Corner reflector signal based fillet weld crack identification system and method

Abstract

The application discloses a corner reflection signal-based corner weld crack identification system and method, belongs to the technical field of tube plate corner weld nondestructive testing, and comprises the following steps: first, initializing and configuring phased array ultrasonic detection parameters, obtaining 0-degree direct reflection and 30-degree oblique reflection scanning signals of a corner weld area and corresponding coordinates; second, constructing a circumferential-depth graph by tomographic sectioning along a sensitive area of a weld root, locking corner reflection candidate signals by comparing the direct reflection and the oblique reflection signals; third, generating an axial-depth graph by tomographic sectioning of the candidate signals along an axial-depth direction, and determining characteristic signals according to positions, amplitudes and depths; and finally, calculating crack sizes through a calibration model, marking crack parameters, and outputting micro-cracks, penetration states and grading marks; the application can accurately identify corner weld root cracks by using corner reflection signals, realize crack positioning, quantification and grading, has high detection precision and strong reliability, and is suitable for efficient detection of tube plate corner weld cracks.

Description

Technical Field

[0001] This invention belongs to the field of non-destructive testing technology for fillet welds in tube sheets, specifically a fillet weld crack identification system and method based on corner reflection signals. Background Technology

[0002] Tube-to-tube sheet fillet welds are critical connection points in pressure-bearing equipment such as heat exchangers and boilers. They are subjected to high temperature and pressure, media corrosion, and alternating loads over a long period of time, which can easily lead to micro-cracks, propagating cracks, and even penetrating cracks at the weld root, affecting the safe operation of the equipment. Therefore, high-precision non-destructive testing is required.

[0003] Currently, fillet weld crack detection mostly employs conventional ultrasonic or phased array single-angle scanning, which has certain drawbacks: conventional ultrasonic beams have limited coverage and low sensitivity for identifying micro-cracks at the root; single-angle phased array scanning is easily affected by weld structure interference, making it difficult to distinguish root crack reflections from geometric clutter, resulting in a high false positive rate. Furthermore, existing technologies lack targeted extraction and analysis of corner reflection signals, making it impossible to accurately pinpoint the root crack location, quantify crack size, and classify crack condition, thus failing to meet the demands of industrial applications for detection accuracy, reliability, and efficiency.

[0004] In summary, there is an urgent need for a detection method that can accurately identify and quantitatively analyze root cracks in fillet welds to solve the problems of low sensitivity, high false positive rate, and inability to accurately quantify and classify existing technologies. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a fillet weld crack identification system and method based on corner reflection signals. First, the phased array ultrasonic testing parameters are initialized and configured to acquire 0° direct and 30° oblique scanning signals and their corresponding coordinates in the fillet weld area. A circumferential-depth map is constructed by tomographic slicing along the sensitive area at the weld root, and corner reflection candidate signals are identified by comparing the direct and oblique signals. An axial-depth map is generated by tomographic slicing the candidate signals along the axial-depth direction, and feature signals are determined based on position, amplitude, and depth. Finally, the crack size is calculated using a calibration model, crack parameters are labeled, and microcrack, penetration status, and classification indicators are output. This invention utilizes corner reflection signals to accurately identify cracks at the root of fillet welds, achieving crack location, quantification, and classification. It boasts high detection accuracy and reliability, making it suitable for efficient detection of fillet weld cracks in pipes and tube sheets.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for identifying fillet weld cracks based on corner reflection signals includes:

[0008] S1: Acquire the phased array direct scanning signal at the first angle and the oblique scanning signal at the second angle in the tube-tube sheet fillet weld inspection area, and simultaneously collect the circumferential, axial and radial coordinate parameters of the corresponding scanning positions;

[0009] S2: Based on the second angle oblique scanning signal and coordinate parameters, perform tomographic slicing along the sensitive area at the root of the weld to generate root slice signal data and construct a circumferential-depth map;

[0010] S3: Compare the root slice signal data with the first angle direct scan signal to lock the corner reflection candidate signal and its coordinates that only exist in the sensitive area of ​​the weld root;

[0011] S4: At the coordinates corresponding to the corner reflection candidate signal, perform tomographic slicing along the axial-depth direction to generate an axial-depth map. Determine that the corner reflection candidate signal is a fixed-position corner reflection feature signal formed by the outer wall of the pipe and the crack. Verify that the corner reflection position, corner reflection amplitude and depth meet the preset crack judgment conditions, and output a crack identification completion signal.

[0012] S5: Based on the crack identification completion signal, calculate the crack size using a preset corner reflection amplitude and depth calibration model, mark the calculation results on the circumferential-depth map and axial-depth map, mark the crack circumferential position, axial depth, radial depth and length data at the corresponding positions, and output the initial microcrack depth judgment mark, crack penetration status mark and grade judgment mark.

[0013] Specifically, before acquiring the phased array first-angle direct scanning signal and second-angle oblique scanning signal of the tube sheet fillet weld detection area, the process further includes performing initialization configuration, including: setting the transmission and reception channel parameters of the phased array ultrasonic detector, configuring the focusing rule corresponding to the first-angle direct scanning, configuring the incident angle of the first-angle direct scanning signal beam to 0° vertical incidence, configuring the focusing rule corresponding to the second-angle oblique scanning, configuring the incident angle of the second-angle oblique scanning signal beam to 30° oblique incidence, simultaneously configuring the sampling frequency, gain, pulse repetition frequency and detection mode of the phased array ultrasonic detector, setting the scanning start point coordinates and scanning end point coordinates, configuring the scanning path to scan around the weld circumference, setting the initialization completion flag to true, and outputting the detection system ready signal.

[0014] Specifically, the steps of S1 include:

[0015] S1.1: After receiving the detection system ready signal, move the phased array probe along the scanning path;

[0016] S1.2: At each sampling trigger position, a first-angle direct scanning signal beam and a second-angle oblique scanning signal beam are synchronously transmitted, and the corresponding first echo sequence and second echo sequence are received respectively;

[0017] S1.3: At the same sampling trigger position, the circumferential coordinate value of the phased array probe center is acquired by the encoder, the axial coordinate value is acquired by the axial displacement sensor, and the radial coordinate value is acquired by the radial displacement sensor.

[0018] S1.4: Pack the first echo sequence, second echo sequence, circumferential coordinate value, axial coordinate value and radial coordinate value obtained at the same sampling trigger position into a data frame, and store the data frames corresponding to each sampling trigger position as the original scan data file in ascending order of circumferential coordinate value.

[0019] Specifically, the steps of S2 include:

[0020] S2.1: Read the second echo sequence and circumferential coordinate values ​​of each data frame from the original scan data file;

[0021] S2.2: Set the starting and ending depth values ​​of the sensitive zone at the weld root to determine the slice depth range;

[0022] S2.3: The sliding window method is used to extract the echo amplitude values ​​of each second echo sequence point by point along the depth direction, and the depth values ​​are within the slice depth range.

[0023] S2.4: Arrange the echo amplitude values ​​corresponding to each depth value under the same circumferential coordinate value in ascending order of depth to form the root signal vector of the corresponding circumferential coordinate value;

[0024] S2.5: Combine the root signal vectors of all circumferential coordinate values ​​in ascending order of circumferential coordinate values ​​to generate root slice signal data; the root slice signal data includes slice depth range, circumferential angle distribution, echo position and echo amplitude;

[0025] S2.6: Perform grayscale mapping processing on the root slice signal data to construct a circumferential-depth map with circumferential coordinates as the horizontal axis and depth as the vertical axis.

[0026] Specifically, the steps of S3 include:

[0027] S3.1: Read the first echo sequence of each data frame from the raw scan data file;

[0028] S3.2: Extract the echo amplitude values ​​within the depth range of the sensitive area at the root of the weld from each first echo sequence to form a first amplitude value set;

[0029] S3.3: Extract echo amplitude values ​​with the same circumferential coordinate values ​​and the same depth range from the root slice signal data to form a second amplitude value set;

[0030] S3.4: Compare the first amplitude value and the second amplitude value corresponding to the same spatial location point one by one, and mark the spatial location points that meet the preset conditions as single-angle unique points. The preset conditions are: the first amplitude value is less than the first preset threshold, and the second amplitude value is greater than the second preset threshold.

[0031] S3.5: Retain the circumferential coordinate values ​​and depth values ​​corresponding to all single-angle unique points as corner reflection candidate signal positions, and use the corner reflection candidate signal positions and their corresponding circumferential coordinate values ​​and depth values ​​as corner reflection candidate signals and their coordinates.

[0032] Specifically, the step of performing tomographic slicing along the axial-depth direction at the coordinates corresponding to the corner reflection candidate signal to generate an axial-depth map includes:

[0033] S4.1: Obtain the circumferential coordinates and depth values ​​corresponding to the corner reflection candidate signals;

[0034] S4.2: Extract all data frames whose circumferential coordinate values ​​are within the preset neighborhood range of the circumferential coordinate values ​​of the corner reflection candidate signal from the original scan data file, and read the second echo sequence and axial coordinate value of the corresponding data frame from each extracted data frame;

[0035] S4.3: With the depth value of the corner reflection candidate signal as the center, set a depth slice window. For each extracted data frame, extract the echo amplitude value sequence within the depth range of the depth slice window from the second echo sequence of the data frame, and associate and store the echo amplitude value sequence extracted from each data frame with the axial coordinate value of the corresponding data frame.

[0036] S4.4: Arrange the echo amplitude value sequence after all associated storage in ascending order of axial coordinate values ​​to generate an axial-depth signal data matrix, and perform grayscale mapping processing on the axial-depth signal data matrix to construct an axial-depth map with axial coordinate as the horizontal axis and depth as the vertical axis.

[0037] Specifically, the output process of the crack identification completion signal includes:

[0038] S4.5: Extract the distribution curve of the echo amplitude value along the axial direction corresponding to the depth value of the corner reflection candidate signal from the generated axial-depth signal data matrix;

[0039] S4.6: Locate the global maximum point of the echo amplitude value in the distribution curve;

[0040] S4.7: Obtain the axial coordinate value corresponding to the global maximum point as the corner reflection axial position; obtain the depth value corresponding to the global maximum point as the corner reflection depth value; obtain the echo amplitude value of the global maximum point as the corner reflection amplitude.

[0041] S4.8: Retrieve preset crack judgment conditions, and calculate the absolute value of the axial deviation between the corner reflection axial position and the preset standard axial position, and calculate the absolute value of the depth deviation between the corner reflection depth value and the preset standard depth value; the preset crack judgment conditions include axial position deviation threshold, depth deviation threshold and amplitude threshold;

[0042] S4.9: When the absolute value of the axial deviation is less than or equal to the axial position deviation threshold, the absolute value of the depth deviation is less than or equal to the depth deviation threshold, and the corner reflection amplitude is greater than or equal to the amplitude threshold, the corner reflection candidate signal is determined to be the fixed position corner reflection characteristic signal formed by the outer wall of the pipe and the crack, which satisfies the crack judgment condition, and a crack identification completion signal carrying the corner reflection axial position, corner reflection depth value, and corner reflection amplitude is output.

[0043] Specifically, the step of calculating the crack size based on the crack identification completion signal using a preset corner reflection amplitude and depth calibration model includes:

[0044] S5.1: Receive the crack identification completion signal and extract the corner reflection amplitude and corner reflection depth values ​​carried in the crack identification completion signal;

[0045] S5.2: Load the preset corner reflection amplitude and depth calibration model; the corner reflection amplitude and depth calibration model is obtained by performing phased array detection experiments on tube sheet fillet weld test blocks containing artificial cracks of different depths, and then fitting and calibrating using the least squares method;

[0046] S5.3: Substitute the extracted corner reflection amplitude and corner reflection depth values ​​into the corner reflection amplitude and depth calibration model to perform forward calculation, solve for the corresponding crack radial depth prediction value, and determine the crack radial depth prediction value as the radial depth component in the crack size calculation result.

[0047] Specifically, the step of calculating the crack size using a preset corner reflection amplitude and depth calibration model also includes a step of calculating the crack axial length value:

[0048] B1: Obtain the axial-depth map and the axial-depth signal data matrix;

[0049] B2: Extract the echo amplitude value sequence corresponding to the corner reflection depth value from the axial-depth signal data matrix along the axial direction;

[0050] B3: Set a crack axial endpoint determination threshold, and mark the position where the echo amplitude value in the echo amplitude value sequence is first greater than or equal to the crack axial endpoint determination threshold as the crack initiation axial position;

[0051] B4: Mark the position where the last echo amplitude value in the echo amplitude value sequence is greater than or equal to the crack axial endpoint determination threshold as the crack termination axial position;

[0052] B5: Calculate the difference between the axial position of crack termination and the axial position of crack initiation, and use the difference as the crack axial length value;

[0053] B6: The crack axial length value is determined as the axial length component in the crack size calculation result, and together with the radial depth component, it is used as the crack size calculation result.

[0054] A fillet weld crack identification system based on corner reflection signals includes:

[0055] The signal acquisition module is used to simultaneously acquire the first angle direct scanning signal, the second angle oblique scanning signal and the corresponding circumferential, axial and radial coordinate parameters of the tube sheet fillet weld inspection area, and package and store them as raw scanning data files;

[0056] The root slicing module generates root slice signal data based on the second angle oblique scanning signal and coordinate parameters. It constructs a circumferential-depth map through grayscale mapping. At the same time, it performs axial-depth tomographic slicing on the corner reflection candidate signal to generate an axial-depth signal data matrix and construct an axial-depth map.

[0057] The candidate signal locking module is used to compare the root slice signal data with the first angle direct scan signal, filter the corner reflection candidate signal positions, and then cluster and merge the corner reflection candidate signals to lock the effective corner reflection candidate signals and their coordinates.

[0058] The crack feature determination module is used to locate the global maximum point to obtain the corner reflection position, amplitude and depth, retrieve the preset crack determination conditions, determine the corner reflection candidate signal as the corner reflection feature signal of the fixed position formed by the pipe outer wall and the crack, and output the crack identification completion signal.

[0059] The crack size calculation module is used to load a preset corner reflection amplitude and depth calibration model, calculate the crack radial depth and axial length, and output the initial microcrack depth judgment mark, crack penetration status mark and grade judgment mark.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] 1. This invention proposes a method for identifying fillet weld cracks based on corner reflection signals. By combining dual-angle phased array scanning with tomographic slicing technology, it accurately extracts corner reflection signals from the sensitive area at the weld root, effectively separating crack reflections from geometric clutter, reducing the false positive rate, and significantly improving the sensitivity and accuracy of identifying and locating micro-cracks at the fillet weld root. Simultaneously, based on density clustering to screen candidate signals and dual-dimensional map cross-validation, the reliability of crack identification is further enhanced, making it suitable for complex weld structure inspection scenarios.

[0062] 2. This invention proposes a method for identifying fillet weld cracks based on corner reflection signals. It relies on a preset calibration model to achieve accurate quantification of crack radial depth and axial length, and simultaneously completes microcrack depth determination, penetration state identification and graded evaluation. It can intuitively label key crack parameters, and the detection process is highly efficient and automated. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the fillet weld crack identification method based on corner reflection signals according to the present invention;

[0064] Figure 2 This is a flowchart illustrating the principle of the fillet weld crack identification method based on corner reflection signals of the present invention.

[0065] Figure 3 This is a diagram illustrating the architecture of the fillet weld crack identification system based on corner reflection signals according to the present invention. Detailed Implementation

[0066] Example 1:

[0067] Please see Figure 1 and Figure 2 The present invention provides an embodiment of a method for identifying fillet weld cracks based on corner reflection signals, comprising the following steps:

[0068] S1: Acquire the phased array direct scanning signal at the first angle and the oblique scanning signal at the second angle in the tube-tube sheet fillet weld inspection area, and simultaneously collect the circumferential, axial and radial coordinate parameters of the corresponding scanning positions;

[0069] Before acquiring the phased array direct scanning signal and the second oblique scanning signal of the tube sheet fillet weld inspection area, the process includes performing initialization configuration, including: setting the transmission and reception channel parameters of the phased array ultrasonic detector, configuring the focusing rule corresponding to the first angle direct scanning, configuring the incident angle of the first angle direct scanning signal beam to 0° vertical incidence, configuring the focusing rule corresponding to the second angle oblique scanning, configuring the incident angle of the second angle oblique scanning signal beam to 30° oblique incidence, simultaneously configuring the sampling frequency, gain, pulse repetition frequency and detection mode of the phased array ultrasonic detector, setting the scanning start point coordinates and scanning end point coordinates, configuring the scanning path to scan around the weld circumference, setting the initialization completion flag to true, and outputting the detection system ready signal.

[0070] Furthermore, this embodiment takes the fillet weld at the connection between the heat exchange tube and the tube sheet of a power plant boiler as a typical inspection object. This location is subjected to high temperature and high pressure, media corrosion, and alternating loads for a long time. The root of the weld is the weakest area where stress concentration is most severe and microcracks are most likely to initiate and gradually propagate. Therefore, accurate identification and size assessment of root cracks are crucial. Before the inspection begins, the initialization and parameter configuration of the entire phased array ultrasonic testing system need to be completed. The specific execution process of the initialization configuration is as follows:

[0071] First, the operator physically connects the phased array ultrasonic testing instrument to the dedicated linear or area array probes, confirming that the probe interfaces, signal transmission cables, encoders, and displacement sensors are properly connected without looseness, signal interference, or poor contact. Then, the operator enters the instrument parameter setting interface and configures the basic parameters of the transmitting and receiving channels, including the number of channels, transmitting voltage, receiving bandwidth, signal filtering type, and filtering range, to ensure stable ultrasonic transmission energy and pure received signals, reducing environmental noise and clutter interference. Next, the operator configures the focusing rule corresponding to the first-angle direct-scan scan. This scan mode uses a 0° vertical incidence method, with the sound beam direction consistent with the normal direction of the outer wall of the pipe. It is mainly used to obtain the basic reflection signal of the entire weld area, providing preliminary imaging of the overall weld structure, base material boundary, and weld fusion line. The focusing rule requires specifying the beam focal depth, focusing aperture, and number of focal points to ensure that the vertically incident sound beam can uniformly cover the weld cross-section without blind spots or signal loss.

[0072] Then, the focusing rule corresponding to the second angle oblique beam scanning is configured. This scanning mode adopts a 30° oblique incidence method, in which the sound beam is incident at an oblique angle to the sensitive area at the root of the weld. Utilizing the high reflection sensitivity of the oblique beam to micro-cracks, incomplete penetration, and root opening defects at the root, it specifically captures the corner reflection signals that only appear in the root area. This is the key signal source for identifying root cracks. The oblique incidence focusing rule needs to be precisely designed according to the pipe wall thickness, weld size, and depth range of the sensitive area at the root, including the sound beam incident angle, refraction path, and focal point distribution, to ensure that the sound beam can accurately cover the narrow area at the root of the weld and avoid the sound beam deviation leading to missed detection or weak signal.

[0073] After configuring the focusing rules for both scanning modes, the sampling frequency, gain, pulse repetition frequency, and detection mode of the phased array ultrasonic detector are uniformly set. The sampling frequency needs to be sufficiently high to ensure complete capture of the detailed features of the high-frequency ultrasonic echo signal, avoiding signal aliasing and distortion. The gain value needs to be reasonably set according to the actual detection distance and signal attenuation to ensure that weak reflection signals can be clearly identified while avoiding saturation and overflow of strong reflection signals. The pulse repetition frequency needs to balance scanning speed and signal stability, ensuring scanning efficiency while avoiding interference between adjacent pulse signals. Full-wave detection or forward detection is used to ensure the complete preservation of the amplitude, phase, and time characteristics of the echo signal. Next, the coordinates of the scanning start point and end point are set. The coordinate system is based on the central axis of the tube, with the circumferential direction representing the direction around the circumference of the tube and the axial direction representing the length of the tube. The scanning direction extends radially from the outer wall of the pipe towards the central axis. The scanning start point is selected at the zero-degree position of the weld circumference, and the scanning end point is set at the 360° position of the circumference to ensure that the probe can completely circle the weld, achieving a full circumferential scan without omissions. The scanning path is configured to scan around the weld circumference. The operator mounts the probe on a dedicated scanning bracket equipped with a precision circumferential encoder, axial displacement sensor, and radial displacement sensor, which can record the probe's position coordinates in three dimensions in real time, continuously, and with high precision. This ensures that each sampled signal corresponds to a unique and accurate spatial position, avoiding signal misalignment, imaging distortion, and misjudgment of defect locations due to positioning deviations. Once all parameters are configured, the equipment is connected normally, and the calibration is qualified, the system automatically sets the initialization completion flag to true and outputs a detection system ready signal, prompting the operator to begin the formal scanning.

[0074] The specific steps of S1 include:

[0075] S1.1: After receiving the detection system ready signal, move the phased array probe along the scanning path;

[0076] Furthermore, once the detection system is ready, the operator initiates the scanning control program, controlling the scanning bracket to move the phased array probe along the preset circumferential scanning path at a uniform speed and smoothly. During the probe's movement, the probe is always kept in close contact with the outer wall surface of the tube, with uniform pressure and good coupling. The coupling agent is a high-viscosity, low-attenuation special coupling agent adapted to the ultrasonic frequency, and is applied with uniform thickness, no bubbles, and no gaps to ensure that the ultrasonic waves can efficiently penetrate the workpiece surface, reduce reflection loss at the coupling interface, and ensure stable transmitted sound beam energy and clear received echo signals.

[0077] S1.2: At each sampling trigger position, a first-angle direct scanning signal beam and a second-angle oblique scanning signal beam are synchronously transmitted, and the corresponding first echo sequence and second echo sequence are received respectively;

[0078] Furthermore, the first-angle direct-scan signal beam is a 0° vertically incident ultrasonic beam, which is synchronously excited by the probe array elements according to the vertical focusing law. The beam is incident radially and vertically to the outer wall of the tube, penetrates the tube wall and then enters the weld area. When it encounters reflective surfaces such as the base material interface, weld fusion line, internal defects, and tube sheet interface, it generates reflected echoes. The reflected echoes return along the original path and are received by the probe array elements to form the first echo sequence. This sequence contains information such as the overall weld structure, base material boundary, fusion line reflection, and reflection of large internal defects. The signal coverage is wide and reflects the overall structural characteristics.

[0079] Furthermore, the second-angle oblique scanning signal beam is a 30° oblique incident ultrasonic beam, which is excited by the probe array elements in groups according to the oblique focusing law. The beam is incident on the outer wall of the pipe at a 30° oblique angle, and after being refracted by the pipe wall, it is incident on the sensitive area at the root of the weld at a specific angle. This area is a high-incidence area of ​​cracks. The oblique beam has extremely high reflection sensitivity to root micro-opening cracks, sharp corner defects, incomplete penetration, etc. When encountering such defects, strong corner reflection echoes are generated. The reflected echoes return along the refraction path and are received by the probe array elements to form a second echo sequence. This sequence mainly contains the reflection signal of the weld root area and has a specific identification capability for root micro-defects.

[0080] Furthermore, after the two sound beams are emitted, the receiving channel of the detector is turned on simultaneously to receive the first echo sequence and the second echo sequence respectively. During the receiving process, the signal is amplified, filtered, and digitally converted in real time to convert the analog ultrasonic signal into a digital signal, ensuring that the characteristics of the signal amplitude, time, and phase are completely preserved without distortion or loss.

[0081] S1.3: At the same sampling trigger position, the circumferential coordinate value of the phased array probe center is acquired by the encoder, the axial coordinate value is acquired by the axial displacement sensor, and the radial coordinate value is acquired by the radial displacement sensor.

[0082] Furthermore, the circumferential coordinate values ​​are acquired by a high-precision circumferential encoder. The encoder is linked with the circumferential drive mechanism of the scanning bracket, which can measure the angular position of the probe along the circumferential direction of the weld in real time. The measurement accuracy reaches the sub-degree level, which can accurately distinguish the small differences between adjacent circumferential positions, ensuring that the circumferential positioning is without deviation or jump. The circumferential angle of each sampling position can be uniquely and accurately recorded.

[0083] Furthermore, the axial coordinate values ​​are acquired by a high-precision axial displacement sensor, which is installed on the axial guide rail of the scanning bracket. The sensor measures the displacement of the probe along the length of the tube in real time with a measurement accuracy at the micrometer level. It can accurately capture the minute axial movement of the probe and avoid misalignment of the root sensitive area slice and misjudgment of the axial position of the crack due to axial positioning errors.

[0084] Furthermore, the radial coordinate values ​​are acquired by a high-precision radial displacement sensor. The sensor monitors the changes in the radial distance between the probe and the outer wall surface of the pipe in real time, compensating for radial position deviations caused by pipe roundness errors, surface flatness differences, and changes in probe contact pressure. This ensures that the radial distance from the probe center to the weld root is constant, and that the ultrasonic beam incident depth is stable and the sensitive area at the root is accurately covered.

[0085] S1.4: Pack the first echo sequence, second echo sequence, circumferential coordinate value, axial coordinate value and radial coordinate value obtained at the same sampling trigger position into a data frame, and store the data frames corresponding to each sampling trigger position as the original scan data file in ascending order of circumferential coordinate value.

[0086] S2: Based on the second angle oblique scanning signal and coordinate parameters, perform tomographic slicing along the sensitive area at the root of the weld to generate root slice signal data and construct a circumferential-depth map;

[0087] The specific steps of S2 include:

[0088] S2.1: Read the second echo sequence and circumferential coordinate values ​​of each data frame from the original scan data file;

[0089] Furthermore, the system reads the stored raw scan data file, parses the header information of each data frame sequentially according to the storage order, identifies and extracts the complete second echo sequence data and corresponding circumferential coordinate values ​​from each data frame. The reading process strictly follows the data format specifications to ensure accurate, complete, and error-free data parsing. The second echo sequence data contains the echo amplitude information at all time points after the obliquely incident ultrasonic beam at each sampling position, corresponding to the reflected signals at different depths along the entire propagation path of the ultrasonic beam from transmission to reception; the circumferential coordinate values ​​are the values ​​of the second echo sequence data at that sampling position. The precise circumferential angle corresponding to the center is used to identify the circumferential distribution position of the signal. During the reading process, the system performs preliminary verification of the data, checking whether the second echo sequence data is complete, whether there are any abnormal abrupt changes, or whether there are any missing values, and checking whether the circumferential coordinate values ​​are continuous, whether there are any jumps, or whether they exceed the preset range. Once abnormal data is found, it is marked and removed to avoid abnormal data interfering with subsequent tomographic slice results, ensuring that the read data is reliable and effective. After reading all data frames, the system establishes a one-to-one correspondence between the second echo sequence data and the circumferential coordinate values, forming an ordered set of signal-position data pairs.

[0090] S2.2: Set the starting and ending depth values ​​of the sensitive zone at the weld root to determine the slice depth range;

[0091] Furthermore, the depth range of the sensitive area at the weld root is precisely set according to actual working condition parameters such as pipe wall thickness, weld structure size, tube sheet connection form, and ultrasonic beam refraction path. This ensures that the slicing depth range can completely cover all areas at the weld root that may develop cracks, without omitting the root crack sensitive depth or including irrelevant signals from non-root areas, reducing invalid data interference, and improving slicing processing efficiency and signal targeting.

[0092] Furthermore, the starting depth value is set as the minimum radial depth at which the ultrasonic beam enters the sensitive area at the root of the weld, corresponding to the junction between the inner wall of the pipe and the root of the weld. This junction is the starting point for the root crack initiation and also the depth at which the oblique beam begins to enter the sensitive area. The ending depth value is set as the maximum radial depth at which the ultrasonic beam leaves the sensitive area at the root of the weld, corresponding to the connection interface between the root of the weld and the tube sheet. This connection interface is the limit depth to which the root crack may extend.

[0093] Furthermore, after determining the starting depth value and the ending depth value, the radial depth range between the two is the slice depth range of the sensitive area at the root of the weld. The slice depth range limits the signal extraction range of the tomographic slice processing, ensuring that only the oblique echo signal in the sensitive area at the root is extracted, effectively shielding the reflected signal from areas outside the root, and improving the identification of root crack signals.

[0094] S2.3: The sliding window method is used to extract the echo amplitude values ​​of each second echo sequence point by point along the depth direction. The sliding window method is the prior art in this field and is not an inventive solution of this application. It will not be described in detail here.

[0095] Furthermore, for each second echo sequence, the sliding window starts from the beginning of the sequence and slides point by point along the depth direction. At each step, it is determined whether the depth value corresponding to the center of the window is within the slice depth range: if it is within the slice depth range, the echo amplitude value corresponding to that depth point is extracted and the valid signal is retained; if it is outside the slice depth range, the depth point is skipped and the invalid signal is not extracted.

[0096] S2.4: Arrange the echo amplitude values ​​corresponding to each depth value under the same circumferential coordinate value in ascending order of depth to form the root signal vector of that circumferential coordinate value;

[0097] Further, the specific steps of S2.4 include: First, the system filters out the effective echo amplitude values ​​corresponding to all sampling positions with the same circumferential coordinate values. These effective echo amplitude values ​​all come from the root sensitive area reflection signals at different depth positions within the same circumferential angle, belonging to the depth direction signal set of the same circumferential position. Then, the filtered effective echo amplitude values ​​are rearranged in ascending order of radial depth values ​​to form ordered echo amplitude values, ensuring that the arrangement order is completely consistent with the actual physical depth direction. Positions with smaller depths are arranged at the front of the vector, and positions with larger depths are arranged at the back of the vector. Each arrangement position corresponds to a unique and accurate radial depth value. After sorting, the ordered echo amplitude values ​​corresponding to all depths under the same circumferential coordinate value are combined to form a one-dimensional array, which is the root signal vector corresponding to that circumferential coordinate value. Each element in the root signal vector contains two related pieces of information: the element position corresponds to the radial depth value, and the element value corresponds to the echo amplitude value at that depth position. The root signal vector completely reflects the reflection intensity distribution characteristics of the weld root sensitive area from shallow to deep under that circumferential angle.

[0098] S2.5: Combine the root signal vectors of all circumferential coordinate values ​​in ascending order of circumferential coordinate values ​​to generate root slice signal data; the root slice signal data includes slice depth range, circumferential angle distribution, echo position and echo amplitude;

[0099] Furthermore, the specific steps of S2.5 include: First, the system collects root signal vectors generated at all sampling locations. Each root signal vector corresponds to a unique circumferential coordinate value. The length of the root signal vector is uniformly equal to the number of depth points within the slice depth range, ensuring that all vector dimensions are consistent and can be directly spliced. Subsequently, in ascending order of circumferential coordinate values, all root signal vectors are sequentially arranged and spliced ​​as row vectors of a matrix. Vectors with smaller circumferential angles are arranged at the top of the matrix, and vectors with larger circumferential angles are arranged at the bottom of the matrix. The row dimension of the matrix corresponds to the circumferential angle distribution of the weld, and the column dimension of the matrix corresponds to the radial depth distribution. The element at the intersection of the row and column is the echo amplitude at the corresponding circumferential and depth positions. The generated root slice signal data is a structured, standardized, and directly imageable two-dimensional data set, which fully contains the slice depth range information of the root sensitive area, the full circumferential angle distribution information, the echo position information of each signal point, and the echo amplitude information of each position. The data dimensions are clear, the information is complete, and the correlation is strong.

[0100] S2.6: Perform grayscale mapping processing on the root slice signal data to construct a circumferential-depth map with circumferential coordinates as the horizontal axis and depth as the vertical axis.

[0101] Furthermore, the specific steps of S2.6 include: the system first determines the maximum and minimum values ​​of the echo amplitude in the root slice signal data, establishes a linear mapping relationship between amplitude values ​​and gray levels, where a larger amplitude signal corresponds to a brighter pixel gray level, closer to white, and a smaller amplitude signal corresponds to a darker pixel gray level, closer to black, with intermediate amplitudes corresponding to transitional gray levels. The mapping process ensures that amplitude differences are proportional to gray level differences, and signal strength changes can be clearly distinguished by gray level brightness; subsequently, according to the correspondence between matrix rows and columns and image pixel coordinates, each of the root slice signal data... The amplitude values ​​of the elements are converted into pixel values ​​of the corresponding gray levels. The matrix rows correspond to the horizontal axis of the image, with the angle increasing from left to right in the circumferential direction. The matrix columns correspond to the vertical axis of the image, with the depth increasing from top to bottom. After the pixel values ​​are filled, a complete grayscale image is formed. The final circumferential-depth map is a two-dimensional visualization image. The horizontal axis is the circumferential coordinate of the weld, and the vertical axis is the radial depth. The grayscale brightness of the image directly reflects the echo reflection intensity at the corresponding position: the grayscale of normal base material and complete weld area is uniformly dark; the grayscale of root cracks, sharp corner defects, and corner reflection positions is significantly brighter, forming high-brightness signal areas.

[0102] S3: Compare the root slice signal data with the first angle direct scan signal to lock the corner reflection candidate signal and its coordinates that only exist in the sensitive area of ​​the weld root;

[0103] The specific steps of S3 include:

[0104] S3.1: Read the first echo sequence of each data frame from the raw scan data file;

[0105] S3.2: Extract the echo amplitude values ​​within the depth range of the sensitive area at the root of the weld from each first echo sequence to form a first amplitude value set;

[0106] Furthermore, the system processes each first echo sequence along the depth direction according to the slice depth range of the sensitive area at the weld root set in S2.2. It uses the same sliding window method and point-by-point extraction rules as the oblique beam signal truncation to accurately extract the echo amplitude values ​​of each first echo sequence whose depth values ​​are within the slice depth range. After extraction, the system summarizes and integrates the direct echo amplitude values ​​of the root sensitive area corresponding to all sampling positions to form a first amplitude value set. The first amplitude value set contains the direct scan echo amplitude values ​​of all positions within the depth range of the root sensitive area and the entire circumferential range of the weld. The data dimension and spatial distribution are completely consistent with the root slice signal data, and can be directly compared point-to-point with the oblique beam signal.

[0107] S3.3: Extract echo amplitude values ​​with the same circumferential coordinate values ​​and the same depth range from the root slice signal data to form a second amplitude value set;

[0108] Furthermore, the system extracts echo amplitude values ​​from the root slice signal data generated in S2.5 that perfectly match the spatial locations of the first amplitude value set, ensuring that the second amplitude value set is completely consistent with the first amplitude value set in terms of circumferential distribution, depth range, and spatial location, achieving a strict one-to-one correspondence. During the extraction process, the system selects oblique echo amplitude values ​​corresponding to each circumferential position and each depth point from the matrix according to the principle of matching circumferential coordinate values ​​one by one and corresponding depth range points one by one. The selection rules, extraction range, and data dimensions are completely synchronized with the first amplitude value set. After the selection is completed, the oblique echo amplitude values ​​of all matching positions are summarized and integrated to form the second amplitude value set. This set is the same size as the first amplitude value set, and the elements correspond one-to-one, with each element corresponding to the oblique echo amplitude value at the same circumferential position and the same depth position.

[0109] S3.4: Compare the first amplitude value and the second amplitude value corresponding to the same spatial location point one by one, and mark the spatial location points that meet the preset conditions as single-angle unique points. The preset conditions are: the first amplitude value is less than the first preset threshold, and the second amplitude value is greater than the second preset threshold.

[0110] Furthermore, the specific steps in S3.4 include:

[0111] (1) Confirm that the first amplitude value set and the second amplitude value set are arranged in the same spatial position order. Each data element in each set corresponds to a unique spatial position point, and the elements between the two sets correspond one-to-one in position order, without any misalignment or omission.

[0112] (2) Set a first preset threshold and a second preset threshold. The first preset threshold is the weak reflection judgment threshold of the direct scanning signal. It needs to be calibrated according to the background noise level of the direct scanning of the crack-free standard test block. In this embodiment, it is set to 20, which means that when the amplitude value of the direct scanning signal is less than 20, it is judged as a weak reflection signal. The second preset threshold is the strong reflection judgment threshold of the oblique scanning signal. It needs to be calibrated according to the background noise level of the oblique scanning of the crack-free standard test block and the minimum effective reflection signal intensity of the crack. In this embodiment, it is set to 60, which means that when the amplitude value of the oblique scanning signal is greater than 60, it is judged as a strong reflection signal. At the same time, initialize the spatial location point traversal counter. The initial value of the counter is set to 1, which is used to mark the sequence number of the spatial location point currently being processed. The maximum value of the counter is equal to the total number of spatial location points contained in a single amplitude value set, ensuring that the traversal process can cover all location points without omission.

[0113] (3) Traverse all spatial location points point by point, and extract the first amplitude value and the second amplitude value corresponding to each location point in turn. The traversal process starts from the first spatial location point with the index 1 and proceeds in order from small to large. Each traversal only processes one spatial location point to ensure that the processing process is orderly and clear. For the spatial location point currently traversed, read the amplitude value of the direct scanning signal corresponding to the location point from the set of first amplitude values ​​and record it as the current first amplitude value. At the same time, read the amplitude value of the oblique scanning signal corresponding to the location point from the set of second amplitude values ​​and record it as the current second amplitude value. The reading process is strictly executed according to the position index correspondence to ensure that the two amplitude values ​​read are indeed from the same spatial location point.

[0114] (4) For the current first amplitude value and the current second amplitude value corresponding to the current spatial position point, perform two preset condition judgment operations in sequence. The judgment process is strictly executed according to the condition order. Only when the previous condition is satisfied will the next condition be judged. If any condition is not satisfied, the spatial position point is skipped directly and not marked. Specifically, the first judgment condition is that the current first amplitude value is less than the first preset threshold. The current first amplitude value is compared with the first preset threshold set to 20 to determine whether the current first amplitude value is less than 20. The second judgment condition is that the current second amplitude value is greater than the second preset threshold. The current second amplitude value is compared with the second preset threshold set to 60 to determine whether the current second amplitude value is greater than 60.

[0115] (5) Based on the judgment results of the two conditions, perform the corresponding marking operation to determine whether the current spatial position point is a single-angle unique point. If the current spatial position point satisfies both judgment conditions at the same time, that is, the current first amplitude value is less than 20 and the current second amplitude value is greater than 60, then immediately mark the spatial position point as a single-angle unique point. After marking, record the circumferential coordinate information and depth coordinate information corresponding to the position point for subsequent use. If the current spatial position point does not satisfy any judgment condition, that is, the current first amplitude value is greater than or equal to 20, or the current second amplitude value is less than or equal to 60, then do not perform any marking operation, directly determine that the position point is not a single-angle unique point, skip the position point, and continue to process the next position point;

[0116] (6) After processing the current spatial location point, increment the value of the traversal counter by 1, and determine whether the current value of the counter exceeds the total number of spatial location points. If the counter value does not exceed the total number, it means that there are still unprocessed spatial location points. Return to (3) and continue to traverse the next spatial location point, repeating the subsequent steps. If the counter value is equal to or exceeds the total number, it means that all spatial location points have been traversed and judged. At this time, end the point-by-point comparison process, summarize all spatial location information marked as single-angle unique points, and complete the entire single-angle unique point screening work.

[0117] S3.5: Retain the circumferential coordinate values ​​and depth values ​​corresponding to all single-angle unique points as corner reflection candidate signal positions, and use the corner reflection candidate signal positions and their corresponding circumferential coordinate values ​​and depth values ​​as corner reflection candidate signals and their coordinates.

[0118] Furthermore, after comparing and marking all spatial locations, the system filters and retains the location information of all points marked as unique single-angle reflection points. It extracts the precise circumferential coordinates and radial depth values ​​corresponding to each unique single-angle reflection point; these locations are the initially identified corner reflection candidate signal locations. Each corner reflection candidate signal location contains unique circumferential coordinates and radial depth values, accurately identifying the spatial location of suspected crack corner reflection signals within the root sensitive area. The corresponding signal is the corner reflection candidate signal. The system binds and stores the corner reflection candidate signals with their corresponding coordinate information, forming a corner reflection candidate signal set. This provides accurate location and signal data for subsequent corner reflection candidate signal clustering and merging, axial-depth slice imaging, and crack feature verification.

[0119] Furthermore, after retaining the circumferential coordinates and depth values ​​corresponding to all unique points at a single angle as candidate locations for corner reflection signals, the process further includes performing a corner reflection candidate signal clustering and merging step:

[0120] A1: Obtain the circumferential coordinates and depth values ​​of all single-angle unique points, with each single-angle unique point serving as a spatial data point;

[0121] Furthermore, due to the spatial diffusion, signal noise interference, and sampling interval dispersion of ultrasonic detection signals, the corner reflection candidate signals generated by a single crack may form multiple adjacent single-angle unique points in space. These points actually correspond to the same crack defect. If they are directly treated as independent candidate signals, it will lead to redundancy in the number of candidate signals, dispersion in location, and complexity in subsequent analysis. Therefore, it is necessary to treat all single-angle unique points as independent spatial data points and perform clustering and merging preprocessing. Multiple adjacent single-angle unique points corresponding to the same crack are merged into a candidate defect cluster, eliminating isolated noise points, simplifying the subsequent analysis process, and improving the accuracy of candidate signals. The system obtains the complete location information of all single-angle unique points. Each single-angle unique point contains independent circumferential coordinate values ​​and radial depth values. In two-dimensional space, each single-angle unique point corresponds to a unique spatial data point, and all spatial data points together constitute the candidate signal spatial dataset.

[0122] A2: A density-based spatial clustering algorithm is used to cluster the spatial data points. The density-based spatial clustering algorithm is existing technology in this field and is not an inventive solution of this application. It will not be described in detail here.

[0123] Furthermore, the core logic of the density-based spatial clustering algorithm is as follows: regions with densely distributed data points belong to the same cluster, while regions with sparse and isolated data points belong to noise points. All spatial data points are traversed, and each unvisited data point is used as the core to search for all data points within its neighborhood. If the number of data points in the neighborhood meets the density requirement, these points are merged into a candidate defect cluster; if the number of data points in the neighborhood is insufficient or there are no neighboring points, they are marked as noise points and removed.

[0124] Furthermore, the density-based spatial clustering algorithm can effectively merge multiple adjacent single-angle unique points generated by the same crack into a candidate defect cluster, restore the actual spatial distribution of the crack, and at the same time eliminate isolated single-angle unique points caused by signal noise and sampling errors, thereby reducing invalid candidate signals.

[0125] A3: Set a neighborhood radius threshold and a minimum number of points threshold to classify single-angle unique points whose point spacing meets the neighborhood radius threshold and whose number of points in the neighborhood is greater than or equal to the minimum number of points threshold into the same candidate defect cluster;

[0126] Furthermore, the specific steps in A3 include:

[0127] (1) Extract all single-angle unique points from the screening results. Each single-angle unique point contains the corresponding circumferential position information and radial depth information. All coordinate information is uniformly converted to the actual physical distance unit to avoid distance calculation deviation due to inconsistent units.

[0128] (2) Set a neighborhood radius threshold to define the maximum spatial distance between two single-angle unique points that can be regarded as adjacent points. It is the core parameter for determining whether points belong to the same local area. In this embodiment, it is set to 1.5mm, which means that when the spatial distance between any two single-angle unique points does not exceed 1.5mm, the two points are determined to be within the adjacent neighborhood.

[0129] (3) Set a minimum number of points threshold to define the minimum number of single-angle unique points required to constitute a valid candidate defect cluster. This is used to distinguish between valid defect areas and isolated noise points. In this embodiment, it is set to 3, which means that a local neighborhood contains at least three or more single-angle unique points before it can be identified as a valid candidate defect cluster. Otherwise, it is identified as an isolated noise area.

[0130] (4) Assign an independent processing status identifier to each valid single-angle unique point. The initial status is uniformly set to unvisited, which means that the point has not yet participated in the neighborhood search and clustering judgment. At the same time, initialize the candidate defect cluster number counter. The initial value of the counter is set to 1, which is used to assign a unique number to each newly identified candidate defect cluster to ensure that different defect clusters can be clearly distinguished.

[0131] (5) Traverse all single-angle unique points that are in an unvisited state, and select unvisited points as the current core points in turn. Each time, only one unvisited point is selected as the core point. After the selection is completed, the processing status of the core point is immediately updated to visited to prevent the point from being selected repeatedly in the future.

[0132] (6) Taking the current core point as the center, traverse all valid single-angle unique points, calculate the spatial distance between each point to be judged and the core point, and determine whether the distance meets the neighborhood radius threshold requirement. The spatial distance calculation comprehensively considers the distance component in the circumferential direction and the distance component in the depth direction. After the calculation is completed, the actual spatial distance value between the two points is obtained. The actual spatial distance value is compared with the neighborhood radius threshold set to 1.5mm. If the actual spatial distance is less than or equal to 1.5mm, it is determined that the point to be judged is within the neighborhood of the core point and is included in the candidate neighborhood point set. If the actual spatial distance is greater than 1.5mm, it is determined that the point to be judged is not within the neighborhood and is directly excluded and not included in the candidate neighborhood point set.

[0133] (7) Count the total number of single-angle unique points in the candidate neighborhood point set, compare the number with the minimum point threshold set to 3, and determine whether the current neighborhood range meets the basic conditions for forming a candidate defect cluster. If the number of points in the candidate neighborhood point set is greater than or equal to 3, the neighborhood range is determined to be a valid defect area, and the neighborhood range needs to be further expanded and associated points merged. If the number of points in the candidate neighborhood point set is less than 3, the core point is determined to be an isolated noise point and does not constitute a valid candidate defect cluster. The point is skipped directly, and the next unvisited point is traversed.

[0134] (8) Select each unvisited neighboring point from the candidate neighboring point set in turn, update it as a new core point, and repeat (6)-(7) to continuously search for single-angle unique points that are adjacent to the current neighboring point and meet the distance condition until no new neighboring points that meet the condition can be found. During the entire expansion process, the processing status of all single-angle unique points included in the neighborhood range is updated to visited to avoid repeated inclusion in different defect clusters.

[0135] (9) All the associated single-angle unique points collected after the neighborhood expansion are uniformly classified into the same candidate defect cluster, and the defect cluster is assigned the number corresponding to the current candidate defect cluster number counter. After the assignment is completed, the value of the candidate defect cluster number counter is incremented by 1 to prepare for the numbering of the next newly identified candidate defect cluster.

[0136] (10) Repeat (5)-(9) to continue traversing the remaining unvisited single-angle unique points and complete the division of all valid candidate defect clusters in turn until the processing status of all single-angle unique points is updated to visited. After the traversal is completed, summarize all divided candidate defect clusters. Each defect cluster contains a set of single-angle unique points that are mutually related and meet the point number requirements. Isolated points that are not included in any defect cluster are judged as noise points.

[0137] A4: Calculate and extract the geometric center coordinates of each candidate defect cluster, use them as the corner reflection candidate signal position of the corresponding candidate defect cluster, and remove isolated single-angle unique points that cannot be classified into any candidate defect cluster as noise points.

[0138] Furthermore, the geometric center coordinates are calculated as follows: the arithmetic mean of the circumferential coordinates of all single-angle unique points within the candidate defect cluster is taken to obtain the circumferential coordinates of the cluster center; the arithmetic mean of the radial depth values ​​of all single-angle unique points within the cluster is taken to obtain the depth value of the cluster center. The combination of the two is the geometric center coordinates of the candidate defect cluster. The geometric center coordinates represent the center position of the suspected crack corner reflection signal, which is accurate and highly representative.

[0139] Furthermore, for isolated single-angle unique points that cannot be classified into any candidate defect cluster, due to their sparse spatial distribution and lack of adjacent effective signal points for support, they are likely invalid signal points caused by ultrasonic signal noise, sampling errors, and small clutter reflections, and do not possess crack signal characteristics. Therefore, they are directly eliminated as noise points and not included in the corner reflection candidate signal range, effectively reducing invalid candidate signals and avoiding subsequent misjudgments.

[0140] A5: Outputs the circumferential coordinates and depth values ​​corresponding to the position of each representative corner reflection candidate signal, replacing the original single-angle unique point coordinates.

[0141] S4: At the coordinates corresponding to the corner reflection candidate signal, perform tomographic slicing along the axial-depth direction to generate an axial-depth map. Determine that the corner reflection candidate signal is a fixed-position corner reflection feature signal formed by the outer wall of the pipe and the crack. Verify that the corner reflection position, corner reflection amplitude and depth meet the preset crack judgment conditions, and output a crack identification completion signal.

[0142] The step of performing tomographic slicing along the axial-depth direction at the coordinates corresponding to the corner reflection candidate signal to generate an axial-depth map includes:

[0143] S4.1: Obtain the circumferential coordinates and depth values ​​corresponding to the corner reflection candidate signals;

[0144] It should be noted that the obtained circumferential coordinate values ​​determine the circumferential position of the slicing process, ensuring that slicing is only performed in the area near the circumferential angle corresponding to the candidate signal; the obtained depth values ​​determine the depth center position of the slicing process, ensuring that the slicing focuses on covering the depth range of the root sensitive area corresponding to the candidate signal, ensuring that the slicing area accurately focuses on the suspected crack location, without deviation or omission.

[0145] S4.2: Extract all data frames whose circumferential coordinate values ​​are within the preset neighborhood range of the corner reflection candidate signal's circumferential coordinate values ​​from the original scan data file, and read the second echo sequence and axial coordinate value of each extracted data frame;

[0146] Furthermore, the specific steps of S4.2 include:

[0147] (1) Read the circumferential coordinate reference value corresponding to the current corner reflection candidate signal to be processed from the clustering processing result, which represents the core circumferential position of the candidate defect;

[0148] (2) The preset neighborhood range needs to be calibrated by combining the actual circumferential coverage width of the phased array ultrasonic probe, the scanning sampling interval and the circumferential signal diffusion characteristics of the crack. In this embodiment, it is set to 5°, which means that the candidate signal circumferential coordinate reference value is the center, extending 5° in the direction of decreasing circumferential angle and 5° in the direction of increasing circumferential angle, forming a circumferential neighborhood range with a span of 10°.

[0149] (3) Based on the circumferential coordinate reference value of the corner reflection candidate signal, subtract the preset neighborhood range value set to 5° from the reference value to obtain the lower limit value of the circumferential coordinate, which represents the minimum circumferential angle boundary of the screening interval; then add the preset neighborhood range value set to 5° to the reference value to obtain the upper limit value of the circumferential coordinate, which represents the maximum circumferential angle boundary of the screening interval; after the calculation is completed, it is clear that the circumferential coordinate values ​​of all data frames to be extracted must be greater than or equal to the lower limit value of the circumferential coordinate, and less than or equal to the upper limit value of the circumferential coordinate.

[0150] (4) Open the original scan data file containing complete scan data, read the header information of all data frames in sequence according to the preset data frame storage order in the file, establish a complete data frame index list, each index item corresponds to an independent data frame, record the storage location of the data frame in the file, the circumferential coordinate value identifier and the data frame integrity verification information; initialize the data frame traversal counter, the initial value of the counter is set to 1, which is used to mark the sequence number of the data frame currently being checked, and the maximum value of the counter is equal to the total number of data frames contained in the original scan data file, to ensure that the traversal process covers all data frames without omission;

[0151] (5) Start traversing all data frames in the original scan data file frame by frame, extract the circumferential coordinate value of each data frame in turn, and perform interval matching judgment. The traversal process is carried out in the order of the data frame index list. Only one data frame is processed at a time to ensure that the processing process is orderly and clear. For the data frame currently traversed, read the actual circumferential coordinate value corresponding to the data frame from the data frame header information.

[0152] (6) Compare the circumferential coordinate value of the current data frame with the lower limit and upper limit of the circumferential coordinate to determine whether the data frame is within the preset neighborhood. If the circumferential coordinate value of the current data frame is greater than or equal to the lower limit of the circumferential coordinate and is simultaneously less than or equal to the upper limit of the circumferential coordinate, then the data frame is determined to be the target data frame and included in the extraction range. If the circumferential coordinate value of the current data frame is less than the lower limit of the circumferential coordinate or greater than the upper limit of the circumferential coordinate, then the data frame is determined to be outside the preset neighborhood and is skipped directly and not included in the extraction range.

[0153] (7) Extract the complete second echo sequence corresponding to the oblique scanning mode from the signal data segment of the data frame. The second echo sequence contains the oblique ultrasonic echo amplitude information corresponding to all depths at the current scanning position. At the same time, extract the actual axial coordinate value corresponding to the data frame from the position data segment of the data frame. This value accurately identifies the scanning position of the probe in the tube length extension direction. The reading process must ensure that the second echo sequence is complete and without missing parts, and the axial coordinate value is accurate and without deviation.

[0154] (8) The second echo sequence read is used as the signal data part and the axial coordinate value is used as the position data part. The two are bound one-to-one to ensure that the corresponding oblique echo signal can be quickly matched by the axial coordinate value when used next.

[0155] (9) After completing the inspection and processing of the current data frame, increment the value of the traversal counter by 1, and determine whether the current value of the counter exceeds the total number of data frames in the original scanned data file. If the counter value does not exceed the total number of data frames, it means that there are still unchecked data frames. Return to (5) to continue traversing the next data frame and repeat the subsequent inspection, judgment, reading and storage steps. If the counter value is equal to or exceeds the total number of data frames, it means that all data frames have been traversed and processed. At this time, the frame-by-frame filtering process ends. All associated data items stored in the temporary cache area are the target data frame set extracted this time. The temporary cache area adopts sequential storage method and is arranged according to the original circumferential order of the target data frames to maintain the order of the data.

[0156] S4.3: With the depth value of the corner reflection candidate signal as the center, set a depth slice window. For each extracted data frame, extract the echo amplitude value sequence within the depth range of the depth slice window from the second echo sequence of the data frame, and associate and store the echo amplitude value sequence extracted from each data frame with the axial coordinate value of the data frame.

[0157] Furthermore, the specific steps of S4.3 include:

[0158] (1) Read the radial depth value corresponding to the current candidate signal to be processed from the clustering results of the corner reflection candidate signal. This value is the center reference of the depth slice window and represents the core position of the candidate defect in the radial depth direction.

[0159] (2) The half-width of the depth slice window needs to be calibrated and set in combination with the radial depth resolution of the phased array ultrasonic detection, the actual range of the sensitive area at the root of the weld and the signal distribution characteristics in the direction of crack depth. In this embodiment, it is set to 1.2mm, which means that the depth reference value is centered, extending 1.2mm in the shallow depth direction and 1.2mm in the deep depth direction, forming a depth slice window with a span of 2.4mm.

[0160] (3) Based on the depth reference value of the corner reflection candidate signal, subtract the half-width value of the depth slice window set to 1.2mm from the reference value to obtain the shallow depth boundary value of the depth slice window, which represents the minimum radial depth boundary of the slice window; then add the half-width value of the depth slice window set to 1.2mm to the reference value to obtain the deep depth boundary value of the depth slice window, which represents the maximum radial depth boundary of the slice window; after the calculation is completed, it is clear that all echo amplitude values ​​to be intercepted must be greater than or equal to the shallow depth boundary value, and less than or equal to the deep depth boundary value.

[0161] (4) Retrieve all the target data frame sets obtained by filtering. Each data frame in the target data frame set contains a complete second echo sequence and the corresponding axial coordinate value. Initialize the data frame traversal counter. The initial value of the counter is set to 1, which is used to mark the sequence number of the data frame currently being processed. The maximum value of the counter is equal to the total number of data frames contained in the target data frame set.

[0162] (5) For the data frame currently being traversed, read the second echo sequence stored inside the data frame completely. The second echo sequence contains the amplitude information of oblique ultrasonic echoes corresponding to all radial depths from shallow to deep.

[0163] (6) Initialize the echo sequence depth traversal counter, which is used to traverse each depth position in the second echo sequence point by point. The initial value of the counter is set to 1, and the maximum value of the counter is equal to the total number of depth sampling points contained in the second echo sequence, to ensure that the traversal covers all depth positions in the sequence without omission. At the same time, initialize the temporary storage area of ​​the current data frame's intercepted amplitude value, which is used to temporarily store the valid echo amplitude value intercepted this time. In the initial state, the temporary storage area of ​​the intercepted amplitude value is empty.

[0164] (7) Starting from the sequence number corresponding to the depth traversal counter of the echo sequence, read the radial depth value corresponding to the current depth sampling point, and then read the echo amplitude value corresponding to the depth position. Compare the read radial depth value with the shallow depth boundary value and deep depth boundary value calculated in (3) to determine whether the depth position is within the depth slice window.

[0165] (8) If the radial depth value of the current depth sampling point is greater than or equal to the shallow depth boundary value and simultaneously less than or equal to the deep depth boundary value, then the echo amplitude value is determined to be a valid signal and stored in the temporary storage area of ​​the truncated amplitude value of the current data frame; if the radial depth value of the current depth sampling point is less than the shallow depth boundary value or greater than the deep depth boundary value, then the echo amplitude value is determined to be an invalid signal and is skipped directly without being stored in the temporary storage area of ​​the truncated amplitude value.

[0166] (9) Update the echo sequence depth traversal counter. After completing the judgment and processing of the current depth sampling point, increment the counter value by 1. Determine whether the current counter value exceeds the total number of depth sampling points of the second echo sequence. If the counter value does not exceed the total number of depth sampling points, it means that there are still unprocessed depth sampling points. Return to (7) and continue to traverse the next depth sampling point. If the counter value is equal to or exceeds the total number of depth sampling points, it means that the second echo sequence of the current data frame has completed full depth traversal. At this time, extract all echo amplitude values ​​stored in the amplitude value temporary storage area, which is the echo amplitude value sequence within the depth slice window range corresponding to the current data frame.

[0167] (10) Read the pre-stored axial coordinate value from the current data frame. This value accurately identifies the scanning position of the probe in the direction of tube length extension. During the reading process, it is necessary to confirm that the value is accurate and without deviation. The echo amplitude value sequence in the temporary storage area of ​​the intercepted amplitude value is used as signal data, and the axial coordinate value is used as position data. The two are bound together as a set of associated data.

[0168] (11) The global result buffer adopts a sequential storage structure. The associated data is stored in the original traversal order of the target data frame to ensure that the data order in the global result buffer is consistent with the scanning axis position order and there is no disorder. After the storage is completed, the temporary storage area of ​​the current data frame's intercepted amplitude value is cleared to prepare for processing the next data frame.

[0169] (12) Update the data frame traversal counter. After completing the current data frame capture and associated storage, increment the data frame traversal counter value by 1. Determine whether the current counter value exceeds the total number of data frames in the target data frame set. If the counter value does not exceed the total number of data frames, it means that there are still unprocessed target data frames. Return to (5) and continue traversing the next data frame. If the counter value is equal to or exceeds the total number of data frames, it means that all target data frames have been processed. At this time, all associated data stored in the global result buffer is the complete result of this depth slice capture and associated storage.

[0170] S4.4: Arrange the echo amplitude value sequence after all associated storage in ascending order of axial coordinate values ​​to generate an axial-depth signal data matrix, and perform grayscale mapping processing on the axial-depth signal data matrix to construct an axial-depth map with axial coordinate as the horizontal axis and depth as the vertical axis.

[0171] Furthermore, the system integrates and sorts all the associated stored echo amplitude value sequences to construct a two-dimensional axial-depth signal data matrix. This achieves structured storage of the axial-depth two-dimensional signal distribution at candidate locations. Each element in the matrix corresponds to a unique axial position and radial depth position, and the element value is the echo amplitude value at the corresponding position. The specific process includes: First, the system collects all echo amplitude value sequences corresponding to all axial positions. The length of each sequence is uniformly equal to the number of depth points within the depth slice window, ensuring that all sequences have consistent dimensions and can be directly spliced. Then, in ascending order of axial coordinate values, all echo amplitude value sequences are sequentially arranged as row vectors of the matrix and spliced. Sequences with smaller axial positions are placed at the top of the matrix, and sequences with larger axial positions are placed at the bottom. Large sequences are arranged below the matrix, with the row dimension corresponding to the axial distribution and the column dimension corresponding to the radial depth distribution. The cross-element of the row and column is the echo amplitude at the corresponding axial and depth positions. After generating the axial-depth signal data matrix, the system uses the same grayscale mapping processing method as the circumferential-depth map to convert the echo amplitude values ​​in the matrix into pixels of different grayscale levels: the larger the amplitude, the brighter the grayscale, and the smaller the amplitude, the darker the grayscale. After mapping, an axial-depth map is constructed, with the horizontal axis representing the axial coordinate and the vertical axis representing the radial depth. This visually presents the reflection intensity distribution at different axial positions and depths of the candidate corner reflection signal location. The crack area will show a continuous, concentrated, and bright signal band, clearly reflecting the axial extension shape, length range, and reflection intensity changes of the crack.

[0172] The process of outputting the crack identification completion signal includes:

[0173] S4.5: Extract the distribution curve of the echo amplitude value along the axial direction corresponding to the depth value of the corner reflection candidate signal from the generated axial-depth signal data matrix;

[0174] Furthermore, the system locates the column corresponding to the depth value of the corner reflection candidate signal in the axial-depth signal data matrix, and extracts the echo amplitude value corresponding to all axial positions in the column. These values ​​reflect the root oblique reflection intensity at the same depth but different axial positions. The extracted echo amplitude values ​​are arranged in ascending order of axial coordinate values ​​to form a one-dimensional axial distribution curve. The horizontal axis of the curve is the axial coordinate, and the vertical axis is the echo amplitude value. The shape of the curve directly reflects the strength change of the corner reflection signal along the axial direction: the signal amplitude is high in the crack area and the curve forms a clear peak; the signal amplitude is low in the non-crack area and the curve is flat without peaks.

[0175] S4.6: Locate the global maximum point of the echo amplitude value in the distribution curve;

[0176] Furthermore, the global maximum point of the corner reflection candidate signal corresponds to the core location where the crack reflection is strongest and most concentrated, i.e., the crack center location. The signal amplitude, axial coordinate, and depth value at this location are key parameters for verifying crack characteristics and determining the existence of cracks.

[0177] Furthermore, the system iterates through all amplitude data points on the axial distribution curve, compares the amplitude of each point one by one, and selects the data point with the largest amplitude value, which is the global maximum point of the echo amplitude value. This point is the peak position of the corner reflection candidate signal intensity, representing the core characteristic position of the crack corner reflection, with the most significant, stable, and representative signal characteristics of the crack.

[0178] S4.7: Obtain the axial coordinate value corresponding to the global maximum point as the corner reflection axial position; obtain the depth value corresponding to the global maximum point as the corner reflection depth value; obtain the echo amplitude value of the global maximum point as the corner reflection amplitude.

[0179] S4.8: Retrieve preset crack judgment conditions, and calculate the absolute value of the axial deviation between the corner reflection axial position and the preset standard axial position, and calculate the absolute value of the depth deviation between the corner reflection depth value and the preset standard depth value; the preset crack judgment conditions include axial position deviation threshold, depth deviation threshold and amplitude threshold;

[0180] Furthermore, the specific steps of S4.8 include:

[0181] (1) Read the axial position of the corner reflection corresponding to the current corner reflection candidate signal to be determined from the axial depth analysis results. This value represents the position coordinate of the center of the crack corner reflection signal actually detected in the axial direction of the pipe. At the same time, read the corner reflection depth value corresponding to the current corner reflection candidate signal to be determined. This value represents the position coordinate of the center of the crack corner reflection signal actually detected in the radial depth direction.

[0182] (2) Retrieve the standard reference parameters corresponding to the preset crack judgment conditions, including the preset standard axial position and the preset standard depth value, as the reference basis for deviation calculation;

[0183] (3) Set the axial position deviation threshold, depth deviation threshold and amplitude threshold to form a complete set of judgment criteria. The axial position deviation threshold is used to define the maximum allowable deviation range between the actual corner reflection axial position and the standard axial position. It needs to be calibrated in combination with the axial positioning accuracy of phased array ultrasonic testing, the scanning sampling interval and the crack axial position signal fluctuation characteristics. In this embodiment, it is set to 0.8 mm. The depth deviation threshold is used to define the maximum allowable deviation range between the actual corner reflection depth value and the standard depth value. It needs to be calibrated in combination with the radial depth resolution of phased array ultrasonic testing, the thickness of the sensitive area at the weld root and the crack depth position signal fluctuation characteristics. In this embodiment, it is set to 0.6 mm. The amplitude threshold is used to define the minimum echo amplitude value required for effective identification of the crack corner reflection signal. It needs to be calibrated in combination with the background noise level of the crack-free standard test block and the minimum effective reflection signal intensity of the crack. In this embodiment, it is set to 60.

[0184] (4) Subtract the preset standard axial position from the axial position of the corner reflection to obtain the axial position difference, which represents the offset direction of the actual position relative to the standard position. Then take the absolute value of the axial position difference to eliminate the influence of positive and negative directions and obtain the absolute value of the axial deviation. The absolute value of the axial deviation only reflects the distance between the actual axial position and the standard axial position and does not reflect the offset direction. After the calculation is completed, confirm that the absolute value of the axial deviation is non-negative and the format is correct and without abnormalities.

[0185] (5) Subtract the preset standard depth value from the corner reflection depth value to obtain the depth difference value, which represents the offset direction of the actual depth relative to the standard depth. Then take the absolute value of the depth difference value to eliminate the influence of positive and negative directions and obtain the absolute value of the depth deviation. This value only reflects the distance between the actual depth value and the standard depth value and does not reflect the offset direction. After the calculation is completed, confirm that the absolute value of the depth deviation is a non-negative value and that the format is correct and there are no abnormalities.

[0186] (6) The calculated absolute values ​​of axial deviation and depth deviation, as well as the corner reflection characteristic parameters, the retrieved standard reference parameters, and the determined threshold parameters are uniformly organized into a structured data set to ensure that all data are complete and clearly correlated.

[0187] S4.9: When the absolute value of the axial deviation is less than or equal to the axial position deviation threshold, the absolute value of the depth deviation is less than or equal to the depth deviation threshold, and the corner reflection amplitude is greater than or equal to the amplitude threshold, the corner reflection candidate signal is determined to be a fixed position corner reflection feature signal formed by the outer wall of the pipe and the crack, which meets the crack judgment condition, and a crack identification completion signal carrying the corner reflection axial position, corner reflection depth value, and corner reflection amplitude is output.

[0188] Furthermore, the system compares the calculated absolute values ​​of axial deviation, depth deviation, and corner reflection amplitude with preset thresholds for axial position deviation, depth deviation, and amplitude, respectively. Only when all three conditions are met simultaneously can the corner reflection candidate signal be confirmed as a characteristic signal generated by a crack. The specific judgment logic is as follows: First, if the absolute value of axial deviation does not exceed the axial position deviation threshold, it indicates that the axial position of the corner reflection is stable within the crack characteristic position range without significant deviation, conforming to the fixed-position reflection characteristics of a crack. Second, if the absolute value of depth deviation does not exceed the depth deviation threshold, it indicates that the corner reflection depth matches the sensitive depth of the crack at the weld root, belonging to a high-incidence area of ​​cracks. Third, if the corner reflection amplitude is not lower than the amplitude threshold, it indicates that the intensity of the reflected signal meets the minimum requirements for corner reflection of a crack, eliminating weak reflection clutter interference. When all three conditions are met simultaneously, it is fully proven that the corner reflection candidate signal possesses the fixed-position corner reflection characteristics formed by the pipe outer wall and the crack, and is a characteristic signal generated by a real crack, rather than structural clutter, pseudo-signals, or non-crack reflection signals. Therefore, it is judged as a valid crack signal, satisfying the crack judgment conditions. Once the determination is complete, the system outputs a crack identification completion signal, which carries key feature parameters, including the axial position of the corner reflection, the corner reflection depth, and the corner reflection amplitude.

[0189] S5: Based on the crack identification completion signal, calculate the crack size using a preset corner reflection amplitude and depth calibration model, mark the calculation results on the circumferential-depth map and axial-depth map, mark the crack circumferential position, axial depth, radial depth and length data at the corresponding positions, and output the initial microcrack depth judgment mark, crack penetration status mark and grade judgment mark.

[0190] The step of calculating the crack size based on the crack identification completion signal using a preset corner reflection amplitude and depth calibration model includes:

[0191] S5.1: Receive the crack identification completion signal and extract the corner reflection amplitude and corner reflection depth values ​​carried in the crack identification completion signal;

[0192] S5.2: Load the preset corner reflection amplitude and depth calibration model; the corner reflection amplitude and depth calibration model is obtained by performing phased array detection experiments on tube sheet fillet weld test blocks containing artificial cracks of different depths, and then fitting and calibrating using the least squares method;

[0193] Furthermore, the construction process of the corner reflection amplitude and depth calibration model is as follows: Multiple sets of standard test blocks for tube sheet fillet welds are fabricated, identical in material, size, and weld structure to the actual workpiece. Artificial cracks of varying radial depths and axial lengths are machined at the weld root of the test blocks, ensuring precise and measurable crack dimensions. Using the same phased array ultrasonic testing system, parameter configuration, and scanning method as the actual testing, all test blocks containing artificial cracks are ultrasonically scanned, and the corner reflection amplitude and depth values ​​corresponding to each artificial crack are collected. The test data from all test blocks are compiled to form a dataset corresponding to the actual crack size and reflection characteristic parameters. The least squares method is used to fit and analyze the dataset, eliminating the influence of detection errors and noise interference, establishing the optimal fitting relationship between the corner reflection amplitude and depth values ​​and the crack radial depth, and finally obtaining the calibration model. After the model is constructed, it is tested with multiple sets of verification test blocks to ensure that the model's calculation accuracy meets engineering requirements and the error is within the allowable range. The least squares method is existing technology in this field and is not an inventive solution of this application; therefore, it will not be elaborated upon here.

[0194] S5.3: Substitute the extracted corner reflection amplitude and corner reflection depth values ​​into the corner reflection amplitude and depth calibration model to perform forward calculation, solve for the corresponding crack radial depth prediction value, and determine the crack radial depth prediction value as the radial depth component in the crack size calculation result.

[0195] Furthermore, the specific steps in S5.3 include:

[0196] (1) Read the corner reflection amplitude corresponding to the crack to be calculated from the crack identification completion signal. The corner reflection amplitude is the maximum echo amplitude of the crack corner reflection signal, which directly reflects the reflection intensity level of the crack. At the same time, read the corner reflection depth value corresponding to the crack to be calculated, which represents the core distribution position of the crack in the depth direction. After reading, confirm that the corner reflection amplitude is a valid value greater than zero and confirm that the corner reflection depth value is a reasonable physical depth value within the sensitive area of ​​the weld root.

[0197] (2) Retrieve and load the preset corner reflection amplitude and depth calibration model. This calibration model is obtained by conducting phased array ultrasonic testing experiments on tube sheet fillet weld test blocks with artificial cracks of different depths and fitting calibration using the least squares method. The model internally solidifies the quantitative mapping relationship between corner reflection amplitude, corner reflection depth and crack radial depth. The retrieval process reads the model file from the system model storage area. During loading, the integrity and consistency of the model file are automatically verified to confirm that the model is not damaged, not tampered with, and has no missing parameters. After loading, the model is in a ready state for executing forward calculation.

[0198] (3) Clarify the internal fitting parameters of the corner reflection amplitude and depth calibration model. All parameters are determined by standard test block experiments and least squares fitting. The values ​​are clear and solidified in the model. The calibration model adopts a binary linear fitting structure, which includes four core fitting parameters, namely amplitude coefficient, depth coefficient, constant term coefficient and fitting residual threshold. Among them, the amplitude coefficient is determined by fitting multiple sets of artificial crack test block detection data. In this embodiment, it is set to 0.02. The depth coefficient is determined by fitting the same batch of artificial crack test block detection data. In this embodiment, it is set to 0.85. The constant term coefficient is determined by the intercept calculation of the fitting process. In this embodiment, it is set to 0.03. The fitting residual threshold is set to 0.05.

[0199] (4) Verify whether the extracted corner reflection amplitude is a normalized value. If it is not normalized, the conversion is completed according to the system's preset normalization rule. The normalization rule establishes a linear mapping relationship based on the maximum and minimum values ​​of all echo amplitudes in the root slice signal data, and converts the original sampled amplitude to a dimensionless range of 0 to 100. Verify whether the unit of the corner reflection depth value is millimeters, confirm that the numerical accuracy is retained to two decimal places, adapt to the model input accuracy requirements, and after the format adaptation is completed, it can be directly sent to the calibration model.

[0200] (5) After the calibration model receives the corner reflection amplitude and corner reflection depth values ​​after the adaptation is completed, it performs calculations according to the internally fixed binary linear fitting logic. The calculation process first multiplies the corner reflection amplitude with the amplitude coefficient set to 0.02 to obtain the amplitude contribution component; then multiplies the corner reflection depth value with the depth coefficient set to 0.85 to obtain the depth contribution component; then adds the amplitude contribution component and the depth contribution component with the constant term coefficient set to 0.03 to obtain the preliminary calculated value of the crack radial depth; finally, the preliminary calculated value is corrected with the fitting residual threshold set to 0.05 to eliminate the small systematic errors in the experimental fitting process. After the correction is completed, the final predicted value of the crack radial depth is obtained.

[0201] (6) Verify the rationality of the crack radial depth prediction value output by the model to ensure that the calculation result conforms to the physical law and the actual detection, including: verifying whether the crack radial depth prediction value is a positive value greater than zero, and eliminating abnormal results of negative or zero values; verifying whether the crack radial depth prediction value is less than or equal to the nominal value of the pipe wall thickness, and avoiding unreasonable values ​​that exceed the pipe wall thickness; verifying whether the accuracy of the crack radial depth prediction value is retained to two decimal places, which meets the requirements of engineering measurement accuracy; if the verification passes, the prediction value is confirmed as a valid calculation result; if the verification fails, the model calculation is determined to be abnormal, and the input parameters or model status need to be rechecked.

[0202] (7) The crack radial depth prediction value that has passed the verification is formally determined as the radial depth component in the crack size calculation result, and the quantitative calculation of the crack radial size is completed. It is clear that this component represents the actual radial distance of the crack extending from the surface of the weld root to the inside of the pipe wall thickness. It is the core indicator for assessing the severity of the crack and the risk of propagation. The radial depth component is associated and bound with the crack axial length component obtained in subsequent calculations to form a complete set of crack size calculation results.

[0203] The calculation of crack size using a preset corner reflection amplitude and depth calibration model also includes a step of calculating the crack axial length:

[0204] B1: Obtain the axial-depth map and the axial-depth signal data matrix;

[0205] Furthermore, the crack axial length is the distance the crack extends along the pipe axis, reflecting the axial expansion range of the crack. Together with the radial depth, it determines the overall size and risk level of the crack. The calculation of the axial length requires accurate determination based on the axial-depth map and the corresponding axial-depth signal data matrix, using the axial distribution range of the crack reflection signal.

[0206] Furthermore, the system acquires the generated axial-depth map and the corresponding axial-depth signal data matrix. The axial-depth signal data matrix contains complete axial position, radial depth, and echo amplitude information of the crack region. The axial-depth map intuitively presents the axial distribution pattern of the crack signal. The combination of the two provides data and visualization basis for axial length calculation.

[0207] B2: Extract the echo amplitude value sequence corresponding to the corner reflection depth value along the axial direction from the axial-depth signal data matrix;

[0208] Furthermore, the system locates the depth column corresponding to the corner reflection depth value from the axial-depth signal data matrix, extracts the echo amplitude values ​​corresponding to all axial positions in the depth column along the axial direction, and forms an axial echo amplitude value sequence. The echo amplitude value sequence reflects the distribution change of the reflected signal intensity along the axial direction at the crack core depth, and the effective signal range of the echo amplitude value sequence corresponds to the axial extension range of the crack.

[0209] B3: Set a crack axial endpoint determination threshold, and mark the position where the echo amplitude value in the echo amplitude value sequence is first greater than or equal to the crack axial endpoint determination threshold as the crack initiation axial position;

[0210] Furthermore, the crack axial endpoint determination threshold is an amplitude threshold that distinguishes the effective reflection signal of the crack from the background noise signal. It is used to accurately identify the start and end positions of the crack axial extension. The value of the crack axial endpoint determination threshold is reasonably set according to the background noise level and the minimum identifiable signal strength of the crack. It is determined through signal calibration in crack-free areas to ensure effective differentiation between crack signals and noise. In this embodiment, the crack axial endpoint determination threshold is set to 25. After setting the crack axial endpoint determination threshold, the system traverses the axial echo amplitude value sequence, starting from the end with the smallest axial coordinate value, and checks the echo amplitude value at each position one by one. The position where the amplitude value is first greater than or equal to the crack axial endpoint determination threshold is the starting point of the effective reflection signal of the crack, and is marked as the crack initiation axial position. The crack initiation axial position represents the starting end of the crack axial extension.

[0211] B4: Mark the position where the last echo amplitude value in the echo amplitude value sequence is greater than or equal to the crack axial endpoint determination threshold as the crack termination axial position;

[0212] Furthermore, the system continues to traverse the axial echo amplitude value sequence, checking backwards from the end with the largest axial coordinate value, or traversing forwards to the end, to find the position where the last amplitude value is greater than or equal to the crack axial endpoint determination threshold. This position is the termination point of the effective crack reflection signal and is marked as the crack termination axial position, representing the termination end of the crack's axial extension. The region between the starting axial position and the termination axial position is the axial distribution range of the effective crack reflection signal, corresponding to the actual axial extension area of ​​the crack.

[0213] B5: Calculate the difference between the axial position of crack termination and the axial position of crack initiation, and use this difference as the crack axial length value;

[0214] Furthermore, the system subtracts the axial coordinate value corresponding to the crack initiation axial position from the axial coordinate value corresponding to the crack termination axial position, and calculates the difference between the two. This difference is the crack axial length value, which accurately quantifies the actual distance the crack extends along the pipe axis, and the unit is consistent with the axial coordinate unit.

[0215] B6: The crack axial length value is determined as the axial length component in the crack size calculation result, and together with the radial depth component, it is used as the crack size calculation result.

[0216] Furthermore, the calculation results are annotated on the circumferential-depth map and the axial-depth map, and the circumferential position, axial depth, radial depth, and length data of the crack are marked at the corresponding positions. The initial microcrack depth determination identifier, crack penetration state identifier, and grading determination identifier are output, including:

[0217] S5.4: Obtain the predicted radial depth and axial length of the crack from the crack size calculation results, and obtain the circumferential and axial coordinate values ​​corresponding to the corner reflection candidate signals;

[0218] S5.5: In the circumferential-depth map, a label rectangle is drawn with the circumferential coordinate value corresponding to the corner reflection candidate signal as the horizontal axis position of the center point and the vertical axis span as the crack radial depth prediction value.

[0219] Furthermore, during annotation, the circumferential coordinate value corresponding to the corner reflection candidate signal is used as the center point of the horizontal axis, and the crack radial depth prediction value is used as the span in the vertical axis direction, that is, extending from the root surface to the deepest position of the crack, to draw a rectangular annotation box. The center of the rectangle corresponds to the circumferential position of the crack, and the length of the vertical axis of the rectangle corresponds to the radial depth of the crack. The rectangular area covers the distribution range of the crack in the circumferential-depth plane, making the annotation clear, intuitive and easy to identify.

[0220] S5.6: In the axial-depth diagram, a label rectangle is drawn with the axial coordinate value corresponding to the corner reflection candidate signal as the horizontal axis position of the center point, the crack axial length value as the horizontal axis span, and the crack radial depth prediction value as the vertical axis span.

[0221] Furthermore, the axial-depth map visually displays the axial position, axial length, and radial depth distribution of the crack. The system draws annotation rectangles on this map, comprehensively presenting the three-dimensional size distribution characteristics of the crack. During annotation, the axial coordinate value corresponding to the corner reflection candidate signal is used as the center point of the horizontal axis, the axial length of the crack is used as the span in the horizontal direction (from the crack initiation end to the termination end), and the predicted radial depth of the crack is used as the span in the vertical direction. A rectangular annotation box is drawn, covering the actual distribution range of the crack in the axial-depth plane, clearly identifying the axial length, radial depth, and center position of the crack. The visualization effect is intuitive and accurate.

[0222] S5.7: Obtain the nominal value of the pipe wall thickness for the fillet weld of the tube sheet, calculate the ratio of the predicted crack radial depth to the nominal value of the pipe wall thickness, and record it as the depth ratio.

[0223] Furthermore, the nominal pipe wall thickness is the standard wall thickness specified during pipe design. It serves as a fixed, known parameter, reflecting the pipe's wall thickness load-bearing capacity and safety margin. The depth ratio is the relative ratio of the crack radial depth to the pipe wall thickness. It is a core indicator for assessing crack severity, penetration risk, and safety level. The larger the ratio, the more severe the crack, the higher the penetration risk, and the lower the safety level. The system reads the nominal pipe wall thickness corresponding to the fillet weld of the pipe-to-tube sheet, divides the predicted crack radial depth by the nominal wall thickness, and calculates the depth ratio. The ratio is a dimensionless value, ranging from 0 to 1. When the ratio is close to 1, the crack is close to penetrating the pipe wall, posing an extremely high risk.

[0224] S5.8: Compare the depth ratio with a preset first depth ratio threshold. If the depth ratio is less than the first depth ratio threshold, output the initial microcrack depth determination flag.

[0225] Furthermore, the first depth ratio threshold is a critical ratio that distinguishes between initial microcracks and conventional cracks. A relatively small threshold value corresponds to a microcrack with shallow radial depth, in its early stages of initiation, and with a low risk of propagation. In this embodiment, the first depth ratio threshold is set to 0.2. The system compares the calculated depth ratio with the first depth ratio threshold: if the depth ratio is less than the first depth ratio threshold, it indicates that the crack has a shallow radial depth, exists only on the surface of the weld root, and is an initially nascent microcrack with a low risk of propagation. Therefore, an initial microcrack depth determination indicator is output, indicating that the crack is a microcrack and requires close attention and regular monitoring, but does not currently pose a significant safety risk.

[0226] S5.9: Compare the predicted radial depth of the crack with the nominal wall thickness of the pipe. If the predicted radial depth of the crack is greater than or equal to the nominal wall thickness of the pipe, output a crack penetration status indicator.

[0227] Furthermore, cracks penetrating the pipe wall can lead to media leakage, equipment failure, and safety accidents, representing the most serious crack condition requiring immediate shutdown for repair and risk elimination. The system directly compares the predicted radial depth of the crack with the nominal pipe wall thickness. If the predicted radial depth is greater than or equal to the nominal wall thickness, it indicates that the crack has completely penetrated the pipe wall, classifying it as a penetrating crack with extremely high risk. Therefore, a crack penetration status indicator is output, suggesting that the crack is a penetrating crack and immediate emergency measures are required.

[0228] S5.10: Output the corresponding classification judgment identifier based on the preset classification interval where the depth ratio is located.

[0229] Furthermore, the preset grading intervals are numerical ranges for crack risk levels based on the ratio of the predicted crack depth to the nominal pipe wall thickness. Each interval corresponds to a crack risk level, with risk levels ranging from low to high, comprehensively covering all severity levels from microcracks to penetrating cracks. Specifically, the intervals are divided into four categories: depth ratio less than 0.2, depth ratio greater than or equal to 0.2 and less than 0.5, depth ratio greater than or equal to 0.5 and less than 0.8, and depth ratio greater than or equal to 0.8. The system outputs the corresponding grading judgment label. Based on the specific value of the depth ratio, the system determines which preset grading interval it falls into and then outputs the corresponding grading judgment label for that interval, clearly indicating the crack risk level, such as low risk, medium risk, high risk, and extremely high risk. This allows operators to quickly grasp the crack risk level and formulate targeted treatment strategies, including regular monitoring for low-risk cracks, focused attention for medium-risk cracks, planned maintenance for high-risk cracks, and emergency shutdown for extremely high-risk cracks.

[0230] Example 2:

[0231] Please see Figure 3 Another embodiment of the present invention provides a fillet weld crack identification system based on corner reflection signals, comprising:

[0232] The signal acquisition module 10 is used to initialize and configure the parameters of the phased array ultrasonic detector, set the transmission and reception channels, focusing rules, scanning paths and coordinate acquisition methods, scan around the weld seam in the circumferential direction, and simultaneously acquire the first angle direct scanning signal, the second angle oblique scanning signal and the corresponding circumferential, axial and radial coordinate parameters of the tube sheet fillet weld detection area, and package and store them as the original scanning data file.

[0233] The root slicing module 20, based on the second angle oblique scanning signal and coordinate parameters, sets the depth range of the sensitive area at the weld root, extracts the root echo amplitude value using the sliding window method, generates root slice signal data, and constructs a circumferential-depth map through grayscale mapping; at the same time, it performs axial-depth tomographic slicing on the corner reflection candidate signal, generates an axial-depth signal data matrix, and constructs an axial-depth map;

[0234] The candidate signal locking module 30 is used to compare the root slice signal data with the first angle direct scan signal, and filter out the single-angle unique points that only exist in the sensitive area of ​​the weld root as the corner reflection candidate signal positions; then, a density-based spatial clustering algorithm is used to cluster and merge the corner reflection candidate signals, remove noise points and extract the geometric center coordinates of the candidate defect clusters, accurately lock the effective corner reflection candidate signals and their coordinates, and reduce the impact of clutter interference on crack identification.

[0235] The crack feature determination module 40 is used to extract the distribution curve of the corner reflection candidate signal, locate the global maximum point to obtain the corner reflection position, amplitude and depth, retrieve the preset crack determination conditions, verify whether the axial deviation, depth deviation and amplitude meet the standards, determine that the corner reflection candidate signal is the fixed position corner reflection feature signal formed by the pipe outer wall and the crack, output the crack identification completion signal, and complete the qualitative identification of the crack.

[0236] The crack size calculation module 50 is used to receive the crack identification completion signal, load the preset corner reflection amplitude and depth calibration model, substitute the parameters to calculate the crack radial depth; determine the crack start and end positions through the axial-depth signal data matrix, and calculate the axial length; finally, mark the crack parameters in the map, and output the initial microcrack depth judgment mark, crack penetration status mark and grading judgment mark in combination with the pipe wall thickness, so as to realize the quantitative output of crack and the grading of status.

[0237] 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 present invention. All of these variations are within the protection scope of the present invention.

Claims

1. A method for identifying fillet weld cracks based on corner reflection signals, characterized in that, include: S1: Acquire the phased array direct scanning signal at the first angle and the oblique scanning signal at the second angle in the tube-tube sheet fillet weld inspection area, and simultaneously collect the circumferential, axial and radial coordinate parameters of the corresponding scanning positions; S2: Based on the second angle oblique scanning signal and coordinate parameters, perform tomographic slicing along the sensitive area at the root of the weld to generate root slice signal data and construct a circumferential-depth map; S3: Compare the root slice signal data with the first angle direct scan signal to lock the corner reflection candidate signal and its coordinates that only exist in the sensitive area of ​​the weld root; S4: At the coordinates corresponding to the corner reflection candidate signal, perform tomographic slicing along the axial-depth direction to generate an axial-depth map. Determine that the corner reflection candidate signal is a fixed-position corner reflection feature signal formed by the outer wall of the pipe and the crack. Verify that the corner reflection position, corner reflection amplitude and depth meet the preset crack judgment conditions, and output a crack identification completion signal. S5: Based on the crack identification completion signal, calculate the crack size using the preset corner reflection amplitude and depth calibration model, mark the calculation results on the circumferential-depth map and axial-depth map, mark the crack circumferential position, axial depth, radial depth and length data at the corresponding positions, and output the initial microcrack depth judgment mark, crack penetration status mark and grade judgment mark. Before acquiring the phased array first-angle direct scanning signal and second-angle oblique scanning signal of the tube sheet fillet weld detection area, the initialization configuration is performed, including: setting the transmission and reception channel parameters of the phased array ultrasonic detector, configuring the focusing rule corresponding to the first-angle direct scanning, configuring the incident angle of the first-angle direct scanning signal beam to 0° vertical incidence, configuring the focusing rule corresponding to the second-angle oblique scanning, configuring the incident angle of the second-angle oblique scanning signal beam to 30° oblique incidence, simultaneously configuring the sampling frequency, gain, pulse repetition frequency and detection mode of the phased array ultrasonic detector, setting the scanning start point coordinates and scanning end point coordinates, configuring the scanning path to scan around the weld circumference, setting the initialization completion flag to true, and outputting the detection system ready signal; The specific steps of S3 include: S3.1: Read the first echo sequence of each data frame from the original scan data file; the data frame is composed of the first echo sequence, the second echo sequence, circumferential coordinate value, axial coordinate value and radial coordinate value obtained from the same sampling trigger position; the first echo sequence is the echo sequence corresponding to the phased array first angle direct scan signal; S3.2: Extract the echo amplitude values ​​within the depth range of the sensitive area at the root of the weld from each first echo sequence to form a first amplitude value set; S3.3: Extract echo amplitude values ​​with the same circumferential coordinate values ​​and the same depth range from the root slice signal data to form a second amplitude value set; S3.4: Compare the first amplitude value and the second amplitude value corresponding to the same spatial location point one by one, and mark the spatial location points that meet the preset conditions as single-angle unique points. The preset conditions are: the first amplitude value is less than the first preset threshold, and the second amplitude value is greater than the second preset threshold. S3.5: Retain the circumferential coordinate values ​​and depth values ​​corresponding to all single-angle unique points as corner reflection candidate signal positions, and use the corner reflection candidate signal positions and their corresponding circumferential coordinate values ​​and depth values ​​as corner reflection candidate signals and their coordinates.

2. The method for identifying fillet weld cracks based on corner reflection signals as described in claim 1, characterized in that, The specific steps of S1 include: S1.1: After receiving the detection system ready signal, move the phased array probe along the scanning path; S1.2: At each sampling trigger position, a first-angle direct scanning signal beam and a second-angle oblique scanning signal beam are synchronously transmitted, and the corresponding first echo sequence and second echo sequence are received respectively; S1.3: At the same sampling trigger position, the circumferential coordinate value of the phased array probe center is acquired by the encoder, the axial coordinate value is acquired by the axial displacement sensor, and the radial coordinate value is acquired by the radial displacement sensor. S1.4: Pack the first echo sequence, second echo sequence, circumferential coordinate value, axial coordinate value and radial coordinate value obtained at the same sampling trigger position into a data frame, and store the data frames corresponding to each sampling trigger position as the original scan data file in ascending order of circumferential coordinate value.

3. The method for identifying fillet weld cracks based on corner reflection signals as described in claim 2, characterized in that, The specific steps of S2 include: S2.1: Read the second echo sequence and circumferential coordinate values ​​of each data frame from the original scan data file; S2.2: Set the starting and ending depth values ​​of the sensitive zone at the weld root to determine the slice depth range; S2.3: The sliding window method is used to extract the echo amplitude values ​​of each second echo sequence point by point along the depth direction, and the depth values ​​are within the slice depth range. S2.4: Arrange the echo amplitude values ​​corresponding to each depth value under the same circumferential coordinate value in ascending order of depth to form the root signal vector of the corresponding circumferential coordinate value; S2.5: Combine the root signal vectors of all circumferential coordinate values ​​in ascending order of circumferential coordinate values ​​to generate root slice signal data; the root slice signal data includes slice depth range, circumferential angle distribution, echo position and echo amplitude; S2.6: Perform grayscale mapping processing on the root slice signal data to construct a circumferential-depth map with circumferential coordinates as the horizontal axis and depth as the vertical axis.

4. The method for identifying fillet weld cracks based on corner reflection signals as described in claim 3, characterized in that, The step of performing tomographic slicing along the axial-depth direction at the coordinates corresponding to the corner reflection candidate signal to generate an axial-depth map includes: S4.1: Obtain the circumferential coordinates and depth values ​​corresponding to the corner reflection candidate signals; S4.2: Extract all data frames whose circumferential coordinate values ​​are within the preset neighborhood range of the circumferential coordinate values ​​of the corner reflection candidate signal from the original scan data file, and read the second echo sequence and axial coordinate value of the corresponding data frame from each extracted data frame; S4.3: With the depth value of the corner reflection candidate signal as the center, set a depth slice window. For each extracted data frame, extract the echo amplitude value sequence within the depth range of the depth slice window from the second echo sequence of the data frame, and associate and store the echo amplitude value sequence extracted from each data frame with the axial coordinate value of the corresponding data frame. S4.4: Arrange the echo amplitude value sequence after all associated storage in ascending order of axial coordinate values ​​to generate an axial-depth signal data matrix, and perform grayscale mapping processing on the axial-depth signal data matrix to construct an axial-depth map with axial coordinate as the horizontal axis and depth as the vertical axis.

5. The method for identifying fillet weld cracks based on corner reflection signals as described in claim 4, characterized in that, The process of outputting the crack identification completion signal includes: S4.5: Extract the distribution curve of the echo amplitude value along the axial direction corresponding to the depth value of the corner reflection candidate signal from the generated axial-depth signal data matrix; S4.6: Locate the global maximum point of the echo amplitude value in the distribution curve; S4.7: Obtain the axial coordinate value corresponding to the global maximum point as the corner reflection axial position; obtain the depth value corresponding to the global maximum point as the corner reflection depth value; obtain the echo amplitude value of the global maximum point as the corner reflection amplitude. S4.8: Retrieve preset crack judgment conditions, and calculate the absolute value of the axial deviation between the corner reflection axial position and the preset standard axial position, and calculate the absolute value of the depth deviation between the corner reflection depth value and the preset standard depth value; the preset crack judgment conditions include axial position deviation threshold, depth deviation threshold and amplitude threshold; S4.9: When the absolute value of the axial deviation is less than or equal to the axial position deviation threshold, the absolute value of the depth deviation is less than or equal to the depth deviation threshold, and the corner reflection amplitude is greater than or equal to the amplitude threshold, the corner reflection candidate signal is determined to be the fixed position corner reflection characteristic signal formed by the outer wall of the pipe and the crack, which satisfies the crack judgment condition, and a crack identification completion signal carrying the corner reflection axial position, corner reflection depth value, and corner reflection amplitude is output.

6. The method for identifying fillet weld cracks based on corner reflection signals as described in claim 5, characterized in that, The step of calculating the crack size based on the crack identification completion signal using a preset corner reflection amplitude and depth calibration model includes: S5.1: Receive the crack identification completion signal and extract the corner reflection amplitude and corner reflection depth values ​​carried in the crack identification completion signal; S5.2: Load the preset corner reflection amplitude and depth calibration model; the corner reflection amplitude and depth calibration model is obtained by performing phased array detection experiments on tube sheet fillet weld test blocks containing artificial cracks of different depths, and then fitting and calibrating using the least squares method; S5.3: Substitute the extracted corner reflection amplitude and corner reflection depth values ​​into the corner reflection amplitude and depth calibration model to perform forward calculation, solve for the corresponding crack radial depth prediction value, and determine the crack radial depth prediction value as the radial depth component in the crack size calculation result.

7. The method for identifying fillet weld cracks based on corner reflection signals as described in claim 6, characterized in that, The calculation of crack size using a preset corner reflection amplitude and depth calibration model also includes a step of calculating the crack axial length: B1: Obtain the axial-depth map and the axial-depth signal data matrix; B2: Extract the echo amplitude value sequence corresponding to the corner reflection depth value from the axial-depth signal data matrix along the axial direction; B3: Set a crack axial endpoint determination threshold, and mark the position where the echo amplitude value in the echo amplitude value sequence is first greater than or equal to the crack axial endpoint determination threshold as the crack initiation axial position; B4: Mark the position where the last echo amplitude value in the echo amplitude value sequence is greater than or equal to the crack axial endpoint determination threshold as the crack termination axial position; B5: Calculate the difference between the axial position of crack termination and the axial position of crack initiation, and use the difference as the crack axial length value; B6: The crack axial length value is determined as the axial length component in the crack size calculation result, and together with the radial depth component, it is used as the crack size calculation result.

8. A fillet weld crack identification system based on corner reflection signals, used to implement the fillet weld crack identification method based on corner reflection signals as described in any one of claims 1-7, characterized in that, include: The signal acquisition module is used to simultaneously acquire the first angle direct scanning signal, the second angle oblique scanning signal and the corresponding circumferential, axial and radial coordinate parameters of the tube sheet fillet weld inspection area, and package and store them as raw scanning data files; The root slicing module generates root slice signal data based on the second angle oblique scanning signal and coordinate parameters. It constructs a circumferential-depth map through grayscale mapping. At the same time, it performs axial-depth tomographic slicing on the corner reflection candidate signal to generate an axial-depth signal data matrix and construct an axial-depth map. The candidate signal locking module is used to compare the root slice signal data with the first angle direct scan signal, filter the corner reflection candidate signal positions, and then cluster and merge the corner reflection candidate signals to lock the effective corner reflection candidate signals and their coordinates. The crack feature determination module is used to locate the global maximum point to obtain the corner reflection position, amplitude and depth, retrieve the preset crack determination conditions, determine the corner reflection candidate signal as the corner reflection feature signal of the fixed position formed by the pipe outer wall and the crack, and output the crack identification completion signal. The crack size calculation module is used to load a preset corner reflection amplitude and depth calibration model, calculate the crack radial depth and axial length, and output the initial microcrack depth judgment mark, crack penetration status mark and grade judgment mark.

Images