Drill rod fatigue crack ultrasonic weak signal enhancement method based on sparse wavelet reconstruction

Through multimodal signal fusion and adaptive sparse wavelet reconstruction methods, the signal attenuation and noise interference problems of microcrack detection of downhole drilling pipes are solved, and the precise positioning and quantitative evaluation of root defects of drilling pipes are realized, which improves detection efficiency and accuracy.

CN120427736APending Publication Date: 2025-08-05CHINA COAL TECH & ENG GRP CHONGQING RES INST CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510581794.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-05-07
Publication Date
2025-08-05

AI Technical Summary

Technical Problem

The existing drill pipe detection technology has low detection efficiency under complex underground working conditions, making it difficult to effectively identify microcracks, and the model's cross-domain generalization capabilities are insufficient, which cannot meet the real-time detection needs.

Method used

Multimodal signal fusion and adaptive sparse wavelet reconstruction methods are adopted, combined with improved dynamic threshold function and singular value decomposition, a dual-path anti-migration network is designed to achieve accurate signal positioning and quantitative evaluation.

Benefits of technology

It significantly improves the signal-to-noise ratio, realizes millisecond real-time detection of microcracks of 0.2mm level, and has a low false alarm rate, filling the technical gap in intelligent detection and early warning of mechanical components under complex working conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120427736A_ABST
    Figure CN120427736A_ABST
Patent Text Reader

Abstract

The invention relates to a drill rod fatigue crack ultrasonic weak signal enhancement method based on sparse wavelet reconstruction, and belongs to the field of drill rod crack nondestructive testing. A multi-array-element ultrasonic data test is adopted for detecting the defect state of the drill rod, target features are enhanced through wavelet denoising result analysis, and crack recognition features are obtained based on a self-adaptive threshold value. The method comprises the following steps: carrying out wavelet decomposition and threshold processing on a detected signal, removing a noise signal, and reconstructing a wavelet coefficient of each layer after decomposition so as to obtain a de-noised ultrasonic detection signal. According to the method, the problem of weak crack signal decoupling under the strong noise background is solved, and accurate positioning and quantitative evaluation of the root defect of the drill rod thread are realized, so that the reliability of coal mine underground drill rod microcrack detection is improved, and a core technical support is provided for constructing an intelligent gas extraction safety monitoring system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of nondestructive detection of drill pipe cracks and relates to a method for enhancing ultrasonic weak signals of drill pipe fatigue cracks based on sparse wavelet reconstruction. Background Art

[0002] During coal mine gas extraction operations, drill pipes, as critical load-bearing components underground, are subjected to constant alternating loads and complex geological stresses. Microcracks are prone to developing at the threaded joints and on the rod surface. According to statistics, over 65% of drill pipe breakages are caused by the propagation of microcracks measuring 0.2-0.5 mm at the thread root. However, existing inspection technologies have a defect detection rate of less than 40% under complex underground conditions. This technical bottleneck poses a serious threat to coal mine safety and production, resulting in economic losses exceeding 10 billion yuan annually. Among traditional nondestructive testing technologies, eddy current testing (EDT) suffers from magnetic field distortion caused by the uneven surface of drill pipe threads, resulting in a false detection rate of 15-30%, and it cannot penetrate the mud overburden. X-ray testing, while offering high resolution, is burdened by bulky equipment and high radiation protection requirements, making it difficult to detect in situ downhole. Conventional ultrasonic testing, while effective for detecting cracks >1 mm in laboratory conditions, plummets to below 35% in actual operations due to mud attenuation (signal attenuation >60 dB / m), strong mechanical vibration (peak acceleration ≥15 g), and temperature gradients (temperature differences of up to 80°C at a depth of 3000 m). Existing research has shown that ultrasonic echo signals propagating through drill pipe threads exhibit multipath reflection and significant dispersion. The spectral overlap between the inherent thread reflection signal and the crack signature signal is as high as 70%, making it difficult to effectively separate them using traditional threshold filtering methods.

[0003] In recent years, machine learning-based intelligent detection algorithms have demonstrated promising results in laboratory simulations. Convolutional neural network (CNN) and support vector machine (SVM) models have achieved crack identification accuracy exceeding 85% on artificial standard test blocks. However, the distribution differences between downhole measured data and laboratory data lead to a sharp decline in model generalization performance, manifested by a sharp drop in the AUC value from 0.85 to below 0.7 due to feature space offset. This is primarily due to three technical bottlenecks: First, signal preprocessing suffers from high redundancy in extracting time-frequency domain features. Traditional wavelet basis function selection lacks adaptability to operating conditions and fails to consider the nonlinear effects of sudden changes in the acoustic impedance of drill pipe materials on signal propagation characteristics. Second, cross-domain transfer learning mechanisms are imperfect. Existing domain adversarial networks (DANNs) achieve only superficial feature matching during the laboratory-field data alignment process, failing to effectively decouple operating noise from crack-sensitive features. Third, real-time processing capabilities are insufficient. Existing systems consume >500ms to process a single frame of signal, failing to meet the millisecond-level response speed requirements of continuous gas extraction operations.

[0004] Domestic and foreign research teams have attempted to improve detection reliability through multi-sensor fusion solutions, such as the vibration-acoustic emission-temperature three-source synchronous acquisition system. However, such methods face the problem of coordinated optimization of weight distribution and noise suppression when fusing feature levels. Measured data show that the signal-to-noise ratio (SNR) after multi-source data fusion is less than 3dB higher. Although the dynamic control technology of negative pressure in gas extraction can improve the desorption efficiency of coalbed methane, its technical route focuses on increasing the extraction volume and does not involve the dimension of drill pipe structure health monitoring. Shock wave permeability enhancement technology uses multi-stage pulses to penetrate the coal body to form a fracture network, but this technology only focuses on the gas desorption rate improvement index and lacks a real-time assessment method for the service status of the drill pipe. Although existing drilling trajectory measurement systems (such as a combination of non-magnetic drill pipe and magnetoresistive sensor) can achieve a drilling spatial positioning accuracy of ±0.5°, they do not integrate crack diagnosis functions, resulting in blind spots in the safety assessment of drilling operations.

[0005] At the signal processing algorithm level, existing research primarily adheres to the classic wavelet threshold denoising framework, but fails to address three key issues: First, traditional hard thresholding functions introduce pseudo-Gibbs phenomena during noise suppression, resulting in >12% distortion of the rising edge of crack echoes; second, a fixed number of decomposition layers (typically 3-4) struggles to adapt to the multi-scale characteristics of drill pipe ultrasonic signals, leading to a loss of up to 25% in high-frequency detail components; third, threshold selection criteria (such as the Sqtwolog fixed threshold) fail to account for the dynamic changes in the local signal-to-noise ratio (SNR), which can lead to over-suppression of the effective signal in areas with strong background noise. Recent research suggests that a combined denoising scheme based on variational mode decomposition (VMD) and singular value decomposition (SVD) can improve the SNR to 15dB, but the computational complexity is as high as O(N^2), making it inadequate for real-time processing in embedded systems. Furthermore, existing feature enhancement methods, which often rely on time-domain envelope detection or spectral peak extraction, fail to fully exploit the phase coupling information of crack-sensitive features in the joint time-frequency domain, resulting in a single-dimensional feature representation.

[0006] Another prominent challenge facing this technical field is the limitation of model migration capabilities. The ultrasonic detection model constructed in the laboratory environment relies on ideal working condition data for training, while the multi-physical field coupling interference introduced by the downhole environment (including electromagnetic interference, stress wave crosstalk, and mud medium heterogeneity) causes a significant shift in the feature distribution. The existing transfer learning method adopts a global domain alignment strategy and does not establish a fine-grained feature correspondence. The measured data shows that the accuracy of the model on the cross-working condition test set is attenuated by 19%. What is more serious is that the traditional method does not take into account the time-varying effects of acoustic characteristics caused by material fatigue damage during the service life of the drill pipe, resulting in the need for frequent recalibration of the model and high maintenance costs. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a method for enhancing the ultrasonic weak signal of drill pipe fatigue cracks based on sparse wavelet reconstruction. Aiming at the core problems of severe signal attenuation, strong noise interference, insufficient cross-domain generalization ability of the model and poor real-time performance in ultrasonic detection of drill pipe microcracks in complex working conditions in coal mines, a technical solution combining multimodal signal fusion and adaptive sparse wavelet reconstruction is proposed to overcome the problem of decoupling weak crack signals under strong noise background, realize accurate positioning and quantitative evaluation of drill pipe thread root defects, thereby improving the reliability of drill pipe microcrack detection in coal mines, and providing core technical support for the construction of an intelligent gas extraction safety monitoring system.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A method for enhancing weak ultrasonic signals of fatigue cracks in drill pipes based on sparse wavelet reconstruction uses multi-element ultrasonic data testing to detect drill pipe defects. The target features are enhanced through wavelet denoising analysis, and crack identification features are obtained based on an adaptive threshold. The detected signal is decomposed and thresholded using wavelet decomposition to remove noise. Each layer of wavelet coefficients is then reconstructed to obtain a denoised ultrasonic detection signal. The method specifically includes the following steps:

[0010] S1: Wavelet decomposition: Obtain the drill pipe ultrasonic testing signal and perform wavelet decomposition on the signal;

[0011] S2: Selection of threshold function: Select an appropriate threshold function to retain the mutation characteristics of the ultrasonic signal after wavelet decomposition and remove the noise that affects the effective signal;

[0012] S3: Signal reconstruction: Perform singular value decomposition on the ultrasonic signal after wavelet denoising and filtering in step S2, and then reconstruct each layer of wavelet coefficients after decomposition;

[0013] S4: using the time domain envelope method to extract the contour of the ultrasonic signal reconstructed in step S3, thereby obtaining a denoised ultrasonic detection signal.

[0014] Furthermore, in step S1, wavelet decomposition specifically includes: selecting a 'db5' wavelet basis function with vanishing moment, high regularity, compact support and similarity to the echo signal, and selecting an appropriate wavelet decomposition layer number based on the selected wavelet basis function.

[0015] Furthermore, in step S2, the selection of the threshold function specifically includes: a heuristic threshold method, a maximum-minimum threshold method, an unbiased risk estimation threshold method, or a fixed threshold method.

[0016] The heuristic threshold method is specifically as follows: Assuming that the wavelet coefficient is W t , L is N W tSum of squares, η = (LN) / N, γ = (log2 N) 3 / 2 / N, the mathematical expression of the heuristic threshold is:

[0017]

[0018] Among them, δ h represents the heuristic threshold, δ represents the variance of the front segment ultrasound sequence signal, and δ1 represents the variance of the back segment ultrasound sequence signal.

[0019] The expression of the maximum and minimum threshold method is:

[0020]

[0021] Among them, δ mina represents the maximum and minimum thresholds, σ represents the variance, and m represents the number of sequences of ultrasonic signals.

[0022] The unbiased risk estimation threshold method specifically includes: arranging the wavelet coefficients in square order to form a set W:

[0023] W=[W1,W2,…,W m ], W1≤W2≤…≤W m

[0024] Among them, W m represents the mth wavelet coefficient;

[0025] Suppose a set G whose elements are calculated by the following formula:

[0026]

[0027] Among them, i∈[1,m], k∈[1,i], sort the elements in the set G so that its minimum value is the risk value, and find the elements in the set g respectively. i Minimum i value and W i , calculate the threshold for unbiased risk estimate

[0028]

[0029] Where σ is the noise variance of the signal.

[0030] The expression of the fixed threshold method is:

[0031]

[0032] Where δ is the ultrasonic discrimination threshold, σ is the variance, and N is the number of ultrasonic signal wavelet coefficients.

[0033] Furthermore, in step S3, singular value decomposition specifically includes: for a vibration signal actually collected, k modal components IMF1, IMF2, ..., IMF are obtained after variational mode decomposition (VMD). k , it can be formed into an n×k dimensional matrix A, where n is the number of sampling points of the signal; since A contains normal signals and noise signals, it can be expressed as

[0034] A=A S +A N (15)

[0035] Among them, A S represents a normal signal of dimension n×k, A N is the noise component of n×k dimensions;

[0036] Perform singular value decomposition on matrix A and get

[0037] A=UΣV T (16)

[0038] Among them, U is an n×n left singular matrix; V is a k×k right singular matrix; U and V are both unitary matrices, satisfying U T U=E,V T V = E, where E is the unit matrix; Σ is a diagonal matrix of dimension n × k, called a singular value matrix, whose main diagonal elements are composed of singular values from large to small, which can be expressed as

[0039]

[0040] Among them, σ i is the i-th singular value, i = 1, 2, …, r, r is the rank of the matrix Σ, and r = min(n, k); the formula is expanded to obtain:

[0041]

[0042] Among them, μ n is the nth left singular value, v k is the kth right singular value;

[0043] Furthermore, there are

[0044]

[0045] Each singular value σ in the matrix Σ i ,i=1,2,…,r represents the size of each matrix The importance of the original matrix A is usually determined by the first 10% or even 1% of the singular values in the matrix Σ. Singular values close to 0 can be considered irrelevant or noise components. Therefore, the first p items can be selected and superimposed according to the cumulative energy contribution rate to obtain an approximate matrix of the original matrix A.

[0046] The beneficial effects of the present invention are as follows: the present invention achieves technological breakthroughs through the following three innovations: first, constructing an adaptive wavelet packet decomposition hierarchical selection mechanism, combining an improved dynamic threshold function (integrating soft and hard threshold characteristics) with a singular value decomposition (SVD) denoising algorithm, while retaining the crack echo mutation characteristics, suppressing the pseudo-Gibbs phenomenon, and improving the signal-to-noise ratio by 8-10dB; secondly, designing a dual-path adversarial migration network, through interactive training of source domain and target domain feature extractors, combined with a dynamic domain discriminator to achieve multi-granularity feature alignment, solve the distribution offset problem between laboratory data and downhole measured data, and the model's cross-condition accuracy reaches 92.3%, an improvement of 19% over the baseline; finally, developing an embedded high-frequency signal processing unit, integrating a wavelet packet-Hilbert time-frequency joint analysis module and a multi-scale attention mechanism, strengthening the phase coupling feature extraction in the 15-25kHz sensitive frequency band, and realizing millisecond-level real-time detection of 0.2mm microcracks (single-frame processing takes <50ms), with a false alarm rate of <3.2%.

[0047] The method of the present invention overcomes the problem of decoupling weak crack signals under a strong noise background, supports the precise positioning and quantitative evaluation of defects at the root of drill pipe threads, significantly reduces the unplanned downtime rate by more than 80%, fills the technical gap in intelligent detection and early warning of mechanical components under complex working conditions, and can be expanded to structural health monitoring in the fields of oil drill pipes, geological exploration tools, etc.

[0048] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0050] Figure 1 This is the flow chart of drill pipe ultrasonic response signal processing and feature extraction;

[0051] Figure 2 Schematic diagram of signal time domain envelope detection. DETAILED DESCRIPTION

[0052] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0053] When using ultrasonic signals to detect defects in drill pipe threads, the propagation, reflection, scattering, and attenuation of ultrasound waves are key physical mechanisms. When ultrasound waves enter the drill pipe material, they propagate as mechanical waves, causing tiny vibrations within the material. Since drill pipe is typically made of metal, the speed of sound is high, allowing ultrasound waves to propagate efficiently within the material. However, when ultrasound waves encounter geometric variations within the material, such as threads and defects, the propagation path and characteristics of the waves are significantly altered. The thread structure itself has a complex effect on ultrasound propagation. The thread geometry causes multiple reflections and scattering of the sound waves, complicating the propagation of the ultrasonic signal. Specifically, the spacing and depth of the threads cause reflections and transmission of the sound waves, resulting in a periodic signal signature. Since each thread causes partial reflection of the ultrasound wave, these reflected waves superimpose to form a complex signal pattern. The presence of defects in the threads further alters the propagation and reflection characteristics of the ultrasound wave. Defects are discontinuities in the material, such as cracks, holes, or other structural damage, which can cause significant reflection of the ultrasound wave. The size and location of the defect are important factors affecting the intensity and characteristics of the reflected signal. Smaller defects (e.g., 2mm in length) offer less obstruction to ultrasonic waves, resulting in a relatively weak reflected signal. However, the characteristics of the reflected signal are still detectable because the acoustic wave reflection caused by the defect and the thread reflection form a different superimposed signal. Larger defects (e.g., 4mm in length) produce a stronger reflected signal. This is because larger defects obstruct more of the acoustic wave propagation path, resulting in higher energy reflected acoustic waves, which significantly enhances the reflected signal. The location of the defect has a significant impact on the characteristics of the reflected signal. When the defect is located at the third thread, the ultrasonic wave will reflect the sound wave early after propagating there, significantly enhancing the reflected signal. In this case, the defect's impact on the waveform is direct and significant. In contrast, when the defect is located at the fifth thread, the ultrasonic wave has already undergone multiple reflections and scattering from the previous threads, making the signal more complex. Despite this, the effects of defect location and length on the reflected signal strength can still be observed, but the signal pattern is more complex, requiring more sophisticated analysis to identify and distinguish.

[0054] In order to achieve decoupling and noise reduction of weak crack signals of drill pipes, an embodiment of the present invention provides a method for processing ultrasonic signals of drill pipes, such as Figure 1 The implementation steps are as follows:

[0055] S1: Select appropriate wavelet basis functions and wavelet decomposition layers to preprocess the drill pipe ultrasonic response signal.

[0056] S2: Combined with an appropriate threshold function, the mutation characteristics of the ultrasonic signal are retained and the noise that affects the effective signal is removed.

[0057] S3: After the detection signal is decomposed by wavelet and threshold processed, the noise signal is removed, and then the wavelet coefficients of each layer after decomposition are reconstructed.

[0058] S4: Use the time domain envelope method to extract the contour of the ultrasonic signal after noise cancellation.

[0059] The key to ultrasonic signal processing for detecting defects in pipes like drill pipe is to address the issue of signal contamination by scattered noise after receiving the signal, thereby effectively extracting the defect echo from the measured signal. Wavelet transforms, S-transforms, Hilbert-Huang transforms, and spectrum separation methods (SSP) are currently the most commonly used methods for ultrasonic signal processing. Signal processing research indicates that wavelet transforms and improved algorithms hold great potential for ultrasonic signal processing of drill pipe cracks.

[0060] As we all know, the ultrasonic signal in the drill pipe is a non-stationary and asymmetric signal. For the noise in the ultrasonic echo, the noise reduction method based on wavelet transform is an effective pre-processing method. The basic mathematical principle of wavelet transform is as follows: Assume that there exists a function ψ(t) and ψ(t)∈L 2 (R)(L 2 (R) is a square integrable real number space), and after Fourier transform, we get the function When the function Satisfy the following formula:

[0061]

[0062] Among them, C ν is the Fourier transform sequence, ω is the wavelet basis function resolution;

[0063] Then ψ(t) can be called the wavelet basis function. After scaling the wavelet basis function ψ(t), the wavelet sequence can be obtained:

[0064]

[0065] Where a is the scale adjustment parameter, which is a non-zero real number; b is the translation adjustment parameter, which is an arbitrary real number.2 (R), then the continuous wavelet transform (CWT) of f(t) is:

[0066]

[0067] The inverse transformation formula is:

[0068]

[0069] Among them, C ψ is the constant of the wavelet function, ψ(a,b) is the coefficient after wavelet transformation;

[0070] Piezoelectric ultrasonic data is discrete. When performing wavelet transform, the wavelet must be discretized, that is, the proportional adjustment parameter a and the translation adjustment parameter b are discretized. When discretized, set a>0, and the allowed conditions become:

[0071]

[0072] The discretization formulas for a and b are Where j∈Z, a0>1, then the corresponding wavelet transform ψ j,k (t) is written as:

[0073]

[0074] For any function f(t), the wavelet transform can be expressed as:

[0075]

[0076] in, is ψ j,k The dual sequence of (t). The inverse transform formula of discrete wavelet transform is:

[0077]

[0078] Where c is a constant. Wavelet denoising is a new noise reduction method developed based on wavelet analysis. By exploiting the difference in amplitude between the valid signal and noise after wavelet analysis, an amplitude threshold is set to separate the noise from the valid signal, filtering out all noise and retaining only the valid signal.

[0079] Wavelet threshold denoising is divided into four steps:

[0080] (1)Wavelet decomposition

[0081] According to the characteristics of the detection signal, a suitable wavelet basis is selected. Based on the asymmetric characteristics of the piezoelectric ultrasonic detection echo signal, the 'db5' wavelet basis function with vanishing moment, large regularity, tight support and similarity to the echo signal is selected. Based on the selected wavelet basis function, the appropriate number of wavelet decomposition layers is selected.

[0082] (2) Threshold function

[0083] A good threshold function should be able to retain the mutation characteristics of the ultrasonic signal and remove the noise that affects the effective signal. Currently, most people use hard threshold or soft threshold or a combination of the two.

[0084] (3) Selection of threshold

[0085] In the wavelet denoising process, the selection of threshold is a key step. If the threshold is too small, the ultrasonic signal may be eliminated. If the threshold is too large, the noise signal may not be eliminated. Different wavelet threshold methods are selected for different signals to be denoised. There are four commonly used threshold selection algorithms:

[0086] 1) Heuristic threshold (Heursure), the principle is: assuming that the wavelet coefficient is Wt, L is the sum of N squares of Wt, η = (LN) / N, γ = (log2 N) 3 / 2 / N, then the mathematical expression of the Heursure rule is:

[0087]

[0088] 2) Minimaxi threshold, the formula is as follows:

[0089]

[0090] 3) Unbiased risk estimation threshold (Rigrsure), the basic principle of which is to arrange the wavelet coefficients in square order to form a set.

[0091] W=[W1,W2,...W m ], W1≤W2≤...≤W m

[0092] Suppose a set G whose elements are calculated by the following formula:

[0093]

[0094] Among them, i∈[1,m], k∈[1,i], sort the elements in the set G so that its minimum value is the risk value, and find the i value and W of gmin in the set respectively. i , calculate the threshold for unbiased risk estimate

[0095]

[0096] Where σ is the noise variance of the signal.

[0097] 4) Fixed threshold method (Sqtwolog), its expression is shown in the following formula.

[0098]

[0099] (4) Signal reconstruction

[0100] After performing wavelet denoising filtering on the signal, singular value decomposition can decompose any matrix into the product of the left singular matrix, the singular value matrix and the right singular matrix. It can achieve the separation of feature space and noise space to a large extent and avoid the distortion of the original signal. For a vibration signal collected in practice, the k modal components IMF1, IMF2, ..., IMF are obtained after the variational mode decomposition (VMD). k It can be formed into an n×k dimensional matrix A, where n is the number of sampling points of the signal. Since A contains normal signals and noise signals, it can be further expressed as

[0101] A=A S +A N

[0102] Among them, A S represents a normal signal of dimension n×k, A N is the noise component of n×k dimensions.

[0103] Perform singular value decomposition on matrix A and get

[0104] A=UΣV T

[0105] Among them, U is an n×n left singular matrix; V is a k×k right singular matrix; U and V are both unitary matrices, satisfying U T U=E,V T V = E, where E is the identity matrix. Σ is a diagonal matrix of dimension n × k, called a singular value matrix, whose main diagonal elements are composed of singular values from large to small, which can be expressed as

[0106]

[0107] Where r is the rank of the matrix Σ, and r = min(n,k). The formula is expanded to:

[0108]

[0109] Furthermore, there are

[0110]

[0111] Each singular value σ in the matrix Σ i ,i=1,2,…,r represents the size of each matrix The importance of the original matrix A. Usually, more than 90% of the energy information in the original matrix A is contained in the first 10% or even 1% of the singular values in the matrix Σ. The singular values close to 0 can be considered as irrelevant or noise components. Therefore, the first p items can be selected according to the cumulative energy contribution rate and superimposed to obtain an approximate matrix of the original matrix A. In this embodiment, the cumulative energy contribution rate is 90%, and there is

[0112]

[0113] The approximate matrix with dimension n×p is obtained as

[0114]

[0115] In addition, the signal time domain envelope method can well observe the characteristics of the crack echo. Intuitively speaking, from the time domain, the signal envelope is the outline of the signal waveform. In essence, the signal envelope is the baseband part of the bandpass signal. Figure 2 As shown, a real bandpass signal is denoted as x(t). By moving the center frequency of its spectrum to zero frequency, the complex envelope of the signal, that is, the phase component of the signal, will be obtained. The signal can be processed according to this method.

[0116] Example:

[0117] This embodiment uses a multi-element ultrasonic data test to detect the defect status of the drill pipe, analyzes the wavelet denoising results, enhances the target features, and obtains crack identification features based on an adaptive threshold. After the detected signal is decomposed by wavelet and thresholded, the noise signal is removed, and then each layer of wavelet coefficients after decomposition is reconstructed to obtain the denoised ultrasonic detection signal. It should be pointed out that the basis function of the wavelet denoising used in the present invention is the Daubechies wavelet (dbN), because it has a tightly supported orthogonal wavelet, is suitable for processing sharp mutations or non-stationary signals, and has high regularity, that is, it can better approximate a smooth function. In addition, a 5-layer decomposition wavelet is used, and a soft threshold function is selected for processing.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for enhancing ultrasonic weak signals of drill pipe fatigue cracks based on sparse wavelet reconstruction, characterized in that: The method specifically comprises the following steps: S1: Wavelet decomposition: Obtain the drill pipe ultrasonic testing signal and perform wavelet decomposition on the signal; S2: Selection of threshold function: The threshold function is selected to retain the mutation characteristics of the ultrasonic signal after wavelet decomposition and remove the noise that affects the effective signal; S3: Signal reconstruction: Perform singular value decomposition on the ultrasonic signal after wavelet denoising and filtering in step S2, and then reconstruct each layer of wavelet coefficients after decomposition; S4: using the time domain envelope method to extract the contour of the ultrasonic signal reconstructed in step S3, thereby obtaining a denoised ultrasonic detection signal.

2. The method for enhancing ultrasonic weak signals of fatigue cracks in drill pipes according to claim 1, characterized in that: In step S1, the wavelet decomposition specifically includes: selecting the 'db5' wavelet basis function, and selecting the number of wavelet decomposition layers based on the selected wavelet basis function.

3. The method for enhancing ultrasonic weak signals of fatigue cracks in drill pipes according to claim 1, characterized in that: In step S2, the selection of the threshold function specifically includes: a heuristic threshold method, a maximum-minimum threshold method, an unbiased risk estimation threshold method, or a fixed threshold method.

4. The method for enhancing ultrasonic weak signals of fatigue cracks in drill pipes according to claim 3, characterized in that: In step S2, the heuristic threshold method is specifically as follows: assuming that the wavelet coefficient is W t , L is N W t Sum of squares, η = (LN) / N, γ = (log2 N) 3 / 2 / N, the mathematical expression of the heuristic threshold is: Among them, δ h represents the heuristic threshold, δ represents the variance of the front segment ultrasound sequence signal, and δ1 represents the variance of the back segment ultrasound sequence signal.

5. The method for enhancing ultrasonic weak signals of fatigue cracks in drill pipes according to claim 3, characterized in that: In step S2, the expression of the maximum and minimum threshold method is: Among them, δ mina represents the maximum and minimum thresholds, σ represents the variance, and m represents the number of sequences of ultrasonic signals.

6. The method for enhancing ultrasonic weak signals of fatigue cracks in drill pipes according to claim 3, characterized in that: In step S2, the unbiased risk estimation threshold method specifically includes: arranging the wavelet coefficients in square order to form a set W: W=[W1,W2,...,W m ],W1≤W2≤...≤W m Among them, W m represents the mth wavelet coefficient; Suppose a set G whose elements are calculated by the following formula: Among them, i∈[1,m], k∈[1,i], sort the elements in the set G so that its minimum value is the risk value, and find the elements in the set g respectively. i Minimum i value and W i , calculate the threshold for unbiased risk estimate Where σ is the noise variance of the signal.

7. The method for enhancing ultrasonic weak signals of fatigue cracks in drill pipes according to claim 3, characterized in that: In step S2, the expression of the fixed threshold method is: Where δ is the ultrasonic discrimination threshold, σ is the variance, and N is the number of ultrasonic signal wavelet coefficients.

8. The method for enhancing ultrasonic weak signals of drill pipe fatigue cracks according to claim 1, characterized in that: In step S3, singular value decomposition specifically includes: for a vibration signal actually collected, k modal components IMF1, IMF2, ..., IMF are obtained after variational mode decomposition. k , which is formed into an n×k dimensional matrix A, where n is the number of sampling points of the signal; A includes normal signals and noise signals, which can be expressed as A=A S +A N (15) Among them, A S represents a normal signal of dimension n×k, A N is the noise component of n×k dimensions; Perform singular value decomposition on matrix A and get A=UΣV T (16) Among them, U is an n×n left singular matrix; V is a k×k right singular matrix; U and V are both unitary matrices, satisfying U T U=E,V T V = E, where E is the identity matrix; Σ is a diagonal matrix of dimension n × k, called a singular value matrix, whose main diagonal elements are composed of singular values from large to small, expressed as Among them, σ i is the i-th singular value, i = 1, 2, …, r, r is the rank of the matrix Σ, and r = min(n, k); the formula is expanded to obtain: Among them, μ n is the nth left singular value, v k is the kth right singular value; Furthermore, there are Each singular value σ in the matrix Σ i ,i=1,2,…,r represents the size of each matrix According to the cumulative energy contribution rate, the first p items are selected for superposition to obtain an approximate matrix of the original matrix A.

Citation Information

Cited By

  • Ultrasonic identification method for alpha phase of titanium alloy strip and application

    CN121899263A