A system and method for detecting cracks in fillet welds based on a double scanning ratio

Abstract

The present application belongs to the field of weld crack detection, and particularly relates to a fillet weld crack detection system and method based on double scanning comparison. The system performs 0° direct scanning and 30° oblique scanning on the same pipe head fillet weld to obtain two groups of data; the 0° scanning data are used to generate a reference defect map for calibrating welding defects and to demarcate a crack sensitive area; the 30° scanning data are subjected to root tomography processing to retain echo signals in a bidirectional preset depth interval with the weld root as the reference, thereby forming a root sensitive signal map; spatial comparison is performed to exclude confirmed defects, thereby obtaining residual signals; and then, according to the corner reflection criterion, signals located at the junction of the pipe outer wall and the weld root and corresponding to a position without defects are determined as root penetrating crack signals. The present application realizes reliable separation of crack and welding defect signals, supports crack depth grading and expansion pipe coupling self-adaptation, has high detection precision, and reduces misjudgment and missed detection.

Description

Technical Field

[0001] This invention belongs to the field of weld crack detection, and particularly relates to a fillet weld crack detection system and method based on dual-scan comparison. Background Technology

[0002] During the long-term service of fillet welds in tube sheets of heat exchangers in chemical plants, penetrating cracks that originate at the weld root and propagate from the outer wall to the inner wall of the tube have become the primary cause of leaks and unplanned shutdowns. These cracks originate at the weld root on the outer wall of the tube and gradually extend from the outside to the inside along the tube wall thickness. In the early stages of propagation, the cracks do not penetrate the tube wall and cannot be detected visually or by penetrant testing. Even when the cracks have expanded extensively along the circumference or are close to penetrating, because the crack direction is nearly parallel to the X-ray beam, no identifiable image can be formed on the film for rod anode X-ray inspection, resulting in missed detection.

[0003] In practice, phased array ultrasonic testing technology has been introduced to solve this problem. However, using a single scanning process presents certain challenges: if only a 0° direct beam scan is used, the sound beam is emitted axially along the pipe wall perpendicular to the probe surface. This can effectively detect manufacturing defects such as porosity, slag inclusions, and incomplete penetration inside the weld, as well as corrosion and wear generated during use. However, it has almost no response to root penetrating cracks extending from the outer wall to the inner wall, and no crack characteristic signals appear on the spectrum, resulting in pipe heads with large-area cracks still being judged as qualified. If only a 30° oblique beam scan is used, although reflections can be formed in some local areas of the crack, the crack itself is tortuous, the reflection points are discontinuous, the amplitude varies, and the position is irregular. The spectrum is simultaneously filled with welding defect signals, weld contour reflection signals, and structural noise, resulting in severe signal aliasing, making it impossible to accurately separate and identify crack characteristic signals.

[0004] In existing technologies, some solutions attempt to combine phased array multi-angle scanning with manual interpretation. However, when faced with penetrating cracks at the weld root, a detection logic that can systematically separate and compare crack signals with welding defect signals is still lacking. Other solutions rely on complex signal processing and pattern recognition algorithms, hoping to directly classify crack signals from a single oblique beam scan pattern. However, due to the unique nature of crack reflection mechanisms—that is, only the corner formed by the outer wall of the pipe and the crack can produce a deterministic strong reflection—pure signal feature extraction cannot reliably eliminate interference from welding defects, resulting in a high risk of misjudgment and missed detection. Furthermore, for complex on-site conditions such as changes in coupling conditions caused by the inner wall indentation after pipe expansion and limitations in beam direction, existing technologies lack a standardized method that can adaptively adjust the scanning strategy and maintain high detection reliability under different operating conditions.

[0005] In summary, the technical problem that the existing technology urgently needs to solve is: how to perform spatial comparison and corner reflection physical criteria discrimination on the root sensitive signal obtained by 30° oblique scanning based on the spatial location of welding defects calibrated by 0° direct scanning, so as to reliably separate the root penetrating crack characteristic signal from the mixed signal. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a fillet weld crack detection system and method based on dual-scan comparison. The system performs 0° direct scanning and 30° oblique scanning on the same pipe head fillet weld to obtain two sets of raw data. Based on the 0° direct scanning data, a 0° reference defect map is generated, calibrating the spatial location of all welding defects and delineating the crack-sensitive zone. Based on the 30° oblique scanning data, root tomography is performed, retaining only the echo signals within a depth range from 1mm towards the weld surface to 1.5mm towards the inner wall of the pipe, forming a root-sensitive signal spectrum. The two spectrum signals are spatially compared point-by-point, excluding signals confirmed as welding defects to obtain the remaining signal set. Then, based on the corner reflection physical criterion, the signal located at the junction of the outer wall of the pipe and the weld root, where no defect is displayed at the corresponding position, is determined to be a root penetrating crack signal. This invention achieves reliable separation of crack signals and welding defect signals, supports crack depth grading and adaptive adjustment of expansion coupling, has high detection accuracy, and reduces false positives and missed detections.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A fillet weld crack detection system based on dual-scan comparison includes:

[0009] The scanning data acquisition module is configured to perform 0° direct scanning and 30° oblique scanning on the same pipe head fillet weld, respectively, to acquire the original detection data of 0° direct scanning and 30° oblique scanning.

[0010] The reference defect map generation module is configured to generate a 0° reference defect map with the spatial location of welding defects calibrated by three-dimensional tomographic imaging based on the original detection data of the 0° direct scan, and to delineate the crack sensitive area without welding defects.

[0011] The spectrum generation module is configured to extract echo signals within a preset root depth range through root tomography processing based on the original detection data from the 30° oblique scanning, and generate a root sensitive signal spectrum.

[0012] The spatial comparison module is configured to compare the first spatial position coordinates of each signal in the root sensitive signal spectrum with the second spatial position coordinates of the defects already marked in the 0° reference defect map, exclude confirmed welding defect signals with overlapping positions, and retain signals with non-overlapping positions as the remaining signal set.

[0013] The crack detection module is configured to determine whether each signal in the remaining signal set simultaneously satisfies the following conditions: the reflection point is located in the junction area between the outer wall of the pipe and the root of the weld, and the corresponding reflection point has no welding defect signal in the 0° reference defect diagram. If these conditions are met, it is determined to be a corner reflection signal, and the root penetrating crack of the corresponding pipe head is confirmed.

[0014] Specifically, in the scanning data acquisition module, the sound beam of the 0° direct scanning is perpendicular to the probe surface and emitted along the pipe wall axis; the sound beam of the 30° oblique scanning is directed at the weld root region at a 30° tilt angle.

[0015] The reference defect map generation module is also configured to generate a 0° direct G-scan map and a D-scan map on the original detection data of the 0° direct scan, and perform three-dimensional sectional analysis and slicing to generate the 0° reference defect map, and delineate the crack sensitive area according to the location area where there is no welding defect signal in the 0° reference defect map.

[0016] The spectrum generation module is also configured to generate a 30° oblique G-scan image on the raw detection data of the 30° oblique scanning, and to perform root tomography slicing processing on the 30° oblique G-scan image.

[0017] Specifically, the root tomographic sectioning process is configured as follows: using the weld root position at the junction of the outer wall of the pipe and the weld as a depth reference, the position extending a first preset distance from the depth reference towards the weld surface is determined as a first depth boundary, and the position extending the first preset distance from the depth reference towards the inner wall of the pipe is determined as a second depth boundary, retaining echo signals whose depth positions fall between the first depth boundary and the second depth boundary; and, the position extending a second preset distance from the depth reference towards the inner wall of the pipe is determined as a third depth boundary, retaining echo signals whose depth positions fall between the second depth boundary and the third depth boundary; generating the root sensitive signal spectrum from all retained echo signals, and discarding the unretained echo signals as weld contour reflection signals and upper defect signals.

[0018] Specifically, the spatial comparison module is further configured to perform point-by-point spatial difference calculation on the first spatial position coordinates of each signal in the root sensitive signal spectrum and the second spatial position coordinates of each defect already marked in the 0° reference defect map, and determine that the positions coincide when the spatial difference is less than a preset overlap threshold.

[0019] The fillet weld crack detection system also includes a depth grading module, configured to extract echo amplitude values ​​from signal data determined to be corner reflection signals, input the echo amplitude values ​​into a pre-calibrated amplitude-crack depth mapping lookup table based on a precision wire-cut groove test block to obtain the corresponding crack depth estimate; and compare the crack depth estimate with a preset first crack depth threshold and a preset second crack depth threshold. If the crack depth estimate is greater than or equal to the first crack depth threshold, the crack is classified as a third crack level; if the crack depth estimate is less than the first crack depth threshold but greater than or equal to the second crack depth threshold, the crack is classified as a second crack level; if the crack depth estimate is less than the second crack depth threshold, the crack is classified as a first crack initiation level.

[0020] Specifically, the fillet weld crack detection system also includes: a coupling adaptation module, configured as follows:

[0021] The reflected echo signal of the inner wall interface of the tube was obtained by the phased array probe before the start of the 30° oblique scanning, and the amplitude of the reflected echo of the inner wall interface was extracted from it. The amplitude of the reflected echo of the inner wall interface was divided by the peak-to-peak amplitude of the transmitted calibration pulse to obtain the amplitude ratio of the reflected echo of the interface.

[0022] A fast Fourier transform is performed on the reflected echo signal from the inner wall of the tube to obtain the frequency domain amplitude distribution. The sum of the products of the amplitude of each frequency component and its corresponding frequency in the frequency domain amplitude distribution is calculated and divided by the sum of the amplitudes of each frequency component to obtain the centroid frequency. The absolute value of the difference between the centroid frequency and the nominal center frequency of the probe is taken as the spectral centroid offset. The sum of the amplitude ratio of the reflected echo from the interface multiplied by a first preset weight and the spectral centroid offset multiplied by a second preset weight is taken as the coupling state parameter.

[0023] The coupling state parameter is compared with a preset coupling quality threshold; if the coupling state parameter is less than the preset coupling quality threshold, it is determined that the upward oblique coupling condition has deteriorated, and a first adaptive adjustment or a second adaptive adjustment is performed.

[0024] If the coupling state parameter is greater than or equal to the preset coupling quality threshold, the upward oblique coupling condition is determined to be normal, and the 30° oblique scanning is continued with the current upward oblique beam direction and the current ultrasound probe.

[0025] Specifically, generating the 0° baseline defect map and delineating the crack-sensitive region includes:

[0026] The raw detection data from the 0° direct beam scan is subjected to a three-dimensional Fourier transform or a synthetic aperture focusing algorithm to reconstruct the A-scan signal sequence acquired along the pipe wall axial direction into a 0° direct beam G-scan image and a D-scan image; wherein, the 0° direct beam G-scan image is a cross-sectional view of the weld perpendicular to the direction of the sound beam, and the D-scan image is a longitudinal cross-sectional view of the weld along the direction of sound beam propagation;

[0027] The two-dimensional image sequence consisting of the 0° direct G-scan image and the D-scan image is subjected to three-dimensional segmentation and slicing. Specifically, a three-dimensional voxel grid is constructed with the probe coupling surface as the origin of the spatial coordinate system and along the radial, axial and circumferential directions of the weld seam with a preset voxel size. The echo amplitude of each A-scan signal sampling point is mapped to the corresponding voxel according to the sound path and the sound beam diffusion angle to form a three-dimensional amplitude matrix. The three-dimensional amplitude matrix is ​​traversed using a preset amplitude threshold. Voxels with amplitudes exceeding the preset amplitude threshold are marked as defect candidate voxels. A three-dimensional connected component labeling algorithm is applied to cluster spatially adjacent defect candidate voxels into independent defect objects.

[0028] Specifically, generating the 0° reference defect map and delineating the crack-sensitive region further includes:

[0029] For each independent defect object, its bulk centroid coordinates are extracted as the second spatial position coordinates of the corresponding independent defect object, and the maximum amplitude, volume and morphological characteristics of the corresponding independent defect object are recorded as attribute parameters; the second spatial position coordinates and attribute parameters of all the labeled independent defect objects are integrated to form the 0° reference defect map;

[0030] The echo signal of the original 0° direct scan detection data is obtained within the full thickness range of the weld. Based on the acoustic impedance difference between the weld metal and the pipe wall base material, the actual boundary contour between the outer wall of the pipe and the molten metal of the weld is extracted by the interface echo recognition algorithm. The actual boundary contour is offset into the weld by a preset molten zone thickness value to obtain the boundary of the welding molten zone.

[0031] In the 0° reference defect diagram, the spatial location area that is within the boundary of the weld melting zone and is not occupied by any of the independent defect objects is determined as the location area where there is no weld defect signal. The spatially connected set of all location areas where there is no weld defect signal that extend continuously along the outer wall of the pipe to the weld root boundary line is defined as the crack sensitive area.

[0032] Specifically, the corresponding estimated crack depth is obtained, including:

[0033] In the original detection data of the 30° oblique scanning, the A-scan signal corresponding to the corner reflection signal is located; based on the spatial coordinates of the reflection point of the corner reflection signal, combined with the incident angle of the sound beam of the 30° oblique scanning, the longitudinal wave velocity of the wedge material and the transverse wave velocity of the pipe wall material under test, the propagation time of the wedge segment and the pipe wall segment along the sound beam path is calculated and summed to obtain the round-trip propagation time of the ultrasonic wave from the probe chip to the probe chip via the reflection point, and the round-trip propagation time is determined as the target time on the A-scan signal;

[0034] Centered on the target time, the time interval defined by extending forward by a first preset search half-width and backward by a second preset search half-width is taken as the target search time interval; both the first preset search half-width and the second preset search half-width are taken as the half-width of the ultrasonic pulse emitted by the phased array probe used in the 30° oblique scanning in the pipe wall material.

[0035] Specifically, obtaining the corresponding estimated crack depth also includes:

[0036] Within the target search time interval, the signal amplitude of the A scan signal is obtained point by point. The signal amplitude corresponding to the sampling point with the largest absolute amplitude among all the obtained sampling points is extracted as the echo amplitude of the corner reflection signal.

[0037] The echo amplitude value is input into a pre-established amplitude-crack depth mapping lookup table, and the corresponding crack depth estimate is obtained by matching the echo amplitude value range.

[0038] If the echo amplitude value is greater than the maximum echo amplitude value in the amplitude-crack depth mapping lookup table, then the estimated crack depth value is taken as the corresponding maximum crack depth value in the lookup table, and an over-range marker signal is output.

[0039] If the echo amplitude value is less than the minimum echo amplitude value in the amplitude-crack depth mapping lookup table, then the estimated crack depth value is set to 0.

[0040] Specifically, a three-dimensional connected component labeling algorithm is applied to cluster spatially adjacent defect candidate voxels into independent defect objects, including:

[0041] Traverse each voxel in the three-dimensional amplitude matrix and obtain the voxel amplitude of that voxel. When the voxel amplitude of that voxel is greater than the preset amplitude threshold, mark that voxel as a defect candidate voxel; when the voxel amplitude of that voxel is not greater than the preset amplitude threshold, do not mark that voxel.

[0042] Create a defect object set to store independent defect objects, and set a merge flag that is initially in an unmerged state for each defect candidate voxel in the three-dimensional magnitude matrix;

[0043] According to the row index order, column index order, and layer index order of the voxels in the three-dimensional amplitude matrix, the following merging operation is performed on each defect candidate voxel in sequence: Check whether there are defect candidate voxels with a merged state among the 26 neighboring voxels centered on the current defect candidate voxel. If so, select the independent defect object to which the defect candidate voxel with a merged state belongs from the 26 neighboring voxels, add the current defect candidate voxel to the independent defect object, and update the merging mark of the current defect candidate voxel to the merged state; if not, create a new independent defect object in the defect object set, add the current defect candidate voxel to the new independent defect object, and update the merging mark of the current defect candidate voxel to the merged state.

[0044] When all candidate defect voxels are in a merged state, each independent defect object stored in the defect object set is a clustered defect object group, and each independent defect object consists of all candidate defect voxels that are spatially adjacent to each other.

[0045] A method for detecting fillet weld cracks based on dual-scan comparison includes:

[0046] For the same pipe head fillet weld, perform 0° direct beam scanning and 30° oblique beam scanning respectively to obtain the original inspection data of 0° direct beam scanning and 30° oblique beam scanning.

[0047] Based on the original detection data from the 0° direct scanning, a 0° reference defect map with the spatial location of welding defects is generated by three-dimensional tomographic imaging, and a crack-sensitive area without welding defects is delineated.

[0048] Based on the original detection data from the 30° oblique scanning, echo signals within a preset root depth range are extracted through root tomography to generate a root sensitive signal spectrum.

[0049] The first spatial position coordinates of each signal in the root sensitive signal spectrum are compared with the second spatial position coordinates of the defects already marked in the 0° reference defect map. The confirmed welding defect signals with overlapping positions are excluded, and the signals with non-overlapping positions are retained as the remaining signal set.

[0050] For each signal in the remaining signal set, determine whether it simultaneously satisfies the following conditions: the reflection point is located in the junction area between the outer wall of the pipe and the root of the weld, and the corresponding reflection point has no welding defect signal in the 0° reference defect diagram. If these conditions are met, it is determined to be a corner reflection signal, and it is determined that there is a root penetrating crack in the corresponding pipe head.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] This invention acquires data through dual scanning using 0° direct scanning and 30° oblique scanning. It establishes a dual-map comparison framework: a 0° baseline defect map identifies all welding manufacturing and usage defects, while a 30° root-sensitive signal map focuses on the root-sensitive area. By comparing spatial locations point-by-point, welding defects already identified in the 0° baseline defect map are systematically excluded from the root-sensitive signal map, forming a set containing only unlabeled residual signals. Based on the physical criterion that pipe outer wall cracks generate high-amplitude, positional reflected echoes when forming corner structures, corner reflection condition determination is performed on the residual signals, thereby identifying root-penetrating crack characteristic signals. The signal is separated from the welding defect signal in both spatial location and physical mechanism, eliminating the interference of welding defects on crack identification. At the same time, the weld contour reflection signal and the upper defect signal are removed by root tomographic slicing, avoiding misjudgment caused by signal aliasing. Furthermore, the crack depth is quantitatively graded based on the amplitude-crack depth mapping relationship, and adaptive adjustment of beam direction switching and micro probe compensation is provided for the coupling deterioration caused by tube expansion and indentation, ensuring the consistency of detection and engineering applicability under different working conditions. Finally, high-precision and high-reliability detection of penetrating cracks at the root of tube and tube sheet fillet welds is achieved, significantly reducing the risk of misjudgment and missed detection. Attached Figure Description

[0053] Figure 1 This is a structural diagram of the fillet weld crack detection system based on dual-scan comparison in Embodiment 1 of the present invention;

[0054] Figure 2 This is a flowchart of generating a 0° reference defect map and delineating the crack-sensitive region in Embodiment 2 of the present invention;

[0055] Figure 3 This is a flowchart for obtaining the corresponding crack depth estimate in Embodiment 3 of the present invention. Detailed Implementation

[0056] Example 1:

[0057] Please see Figure 1 One embodiment of the present invention provides a fillet weld crack detection system based on dual-scan comparison, the system comprising:

[0058] The scanning data acquisition module is configured to perform 0° direct scanning and 30° oblique scanning on the same pipe head fillet weld, respectively, to acquire the original detection data of 0° direct scanning and 30° oblique scanning. The sound beam of the 0° direct scanning is perpendicular to the probe surface and emitted along the pipe wall axis, covering the entire thickness range of the weld. The sound beam of the 30° oblique scanning is directed at the root area of ​​the weld at a 30° tilt angle, with the focal point covering the sensitive area from the outer wall of the pipe to the root of the weld.

[0059] The reference defect map generation module is configured to generate a 0° direct-scan G-scan map and a D-scan map on the original 0° direct-scan inspection data, and perform three-dimensional sectional analysis and slicing to identify and mark the spatial positions of all welding manufacturing defects and service defects, generate a 0° reference defect map marked with the coordinates of each defect position, and delineate crack sensitive areas based on the location areas in the 0° reference defect map where no welding defect signals exist.

[0060] The image generation module is configured to generate a 30° oblique beam G-scan image on the original 30° oblique beam G-scan data, and perform root tomographic slicing processing on the 30° oblique beam G-scan image. The root tomographic slicing processing includes: using the weld root position at the junction of the outer wall of the pipe and the weld as a depth reference, determining the position extending a first preset distance from the depth reference towards the weld surface as a first depth boundary, and determining the position extending the first preset distance from the depth reference towards the inner wall of the pipe as a second depth boundary, retaining echo signals whose depth positions fall between the first depth boundary and the second depth boundary; and retaining echo signals whose depth positions fall between the first depth boundary and the second depth boundary. The distance is determined as the third depth boundary, and echo signals whose depth falls between the second depth boundary and the third depth boundary are retained. The root sensitive signal spectrum is generated from all retained echo signals, and the unretained echo signals are discarded as weld contour reflection signals and upper defect signals. In this embodiment, the first preset distance is ±1mm, where "±" represents a bidirectional depth range offset by 1mm from the weld root position to both the weld surface direction and the pipe inner wall direction. The second preset distance is 1.5mm, so that the target depth range is the depth range from 1mm from the weld root position to the weld surface direction to 1.5mm from the pipe inner wall direction.

[0061] The spatial comparison module is configured to extract the first spatial position coordinates of each signal in the root sensitive signal spectrum, and compare each first spatial position coordinate with the second spatial position coordinates of each defect already marked in the 0° reference defect map. Specifically, it performs point-by-point spatial difference calculation, and when the spatial difference is less than a preset overlap threshold, it is determined to be a position overlap. Signals with overlapping positions are excluded as confirmed welding defect signals, and signals with non-overlapping positions are retained to form a set of remaining signals.

[0062] In this embodiment, the preset overlap threshold needs to be determined by those skilled in the art based on the overall spatial resolution of the ultrasonic testing system. Its value should not be less than the maximum positioning error the system may produce, ensuring that the same defect is not incorrectly judged as non-overlapping due to minor measurement deviations. The main sources of positioning error include the size limitations of the three-dimensional voxel grid, the spatial ambiguity of reflection points caused by the beam diffusion angle, and the influence of the stepping accuracy of mechanical scanning. Specifically, this embodiment uses 0.2 mm as the preset overlap threshold. This value is approximately half the half-wavelength of the 7.5 MHz probe in the pipe wall material, and is on the same order of magnitude as the voxel size and scanning stepping. This ensures that signals from the same welding defect, slightly offset only due to resolution limitations, are reliably judged as overlapping, while also effectively distinguishing adjacent defects. In practice, those skilled in the art typically select this threshold within the range of 0.1 mm to 0.5 mm and verify it by comparing the measured positioning deviations of known defects on the test block, ensuring that the comparison results are neither missed nor misjudged.

[0063] The crack detection module is configured to extract the spatial coordinates of the reflection point for each signal in the remaining signal set, and determine whether the following conditions are met simultaneously: the reflection point is located on the outer wall side of the pipe and within a preset neighborhood of the weld root boundary line, and the original detection data of the 0° direct scan shows no defect signal at the reflection point location; if the conditions are met simultaneously, the signal is determined to be a corner reflection signal, and a root penetrating crack is determined to exist at the corresponding pipe head.

[0064] In this embodiment, the setting of the preset neighborhood range needs to be determined by those skilled in the art based on the actual geometric features of the weld root boundary profile and the spatial resolution of the detection system. This neighborhood range is used to define whether the reflection point is close enough to the theoretical boundary line between the outer wall of the pipe and the weld root, thereby satisfying the physical conditions for corner reflection. Its value mainly depends on the extraction accuracy of the boundary profile, the positioning deviation of the weld root molten zone boundary, and the size of the three-dimensional voxel mesh. Since the actual boundary formed during the welding process is not an ideal geometric line segment, but a transition area with a certain width, and the boundary profile extracted by the interface echo recognition algorithm itself has positional uncertainties caused by sampling point intervals and signal interpretation, a reasonable tolerance range needs to be set. In this specific implementation, a preset neighborhood range of 0.3 mm was selected. This value is on the same order of magnitude as the preset overlap threshold and is slightly larger than the radial dimension of a single voxel. This reliably covers the positional fluctuations caused by the boundary contour extraction deviation and the geometric irregularities at the root, ensuring that the corner reflection signal originating from the junction of the pipe outer wall and the weld root is accurately captured, while avoiding misjudging other reflection signals far from the boundary line as crack signals. In practice, those skilled in the art typically combine the measured reflection point coordinates of the artificially simulated crack groove at a known location on the calibration block with the statistical value of the deviation between the measured coordinates and the theoretical boundary line to select this neighborhood range within the range of 0.2 mm to 0.5 mm.

[0065] The physical basis of the corner reflection signal is as follows: When the outer wall of the pipe is uncracked, its surface is a continuous, smooth curved surface formed by mechanical processing. When an ultrasonic oblique beam irradiates this smooth curved surface at a preset incident angle, the beam undergoes specular reflection at the interface of the outer wall of the pipe. The propagation direction of the reflected beam is determined by both the incident angle and the difference in acoustic impedance on both sides of the interface. The reflected beam propagates in a direction away from the probe, and does not form a directional received echo at the probe's receiving position. When the outer wall of the pipe develops a crack that extends from the outside to the inside along the pipe wall thickness direction due to fatigue or stress corrosion, the two opposite surfaces of the crack are respectively... The tube consists of the original smooth surface of the outer wall and the crack fracture surface. These two opposing surfaces intersect at the crack initiation point to form a corner structure with dihedral geometry. When the ultrasonic oblique beam irradiates this corner structure, the beam is first reflected on the crack fracture surface and then reflected a second time on the outer wall surface of the tube. The two consecutive reflections cause the beam to return to the probe in a direction similar to the incident path, thereby forming a reflected echo at the probe with a definite spatial position and an echo amplitude exceeding a preset amplitude threshold. The spatial position of this reflected echo corresponds to the crack initiation point on the outer wall of the tube.

[0066] In this embodiment, the preset incident angle needs to be determined by those skilled in the art based on the waveform conversion requirements of the sound beam in the pipe wall material, the geometric accessibility of the weld root, and the generation mechanism of the corner reflection signal. The primary consideration in selecting this incident angle is to ensure that the sound beam refracted into the pipe wall propagates in the form of a pure transverse wave, avoiding interference from the longitudinal wave component. Therefore, the incident angle must be greater than the first critical angle, i.e., the arcsine of the ratio of the longitudinal wave velocity of the wedge to the longitudinal wave velocity of the pipe wall, and less than the second critical angle, i.e., the arcsine of the ratio of the longitudinal wave velocity of the wedge to the transverse wave velocity of the pipe wall, to ensure that only transverse waves exist in the pipe wall. Secondly, the incident angle also needs to ensure that the obliquely incident sound beam can effectively cover the boundary area between the outer wall of the pipe and the weld root, so as to facilitate the formation of a stable corner reflection at that position. As a specific example, the embodiment uses 30° as the preset incident angle. This angle, with the commonly used combination of plexiglass wedge and carbon steel pipe wall material, can generate pure transverse wave detection conditions. Furthermore, the main lobe of the sound beam, after refraction, points towards the sensitive area at the root of the weld, forming a suitable angle with the pipe wall normal. This allows the incident transverse wave sound beam to return to the probe along a direction close to its original path after two consecutive reflections from the crack surface and the pipe wall surface when there is a corner structure formed by a crack on the outer wall of the pipe, generating a high-amplitude reflected echo that satisfies the physical criteria of corner reflection. In practice, those skilled in the art will calculate the usable incident angle range using Snell's law based on the sound velocity of the selected wedge material, the longitudinal wave sound velocity, and the transverse wave sound velocity of the pipe wall material being tested. They will then process fan-shaped test blocks or simulated crack test blocks to conduct sound field directivity tests, ultimately selecting an angle within the range of 25° to 35° that maximizes the signal-to-noise ratio of the echo at the root region and ensures the most stable corner reflection signal as the preset incident angle.

[0067] As a concrete example, consider the 30° oblique beam scanning inspection of a fillet weld on a carbon steel pipe with an outer diameter of 25 mm and a wall thickness of 2.5 mm. When there are no cracks on the outer wall of the pipe, the 30° oblique beam irradiates the smooth surface of the outer wall of the pipe, resulting in specular reflection. The reflection angle is equal to the incident angle, and the reflected beam propagates towards the side of the pipe plate away from the probe. The echo amplitude received by the phased array probe at this location is only at the noise level and does not form a recognizable reflection signal. However, when there is a penetrating crack on the outer wall of the pipe at the junction of the weld root, originating from the outer wall and extending inward along the wall thickness direction to a depth of approximately 1.2 mm, the crack fracture surface... The smooth outer surface of the pipe forms a dihedral corner structure with an angle of approximately 90° at the crack initiation point. A 30° oblique sound beam first irradiates the crack fracture surface, producing the first reflection. The reflected beam continues to propagate to the adjacent outer surface of the pipe, producing a second reflection. After two reflections, the beam propagation direction is basically parallel to the incident direction but opposite in direction. It returns along the original sound beam path and is received by the probe. On the root sensitive signal spectrum, it appears as a reflected signal located at the junction of the weld root, with the radial coordinate corresponding to the position of the outer wall of the pipe, and the echo amplitude is higher than the preset amplitude threshold. The position of this signal coincides with the actual crack initiation point.

[0068] The depth grading module is configured to extract the echo amplitude value from the signal data determined to be the corner reflection signal, input the echo amplitude value into a pre-calibrated amplitude-crack depth mapping relationship lookup table based on a precision wire-cut groove test block, obtain the corresponding crack depth estimate; and compare the crack depth estimate with a preset first crack depth threshold and a preset second crack depth threshold, wherein the first crack depth threshold is greater than the second crack depth threshold; if the crack depth estimate is greater than or equal to the first crack depth threshold, the crack is classified as a third crack level; if the crack depth estimate is less than the first crack depth threshold but greater than or equal to the second crack depth threshold, the crack is classified as a second crack level; if the crack depth estimate is less than the second crack depth threshold, the crack is classified as a first crack initiation level.

[0069] In this embodiment, the specific steps for extracting the echo amplitude are as follows:

[0070] In the original detection data of the 30° oblique scanning, the A-scan signal corresponding to the corner reflection signal is located. Based on the spatial coordinates of the reflection point of the corner reflection signal, combined with the incident angle of the sound beam during the 30° oblique scanning, the longitudinal wave velocity of the wedge material, and the transverse wave velocity of the tested pipe wall material, the propagation time of the wedge segment and the pipe wall segment along the sound beam path is calculated and summed to obtain the round-trip propagation time of the ultrasonic wave from the probe chip, through the reflection point, and back to the probe chip. This round-trip propagation time is determined as the target time on the A-scan signal. With the target time as the center, the [missing information - likely a data point or data point] is then used to determine the [missing information - likely a data point or data point]. The time interval defined by extending the target time forward by a first preset search half-width and extending it backward by a second preset search half-width is taken as the target search time interval. Both the first preset search half-width and the second preset search half-width are taken as the half-width of the ultrasonic pulse emitted by the phased array probe used in the 30° oblique scan in the pipe wall material. Within the target search time interval, the signal amplitude of the A-scan signal is obtained point by point. The signal amplitude corresponding to the sampling point with the largest absolute amplitude among all the obtained sampling points is extracted as the echo amplitude of the corner reflection signal.

[0071] In this embodiment, the setting of the first crack depth threshold and the second crack depth threshold needs to be determined by those skilled in the art based on the remaining pressure-bearing capacity of the pipe wall under test, the design margin of the equipment, and relevant engineering critical assessment criteria. The core of this grading system is to distinguish the severity of cracks to guide subsequent maintenance decisions. Typically, the second crack depth threshold corresponds to the crack initiation or micro-propagation stage, and its value should be able to reliably distinguish between micro-initiation cracks existing only on the surface and propagating cracks that have extended into the wall thickness, while also being greater than the minimum distinguishable crack depth of the detection system. The first crack depth threshold corresponds to the critical depth at which the crack has approached or reached the point where immediate intervention is required, and its value should be based on the safety requirement that the remaining wall thickness of the pipe can withstand the design pressure, usually related to the safety score of the wall thickness. As a specific example, this embodiment selects 0.5 mm as the second crack depth threshold to classify the first crack initiation level and the shallower propagation crack level. This value is approximately 20% of the 2.5 mm wall thickness, which can clearly define whether the crack has caused substantial wall thickness damage. A 1.5 mm threshold is selected as the first crack depth threshold to classify the shallower crack level and the high-risk crack level requiring immediate treatment. This value is approximately 60% of the wall thickness, close to the maximum allowable local thinning of a typical heat exchanger tube. Exceeding this depth results in insufficient remaining wall thickness, posing a significant risk of leakage or breakage. In practice, those skilled in the art will select the second crack depth threshold within the range of 10% to 30% of the wall thickness and the first crack depth threshold within the range of 40% to 80% of the wall thickness, based on the specific heat exchanger tube's design wall thickness, allowable material stress, corrosion allowance under operating conditions, and heat exchanger tube maintenance standards. The thresholds are then verified through simulated samples containing real fatigue cracks to ensure the engineering rationality and safety reliability of the threshold classification.

[0072] The coupling adaptation module is configured as follows:

[0073] The reflected echo signal from the inner wall interface of the tube, measured by a phased array probe before the start of the 30° oblique scanning, is acquired. The amplitude of the reflected echo from the inner wall interface is extracted, and the amplitude is divided by the peak-to-peak amplitude of the transmitted calibration pulse to obtain the interface reflected echo amplitude ratio. A fast Fourier transform is performed on the reflected echo signal to obtain the frequency domain amplitude distribution. The sum of the products of the amplitude of each frequency component and its corresponding frequency in the frequency domain amplitude distribution is calculated and divided by the sum of the amplitudes of each frequency component to obtain the centroid frequency. The absolute value of the difference between the centroid frequency and the probe's nominal center frequency is used as the spectral centroid offset. The sum of the interface reflected echo amplitude ratio multiplied by a first preset weight and the spectral centroid offset multiplied by a second preset weight is used as the coupling state parameter. In this embodiment, the setting of the first and second preset weights needs to be determined by those skilled in the art based on the sensitivity, stability, and anti-interference ability of the interface reflected echo amplitude ratio and the spectral centroid offset in reflecting coupling quality. The amplitude ratio of the reflected echo at the interface directly characterizes the transmission efficiency of the sound beam at the interface of the inner wall of the tube. It is the most intuitive indicator for evaluating the continuity of the coupling layer. Its value shows a monotonically decreasing trend as the coupling state deteriorates, with a clear response law and obvious physical meaning. Therefore, it should be given a high weight. The spectral centroid offset characterizes the modulation effect of the coupling layer non-uniformity on the acoustic pulse spectrum. Its changing trend is also correlated with the coupling state. However, since the spectrum estimation itself is greatly affected by noise and signal truncation length, the measurement repeatability and signal-to-noise ratio of this indicator are relatively low. Therefore, it should be given a low weight, making it an auxiliary correction term in the evaluation. As a specific example, the embodiment uses 0.7 as the first preset weight and 0.3 as the second preset weight, with the sum of the two being 1, making the coupling state parameter a normalized comprehensive indicator. This weighting ratio ensures that the coupling quality evaluation is based on energy attenuation as the dominant criterion, while incorporating spectral feature changes as supplementary information. While maintaining the stability and reliability of the evaluation results, it improves the sensitivity of identifying early coupling degradation trends. In practice, those skilled in the art will prepare gradient samples with different coupling agent coating thicknesses, collect interface reflection echo signals, calculate the corresponding amplitude ratio and centroid offset, analyze the sensitivity and linearity of the two sets of data to changes in coupling state, and finally select a first preset weight in the range of 0.6 to 0.8, and correspondingly select a second preset weight in the range of 0.4 to 0.2, so that the sum of the two is kept at 1, to ensure the accuracy and robustness of the comprehensive evaluation.

[0074] The coupling state parameter is compared with a preset coupling quality threshold; if the coupling state parameter is less than the preset coupling quality threshold, it is determined that the upward oblique coupling condition has deteriorated, and a first adaptive adjustment or a second adaptive adjustment is performed.

[0075] If the coupling state parameter is greater than or equal to the preset coupling quality threshold, the upward oblique coupling condition is determined to be normal, and the 30° oblique scanning is continued with the current upward oblique beam direction and the current ultrasound probe.

[0076] Furthermore, in this embodiment, the first adaptive adjustment includes:

[0077] Keeping the axial and circumferential placement coordinates of the ultrasonic probe on the inner wall of the tube constant, the currently loaded transmission and reception delay values ​​are read from the phased array instrument delay calculation unit. Using the center point coordinates of the region where the outer wall of the tube meets the weld root as the sound beam deflection target point, and the center point coordinates of the weld root region on the tube sheet side as the new sound beam deflection target point, the sound ray propagation path length from the center of the array element to the new sound beam deflection target point is calculated element by element based on the wedge tilt angle, the longitudinal wave velocity of the wedge, the transverse wave velocity of the tube wall, and Snell's law of refraction. Divide the degree by the corresponding material sound velocity to obtain the one-way propagation time of each array element. Take the maximum value of the one-way propagation time of all array elements as the reference time. Subtract the one-way propagation time of each array element from the reference time to obtain the transmission delay value of each array element. Set the reception delay value of each array element to be equal to the corresponding transmission delay value to form a new set of transmission delay values ​​and reception delay values ​​and load them into the beamformer of the phased array instrument. This changes the deflection direction of the main lobe of the synthesized sound beam from the upward tilting direction pointing to the junction area between the outer wall of the tube and the root of the weld to the downward tilting direction pointing to the root area of ​​the weld on the tube sheet side.

[0078] Simultaneously, the measured contour data of the inner wall of the pipe, obtained by scanning along the axial direction of the pipe with a sampling interval of 0.1 mm by the displacement sensor, is read. This measured contour data is a three-dimensional array of axial position, circumferential angle, and radial coordinates. Each signal in the root-side sensitive signal spectrum is traversed, and the radial coordinate value R_signal, axial coordinate value Z_signal, and circumferential angle value θ_signal of the reflection point of that signal are extracted. Contour points whose axial coordinate is equal to Z_signal and whose circumferential angle is equal to θ_signal are retrieved from the measured contour data of the inner wall of the pipe, and their radial coordinate value R_profile is read. The difference between R_signal and R_profile is calculated. The absolute value D_diff; if D_diff is less than 0.2 mm, the signal is confirmed by the spatial comparison module as not belonging to the confirmed welding defect signal, and the signal is not determined as a corner reflection signal in the crack judgment module, then the signal type of the signal is marked as "inner wall depression geometric interference". The red channel value of the pixel area occupied by the signal in the root sensitive signal spectrum is set to 255, the green channel value is set to 0, and the blue channel value is set to 0 for color marking. A new record is added to the inspection report database. The record contains the axial coordinate value, circumferential angle value, radial coordinate value and the marking reason field of the signal. The marking reason field is written with the string "inner wall depression geometric interference".

[0079] In this embodiment, the second adaptive adjustment includes: creating a text prompt control and a graphical guide control in the human-computer interaction interface; setting the text displayed in the text prompt control to "Current coupling quality is insufficient, please replace with a micro probe and place the micro probe at the bottom of the expansion tube groove"; loading a schematic diagram of the axial cross-section of the inner wall of the tube head to be tested in the graphical guide control, drawing a red arrow pattern at the coordinate position of the lowest point at the bottom of the expansion tube groove in the schematic diagram, with the tip of the arrow pointing to the lowest point coordinate, drawing the text label "Probe Placement Center" next to the arrow pattern, and drawing a directional line segment near the arrow pattern, the starting point of the directional line segment being the coordinate of the lowest point at the bottom of the expansion tube groove, and the ending point being the coordinate of the junction area between the outer wall of the tube and the root of the weld; reading the preset upper limit value of the effective working area length of the micro probe wafer array (8 mm) and the upper limit value of the width (5 mm), and displaying the micro probe specifications "Chip length ≤ 8 mm, width ≤ 5 mm, bottom curvature radius matches the expansion tube groove" below the text prompt control;

[0080] After the detection system receives the confirmation command for replacement completion input by the operator through the human-machine interface, it re-executes the operation of acquiring coupling state parameters: controlling all array elements of the phased array probe to be excited in an equal-amplitude and in-phase manner, transmitting a calibration pulse with a peak-to-peak amplitude of 100 mV, receiving the interface reflected echo signal from the receiving array element, extracting the amplitude of the sampling point with the largest absolute amplitude in the echo signal, dividing the amplitude by 100 mV to obtain the interface reflected echo amplitude ratio, performing a fast Fourier transform on the echo signal to obtain the frequency domain amplitude distribution, calculating the absolute value of the difference between the arithmetic centroid frequency and the probe's nominal center frequency as the spectral centroid offset, and multiplying the interface reflected echo amplitude ratio by 0.7 and the spectral centroid offset by 0.3 to obtain the new coupling state parameter; comparing the new coupling state parameter with the preset coupling quality threshold, when the new coupling state parameter is greater than or equal to the preset coupling quality threshold, displaying the text "Coupling quality qualified" in the human-machine interface, and continuing to perform scanning according to the delay rule corresponding to the upward 30° oblique scanning, using the miniature probe as the ultrasonic transmitting and receiving device.

[0081] In this embodiment, the preset coupling quality threshold is determined as follows: Before the formal testing begins, a calibration tube head is selected. This calibration tube head has the same outer diameter, wall thickness, and material grade as the tube being tested. The inner wall surface of the calibration tube head is machined to remove any expansion bulges, ensuring that the arithmetic mean deviation of the inner wall surface roughness does not exceed 1.6 micrometers and the inner wall roundness deviation does not exceed 0.02 millimeters. On the inner wall surface of this calibration tube head, the coupling calibration operation is repeated twenty times according to the probe clamping pressure and coupling agent coating thickness specified in the normal testing procedure. Each coupling calibration operation performs the following steps: all elements of the phased array probe are excited in an equal-amplitude, in-phase manner, emitting a calibration pulse with a peak-to-peak amplitude of 100 millivolts. The probe receiving element receives the pulse reflected from the interface between the inner wall of the tube and the coupling agent. The interface reflection echo signal is used to extract the amplitude corresponding to the sampling point with the largest absolute amplitude as the interface reflection echo amplitude of the inner wall of the pipe. This amplitude is divided by 100 mV to obtain the interface reflection echo amplitude ratio. A fast Fourier transform is performed on the interface reflection echo signal to obtain the frequency domain amplitude distribution. The arithmetic centroid frequency of this frequency domain amplitude distribution is calculated. The absolute value of the difference between the arithmetic centroid frequency and the nominal center frequency of the probe is used as the spectral centroid offset. The product of the interface reflection echo amplitude ratio multiplied by 0.7 and the product of the spectral centroid offset multiplied by 0.3 are added to obtain the coupling state parameter value for this operation. The twenty coupling state parameter values ​​obtained from twenty coupling calibration operations are used to form a sample set. The arithmetic mean of this sample set is calculated, and the standard deviation of this sample set is calculated. The difference obtained by subtracting three times the standard deviation from the arithmetic mean is used as the preset coupling quality threshold.

[0082] In this embodiment, switching the beam direction of the 30° oblique scanning from upward oblique to downward oblique means that the axial and circumferential placement positions of the ultrasonic probe on the inner wall of the tube remain unchanged. The delay calculation unit of the phased array instrument recalculates the transmission delay value and reception delay value of each array element, so that the deflection direction of the main lobe of the synthesized sound beam changes from the original upward tilting direction from the probe position toward the junction area between the outer wall of the tube and the weld root to the downward tilting direction from the probe position toward the weld root area on the tube sheet side. The recalculated transmission delay value changes the incident angle of the transmitted sound beam at the interface between the wedge and the tube wall. The changed incident angle causes the sound beam refracted into the tube wall to avoid the tube expansion depression area of ​​the inner wall of the tube and enter the tube wall from the flat area of ​​the inner wall without geometric deformation. It propagates in the form of a transverse wave inside the tube wall material. After refraction at the inner interface of the tube wall, the main lobe of the sound beam covers the entire spatial range of the sensitive area at the weld root.

[0083] In this embodiment, the radial zero-point reference of the displacement sensor is determined as follows: Before performing mechanical scanning on the tube head to be tested, the contact of the displacement sensor is moved axially along the inner wall of the tube to a reference axial position at a preset distance from the end face of the tube head. This reference axial position is located outside the range of the tube expansion process and in the area where the inner wall of the tube has not undergone geometric deformation. At this reference axial position, the contact of the displacement sensor contacts the inner wall surface of the tube with a preset contact pressure, and the radial displacement reading of the displacement sensor at this position is recorded. This radial displacement reading is stored as the radial zero-point reference value in the parameter register of the detection system. Subsequently, continuous mechanical scanning along the axial direction is started. The original radial displacement value recorded by the displacement sensor at each axial sampling position is subtracted from the radial zero-point reference value in real time to obtain the radial offset relative to the undeformed surface of the inner wall of the tube. When the radial offset is positive, it indicates that the inner wall of the tube at the current position is concave towards the central axis of the tube, and when it is negative, it indicates that it is convex outward. All axial sampling positions and their corresponding radial offsets, as well as the circumferential angle values ​​corresponding to each axial sampling position, constitute the measured contour data of the inner wall of the tube.

[0084] In this embodiment, the preset distance needs to be determined by those skilled in the art based on the influence area of ​​the tube expansion process of the heat exchanger tube, the structural dimensions of the tube end, and the axial positioning stability of the displacement sensor contact. The core function of this preset distance is to select the radial zero-point reference point in the original tube wall area completely unaffected by the tube expansion process, avoiding the inner wall bulges or depressions caused by tube expansion near the tube end, thereby providing an accurate geometric benchmark for the radial offset calculation of the subsequent measured contour data. The value of the preset distance must simultaneously meet two conditions: first, it must be greater than the maximum axial deformation extension length generated by the tube expansion process on the inner wall of the tube end; second, it must ensure that the displacement sensor contact has sufficient axial travel margin at this position to complete stable contact and zero-point calibration. In specific implementation, for heat exchanger tubes with common outer diameters of 19 mm to 25 mm, the axial length of the tube expansion zone is usually no more than 8 mm. In this embodiment, 10 mm is selected as the preset distance. This value ensures that the reference axial position falls completely within the undeformed smooth tube section between the tube end face and the tube expansion zone, while maintaining an appropriate distance from the tube end face, facilitating the positioning operation of the displacement sensor contact. In practice, those skilled in the art will select the preset distance within the range of 8 mm to 15 mm by measuring the actual axial extension range of the expansion zone of the tube according to the actual tube expansion process parameters and tube head processing drawings. When the detection system is initialized, the roundness and surface roughness of the inner wall of the tube at the reference position will be verified to confirm that it is the original tube wall surface without deformation before zero-position calibration is performed.

[0085] The radial coordinates of the signal in the root sensitive signal spectrum increase from the probe coupling surface as the depth zero point towards the inside of the pipe wall. Subtracting the nominal radius of the pipe inner wall from the radial coordinate value of the signal yields a radial coordinate conversion value based on the undeformed surface of the pipe inner wall. The positive direction of this radial coordinate conversion value is consistent with the positive direction of the radial offset, thereby making the radial position of the signal in the root sensitive signal spectrum and the radial position of the contour point in the measured contour data of the pipe inner wall in the same spatial coordinate system, realizing direct spatial comparison between the two.

[0086] In this embodiment, the micro probe refers to an independent phased array probe, whose wafer array has an effective working length of no more than 8 mm in the circumferential direction and an effective working width of no more than 5 mm in the axial direction. The wafer array of the micro probe contains no less than 16 array elements, and the spacing between array elements is no more than 0.5 mm. The bottom of the outer shell of the micro probe is a concave curved surface, and the radius of curvature of the concave curved surface is the same as the radius of curvature of the bottom of the expansion tube groove, so that when the micro probe is placed at the bottom of the expansion tube groove, the concave curved surface at the bottom of the shell and the bottom surface of the groove achieve curved surface contact. The maximum width of the outer shell of the micro probe is less than the groove width of the expansion tube groove, so that the micro probe can be completely embedded in the expansion tube groove and the two sides of the shell do not interfere with the sidewalls of the groove.

[0087] The second adaptive adjustment is as follows: the human-machine interface of the detection system is switched to a guidance interface, which simultaneously displays a text prompt area and a graphic identification area; the text prompt area displays the specific text content in a preset font size as "The current coupling quality is insufficient. Please replace it with a miniature probe and place the miniature probe at the bottom of the expansion groove"; the graphic identification area displays a schematic diagram of the axial cross-section of the inner wall of the tube to be tested. In this schematic diagram, the outline of the inner wall of the tube, the outline of the expansion groove, and the weld area are all drawn with lines. The lowest point of the bottom of the expansion groove is marked with a red arrow in the schematic diagram, and the tip of the red arrow points to the lowest point of the bottom of the expansion groove. Next to the arrow, the text "Probe Placement Center" is marked. Near the arrow, the direction line segment marks the beam emission direction of the miniature probe, and the direction line segment points to the junction area between the outer wall of the tube and the root of the weld. After the operator completes the replacement and placement of the miniature probe according to the instructions on the guidance interface, the detection system automatically re-executes the operation of acquiring coupling status parameters and comparing the preset coupling quality threshold. After confirming that the current coupling status parameters are greater than or equal to the preset coupling quality threshold, the human-machine interface displays the text confirmation message "Coupling quality qualified". The system uses the miniature probe as the ultrasonic transmitting and receiving device and continues to complete the detection according to the delay rule and scanning parameters of the upward 30° oblique scanning.

[0088] Example 2:

[0089] Please see Figure 2 It should be further explained that, in this embodiment, the generation of the 0° reference defect map and the delineation of the crack-sensitive area include:

[0090] A101. Perform a three-dimensional Fourier transform or synthetic aperture focusing algorithm on the original detection data of the 0° direct beam scan to reconstruct the A-scan signal sequence acquired along the pipe wall axial direction into a 0° direct beam G-scan image and a D-scan image; wherein, the 0° direct beam G-scan image is a cross-sectional view of the weld perpendicular to the direction of the sound beam, and the D-scan image is a longitudinal cross-sectional view of the weld along the direction of sound beam propagation.

[0091] In this embodiment, the delay rule of the synthetic aperture focusing algorithm is determined based on the element spacing of the phased array probe used in the 0° direct scan, the probe's nominal center frequency, and the longitudinal wave velocity of the pipe wall material being tested. For a transceiver combination consisting of the p-th transmitting element and the q-th receiving element, the spatial straight-line distance from the transmitting element to the predetermined target focal point within the weld area is first calculated as the transmitting path, and the spatial straight-line distance from the target focal point to the receiving element is simultaneously calculated as the receiving path. The transmitting and receiving paths are added together to obtain the total propagation distance corresponding to the transceiver combination. Then, the total propagation distance is divided by the longitudinal wave velocity of the pipe wall material being tested to obtain the round-trip propagation time required for the ultrasonic wave to be emitted from the transmitting element, reflected by the target focal point, and received by the receiving element. This round-trip propagation time is the [missing information - likely a specific time value]. The delay value assigned to the transceiver combination; the above delay calculation is performed on all transceiver combinations of the phased array probe to form a delay rule table and load it into the beamformer of the phased array instrument, so that the synthesized sound beams are superimposed in phase at the target focal point; the horizontal axis of the 0° direct G-scan image corresponds to the circumferential angle position of the weld, and the unfolding position of the weld cross-section along the circumferential direction is marked in degrees or radians; the vertical axis corresponds to the depth direction with the probe coupling surface as the zero point, and the vertical distance extending from the probe coupling surface into the pipe wall is marked in millimeters; the horizontal axis of the D-scan image corresponds to the scanning position along the pipe wall axis, and the distance the probe moves along the pipe axis direction is marked in millimeters; the vertical axis corresponds to the depth direction with the probe coupling surface as the zero point, and the vertical distance extending from the probe coupling surface into the pipe wall is marked in millimeters.

[0092] A102. Perform three-dimensional segmentation and slicing on the two-dimensional image sequence composed of the 0° direct G-scan image and the D-scan image. Specifically, with the probe coupling surface as the origin of the spatial coordinate system, construct a three-dimensional voxel grid along the radial, axial, and circumferential directions of the weld with a preset voxel size; map the echo amplitude of each A-scan signal sampling point to the corresponding voxel according to the sound path and beam spread angle to form a three-dimensional amplitude matrix; traverse the three-dimensional amplitude matrix using a preset amplitude threshold, mark voxels with amplitudes exceeding the preset amplitude threshold as defect candidate voxels, and apply a three-dimensional connected component labeling algorithm to cluster spatially adjacent defect candidate voxels into independent defect objects. Specifically:

[0093] A1021. Traverse each voxel in the three-dimensional amplitude matrix and obtain the voxel amplitude of the voxel. When the voxel amplitude of the voxel is greater than the preset amplitude threshold, mark the voxel as a defect candidate voxel; when the voxel amplitude of the voxel is not greater than the preset amplitude threshold, do not mark the voxel.

[0094] A1022. Create a defect object set to store independent defect objects, and set a merge flag initially in an unmerged state for each defect candidate voxel in the three-dimensional amplitude matrix;

[0095] A1023. According to the row index order, column index order, and layer index order of the voxels in the three-dimensional amplitude matrix, perform the following merging operation on each defect candidate voxel in sequence: Check whether there are defect candidate voxels with a merged state among the 26 neighboring voxels centered on the current defect candidate voxel. If so, select the independent defect object to which the defect candidate voxel with a merged state belongs from the 26 neighboring voxels, add the current defect candidate voxel to the independent defect object, and update the merging mark of the current defect candidate voxel to the merged state; if not, create a new independent defect object in the defect object set, add the current defect candidate voxel to the newly created independent defect object, and update the merging mark of the current defect candidate voxel to the merged state.

[0096] In this embodiment, for a defect candidate voxel located at the boundary of the three-dimensional amplitude matrix, the neighboring voxel inspection range is dynamically adjusted according to the actual position of the defect candidate voxel in the row index, column index, and layer index of the three-dimensional amplitude matrix: taking the current defect candidate voxel's row index, column index, and layer index as the center, one index unit is extended upwards and downwards respectively to form the row index range, column index range, and layer index range. The intersection of the above three index ranges with the effective range of the row index, column index, and layer index of the three-dimensional amplitude matrix is ​​obtained to obtain the actual effective subset of the row index, column index, and layer index to be inspected. Each neighboring voxel position in the three-dimensional index space composed of these three effective subsets is traversed, and it is checked whether there is a defect candidate voxel with a merged state at that position. If there is, a merge operation is performed. If a neighboring index position is excluded from the intersection operation because it exceeds the effective index range of the matrix, then that position does not participate in the traversal and merge judgment.

[0097] A1024. When all candidate defect voxels are in a merged state, each independent defect object stored in the defect object set is a clustered defect object group, and each independent defect object is composed of all candidate defect voxels that are spatially adjacent to each other.

[0098] As a concrete example, taking the inspection of fillet welds in a carbon steel pipe with an outer diameter of 25 mm and a wall thickness of 2.5 mm as an example, a phased array probe with a center frequency of 7.5 MHz is used for 0° direct scanning. The longitudinal wave velocity of the pipe wall is 5900 m / s, and the axial scanning step distance is 0.2 mm. When constructing the three-dimensional voxel mesh, the axial voxel size is set to 0.1 mm, the circumferential voxel size corresponds to an arc length of 0.15 mm, and the radial voxel size is set to 0.098 mm. A reference A-scan signal segment containing 500 sampling points is selected within the defect-free area of ​​the pipe base material. The root mean square value of the amplitude of all sampling points within this segment is calculated as the noise level of the pipe base material, which is 0.6% of the full-screen scale. Simultaneously, a reference voxel set containing 1000 voxels is defined within this defect-free area. The arithmetic mean of the voxel amplitudes in this set is calculated as 2% of the full-screen scale, and the standard deviation is 0.3% of the full-screen scale. The preset amplitude threshold is set to 2% plus 6 times 0.3%, which is 3.8% of the full-screen scale. Each voxel in the three-dimensional amplitude matrix is ​​traversed. For a voxel with an amplitude of 12% of the full-screen scale, since its amplitude is greater than 3.8%, this voxel is marked as a defect candidate voxel and assigned a merged label with an initial unmerged state. For a voxel with an amplitude of 1.5% of the full-screen scale, since its amplitude is not greater than 3.8%, this voxel is not marked.

[0099] After creating the defect object set, all candidate voxels in the 3D amplitude matrix are traversed sequentially according to row indices from 0 to 249, column indices from 0 to 166, and layer indices from 0 to 25. When processing a candidate voxel with row index 120, column index 80, and layer index 12, its 26 neighboring voxels are checked. The row index range of these 26 neighboring voxels is 119 to 121, the column index range is 79 to 81, and the layer index range is 11 to 13. Within this neighborhood, a neighboring voxel with row index 119, column index 80, and layer index 12 is found to have been marked as a candidate voxel and its merge mark is in a merged state. The defect identifier of the independent defect object to which this neighboring voxel belongs is 1. Therefore, the current candidate voxel is added to the independent defect object 1, and the merge mark of the current candidate voxel is updated to a merged state. When processing another defect candidate voxel with row index 30, column index 140, and layer index 8, its 26 neighboring voxels are checked. If no defect candidate voxels with a merged state are found within this neighborhood, a new independent defect object with defect identifier 2 is created in the defect object set. The defect candidate voxel is added to this independent defect object 2, and its merge mark is updated to a merged state. The process continues to process subsequent defect candidate voxels in index order until the merge marks of all defect candidate voxels are updated to a merged state. Finally, the defect object set stores two independent defect objects: Independent defect object 1 contains 85 voxels, corresponding to a porosity defect with a volume of 0.8 cubic millimeters; Independent defect object 2 contains 122 voxels, corresponding to a slag inclusion defect with a volume of 1.2 cubic millimeters. Each independent defect object consists of all defect candidate voxels spatially connected through face adjacency, edge adjacency, or corner adjacency. This process addresses the clustering accuracy problem caused by irregular defect morphology, which leads to the same defect being incorrectly segmented into multiple independent objects or adjacent defects being incorrectly merged into a single object.

[0100] In this embodiment, the preset voxel size is determined as follows: the axial voxel size Δz is set to 0.5 times the axial scan step distance Δs; the circumferential voxel size Δθ is set to 0.5 times the arc length of the adjacent scan line at the weld root depth; and the radial voxel size Δr is set to 0.25 times half the wavelength λ / 2 of the ultrasonic emission pulse, where wavelength λ = c_L / f_c, where c_L represents the longitudinal wave velocity of the tested pipe wall material, and f_c represents the nominal center frequency of the phased array probe used for the 0° direct scan. The ratio λ is the longitudinal wave wavelength of the ultrasonic wave propagating in the pipe wall material. The radial voxel size Δr is taken as one-quarter of half the wavelength, thereby directly relating the radial sampling density to the inherent spatial resolution of the ultrasonic pulse, ensuring that the radial accuracy of the three-dimensional voxel grid matches the physical characteristics of the detected sound wave.

[0101] In this embodiment, the echo amplitude of each A-scan signal sampling point is mapped to the corresponding voxel using a voxel intersection length weighted allocation method. Specifically, for the sound beam path corresponding to a certain sampling point, the intersection length L_a of the sound beam path passing through each voxel is calculated, and the echo amplitude A of the sampling point is allocated according to the intersection length ratio, that is, the amplitude allocated to the a-th voxel is A×(L_a / ΣL_a).

[0102] The preset amplitude threshold is calculated as follows: In the three-dimensional amplitude matrix, a defect-free reference voxel set containing only the pipe base material region is manually defined. The arithmetic mean μ_noise and standard deviation σ_noise of the amplitude of all voxels in the reference voxel set are calculated. The preset amplitude threshold T is set as T = μ_noise + 6σ_noise.

[0103] In this embodiment, the calculation method for the arc length of adjacent scan lines at the weld root depth is as follows: The total number of scan lines actually excited and acquired by the phased array probe used for the 0° direct scan in the circumferential direction is obtained. This total number of scan lines is determined by the number of circumferential array elements of the probe and the electronic scanning mode. When using the full-element sequential transmit / receive mode, the total number of scan lines equals the number of circumferential array elements; when using the sparse excitation mode, the total number of scan lines equals the number of circumferential array elements divided by the excitation element step size. The calculation method for the arc length of adjacent scan lines at the weld root depth is as follows: The total number of scan lines is obtained by the phased array probe used for the 0° direct scan in the circumferential direction. The radial coordinates of each point at the weld root boundary line on the actual interface contour between the outer wall of the pipe and the molten metal of the weld are extracted in step A104. The arithmetic mean of these radial coordinates is taken as the radial coordinate of the pipe wall corresponding to the weld root position. The circumference of the weld root depth is calculated with the radial coordinate of the pipe wall as the radius, which is twice pi multiplied by the radial coordinate of the pipe wall. The circumference is divided by the total number of scan lines to obtain the arc length value of the adjacent scan lines at the weld root depth, and the circumferential voxel size is set to 1 / 2 of the arc length value.

[0104] In this embodiment, the geometric definition of the sound beam path is as follows: A three-dimensional conical sound beam region is constructed using the geometric center of a single transmitting element of the phased array probe used for the 0° direct scan as the vertex of a cone, the half-diffusion angle of the main lobe of the ultrasonic pulse emitted by that element in the tube wall material as the half-angle of the cone, and the direction of the normal to the working surface of that element as the direction of the central axis of the cone. The half-diffusion angle is determined by the width of the element, the nominal center frequency of the probe, and the longitudinal wave velocity in the tube wall material. Specifically, the half-diffusion angle is equal to the inverse of the quotient of the longitudinal wave velocity divided by the product of the nominal center frequency of the probe and the width of the element. The chord value; the axial extension range of the three-dimensional conical sound beam region extends downward from the probe coupling surface to the depth of the inner wall of the tube; each voxel in the three-dimensional voxel grid that spatially overlaps with the conical sound beam region is determined as the voxel through which the sound beam path corresponding to the sampling point passes, the spatial intersection volume between the conical sound beam region and each overlapping voxel is calculated, and the spatial intersection volume is used as the intersection length weight of each corresponding voxel, thereby realizing the expansion of the ideal geometric line segment path of a single voxel into a spatial volume path that reflects the actual sound beam diffusion characteristics, and improving the spatial allocation accuracy when the echo amplitude is mapped to voxels.

[0105] A103. For each independent defect object, extract its centroid coordinates as the second spatial position coordinates of the corresponding independent defect object, and record the maximum amplitude, volume and morphological characteristics of the corresponding independent defect object as attribute parameters; integrate the second spatial position coordinates and attribute parameters of all the marked independent defect objects to form the 0° reference defect map; wherein, the second spatial position coordinates are normalized and stored with the weld root and tube sheet interface as the radial reference and the tube end face as the axial reference.

[0106] In this embodiment, the centroid coordinates of the independent defect object are calculated as follows: the arithmetic mean of the three-dimensional coordinates of the N candidate voxels of the independent defect object is calculated, i.e., r_c=(Σr_k) / N, θ_c=(Σθ_k) / N, z_c=(Σz_k) / N, where r_k, θ_k, and z_k are the radial coordinate, circumferential angular coordinate, and axial coordinate of the k-th candidate voxel, respectively, k is the index of the candidate voxel, and k=1, 2, ..., N; r_c represents the coordinate value of the centroid of the independent defect object in the radial direction of the pipe, i.e., the depth position extending from the probe coupling surface into the pipe wall; θ_c represents the angular coordinate value of the centroid of the independent defect object in the circumferential direction of the pipe, i.e., the angular position along the circumferential unfolding direction of the weld; z_c represents the coordinate value of the centroid of the independent defect object in the axial direction of the pipe, i.e., the axial position along the pipe wall axis. Location; the volume is the total number N of candidate voxels contained in the independent defect object multiplied by the volume of a single voxel Δr×Δz×(r_c×Δθ); the morphological features include the maximum size Δr_max in the radial direction, the maximum size Δθ_max in the circumferential direction, the maximum size Δz_max in the axial direction of the independent defect object, as well as the aspect ratio Δr_max / Δz_max and the sphericity parameter; in the normalized storage, the radial reference is the radial coordinate R_root of the weld root and the junction line of the tube sheet, and the normalized radial coordinate is r_norm=r_c-R_root; the axial reference is the axial coordinate Z_end of the tube end face, and the normalized axial coordinate is z_norm=z_c-Z_end; the circumferential reference is the zero-degree scale position θ_zero, and the normalized circumferential coordinate is θ_norm=θ_c-θ_zero.

[0107] A104. Obtain the echo signal of the original 0° direct scan detection data within the full thickness range of the weld. Based on the acoustic impedance difference between the weld metal and the pipe wall base material, extract the actual boundary contour between the outer wall of the pipe and the molten metal of the weld through the interface echo recognition algorithm. Offset the actual boundary contour towards the inside of the weld by a preset molten zone thickness value to obtain the boundary of the weld molten zone. In the 0° reference defect map, determine the spatial location area within the boundary of the weld molten zone that is not occupied by any of the independent defect objects as the location area where there is no welding defect signal. And define the spatial connected set of all location areas where there is no welding defect signal that extends continuously along the outer wall side of the pipe to the weld root boundary line as the crack sensitive area.

[0108] In this embodiment, the execution steps of the interface echo recognition algorithm are as follows: Within the full thickness range of the weld, the A-scan signal is searched radially from the inner wall of the pipe to the outer wall, point by point. The position where the echo amplitude first exceeds the noise level of the pipe base material by 12dB, and the amplitudes of the following three consecutive sampling points show a strictly monotonically decreasing trend, is marked as the interface reflection point between the outer wall of the pipe and the molten metal of the weld. The noise level of the pipe base material is determined by selecting a reference A-scan signal segment without any defect signals in the pipe base material region, calculating the root mean square value of the amplitude of the sampling points within this reference A-scan signal segment, and using this as the noise level of the pipe base material. After performing the search and marking operation on all scan lines, all interface reflection points are connected sequentially along the circumference and aligned sequentially along the axial direction to form the actual boundary contour composed of a three-dimensional point set. The preset melting zone thickness value t_fusion is determined as follows: The measured values ​​t_1, t_2, and t_3 of the fusion zone thickness in the weld root region of three metallographic specimens were obtained from the welding procedure qualification report. The arithmetic mean was taken as the preset fusion zone thickness value t_fusion = (t_1 + t_2 + t_3) / 3. The actual boundary contour was shifted into the weld by translating it into the weld by t_fusion along the normal direction of each point on the actual boundary contour at that point. The line connecting the points after translation constitutes the boundary of the weld fusion zone. Being within the boundary of the weld fusion zone means that the spatial position simultaneously satisfies two conditions: First, the radial coordinate r satisfies R_outer ≤ r ≤ R_fusion, where R_outer is the radial coordinate of the outer wall of the pipe, and R_fusion is the radial coordinate of the boundary of the weld fusion zone; Second, the circumferential coordinate θ and the axial coordinate z fall within the projection range of the closed curve of the actual boundary contour at that depth.

[0109] In this embodiment, the phrase "continuously extending along the outer wall of the pipe to the weld root boundary" refers to performing the following steps:

[0110] Traverse all voxels in the 0° reference defect diagram that satisfy the condition "within the boundary of the welding melting zone and not occupied by any independent defect object", and form an initial seed point set with all traversed voxels. Each voxel in the initial seed point set simultaneously satisfies: the radial coordinate r satisfies R_outer≤r≤R_fusion, and the spatial distance between the spatial position coordinate of the voxel and the second spatial position coordinate of each independent defect object already marked in the 0° reference defect diagram is greater than a preset repulsion radius. The preset repulsion radius is taken as the cube root of the total number of defect candidate voxels of the independent defect object.

[0111] Sequentially take out an unprocessed seed point from the initial seed point set, take the seed point as the center, and check each neighboring voxel within its 26-neighborhood range; for any neighboring voxel, if the neighboring voxel satisfies all the following conditions: first, its radial coordinate r satisfies R_root < r ≤ R_fusion, where R_root is the radial coordinate of the boundary between the weld root and the tube sheet; second, the neighboring voxel is not occupied by any independent defect object in the 0° reference defect image; third, the neighboring voxel has not been marked as belonging to any grown connected region, then mark the neighboring voxel as a member of the current connected region, and add the neighboring voxel to the initial seed point set as a new seed point; repeat the process of taking out seed points and checking neighboring voxels until there are no unprocessed seed points in the initial seed point set.

[0112] In the neighborhood checking process, when the radial coordinate r of a certain neighboring voxel satisfies r ≤ R_root, the neighboring voxel is not included in the current connected region, and the growth along the direction of this neighboring voxel is stopped, while the growth in the directions of other neighboring voxels continues.

[0113] The complete set of all voxels contained in all connected regions obtained after completing the region growing step is delimited as the crack sensitive area; wherein, the crack sensitive area is connected to the boundary of the welding fusion zone on the outer wall side of the pipe, and extends to the boundary line of the weld root on the root side to terminate.

[0114] This embodiment reconstructs the 0° direct-scan A-scan signal into a 0° direct-scan G-scan and D-scan image using a synthetic aperture focusing algorithm, solving the problem that the original A-scan signal cannot intuitively reflect the spatial morphology of defects. It constructs a three-dimensional voxel grid by adaptively determining voxel sizes based on the scanning step and ultrasonic pulse wavelength, and uses a voxel intersection length weighted allocation method to accurately map echo amplitudes to corresponding voxels, solving the problem of insufficient spatial resolution caused by beam diffusion. It calculates an adaptive amplitude threshold based on noise statistical characteristics and uses a three-dimensional connected component labeling algorithm to cluster candidate defect voxels into independent defect objects, solving the problem of missed or over-segmented defects caused by fixed thresholds in low signal-to-noise ratio regions. Finally, it extracts the outer surface of the tube using an interface echo recognition algorithm based on acoustic impedance differences, with a joint criterion of echo amplitude jumps and multi-sampling point monotonicity. The actual boundary contour between the wall and the weld molten metal was obtained, and the boundary of the weld molten zone was translated into the weld interior using the statistical value of the measured molten zone thickness of the metallographic sample as the offset. This solved the problem of complex weld geometry and difficulty in accurately locating the boundary of the molten zone. By traversing the voxels in the 0° reference defect map that are within the boundary of the weld molten zone and not occupied by any independent defect objects, an initial seed point set was constructed. After excluding neighboring voxels of the defect based on the repulsion radius, region growth was performed, which solved the problem of the area around the weld defect being mistakenly included in the crack sensitive area. The finally defined crack sensitive area is connected to the boundary of the weld molten zone on the outer wall side of the pipe and terminates at the weld root boundary line on the root side. This achieved accurate and complete coverage of the potential crack initiation area, providing an accurate spatial reference for the comparison and judgment of the subsequent 30° oblique scanning signal.

[0115] Example 3:

[0116] Please see Figure 3 It should be further explained that the estimated crack depth obtained in this embodiment includes:

[0117] B101. In the original detection data of the 30° oblique scanning, locate the A-scan signal corresponding to the corner reflection signal; based on the spatial coordinates of the reflection point of the corner reflection signal, combined with the incident angle of the sound beam of the 30° oblique scanning, the longitudinal wave velocity of the wedge material and the transverse wave velocity of the pipe wall material under test, calculate the propagation time of the wedge segment and the pipe wall segment along the sound beam path and sum them to obtain the round-trip propagation time of the ultrasonic wave from the probe chip through the reflection point back to the probe chip, and determine the round-trip propagation time as the target time on the A-scan signal.

[0118] In this embodiment, the propagation time of the wedge segment is calculated as follows: The tilt angle α_wedge and center thickness h_wedge of the wedge used in the 30° oblique scanning are obtained. Taking the spatial coordinates of the reflection point as the endpoint and the center of the probe wafer as the starting point, the actual acoustic ray path from the probe wafer to the wedge-tube wall interface is traced within the geometric model of the wedge based on the beam incidence angle and Snell's law of refraction. The propagation distance L_wedge of this path in the wedge material is calculated. L_wedge is then divided by the longitudinal wave velocity c_L_wedge of the wedge material to obtain the one-way travel time of the wedge segment. The propagation time t_wedge = L_wedge / c_L_wedge; the propagation time of the pipe wall segment is calculated as follows: the difference between the radial coordinate of the reflection point and the radial coordinate of the outer wall of the pipe is taken as the propagation distance L_pipe of the pipe wall segment, and L_pipe is divided by the transverse wave velocity c_S_pipe of the pipe wall material being measured to obtain the one-way propagation time t_pipe = L_pipe / c_S_pipe; the round-trip propagation time t_total = 2 × (t_wedge + t_pipe), and t_total is determined as the target time.

[0119] In this embodiment, the acquisition and updating method of the wedge's tilt angle and center thickness is as follows: During the initialization phase of the phased array detection system, the system automatically acquires the tilt angle, center thickness, and longitudinal wave velocity of the currently installed wedge by reading the wedge parameter data block pre-programmed into the electronic storage chip embedded in the wedge. The electronic storage chip is a non-volatile memory soldered inside the wedge shell. Its internal parameter data block is written by the manufacturer during the wedge's factory calibration and includes the actual tilt angle, actual center thickness, and actual longitudinal wave velocity obtained by precise measurement. In this embodiment, all three actual values ​​are obtained by the manufacturer using a coordinate measuring machine. The system is calibrated piece by piece using the ultrasonic pulse reflection method. When the operator replaces the wedge, the phased array detection system, after power-on or triggering the wedge identification refresh command, rereads the parameter data block in the electronic storage chip of the newly installed wedge, extracts the actual tilt angle value, actual center thickness value, and actual longitudinal wave velocity value, and updates the corresponding wedge tilt angle variable, wedge center thickness variable, and wedge longitudinal wave velocity variable in the ray tracing calculation module. This ensures that each ray path calculation uses the real geometric parameters of the currently installed wedge, eliminating the impact of geometric parameter deviations caused by wedge wear, replacement, or manufacturing tolerances on the accuracy of propagation time calculation.

[0120] B102. Taking the target time as the center, the time interval defined by extending forward by a first preset search half-width and backward by a second preset search half-width from the target time is taken as the target search time interval; the first preset search half-width and the second preset search half-width are both taken as the half-width of the ultrasonic pulse emitted by the phased array probe used in the 30° oblique scanning in the pipe wall material.

[0121] In this embodiment, the half-width at half-maximum (WHM) of the ultrasonic pulse in the pipe wall material is obtained as follows: On the precision wire-cut groove test block, an artificially simulated crack groove of known depth is selected, and a 30° oblique scanning setting, the same as that used for crack detection, is used for scanning. The signal segment containing the reflected echo from the artificially simulated crack groove is extracted from the obtained A-scan signal. Within this signal segment, the maximum amplitude of the echo pulse is determined, and the time width corresponding to the amplitude dropping from the rising or falling edge of the maximum amplitude to half the maximum amplitude is measured. This time width is taken as the full width at half-maximum (WHM) of the ultrasonic pulse. Half of the full width at half-maximum is taken as the WHM of the ultrasonic pulse in the pipe wall material, T_pulse_half. Both the first preset search half-width and the second preset search half-width are taken as T_pulse_half. The target search time interval is [t_target-T_pulse_half, t_target+T_pulse_half]. Wherein, t_target represents the theoretical round-trip propagation time of the ultrasonic wave emitted from the probe wafer, returning to the probe wafer via the corner reflection point, i.e., the target time corresponding to the corner reflection echo of the root penetrating crack on the A-scan signal. This time value is obtained by calculating the propagation time of the wedge segment and the pipe wall segment along the actual sound ray path based on the three-dimensional spatial coordinates of the corner reflection signal, combined with the incident angle of the sound beam during the 30° oblique scan, the longitudinal wave velocity of the wedge material, and the transverse wave velocity of the tested pipe wall material, and summing them up. This is used to accurately locate the center of the search interval where the crack echo is located on the time axis of the A-scan signal.

[0122] B103. Within the target search time interval, the signal amplitude of the A scan signal is obtained point by point, and the signal amplitude corresponding to the sampling point with the largest absolute amplitude among all the obtained sampling points is extracted as the echo amplitude of the corner reflection signal.

[0123] In this embodiment, the signal amplitude of the A-scan signal is acquired point by point sampling, and according to the time sampling interval Δt_sample of the original detection data of the 30° oblique scanning, starting from the starting moment of the target search time interval, each sampling moment is determined with Δt_sample as the step size, and the signal amplitude of the A-scan signal corresponding to each sampling moment is read sequentially until the next sampling moment to be read exceeds the ending moment of the target search time interval; the absolute value of all read amplitudes is calculated one by one, the maximum absolute amplitude is searched from all calculated absolute values, and the maximum absolute amplitude is taken as the echo amplitude A_crack of the corner reflection signal.

[0124] B104: inputting the echo amplitude into a pre-established amplitude-crack depth mapping relationship look-up table, and obtaining the corresponding estimated crack depth by matching the echo amplitude interval; if the echo amplitude is greater than the maximum echo amplitude in the amplitude-crack depth mapping relationship look-up table, the estimated crack depth is taken as the corresponding maximum crack depth value in the amplitude-crack depth mapping relationship look-up table, and an over-range marking signal is output; if the echo amplitude is less than the minimum echo amplitude in the amplitude-crack depth mapping relationship look-up table, the estimated crack depth is set as 0.

[0125] In this embodiment, the amplitude-crack depth mapping relationship look-up table is composed of a set of node pairs {A_i, d_i} arranged in ascending order of echo amplitude, i=1, 2, ..., M, where A_i is the echo amplitude corresponding to the i-th node, d_i is the crack depth value corresponding to the i-th node, and A_1<A_2<...<A_M; the step of obtaining the corresponding estimated crack depth by matching the echo amplitude interval is specifically: comparing the echo amplitude A_crack with each node A_i sequentially, if A_crack satisfies A_j≤A_crack≤A_{j+1}, performing linear interpolation according to the crack depth values d_j and d_{j+1} corresponding to nodes A_j and A_{j+1} to obtain the estimated crack depth d_pred=d_j+(d − {j+1}-d_j)×(A_crack−A_j) / (A_{j+1}-A_j); if A_crack>A_M, d_pred=d_M is taken, and a Boolean over-range marking signal is set to true; if A_crack<A_1, d_pred=0 is taken.

[0126] The amplitude-crack depth mapping lookup table is pre-established as follows: A precision wire-cut groove test block with the same material and wall thickness as the fillet weld of the pipe to be tested is prepared. At the junction of the outer wall of the pipe and the root of the weld on the test block, multiple artificial simulated crack grooves of different depths are machined using wire cutting technology. The depth range of the multiple artificial simulated crack grooves of different depths covers from 5% to 80% of the pipe wall thickness, and the depth difference between adjacent artificial simulated crack grooves is no greater than 10% of the pipe wall thickness. For each artificial simulated crack groove with a known depth... Artificial crack grooves are simulated using the same phased array probe, wedge, scanning arrangement, and scanning parameters as those used in crack detection. A 30° oblique scanning is performed, and the echo amplitude values ​​of the corresponding artificial crack grooves are extracted from the obtained data, following the same procedure as extracting echo amplitude values. The known depths of each artificial crack groove are paired with the corresponding extracted echo amplitude values ​​to form calibration data pairs. Cubic spline interpolation is performed on all calibration data pairs to generate an amplitude-crack depth mapping lookup table with echo amplitude values ​​as the index and crack depth as the output value.

[0127] In this embodiment, the cubic spline interpolation process employs natural boundary conditions. Specifically, in the two-dimensional data point sequence comprised of all calibration data pairs, the second derivative of the mapping curve at the first data point is set to zero, and the second derivative of the mapping curve at the last data point is also set to zero. This ensures the interpolation curve exhibits a linear extension trend at both boundaries, preventing non-physical oscillations at the boundaries due to lack of constraint. After generating continuous mapping curves, the minimum echo amplitude is extracted from all calibration data pairs as the lower limit of the echo amplitude range in the lookup table. This minimum value corresponds to the echo amplitude of the data point with the smallest crack depth in the calibration data pair. The maximum value of the echo amplitude is extracted and used as the upper limit of the echo amplitude range in the lookup table. This maximum value corresponds to the echo amplitude value of the data point with the largest crack depth in the calibration data pair. Below the lower limit of the echo amplitude range and above the upper limit of the echo amplitude range in the lookup table, the extrapolation calculation of the cubic spline interpolation curve is not performed. When the echo amplitude value to be queried is less than the lower limit of the echo amplitude range, the crack depth estimate is returned directly as zero. When the echo amplitude value to be queried is greater than the upper limit of the echo amplitude range, the maximum crack depth value is returned directly and an over-range marker signal is output at the same time. This ensures that the crack depth estimate always comes from a reliable mapping relationship within the coverage area of ​​the calibration data.

[0128] In this embodiment, the precision wire-cut groove test block is made of carbon steel, the same material as the pipe being tested, with a wall thickness of 2.5 mm. Using wire-cut electrical discharge machining (EDM), ten artificial simulated crack grooves with depths of 0.125 mm, 0.25 mm, 0.375 mm, 0.5 mm, 0.75 mm, 1.0 mm, 1.25 mm, 1.5 mm, 1.75 mm, and 2.0 mm are machined at the junction of the outer wall of the pipe and the weld root. The depth difference between adjacent grooves is no greater than 0.25 mm, satisfying the condition of not exceeding 10% of the pipe wall thickness (0.25 mm), and the depth range covers from 5% of the pipe wall thickness (0.125 mm) to 80% of the pipe wall thickness (2.0 mm). For the m-th artificial simulated crack groove, its... The known depth is denoted as d_m. A phased array probe with a center frequency of 7.5MHz and a wedge with an inclination angle of 30° are used. A 30° oblique scan is performed with the same probe arrangement and scanning step parameters as the actual pipe head inspection. The echo amplitude value A_m is obtained from the collected data according to the operation of extracting the echo amplitude value, where m=1, 2, ..., 10. After the ten calibration data pairs (d_m, A_m) are processed by cubic spline interpolation, a continuous mapping curve is generated with the echo amplitude value as the horizontal axis and the crack depth as the vertical axis. The continuous mapping curve is discretized with the echo amplitude value step size being 1% of the full screen scale, generating a set of node pairs {A_i, d_i}, which constitutes the amplitude-crack depth mapping relationship lookup table.

[0129] As a concrete implementation example, the fillet weld between a carbon steel tube with an outer diameter of 25 mm and a wall thickness of 2.5 mm and the tube sheet in a heat exchanger of a petrochemical plant is taken as the inspection object. This tube head is listed as a key suspected object during periodic inspections, and it is necessary to confirm whether there is a root penetrating crack. During the inspection, the scanning data acquisition module first performs a 0° direct beam scan and a 30° oblique beam scan on the fillet weld of the tube head. The 0° direct beam scan uses a phased array probe with a center frequency of 7.5 MHz. The sound beam is emitted axially along the tube wall perpendicular to the probe working surface, covering the entire thickness range of the weld from the inner wall to the outer wall of the tube. The 30° oblique beam scan uses the same type of phased array probe with a 30° tilt wedge. The sound beam is directed at the root area of ​​the weld at a 30° tilt angle, and the focal point covers the sensitive area from the outer wall of the tube to the root of the weld. The two sets of raw inspection data are stored separately.

[0130] The baseline defect map generation module executes a synthetic aperture focusing algorithm on the raw detection data from a 0° direct scan. Based on a probe element spacing of 0.5 mm, a center frequency of 7.5 MHz, and a longitudinal wave velocity of 5900 m / s, it calculates the delay rule to reconstruct the A-scan signal sequence into a 0° direct G-scan and D-scan map. Using the probe coupling surface as the origin of the spatial coordinate system, the axial voxel size is set to 0.1 mm, the corresponding arc length of the circumferential voxel size is set to 0.15 mm, and the radial voxel size is set to 0.098 mm to construct a three-dimensional voxel mesh. A voxel intersection length weighted allocation method is used. The echo amplitude of each sampling point was mapped to the corresponding voxel. In the defect-free area of ​​the pipe base material, the noise mean was measured to be 2% of the full-screen scale, with a standard deviation of 0.3%. The preset amplitude threshold was set to 3.8% of the full-screen scale. The three-dimensional amplitude matrix was traversed, and voxels with amplitudes exceeding 3.8% were marked as candidate defect voxels. Using the 26-neighbor connectivity criterion, voxels were merged sequentially according to row index, column index, and layer index. A porosity defect with a volume of approximately 0.8 cubic millimeters was identified in the upper part of the weld, and a slag inclusion defect with a volume of approximately 1.2 cubic millimeters was identified near the root of the weld. After recording the centroid spatial coordinates of the two defects, a 0° reference defect map was generated. A-scan signals were searched radially from the inner wall to the outer wall of the pipe throughout the full thickness of the weld. The criterion was that the echo amplitude first exceeded the noise level of the pipe base material by 12 dB, and the amplitude of the next three consecutive sampling points strictly decreased monotonically. Interface reflection points between the outer wall of the pipe and the molten metal of the weld were extracted. All interface reflection points were connected circumferentially and aligned axially to form the actual interface profile. The measured thicknesses of the molten zone of three metallographic samples (1.2 mm, 1.5 mm, 1.6 ... The arithmetic mean of 1.2 mm, calculated from 3 mm and 1.1 mm, is taken as the preset melting zone thickness. Each point on the actual interface contour is translated 1.2 mm inward along the normal direction to obtain the boundary of the welding melting zone. In the 0° reference defect map, voxels within the boundary of the welding melting zone and not occupied by any independent defect objects are constructed as the initial seed point set. After excluding defect neighborhood voxels with a preset rejection radius, 26 neighborhood region growth is performed to delineate the crack sensitive area located on the outer wall of the pipe, spanning approximately 30° circumferentially, and extending radially from the boundary of the welding melting zone to the weld root interface line.

[0131] In this embodiment, the preset rejection radius needs to be determined by those skilled in the art based on the dimensional characteristics of the identified welding defects, the resolution of the three-dimensional voxel mesh, and the safety margin of the crack-sensitive zone. The core function of this rejection radius is to exclude voxels within a certain range around the confirmed welding defects from the initial seed point set before the region growing algorithm is executed. This prevents the crack-sensitive zone from intruding into the defect itself and its adjacent areas due to excessive growth, ensuring that the defined sensitive zone only covers potential crack initiation locations without welding defects and satisfying geometric continuity conditions. The rejection radius should be slightly larger than the equivalent radius of a single independent defect object, allowing it to completely cover the voxel set of the defect and its boundary transition zone. However, it should not be too large to avoid excessive erosion of the defect-free space available for region growing, leading to discontinuities or omissions in the crack-sensitive zone. As a specific example, in this embodiment, the preset rejection radius is the cube root of the total number of candidate voxels for an independent defect object. This cube root value mathematically represents the equivalent spatial scale of the defect object and has a monotonically corresponding relationship with the defect volume, enabling adaptive determination of the rejection range based on the actual size of each defect. In practice, those skilled in the art can superimpose a fixed micrometer-level expansion margin on the cube root value to compensate for the boundary effect caused by voxel discretization. Finally, by comparing the degree of agreement between the sensitive area delineation results of the known defect distribution area on the test block and the actual crack propagation path, the expansion margin can be verified and adjusted.

[0132] The image generation module generates a 30° oblique G-scan image from the original 30° oblique scanning data and performs root tomographic slicing. Using the weld root position determined by the actual interface contour as the depth reference, the position extending 1 mm from the weld root towards the weld surface is defined as the first depth boundary, the position extending 1 mm from the weld root towards the inner wall of the pipe is defined as the second depth boundary, and the position extending 1.5 mm from the weld root towards the inner wall of the pipe is defined as the third depth boundary. It iterates through each pixel in the 30° oblique G-scan image. If the radial depth coordinate of a pixel is not less than the first depth boundary and not greater than the second depth boundary, or not less than the second depth boundary and not greater than the third depth boundary, the original amplitude of the pixel is retained. Otherwise, the amplitude of the pixel is set to zero. The root sensitive signal map is generated from all the retained pixels. During this process, the weld contour reflection signal and the signal from the pores in the upper part of the weld are effectively removed.

[0133] The spatial comparison module extracts the spatial coordinates of a total of 3 reflected signals from the root sensitive signal spectrum. It calculates the spatial difference point by point with the coordinates of pores and inclusions in the 0° reference defect map. One signal is identified as a confirmed welding defect signal and is excluded because the spatial difference with the location of the inclusion is less than the preset overlap threshold of 0.2 mm. One signal located near the inner wall of the pipe is retained for observation because the spatial difference with any marked defect is greater than 0.2 mm, but the radial position is deviated from the outer wall of the pipe. Another signal located near the junction line of the weld root is retained in the remaining signal set because the spatial difference with any marked defect is greater than 0.2 mm.

[0134] The crack detection module extracts the coordinates of the reflection point of the retained signal, determines that its radial position falls on the outer wall of the pipe and is within a preset neighborhood of 0.3 mm at the weld root boundary line. At the same time, it confirms that no defect signal is displayed at this position when the 0° direct scan is performed, and the amplitude of the signal echo is 68% of the full screen scale, which is higher than the preset amplitude threshold, thus meeting the physical criteria for corner reflection signals. Based on this, it is determined that there is a root penetrating crack in the pipe head.

[0135] In a 30° oblique A-scan signal, the depth grading module traces the sound ray path based on a 30° incident beam angle and Snell's law of refraction. The calculated propagation distance of the wedge segment is 5.8 mm. Dividing this by the longitudinal wave velocity of the wedge segment (2730 m / s) yields a one-way propagation time of 2.12 microseconds. Taking the difference of 0 mm between the radial coordinates of the reflection point and the radial coordinates of the outer wall of the pipe as the propagation distance of the pipe wall segment, the one-way propagation time of the pipe wall segment is 0 microseconds. The round-trip propagation time is calculated to be 2 × 2.12 = 4.24 microseconds. This 4.24 microseconds is determined as the target time. Using 4.24 microseconds as the center, a pre-measured distance is taken on the precision wire-cut groove test block. The full width at half maximum (FWHM) of the probe pulse in the pipe wall material was found to be 0.32 microseconds. A FWHM of 0.16 microseconds was selected as the first and second preset search FWHMs, and the target search time interval was determined to be from 4.08 microseconds to 4.40 microseconds. The time sampling interval for the 30° oblique scanning was 0.01 microseconds. Starting from 4.08 microseconds, the amplitude of the A-scan signal was read sequentially in 0.01 microsecond steps, for a total of 32 sampling points. The absolute value of each amplitude was calculated, and the maximum absolute amplitude was found to be 68% of the full-screen scale. This 68% was taken as the echo amplitude of the corner reflection signal. Input a pre-established amplitude-crack depth mapping lookup table containing 10 node pairs sorted in ascending order of echo amplitude value. These pairs correspond to the following values: echo amplitude 12% corresponds to a crack depth of 0.125 mm; echo amplitude 18% corresponds to 0.25 mm; echo amplitude 26% corresponds to 0.375 mm; echo amplitude 35% corresponds to 0.5 mm; echo amplitude 46% corresponds to 0.75 mm; echo amplitude 58% corresponds to 1.0 mm; echo amplitude 65% corresponds to 1.25 mm; echo amplitude 74% corresponds to 1.5 mm; echo amplitude 82% corresponds to 1.75 mm; and echo amplitude 1.0 mm corresponds to 1.75 mm. A 90% amplitude value corresponds to 2.0 mm. The 68% echo amplitude value falls between the 65% and 74% nodes. Based on the linear interpolation formula, the estimated crack depth is 1.2 + (1.5 - 1.25) × (68% - 65%) / (74% - 65%) = 1.25 + 0.25 × 3% / 9% = 1.33 mm. Comparing the estimated crack depth of 1.33 mm with the first crack depth threshold of 1.5 mm and the second crack depth threshold of 0.5 mm, 1.33 mm is less than 1.5 mm but greater than 0.5 mm. Therefore, this crack is classified as a shallow crack, and it is recommended to take key monitoring measures for this pipe head.

[0136] When testing another tube in the same heat exchanger, the coupling adaptation module performs a coupling calibration operation. All elements of the phased array probe excite and emit calibration pulses with a peak-to-peak amplitude of 100 mV in an equal-amplitude and in-phase manner. The receiving elements receive the reflected echo signal from the interface, extracting the maximum absolute amplitude sampling point amplitude as 42 mV. The amplitude ratio of the interface reflected echo is calculated to be 0.42. Performing a Fast Fourier Transform on the echo signal yields a spectral centroid offset of 0.15 MHz. The coupling state parameter is calculated to be 0.42 × 0.7 + 0.15 × 0.3 = 0.339, which is lower than the preset coupling quality threshold of 0.52. The system determines that the upward oblique projection coupling condition has deteriorated and displays the text prompt "Current..." on the human-machine interface. The coupling quality is insufficient. Please replace it with a miniature probe and place the miniature probe at the bottom of the expansion tube groove. The center position of the bottom of the expansion tube groove is marked with a red arrow. The operator places a miniature probe with a wafer array length of 8 mm, a width of 5 mm, a number of array elements of 16, and an array element spacing of 0.5 mm at the bottom of the expansion tube groove. The radius of curvature of the concave surface at the bottom of the shell is the same as the radius of curvature of the bottom of the groove to achieve surface contact. The system re-performs coupling calibration, and the coupling status parameter is restored to 0.61. The human-machine interface displays "Coupling quality is qualified". The miniature probe is then used to continue to complete the upward 30° oblique scanning. Finally, the test result shows that there is no abnormal reflection signal in the root sensitive signal spectrum, and it is judged to be qualified.

[0137] Example 4:

[0138] Another embodiment of the present invention provides a method for detecting fillet weld cracks based on dual-scan comparison, comprising:

[0139] S1. Perform 0° direct beam scanning and 30° oblique beam scanning on the same pipe head fillet weld, respectively, to obtain the original detection data of 0° direct beam scanning and 30° oblique beam scanning. The sound beam of the 0° direct beam scanning is perpendicular to the probe surface and emitted along the pipe wall axis, covering the full thickness range of the weld. The sound beam of the 30° oblique beam scanning is directed at the root area of ​​the weld at a 30° tilt angle, with the focal point covering the sensitive area from the outer wall of the pipe to the root of the weld.

[0140] S2. Generate a 0° direct-scan G-scan image and a D-scan image on the original detection data of the 0° direct scan, and perform three-dimensional sectional analysis and slicing to identify and mark the spatial positions of all welding manufacturing defects and usage defects. Generate a 0° reference defect image marked with the coordinates of each defect position, and delineate the crack sensitive area according to the location area where there is no welding defect signal in the 0° reference defect image.

[0141] S3. Generate a 30° oblique beam G-scan image on the original 30° oblique beam G-scan data, and perform root tomographic slicing processing on the 30° oblique beam G-scan image; the root tomographic slicing processing includes: using the weld root position at the junction of the outer wall of the pipe and the weld as a depth reference, determining the position extending a first preset distance from the depth reference towards the weld surface as a first depth boundary, and determining the position extending the first preset distance from the depth reference towards the inner wall of the pipe as a second depth boundary, retaining the echo signal whose depth position falls between the first depth boundary and the second depth boundary; and, retaining the position extending the second preset distance from the depth reference towards the inner wall of the pipe as a second depth boundary. The third depth boundary is defined, and echo signals whose depth positions fall between the second depth boundary and the third depth boundary are retained. The root sensitive signal spectrum is generated from all retained echo signals, and the unretained echo signals are discarded as weld contour reflection signals and upper defect signals. In this embodiment, the first preset distance is ±1mm, where "±" represents a bidirectional depth range offset by 1mm from the weld root position to both the weld surface direction and the pipe inner wall direction, respectively. The second preset distance is 1.5mm, so that the target depth range is the depth range from 1mm from the weld root position towards the weld surface direction to 1.5mm towards the pipe inner wall direction.

[0142] S4. Extract the first spatial position coordinates of each signal in the root sensitive signal spectrum, and compare each first spatial position coordinate with the second spatial position coordinates of each defect marked in the 0° reference defect map one by one. Specifically, perform point-by-point spatial difference calculation. When the spatial difference is less than the preset overlap threshold, it is determined to be a position overlap. The signals with overlapping positions are excluded as confirmed welding defect signals, and the signals with non-overlapping positions are retained to form a set of remaining signals.

[0143] S5. For each signal in the remaining signal set, extract the spatial coordinates of its reflection point and determine whether it simultaneously satisfies the following conditions: the reflection point is located on the outer wall side of the pipe and within the preset neighborhood of the weld root boundary line, and the original detection data of the 0° direct scan does not show any defect signal at the reflection point position; if both conditions are met, then the signal is determined to be a corner reflection signal, and it is determined that there is a root penetrating crack in the corresponding pipe head.

[0144] S6. Extract the echo amplitude value from the signal data determined to be the corner reflection signal, input the echo amplitude value into a lookup table of amplitude-crack depth mapping relationship pre-calibrated based on the precision wire-cut groove test block, and obtain the corresponding crack depth estimate; compare the crack depth estimate with a preset first crack depth threshold and a preset second crack depth threshold, wherein the first crack depth threshold is greater than the second crack depth threshold; if the crack depth estimate is greater than or equal to the first crack depth threshold, the crack is classified as a third crack level; if the crack depth estimate is less than the first crack depth threshold but greater than or equal to the second crack depth threshold, the crack is classified as a second crack level; if the crack depth estimate is less than the second crack depth threshold, the crack is classified as a first crack initiation level.

[0145] S7. Acquire the reflected echo signal from the inner wall interface of the tube measured by the phased array probe before the start of the 30° oblique scanning, and extract the amplitude of the reflected echo from the inner wall interface. Divide the amplitude of the reflected echo from the inner wall interface by the peak-to-peak amplitude of the transmitted calibration pulse to obtain the interface reflected echo amplitude ratio. Perform a fast Fourier transform on the reflected echo signal from the inner wall interface to obtain the frequency domain amplitude distribution. Calculate the sum of the products of the amplitude of each frequency component and its corresponding frequency in the frequency domain amplitude distribution, and divide it by the sum of the amplitudes of each frequency component to obtain the centroid frequency. Use the absolute value of the difference between the centroid frequency and the probe's nominal center frequency as the spectral centroid offset. Multiply the interface reflected echo amplitude ratio by a first preset weight and the spectral centroid offset by a second preset weight to obtain the coupling state parameter.

[0146] The coupling state parameter is compared with a preset coupling quality threshold; if the coupling state parameter is less than the preset coupling quality threshold, it is determined that the upward oblique coupling condition has deteriorated, and the first adaptive adjustment or the second adaptive adjustment is performed.

[0147] If the coupling state parameter is greater than or equal to the preset coupling quality threshold, the upward oblique coupling condition is determined to be normal, and the 30° oblique scanning is continued with the current upward oblique beam direction and the current ultrasound probe.

[0148] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of the present invention without departing from the spirit and scope of the claims. All of these variations are within the protection scope of the present invention.

Claims

1. A fillet weld crack detection system based on dual-scan comparison, characterized in that, include: The scanning data acquisition module is configured to perform 0° direct scanning and 30° oblique scanning on the same pipe head fillet weld, respectively, to acquire the original detection data of 0° direct scanning and 30° oblique scanning. The reference defect map generation module is configured to perform a three-dimensional Fourier transform or synthetic aperture focusing algorithm on the raw detection data of the 0° direct beam scan, and reconstruct the A-scan signal sequence acquired along the pipe wall axis into a 0° direct beam G-scan map and a D-scan map; wherein, the 0° direct beam G-scan map is a cross-sectional view of the weld perpendicular to the direction of the sound beam, and the D-scan map is a longitudinal cross-sectional view of the weld along the direction of sound beam propagation; A three-dimensional sectional analysis slice is performed on the two-dimensional image sequence composed of the 0° direct G scan image and the D scan image to generate a 0° reference defect image with the spatial location of welding defects, and to delineate the crack sensitive area without welding defects. The spectrum generation module is configured to extract echo signals within a preset root depth range through root tomography processing based on the original detection data from the 30° oblique scanning, and generate a root sensitive signal spectrum. The spatial comparison module is configured to compare the first spatial position coordinates of each signal in the root sensitive signal spectrum with the second spatial position coordinates of the defects already marked in the 0° reference defect map, exclude confirmed welding defect signals with overlapping positions, and retain signals with non-overlapping positions as the remaining signal set. The crack detection module is configured to determine whether each signal in the remaining signal set simultaneously satisfies the following conditions: the reflection point is located in the junction area between the outer wall of the pipe and the root of the weld, and the corresponding reflection point has no welding defect signal in the 0° reference defect diagram. If these conditions are met, it is determined to be a corner reflection signal, and the root penetrating crack of the corresponding pipe head is confirmed.

2. The fillet weld crack detection system based on dual-scan comparison as described in claim 1, characterized in that, In the scanning data acquisition module, the sound beam of the 0° direct scanning is perpendicular to the probe surface and emitted along the pipe wall axis; the sound beam of the 30° oblique scanning is directed at the weld root region at a 30° tilt angle. The reference defect map generation module is further configured to delineate the crack sensitive area based on the location area where no welding defect signal exists in the 0° reference defect map; The spectrum generation module is also configured to generate a 30° oblique G-scan image on the raw detection data of the 30° oblique scanning, and to perform root tomography slicing on the 30° oblique G-scan image.

3. The fillet weld crack detection system based on dual-scan comparison as described in claim 2, characterized in that, The root sectioning process is configured as follows: using the weld root position at the junction of the outer wall of the pipe and the weld as a depth reference, the position extending a first preset distance from the depth reference towards the weld surface is defined as a first depth boundary, and the position extending the first preset distance from the depth reference towards the inner wall of the pipe is defined as a second depth boundary, retaining echo signals whose depth positions fall between the first depth boundary and the second depth boundary; and, the position extending a second preset distance from the depth reference towards the inner wall of the pipe is defined as a third depth boundary, retaining echo signals whose depth positions fall between the second depth boundary and the third depth boundary; The root sensitive signal spectrum is generated from all retained echo signals, and the unretained echo signals are discarded as weld contour reflection signals and upper defect signals.

4. The fillet weld crack detection system based on dual-scan comparison as described in claim 1, characterized in that, The spatial comparison module is further configured to perform point-by-point spatial difference calculation on the first spatial position coordinates of each signal in the root sensitive signal spectrum and the second spatial position coordinates of each defect marked in the 0° reference defect map, and determine that the positions coincide when the spatial difference is less than a preset overlap threshold. The fillet weld crack detection system also includes a depth grading module, configured to extract echo amplitude values ​​from signal data determined to be corner reflection signals, input the echo amplitude values ​​into a pre-calibrated amplitude-crack depth mapping lookup table based on wire-cut groove test blocks to obtain the corresponding crack depth estimate; and compare the crack depth estimate with a preset first crack depth threshold and a preset second crack depth threshold. If the crack depth estimate is greater than or equal to the first crack depth threshold, the crack is classified as a third crack level; if the crack depth estimate is less than the first crack depth threshold but greater than or equal to the second crack depth threshold, the crack is classified as a second crack level; if the crack depth estimate is less than the second crack depth threshold, the crack is classified as a first crack initiation level.

5. The fillet weld crack detection system based on dual-scan comparison as described in claim 1, characterized in that, The three-dimensional segmentation and slicing of the two-dimensional image sequence composed of the 0° direct-view G scan and D scan is specifically as follows: Using the probe coupling surface as the origin of the spatial coordinate system, a three-dimensional voxel grid is constructed along the radial, axial and circumferential directions of the weld with a preset voxel size; the echo amplitude of each A-scan signal sampling point is mapped to the corresponding voxel according to the sound path and the sound beam spread angle to form a three-dimensional amplitude matrix. The three-dimensional amplitude matrix is ​​traversed using a preset amplitude threshold. Voxels with amplitudes exceeding the preset amplitude threshold are marked as defect candidate voxels. A three-dimensional connected component labeling algorithm is then applied to cluster spatially adjacent defect candidate voxels into independent defect objects.

6. The fillet weld crack detection system based on dual-scan comparison as described in claim 5, characterized in that, Generating the 0° reference defect map and delineating the crack-sensitive region further includes: For each independent defect object, its bulk centroid coordinates are extracted as the second spatial position coordinates of the corresponding independent defect object, and the maximum amplitude, volume and morphological characteristics of the corresponding independent defect object are recorded as attribute parameters; the second spatial position coordinates and attribute parameters of all the labeled independent defect objects are integrated to form the 0° reference defect map; The echo signal of the original 0° direct scan detection data is obtained within the full thickness range of the weld. Based on the acoustic impedance difference between the weld metal and the pipe wall base material, the actual boundary contour between the outer wall of the pipe and the molten metal of the weld is extracted by the interface echo recognition algorithm. The actual boundary contour is offset into the weld by a preset molten zone thickness value to obtain the boundary of the welding molten zone. In the 0° reference defect diagram, the spatial location area that is within the boundary of the weld melting zone and is not occupied by any of the independent defect objects is determined as the location area where there is no weld defect signal. The spatially connected set of all location areas where there is no weld defect signal that extend continuously along the outer wall of the pipe to the weld root boundary line is defined as the crack sensitive area.

7. The fillet weld crack detection system based on dual-scan comparison as described in claim 4, characterized in that, The process of obtaining the corresponding estimated crack depth includes: In the original detection data of the 30° oblique scanning, the A-scan signal corresponding to the corner reflection signal is located; based on the spatial coordinates of the reflection point of the corner reflection signal, combined with the incident angle of the sound beam of the 30° oblique scanning, the longitudinal wave velocity of the wedge material and the transverse wave velocity of the pipe wall material under test, the propagation time of the wedge segment and the pipe wall segment along the sound beam path is calculated and summed to obtain the round-trip propagation time of the ultrasonic wave from the probe chip to the probe chip via the reflection point, and the round-trip propagation time is determined as the target time on the A-scan signal; Centered on the target time, the time interval defined by extending forward by a first preset search half-width and backward by a second preset search half-width is taken as the target search time interval; both the first preset search half-width and the second preset search half-width are taken as the half-width of the ultrasonic pulse emitted by the phased array probe used in the 30° oblique scanning in the pipe wall material.

8. The fillet weld crack detection system based on dual-scan comparison as described in claim 7, characterized in that, The process of obtaining the corresponding estimated crack depth also includes: Within the target search time interval, the signal amplitude of the A scan signal is obtained point by point. The signal amplitude corresponding to the sampling point with the largest absolute amplitude among all the obtained sampling points is extracted as the echo amplitude of the corner reflection signal. The echo amplitude value is input into a pre-established amplitude-crack depth mapping lookup table, and the corresponding crack depth estimate is obtained by matching the echo amplitude value range. If the echo amplitude value is greater than the maximum echo amplitude value in the amplitude-crack depth mapping lookup table, then the estimated crack depth value is taken as the corresponding maximum crack depth value in the lookup table, and an over-range marker signal is output. If the echo amplitude value is less than the minimum echo amplitude value in the amplitude-crack depth mapping lookup table, then the estimated crack depth value is set to 0.

9. A method for detecting fillet weld cracks based on dual-scan comparison, implemented based on the fillet weld crack detection system based on dual-scan comparison as described in any one of claims 1-8, characterized in that, include: For the same pipe head fillet weld, perform 0° direct beam scanning and 30° oblique beam scanning respectively to obtain the original inspection data of 0° direct beam scanning and 30° oblique beam scanning. Based on the original detection data from the 0° direct scanning, a 0° reference defect map with the spatial location of welding defects is generated by three-dimensional tomographic imaging, and a crack-sensitive area without welding defects is delineated. Based on the original detection data from the 30° oblique scanning, echo signals within a preset root depth range are extracted through root tomography to generate a root sensitive signal spectrum. The first spatial position coordinates of each signal in the root sensitive signal spectrum are compared with the second spatial position coordinates of the defects already marked in the 0° reference defect map. The confirmed welding defect signals with overlapping positions are excluded, and the signals with non-overlapping positions are retained as the remaining signal set. For each signal in the remaining signal set, determine whether it simultaneously satisfies the following conditions: the reflection point is located in the junction area between the outer wall of the pipe and the root of the weld, and the corresponding reflection point has no welding defect signal in the 0° reference defect diagram. If these conditions are met, it is determined to be a corner reflection signal, and it is determined that there is a root penetrating crack in the corresponding pipe head.

Images