wherein, In S4, the signal reconstructed by the optimal wavelet basis is subjected to Hilbert transform, and the formula is: generating an envelope signal:

6. The method of claim 1, wherein, In S5, the mean filtering includes the following steps: (1) Design a variable window length filter, wherein the window width f s is the sampling rate, f c is the defect characteristic frequency obtained by spectrum analysis; (2) Perform a sliding window average, formula is:

7. The method of claim 1, wherein, In S6, the threshold processing includes the following steps: (1) Calculate the noise standard deviation, σ n From the envelope signal median estimate, the formula is: n = median(|e f (t)|) / 0.6745; (2) Set dynamic threshold: T h = k1σ n , k1 ∈ [3.0, 4.0] is a high threshold, : T l = k2σ n , k2 ∈ [1.5, 2.5] is a low threshold; (3) Output defect position set: {t i |e f (t i ) > T h}∪{t j |e f (t j ) > T l and the field exists T h point}.

Images