T-shaped welding node crack detection method based on signal sparse representation

Abstract

The invention discloses a signal sparse representation-based T-shaped welding node crack detection method, which comprises the following steps of: detecting a T-shaped welding node by using a transmitting and receiving longitudinal wave probe, and receiving a first echo signal at the moment; then, the receiving probe is moved downwards by a certain distance for detection again, and at the moment, a second echo signal is obtained; the method comprises the following steps: processing echo signals received twice by using a signal sparse representation method to extract diffraction wave arrival time in the corresponding two echo signals, then calculating coordinates of a defect tip by using a biellipse positioning method, and completing quantitative calculation of a crack according to the coordinates of the defect tip to obtain an angle and an extension length of the crack.

Description

Technical Field

[0001] The invention relates to a detection method for cracks in a T-shaped welding node based on sparse representation of signals, and belongs to the field of ultrasonic detection. Background Art

[0002] In order to effectively complete the evaluation of T-shaped welded joint cracks, it is necessary to use non-destructive testing technology to effectively, completely and accurately detect the welded joints. Ultrasonic diffraction time difference method (TOFD) technology has the advantages of high precision, high efficiency, and accurate quantitative analysis. It accurately calculates the defects through the time difference of the diffraction echo at the tip of the defect. TOFD technology has good detection performance for complex or thick structures such as steel structure welds and thick-walled structures, and can accurately complete the quantitative detection of defects.

[0003] TOFD technology also has certain disadvantages. Diffraction waves are usually weaker than direct waves and bottom echoes and are easily covered by noise. For example, when the defect is short, the diffraction echoes may be received by the probe almost at the same time, resulting in echo overlap. At this time, the received echo signal will be an aliased time domain signal containing Gaussian white noise and defect echo overlap, which makes it difficult for inspection personnel to perform subsequent defect evaluation and quantitative calculation. In order to improve the time domain resolution of the signal, the present invention uses a signal sparse representation method to process the aliased time domain signal, and accurately extracts the wave arrival time of each wave in the time domain signal through signal sparse representation, so as to complete the subsequent crack quantification detection.

[0004] Signal sparse representation is mainly divided into two parts. The first part is the selection of the dictionary. The design of the dictionary is crucial. The dictionary needs to represent possible defect echo signals as much as possible so that the echo signals can be represented with as few dictionary atoms as possible. The second part is the selection of the sparse constraint solving algorithm. The sparse constraint solving algorithm can determine whether the sparse solution you get is the global optimal one or whether the sparsity is the best. If the sparse coefficients are fewer, you will get more key information. Whether the algorithm is appropriate determines the quality of the signal sparse representation. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method for detecting cracks in a T-shaped welding node based on sparse representation of signals.

[0006] The technical solution adopted by the present invention is: a detection method for cracks in T-shaped welded joints based on sparse signal representation, using a one-transmit-one-receive longitudinal wave probe to detect T-shaped welded joints, and the propagation path of ultrasonic waves in the structure is as follows: Figure 1As shown, the first echo signal will be received at this time; then the receiving probe is moved downward a certain distance for detection again, and the second echo signal will be obtained at this time; the two received echo signals are processed by using the signal sparse representation method to extract the arrival time of the diffraction waves in the corresponding two echo signals, and then the coordinates of the defect tip are calculated by the double ellipse positioning method, and the quantitative calculation of the crack is completed according to the coordinates of the defect tip to obtain the angle and extension length of the crack; specifically, the following steps are included:

[0007] Step 1: The detection object in this paper is the T-shaped welding node of the orthotropic steel bridge deck. The detection probe is a one-receive and one-transmit longitudinal wave probe. First, the top plate thickness h1, U rib thickness h2 and weld conditions at the T-shaped welding node such as weld throat length and welding opposite side angle β of the orthotropic steel bridge deck are measured. Then, according to the above conditions, a suitable transmitting probe T1, that is, a suitable probe angle θ and a probe frequency f are selected;

[0008] Step 2: Set up a two-dimensional rectangular coordinate system at the T-shaped welding node with the line connecting the lower surface of the top plate and the outer side of the U rib as the coordinate origin, where the x-axis is perpendicular to the lower surface of the top plate, and the y-axis coincides with the lower surface of the top plate. The definition of the coordinates is based on Figure 2 As shown. Since the typical crack starts from the root of the T-type node, the root coordinates are (0, c), and the crack to be detected will extend to any position inside the T-type welding node. The transmitting probe is arranged on the right side of the T-type welding node, that is, the lower surface of the top plate with coordinates (0, a), and the receiving probe R1 is arranged on the lower side of the T-type node, that is, the outer side of the U rib with coordinates (d, e). Perform an A scan on the T-type node, and the ultrasonic wave enters the top plate through the coupling layer at an angle θ, and is reflected on the upper side of the top plate with coordinates (-h, b) and directed to the T-type node. When the ultrasonic wave reaches the node, part of the ultrasonic wave is diffracted by the defect tip (x1, y1), and part of the ultrasonic wave is directly or directed to the inside of the U rib and is received by the probe after reflection, and the first detection time domain signal Y1 is obtained.

[0009] Each coordinate can be obtained by simple calculation, such as a=2h / tan(90°-θ) in the transmitting point (0, a), b=h / tan(90°-θ) in the reflecting point, and c in the root node can be obtained by measurement. The setting of the receiving point (d, e) can also be easily obtained by simple calculation. The receiving probe needs to be set at a reasonable position. It should not be set too far to receive a weak echo signal, nor too close to receive the bottom echo, which is not conducive to signal judgment. Therefore, it is assumed that the ultrasonic wave emitted by the reflection point just enters the U rib from the edge, that is, the right weld toe of the T-shaped welding node. The incident angle ω is deduced from the distance from the right weld toe to the reflection point. According to ω, and the distance O from the weld toe to the inside of the U rib, the x-axis length of the first segment of the bottom echo p1=O*tanβ*tanω / (tanβ-tanω); according to the thickness of the U rib h1 and the known angle, the x-axis length of the second segment of the bottom echo p2=h1cos(2β-ω-90°) / sin(β-ω) can be deduced. Here we can get the coordinates of the receiving point d=p1+p2, e=d*tan(90°-β).

[0010] Step 3: Extract the aliased time domain signal Y1 received in step 2, process the signal using spectrum analysis and Hilbert transform, and extract discrete parameters such as frequency, bandwidth, and signal delay to design parameterized dictionary A.

[0011] The parameterized dictionary used is constructed based on the Gabor function, as shown in formula (6):

[0012]

[0013] Step 4: Use a non-convex sparse regularization method based on the generalized minimax concave penalty function (Gmc) to process the signal Y1 in step 2, and construct the objective function F based on the generalized minimax concave penalty function as shown in formula (1). Substitute the signal Y1 in step 2 and the dictionary A in step 3 into the objective function F. This formula belongs to the saddle point problem. Use the forward-backward algorithm to solve it and get the sparse solution X 01 According to the sparse solution and dictionary, the corresponding delay parameters are found to extract the arrival time t of the diffraction wave at the defect tip. 11 ;

[0014]

[0015] Step 5: Move the receiving probe downward by a distance S. At this time, the coordinates of the receiving probe R2 are (i, j). Repeat steps 2, 3, and 4 to obtain the arrival time t of the defect tip diffraction wave received by the R2 receiving point. 12 ;

[0016] In step 5, the coordinates of the receiving probe R2 are obtained by shifting R1 downward by S, so the coordinates (i, j) are i=d+S*cos(90°-β), j=e+S*sin(90°-β) respectively.

[0017] Step 6: According to TOFD-LWE, i.e., the double ellipse positioning method, the reflection point and the receiving point R1 can be regarded as the two foci of the ellipse, and the defect tip point will always be on the ellipse, which conforms to the geometric property of the ellipse, that is, the sum of any point on the ellipse to the two foci is a constant. The same is true for point R2. Substitute the arrival time of the diffraction wave obtained in step 4 and step 5 into the following formula (2-3). The two formulas can be combined to calculate the defect tip coordinates (x1, y1).

[0018]

[0019] In step 6, the calculation is performed with the reflection point and the receiving point as the focus, and the arrival time of the diffracted wave is obtained from the excitation of the probe, so the time t0 from the ultrasonic wave from the emission point to the reflection point needs to be subtracted. And this initial time t0 = h / C p cosθ, where C p is the speed of sound of longitudinal waves in steel structures.

[0020] Step 7: Based on the node root coordinates (0, c) and the defect tip coordinates (x1, y1) in steps 2 and 6, the length and angle of the crack can be accurately calculated using formula (4-5);

[0021]

[0022] Beneficial effects: The present invention performs TOFD detection and scanning on the T-shaped welding nodes of orthogonal anisotropic steel bridge decks, and performs signal sparse representation on the collected aliased time domain signals containing high noise and defect signals. According to the analysis and extraction of parameters of the aliased time domain signals, a corresponding dictionary is designed and a signal sparse representation algorithm is used to process the time domain signals. While reducing the signal-to-noise ratio of the time domain signal, the wave arrival time of each echo signal of the defect is accurately extracted. Based on the obtained wave arrival time, the double ellipse positioning method is used to calculate the precise coordinates of the crack tip, and the quantitative calculation of the crack is completed. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is the ultrasonic propagation path in TOFD detection of T-shaped welding nodes;

[0024] Figure 2 It is the coordinate definition of TOFD detection of T-shaped welding nodes;

[0025] Figure 3 It is a flow chart of a quantitative detection method of a weld crack of a T-shaped welding node of the present invention;

[0026] Figure 4 It is a part of the atoms of the corresponding dictionary designed according to the aliased time domain signal;

[0027] Figure 5 It is the noise-free time domain signal obtained by simulating TOFD detection;

[0028] Figure 6 It is to use GMC to reconstruct and sparsely represent the signal Y1 with 0dB noise added;

[0029] Figure 7 It is to use GMC to reconstruct and sparsely represent the signal Y2 with 0dB noise added;

[0030] Figure 8 This is a comparison chart of the precise quantitative characterization of cracks. DETAILED DESCRIPTION

[0031] The present invention will be described in more detail below with reference to the accompanying drawings and specific embodiments:

[0032] like Figure 3 As shown in the figure, a detection method for cracks in T-shaped welded joints based on sparse signal representation is presented. In order to further verify the feasibility and effectiveness of the present invention, the TOFD technology is used to detect weld cracks in T-shaped welded joints using the Abaqus finite element simulation software, and the output data, i.e., the aliased time domain signal obtained by the detection, is extracted. Then, the time domain signal is sparsely represented using the Matlab software, the arrival time of the diffraction wave is extracted, and the coordinates of the defect tip are solved using the double ellipse positioning method, and the quantitative calculation of the crack is completed based on the defect coordinates.

[0033] First, an orthotropic steel bridge deck with a 5mm crack model is established. The top plate size is 150mm×16mm, the U rib size is 100mm×8mm, the welding penetration is 80%, the coordinates of the weld heel in the T-shaped welding node are (0,-6.4), the coordinates of the right weld toe are (0,5.6), the weld angle is 105 degrees, and the crack extends from the root to the weld throat as a 15-degree inclined crack of 5mm. The crack coordinate design value is (1.29,-1.57). The probe settings in the TOFD test are a one-transmit-one-receive longitudinal wave probe with a probe frequency of 5MHz, an incident angle of 60 degrees, and a chip size of 6mm.

[0034] According to the design, the coordinates of the two probes are the transmitting probe coordinates (0, 55.4), the reflection point coordinates (-16, 27.7), and the receiving probe coordinates (23.32, -6.25). The probe delay of the transmitting probe, that is, the time from the start of the excitation pulse to the probe emission, is 7us, and the longitudinal wave speed in the material is 5856m / s.

[0035] Secondly, Abaqus software is used for simulation. The finite element model is designed according to the above parameters such as the probe, structure and weld, and then the material parameters such as T-node material, wedge material and coupling agent material are defined. The next step is mesh design. The appropriate mesh size is calculated according to the wave velocity and center frequency, that is, l = C p / 10f0, such as the maximum grid size of the node structure is 0.1mm. The time step is usually set to 1 / 20 of the maximum frequency. The loading of the excitation signal can be achieved by applying a transient excitation pulse modulated by a Gabor function to the transmitting probe to simulate the propagation of ultrasound in the structure. The signal is received by setting the field output at the receiving point outside the U rib to obtain the time domain signal y1 of the ultrasonic detection. Since there will be no noise in the simulation, the obtained time domain signal will be as follows Figure 4 As shown, the circle is the effective echo signal before the bottom wall echo. The following will focus on processing this signal and the aliased time domain signal Y1 with 5dB Gaussian white noise by Gmc algorithm to simulate the sparse representation of the signal under normal noise conditions.

[0036] Next, the aliased time domain signal y1 is processed. First, the spectrum analysis is performed to average the energy of the spectrum diagram and extract the discrete frequency parameters. Secondly, the signal y1 is Hilbert transformed to obtain the upper and lower envelope diagrams of the signal. The area of ​​the envelope diagram is also averaged to extract the signal delay parameters. This averaging method allows us to extract more parameters in energy-intensive or time-domain-intensive signal areas and reduce unnecessary calculations. Finally, based on the extraction and analysis of the incident pulse signal and the bottom echo signal, the attenuation effect will cause the high-frequency components of the signal to be more easily lost than the low-frequency components, so the frequency domain bandwidth will also decrease accordingly, and the time domain bandwidth parameter will gradually increase. Considering the influence of noise, the discrete bandwidth parameter is preferably 0.5 times the incident signal bandwidth to 1.5 times the bottom reflection echo signal bandwidth. The above three discrete parameters are substituted into formula (6) to construct a Gabor dictionary with good correlation with the signal, where some dictionary atoms such as Figure 5 shown.

[0037] After constructing the dictionary, a non-convex regularization method based on the generalized minimax concave penalty function (Gmc) is used to solve the aliased time domain signal Y1 with Gaussian white noise added to obtain the optimal sparse solution and obtain the reconstructed signal and the accurate arrival time of each echo according to the dictionary. Figure 6 The figure shows the result of using the dictionary and Gmc algorithm to perform sparse representation on signal Y1, and extract the diffraction echo t 11 The arrival time is 21.98us.

[0038] According to the above content, a second test is performed, in which the receiving probe is moved down by a distance of S, that is, 5mm, to obtain the second receiving point R2 (28.15, -7.54). A second test can be performed to obtain the second aliased time domain signal y2, and the same dictionary design and signal processing as above can be performed to obtain Figure 7 The reconstructed signal and sparse signal are shown in Figure 1, and the arrival time t of the diffraction echo of the second detection is obtained. 12 .

[0039] Substitute the arrival times of the two diffraction echoes into formula (2-3) for calculation, as shown below:

[0040]

[0041] According to the two formulas, the coordinates of the defect tip (1.29, -1.65) are calculated together, and the coordinates and the node root coordinates (0, -6.4) are substituted into formula (4-5), and finally the following can be obtained: Figure 8 The crack quantification detection fitting results shown in the figure have errors within 3% in both length and angle.

[0042]

[0043] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and deformations of these embodiments within the scope of the principles and technical ideas of the present invention still fall within the protection scope of the present invention.

Claims

1. A method for detecting cracks in a T-shaped welding node based on sparse signal representation, characterized in that: The steps include: Step 1: The detection object is the T-shaped welding node of the orthotropic steel bridge deck. The detection probe is a one-receive and one-transmit longitudinal wave probe. First, the top plate thickness h1, U rib thickness h2 and weld conditions at the T-shaped welding node of the orthotropic steel bridge deck are measured. The weld conditions include the weld throat length and the welding opposite side angle β. Then, according to the above conditions, the probe angle θ and probe frequency f of the transmitting probe T1 are selected; Step 2: At the T-type welding node, a two-dimensional rectangular coordinate system is set with the line connecting the lower surface of the top plate and the outer side of the U rib as the origin of the coordinates, where the x-axis is perpendicular to the lower surface of the top plate, and the y-axis coincides with the lower surface of the top plate; since the typical crack starts from the root of the T-type node, the root coordinates are (0, c), and the crack to be detected will extend to any position inside the T-type welding node. The transmitting probe is arranged on the right side of the T-type welding node, that is, the coordinates of the lower surface of the top plate are (0, a), and the receiving probe R1 point is arranged on the lower side of the T-type node, that is, the coordinates of the outer side of the U rib are (d, e); perform A scanning on the T-type node, and the ultrasonic wave enters the top plate through the coupling layer at an angle θ, and is reflected on the upper side of the top plate with coordinates (-h, b) and is emitted to the T-type node. When the ultrasonic wave is emitted to the node, part of the ultrasonic wave is diffracted by the defect tip (x1, y1), and part of the ultrasonic wave is directly or emitted to the inner side of the U rib and is received by the probe after reflection, and the first detection time domain signal Y1 is obtained; Step 3: Extract the aliased time domain signal Y1 received in step 2, process the signal using spectrum analysis and Hilbert transform, and extract discrete parameters such as frequency, bandwidth, and signal delay to design parameterized dictionary A; Step 4: Use a non-convex sparse regularization method based on a generalized minimax concave penalty function to process the signal Y1 in step 2, and construct an objective function F based on a generalized minimax concave penalty function as shown in formula (1). Substitute the signal Y1 in step 2 and the dictionary A in step 3 into the objective function F. This formula belongs to the saddle point problem. Use the forward-backward algorithm to solve it and get the sparse solution X 01 ; According to the sparse solution and dictionary, find the corresponding delay parameter to extract the arrival time t of the diffraction wave at the defect tip 11 ; Step 5: Move the receiving probe downward by a distance S. At this time, the coordinates of the receiving probe R2 are (i, j). Repeat steps 2, 3, and 4 to obtain the arrival time t of the defect tip diffraction wave received by the R2 receiving point. 12 ; Step 6: According to the double ellipse positioning method, the reflection point and the receiving point R1 can be regarded as the two foci of the ellipse, and the defect tip point will always be on the ellipse, which conforms to the geometric property of the ellipse, that is, the sum of any point on the ellipse to the two foci is a constant; the same is true for point R2; substitute the arrival time of the diffraction wave obtained in step 4 and step 5 into the following formula (2-3) and combine the two formulas to calculate the defect tip coordinates (x1, y1); Step 7: Based on the node root coordinates (0, c) and the defect tip coordinates (x1, y1) in steps 2 and 6, use formula (4-5) to accurately calculate the length and angle of the crack; 2. A method for detecting cracks in T-shaped welding nodes based on sparse signal representation according to claim 1, characterized in that: The coordinates in step 2 are obtained by calculation: in the transmitting point (0, a), a=2h / tan(90°-θ), in the reflection point b=h / tan(90°-θ), c in the root node is obtained according to the measurement, and the setting of the receiving point (d, e) is also obtained by calculation; assuming that the ultrasonic wave emitted by the reflection point just enters the U rib from the edge, that is, the right weld toe of the T-shaped welding node, the incident angle ω is deduced according to the distance from the right weld toe to the reflection point, and the x-axis length p1=O*tanβ*tanω / (tanβ-tanω) of the first segment of the bottom echo is deduced according to ω and the distance O from the weld toe to the inner side of the U rib; the x-axis length p2=h1cos(2β-ω-90°) / sin(β-ω) of the second segment of the bottom echo is deduced according to the U rib thickness h1 and the known angle; the coordinates of the receiving point d=p1+p2, e=d*tan(90°-β) are obtained.

3. According to the method for detecting cracks in T-shaped welding nodes based on sparse signal representation as described in claim 1, it is characterized in that: The parameterized dictionary used in step 3 is constructed based on the Gabor function, and the function is shown in formula (6):

4. A method for detecting cracks in T-shaped welding nodes based on sparse signal representation according to claim 1, characterized in that: In step 5, the coordinates of the receiving probe R2 are obtained by shifting R1 downward by S, so the coordinates (i, j) are i=d+S*cos(90°-β), j=e+S*sin(90°-β) respectively.

5. The method for detecting cracks in T-shaped welding nodes based on sparse signal representation according to claim 1, characterized in that: In step 6, the calculation is performed with the reflection point and the receiving point as the focus, and the arrival time of the diffracted wave is obtained from the excitation of the probe, so it is necessary to subtract the time t0 from the emission point to the reflection point of the ultrasonic wave, t0 = h / C p cosθ, where C p is the speed of sound of longitudinal waves in steel structures.

Images