A thin-wall weld automatic detection method

Abstract

The present application belongs to the field of nondestructive testing, and discloses a kind of thin-walled weld automatic detection method, it includes the following steps: to be detected thin-walled weld implementation automation scanning;Current frame full focus ultrasonic image is carried out depth interval mask and adaptive segmentation, extract background noise characteristic value N and bottom echo intensity characteristic value R;Based on background noise characteristic value N and bottom echo intensity characteristic value R generate signal quality parameter Q=(R‑N) / R0;Signal quality parameter Q with the preset signal failure threshold Q fail And signal recovery threshold Q rec Compare, and judge whether signal is effective, when judging as signal failure, pause weld scanning movement, and probe posture is adjusted;When judging as signal effective, continue to execute scanning from pause position according to preset trajectory.It can realize real-time, quantitative monitoring of signal effectiveness in thin-walled weld automatic detection, and output corresponding control signal to adjust probe working state in time, improve detection efficiency.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of nondestructive testing, and particularly relates to an automatic detection method for thin-wall welds. BACKGROUND

[0002] In ultrasonic automatic detection of steel thin-wall welds, the existing method usually requires an operator to monitor the scanning signals in real time throughout the process. Once an abnormal situation such as signal loss or unstable acoustic coupling occurs in a channel, the scanning needs to be manually stopped immediately, and after the signal abnormality problem is solved, the scanning path needs to be reset and the detection needs to be restarted. This method has obvious shortcomings: it is highly dependent on the operator, and is easily affected by human factors such as the operator's experience, and often the data signal quality is not up to standard until after the scanning is completed. The above problems directly restrict the efficiency and reliability of automatic detection.

[0003] To improve the level of automation, the existing technology attempts to introduce a position encoding and data synchronization mechanism. For example, Chinese patent application CN108956776A discloses an ultrasonic detection method based on an encoder to collect and store scanning data and probe position information in real time, which realizes data traceability of the detection process. However, this method can only conduct post-evaluation by counting the signal abnormal area (such as signal loss area ratio ≤ 5%) in the image after the detection is completed, and does not have the ability to monitor and judge the signal quality in real time, nor can it output control instructions to interrupt or adjust the scanning process when signal abnormalities occur. Therefore, once the signal fails, the entire segment needs to be reworked and rescaned, which seriously affects the detection efficiency.

[0004] Another technology such as Chinese patent application CN114184676A introduces a full-focus imaging technology to improve imaging resolution, but its application scenario is limited to aluminum alloy irregular structure welds, and does not involve an online evaluation mechanism for ultrasonic signal effectiveness. During the entire automatic scanning process, the system cannot dynamically perceive changes in signal quality, nor can it trigger corresponding control responses accordingly.

[0005] Currently, in actual engineering applications, even if automatic scanning equipment is used, the operator still needs to monitor the ultrasonic signal waveform throughout the process. Once signal loss, bottom wave attenuation, or noise abnormalities are found in a channel, the scanning process needs to be manually paused, the cause (such as probe posture deviation, surface contamination, or acoustic interference) needs to be investigated, and after the problem is solved, the starting point needs to be reset and the detection needs to be restarted. This intervention mode that relies on manual operation is not only inefficient, but also highly affected by the skill level and subjective judgment of the operator, which can easily cause deviations in detection information records. More seriously, it is often not until after the entire weld scanning is completed that it is found that the signal quality of the collected data does not meet the analysis requirements, such as too low signal-to-noise ratio, missing bottom wave, or image distortion, which results in invalid detection and repeated operation, greatly restricting the reliability and practicality of the automatic detection system.

[0006] In summary, the prior art generally has the following common defects: (1) lack of real-time, quantitative evaluation mechanism for the effectiveness of ultrasonic detection signals; (2) unable to automatically trigger control response when signal quality deteriorates to avoid invalid data acquisition; (3) detection process is disconnected from signal state, lacking closed-loop linkage capability, resulting in low efficiency, poor reliability, and untraceable data.

[0007] Therefore, there is an urgent need for a thin-walled weld detection method that can monitor signal effectiveness in real time during automatic scanning, dynamically determine signal quality state, and respond to the motion control system in linkage, to realize intelligent ultrasonic detection with high reliability, high efficiency, and full-process traceability. SUMMARY

[0008] The purpose of the present application is to provide a thin-walled weld automatic detection method that can judge in real time whether the scanning signal is effective and output corresponding control signals to adjust the probe working state in time, thereby improving the detection efficiency.

[0009] To solve the above technical problems, the present application adopts the following technical scheme: a thin-walled weld automatic detection method, comprising the following steps:

[0010] Step S1, using a full-focus phased array ultrasonic system to implement automatic scanning of the detected thin-walled weld, specifically including:

[0011] Step S11, configuring the phased array probe and wedge parameters, and setting the full-matrix capture mode;

[0012] Step S12, the scanning device drives the phased array probe to scan the weld area according to the preset scanning trajectory;

[0013] Step S13, cooperatively collecting full-focus ultrasonic images of the weld area through the first full-focus imaging mode and the second full-focus imaging mode which are different in mode;

[0014] Step S2, performing real-time signal effectiveness judgment on the images collected in step S13, specifically including:

[0015] Step S21, performing depth interval mask and adaptive segmentation on the current frame of full-focus ultrasonic images to extract background noise characteristic value N and bottom surface echo intensity characteristic value R;

[0016] Step S22, generating signal quality parameter Q = (R - N) / R0 based on the background noise characteristic value N and the bottom surface echo intensity characteristic value R;

[0017] Wherein R0 is a preset bottom surface echo reference intensity;

[0018] Step S23, comparing the signal quality parameter Q with a preset signal invalidation threshold Qfail and signal recovery threshold Q rec Comparison:

[0019] If Q fail , determine that the signal is invalid;

[0020] If Q rec , determine that the signal is valid;

[0021] Step S3, realize scanning control linkage according to the signal validity determination result, specifically including:

[0022] When determining that the signal is invalid, pause the weld scanning motion, and adjust the probe attitude, and only allow the probe to adjust the attitude in the direction perpendicular to the workpiece surface and the rotation direction around the probe axis until the signal is determined to be valid;

[0023] When determining that the signal is valid, continue to perform scanning according to the preset trajectory from the paused position until the scanning is completed.

[0024] In another embodiment, the full-focus ultrasonic image is generated by performing time-delay stacking in at least two pixel angle sub-zones; the pixel angle sub-zone is an imaging sector defined in the workpiece cross-sectional coordinate system, which is determined by a preset ultrasonic propagation path type and a wedge nominal incident angle; the first pixel angle sub-zone is centered on α1°, covering the lower and root regions of the weld, and this mode is the first full-focus imaging mode; the second pixel angle sub-zone is centered on α2°, covering the upper and near-surface regions of the weld; the opening angle of each pixel angle sub-zone is not less than 30°, and only the pixels falling into the sub-zone participate in the synthesis of the corresponding full-focus image, 30

[0025] In another embodiment, the ultrasonic propagation path corresponding to the first pixel angle sub-zone is a transverse wave-bottom surface reflection-defect-bottom surface reflection-transverse wave path (TT-TT path), which is used to enhance the detection capability of the root defects of the weld.

[0026] In another embodiment, the ultrasonic propagation path corresponding to the second pixel angle sub-zone is a transverse wave-defect-transverse wave path (TT path), which is used to improve the imaging resolution and signal-to-noise ratio of the near-surface region.

[0027] In another embodiment, the depth interval mask and adaptive segmentation in step S21 specifically include:

[0028] Step S211, preset a background noise ROI bg with a depth range of 0.3T to 0.7T of the workpiece thickness T; and a bottom echo ROI bot with a depth range of T±1mm;

[0029] Step S212, Otsu adaptive threshold segmentation is performed on the whole image I to obtain a binary image B, wherein a region with a connected domain area greater than 50 pixels is regarded as a "weld or defect" region;

[0030] Step S213, logical AND operation is performed on the binary image B and the ROI bg and the ROI bot respectively to obtain a background region mask M bg_defect containing defect interference and a bottom surface region mask M bot_defect ;

[0031] Step S214, M bg_defect and M bot_defect are inverted respectively, and logical AND operation is performed on the original ROI bg , the ROI bot , to obtain a pure background noise mask M bg and a pure bottom surface echo mask M bot , which are used for subsequent calculation.

[0032] In another embodiment, the background noise characteristic value N is the root mean square of the pixel amplitude in the pure background noise mask M bg , and the calculation formula is: ,

[0033] wherein M bg (i,j) is a binary mask of the pure background noise region, and is 1 or 0; I(i,j) is the pixel amplitude of the full-focus ultrasonic image at position (i,j).

[0034] In another embodiment, the bottom surface echo intensity characteristic value R is any one of the following statistical quantities of the pixel amplitude in the pure bottom surface echo mask M bot : (1) the arithmetic mean value of the first 1% of the amplitudes; (2) the average value of all pixels with an amplitude greater than a -6 dB threshold; (3) the root mean square value of all pixels; wherein the defect region is excluded by the mask M bot and does not participate in the statistics.

[0035] In another embodiment, the background noise ROI bg avoids the heat affected zone, the bottom surface reflection zone and the surface blind area; and the bottom surface echo ROI bot avoids the weld center, the defect region and the geometric reflector region.

[0036] In another embodiment, the background noise ROI bg is set in the middle layer region of the base material entity far from the weld heat affected zone, and the depth range is 0.3T to 0.7T, so as to ensure that the extracted is the system intrinsic background noise.

[0037] In another embodiment, the bottom echo ROI bot The bottom echo ROI is set at the base metal region close to the weld, with a depth range of T±1 mm and a horizontal position outside the heat affected zone to ensure that the measured R value only reflects the penetration ability of the acoustic wave.

[0038] In another embodiment, the background noise ROI bg The position of the bottom echo ROI bot is fixed according to the workpiece geometry parameters in the initialization stage of the detection program, and is dynamically updated with the probe position during the scanning process.

[0039] In another embodiment, the signal failure threshold and the signal recovery threshold are respectively set as:

[0040] N0 is the background noise reference value.

[0041] In another embodiment, the method for obtaining the bottom echo reference intensity R0 and the background noise reference value N0 includes:

[0042] Step S231: Selecting a calibration region in the base metal area of the workpiece, which is known to be defect-free, flat, and well-coupled;

[0043] Step S232: Using the same full-focus parameters as the formal scanning, changing the probe incident angle by 0.5° steps within a range of ±5° through an electrically controlled rotating mechanism, collecting A-scan waveforms at each angle and recording the bottom echo peak value;

[0044] Step S233: Selecting the angle with the highest bottom echo peak value as the representative angle, and adjusting the bottom echo amplitude at this angle to 80% of the full scale to lock the system gain;

[0045] Step S234: Collecting a fan-scan full-focus image under the gain condition, and extracting the average value of the pixel intensity in the pure bottom echo mask M bot , denoted as R0;

[0046] Step S235: Calculating the root mean square of the pixel amplitude in the pure background noise mask M bg region of the same image, denoted as N0.

[0047] In another embodiment, the probe posture adjustment includes:

[0048] Lifting the probe in a direction perpendicular to the workpiece surface to fine-tune the coupling distance;

[0049] Stepping the probe holder by a small angle around the probe axis, collecting a full-focus image at each step and calculating the Q value;

[0050] If the adjusted Q < Q failThen repeat the lifting and rotating operation to optimize until Q >= Q rec ;

[0051] When Q >= Q rec , the weld scanning motion is resumed, and the preset trajectory is continued from the coordinates recorded at the time of suspension.

[0052] The parameters of the phased array probe and the wedge are determined by simulation optimization to minimize the comprehensive objective function of defect quantitative error and acoustic coupling loss; the simulation optimization method comprises: generating a plurality of combinations of probe and wedge parameters by using an experimental design method; calculating the comprehensive objective function value corresponding to each combination through an ultrasonic propagation simulation model; constructing a response surface model based on the calculation results and performing global optimization to determine the final parameters.

[0053] The beneficial effects of the present application are:

[0054] 1. Improve detection reliability: through cross verification of dual-mode A-scan signals, eliminate the detection blind area of single mode, significantly improve the detection ability of near-surface area defects;

[0055] 2. The reference intensity R0 of the bottom echo and the reference value N0 of the background noise are both taken from the image obtained by real-time scanning, so that the judgment standard is closer to the weld to be detected itself, and the judgment accuracy is higher;

[0056] 3. Quantization and real-time control of signal effectiveness: convert the traditional signal quality judgment relying on human experience into quantitative judgment based on the bottom wave peak value, avoid misjudgment caused by human factors, and significantly improve the automation degree and data effectiveness of thin plate weld detection;

[0057] 4. Prevent invalid data collection: through real-time signal effectiveness monitoring and automatic stop scanning mechanism, prevent invalid data accumulation caused by poor signals, avoid subsequent batch rework, and improve detection efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 is the detection flowchart of the present detection system;

[0059] Figure 2 is the flowchart of the present method;

[0060] Figure 3 is the probe scanning mode schematic diagram under the first full-focus imaging mode;

[0061] Figure 4 is the probe scanning mode schematic diagram under the second full-focus imaging mode;

[0062] Figure 5 is the background noise ROI bg selection and pixel coordinate schematic diagram;

[0063] Figure 6 For bottom echo ROI bot Selecting pixel coordinates diagram. Detailed Implementation

[0064] like Figure 1 As shown, the automatic inspection method for thin-walled welds includes the following steps:

[0065] Step S1 involves using a fully focused phased array ultrasonic system to automatically scan the thin-walled weld under inspection, specifically including:

[0066] Step S11: Configure the phased array probe and wedge parameters, and set the full matrix acquisition mode;

[0067] Step S12: The scanning device drives the phased array probe to scan the weld area according to the preset scanning trajectory;

[0068] Step S13: Acquire full-focus ultrasonic images of the weld area by coordinating a first full-focus imaging mode and a second full-focus imaging mode with different modes;

[0069] Step S2 involves real-time signal validity assessment of the images acquired in step S13, specifically including:

[0070] Step S21: Perform depth interval masking and adaptive segmentation on the current frame of full-focus ultrasound image, and extract background noise feature value N and bottom surface echo intensity feature value R;

[0071] Step S22: Generate signal quality parameter Q = (R - N) / R0 based on background noise characteristic value N and bottom echo intensity characteristic value R;

[0072] Where R0 is the preset bottom surface echo reference intensity;

[0073] Step S23: Compare the signal quality parameter Q with the preset signal failure threshold Q. fail and signal recovery threshold Q rec Compare:

[0074] If Q<Q fail If so, the signal is deemed invalid;

[0075] If Q≥Q rec If so, the signal is considered valid;

[0076] Step S3, based on the signal validity determination result, implements scanning control linkage, specifically including:

[0077] Before the scan begins, the system needs to be adjusted until Q ≥ Q. rec Only after this can the scan begin, during which Q is allowed to drop to Q.fail ≤ Q < Q rec When Q is greater than or equal to Q fail At this time, the scanning can be continued, but when Q < Q fail When it is determined that the signal is invalid, the weld scanning motion is paused, the probe posture is adjusted, and the probe is only allowed to adjust the posture in the direction perpendicular to the workpiece surface and the rotation direction around the probe axis until Q ≥ Q rec When it is determined that the signal is valid, the scanning continues from the paused position according to the preset trajectory;

[0078] More specifically, when the signal quality parameter Q < Q fail , it is determined that the current ultrasonic signal is invalid, and the industrial control computer immediately outputs an enable prohibition instruction to the motion control unit through the real-time bus to stop the servo drive output of the main motion shaft for advancing the scanning device in the weld direction and the transverse direction, while latching the current scanning position information;

[0079] Subsequently, the system enters a signal recovery attempt mode: only the enable state of the auxiliary motion shaft for adjusting the relative posture of the probe and the workpiece is retained, the probe is controlled to perform a slight lifting in the direction perpendicular to the workpiece surface and a small-angle step rotation around its own axis; after each posture adjustment, a full-focus ultrasonic image is re-acquired and the signal quality parameter Q is calculated;

[0080] If the signal quality parameter Q ≥ Q rec , it is determined that the signal validity has been restored, the servo enable of the main motion shaft is reactivated, and the scanning device continues to perform automatic scanning from the latched position;

[0081] If the signal quality parameter Q is always lower than Q rec after multiple posture adjustments, the system determines that the current signal invalidity is not caused by the coupling state between the probe and the workpiece, but may be caused by hardware abnormalities such as ultrasonic instruments, phased array probes, data acquisition links, etc. At this time, the system terminates the enable output of all motion shafts, aborts the current detection process, and prompts "signal abnormality - suspected equipment failure" on the human-machine interface, while saving the current signal state, position information and log data to the detection report for subsequent diagnostic analysis;

[0082] Wherein: the full-focus ultrasonic image is generated by performing delay stacking in at least two pixel angle sub-regions; for example, Figure 3 、 4As shown, the pixel angle sub-zone is an imaging sector defined in the workpiece cross-section coordinate system, which is determined by the preset ultrasonic propagation path type and the nominal wedge incident angle. The first pixel angle sub-zone takes a1° as the center line and covers the lower and root areas of the weld. This mode is the first full-focus imaging mode. The second pixel angle sub-zone takes a2° as the center line and covers the upper and near-surface areas of the weld. The opening angle of each pixel angle sub-zone is not less than 30°, and only the pixels falling into the sub-zone participate in the synthesis of the corresponding full-focus image. 30 < a1 < 50, 50 < a2 < 70, and the optimal values are a1 = 45 and a2 = 60, respectively. The ultrasonic propagation path corresponding to the first pixel angle sub-zone is a transverse wave-bottom reflection-defect-bottom reflection-transverse wave path, which is used to enhance the detection capability of the root defects of the weld. The ultrasonic propagation path corresponding to the second pixel angle sub-zone is a transverse wave-defect-transverse wave path, which is used to improve the imaging resolution and signal-to-noise ratio of the near-surface area.

[0083] As shown in Figure 5 , 6 , the depth interval mask and adaptive segmentation in step S21 specifically include:

[0084] Step S211, preset a background noise ROI bg with a depth range of 0.3T to 0.7T of the workpiece thickness T; and a bottom echo ROI bot with a depth range of T±1mm;

[0085] Step S212, the Otsu adaptive threshold segmentation is performed on the entire image I to obtain a binary image B, wherein the area with a connected domain area greater than 50 pixels is regarded as a "weld or defect" area;

[0086] Step S213, the binary image B is subjected to logical AND operation with the ROI bg and the ROI bot , respectively, to obtain a background area mask M bg_defect containing defect interference and a bottom area mask M bot_defect ;

[0087] Step S214, M bg_defect and M bot_defect are inverted, respectively, and are subjected to logical AND operation with the original ROI bg and the ROI bot to obtain a pure background noise mask M bg and a pure bottom echo mask M bot , which are used for subsequent calculation.

[0088] In another embodiment, the background noise characteristic value N is the root mean square of the pixel amplitude in the pure background noise mask M bg , and the calculation formula is: ,

[0089] wherein: M bg (i,j) is a binary mask of pure bottom echo region, taking value of 1 or 0; I(i,j) is the pixel amplitude of full focus ultrasonic image at position (i,j).

[0090] The bottom echo intensity eigenvalue R is a pure bottom echo mask M bot Any one of the following statistics of inner pixel amplitude: (1) the arithmetic mean of the first 1% of pixel amplitude; (2) the average of all pixels with amplitude greater than -6 dB threshold; (3) the root mean square value of all pixels; wherein the defect region has been masked by M bot Excluded, not involved in statistics. Background noise ROI bg Avoiding the heat affected zone, bottom reflection zone and surface blind area; bottom echo ROI bot Avoiding the weld center, defect region and geometric reflector region. Background noise ROI bg Set in the middle layer region of the base material entity far away from the weld heat affected zone, the depth range is 0.3T to 0.7T, to ensure that the extracted is the intrinsic background noise of the system. Bottom echo ROI bot Set in the base material bottom region close to the weld, the depth range is T±1 mm, and the horizontal position is outside the heat affected zone, to ensure that the measured R value only reflects the penetration ability of the sound wave. Background noise ROI bg The position of the bottom echo ROI bot is fixed according to the workpiece geometric parameters in the initialization stage of the detection program, and is dynamically updated with the probe position in the scanning process. The signal failure threshold and the signal recovery threshold are respectively set as:

[0091] .

[0092] The acquisition method of the bottom echo reference intensity R0 and the background noise reference value N0 includes:

[0093] Step S231, selecting a calibration region in the base material region of the workpiece, which is known to be defect-free, flat and well-coupled;

[0094] Step S232, using the same full focus parameters as the formal scanning, changing the probe incidence angle by 0.5° steps in the range of ±5° through the electrically controlled rotating mechanism, collecting A-scan waveforms at each angle and recording the bottom echo peak value;

[0095] Step S233, selecting the angle with the highest bottom echo peak value as the representative angle, and adjusting the bottom echo amplitude at this angle to 80% of the full scale to lock the system gain;

[0096] Step S234, collecting a fan scanning full focus image under the gain condition, and extracting the pure bottom echo mask M botThe average value of the inner pixel intensity is denoted as R0;

[0097] In step S235, the pure background noise mask M bg The root mean square of the pixel amplitude in the region is calculated and denoted as N0.

[0098] The probe posture adjustment includes:

[0099] The probe is lifted in a direction perpendicular to the workpiece surface to fine-tune the coupling distance;

[0100] The probe frame is rotated by a small angle step by step around the probe axis, and a full-focus image is collected and the Q value is calculated at each step;

[0101] If Q after adjustment is Q fail , the lifting and rotating operations are repeated for optimization until Q≥Q rec .

[0102] When Q≥Q rec , the weld scanning motion is resumed, and the preset trajectory is continued to be executed from the coordinates recorded at the time of suspension.

[0103] The parameters of the phased array probe and the wedge are determined through simulation optimization to minimize the comprehensive objective function of defect quantification error and acoustic wave coupling loss; the simulation optimization method includes: using an experimental design method to generate multiple sets of probe and wedge parameter combinations; calculating the comprehensive objective function value corresponding to each combination through an ultrasonic propagation simulation model; constructing a response surface model based on the calculation results and performing global optimization to determine the final parameters.

[0104] The thin-walled weld full-focus ultrasonic automatic detection system includes a full-focus phased array ultrasonic instrument, a computer, and a servo scanning device; specifically, the full-focus phased array ultrasonic instrument drives the phased array probe to implement automatic scanning of the thin-walled weld in a first full-focus mode at 45° and a second full-focus mode at 60°, and extracts the background noise N and the maximum pixel amplitude R of the weld root from the full-focus image in real time; the computer runs in parallel with the full-focus phased array ultrasonic instrument, generates a signal quality parameter based on R and N, and outputs a signal quality state determination result after comparison with a preset threshold; the servo scanning device receives instructions from the computer, automatically stops scanning motion when receiving a signal quality failure determination, performs precise adjustment of the probe frame until the signal meets the good signal quality determination again, and then continues scanning from the original stop position. The computer obtains the R and N values from the full-focus phased array ultrasonic instrument in real time through a shared memory interface, and the data delay is less than one scanning step period. The servo scanning device has a built-in motion controller for receiving instructions from the computer and driving the servo motor, and the signal quality failure response time is less than the interval between two consecutive samples, and the encoder information of the failure position is recorded automatically; the phased array probe detection parameters are as follows:

[0105]

[0106] The above examples are only for illustrating the technical concept and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and to implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made according to the spirit of the present application should be covered within the protection scope of the present application.

Claims

1. An automatic inspection method for thin-walled welds, characterized in that: It includes the following steps: Step S1 involves using a fully focused phased array ultrasonic system to automatically scan the thin-walled weld under inspection, specifically including: Step S11: Configure the phased array probe and wedge parameters, and set the full matrix acquisition mode; Step S12: The scanning device drives the phased array probe to scan the weld area according to the preset scanning trajectory; Step S13: Acquire full-focus ultrasonic images of the weld area by coordinating a first full-focus imaging mode and a second full-focus imaging mode with different modes; Step S2 involves real-time signal validity assessment of the images acquired in step S13, specifically including: Step S21: Perform depth interval masking and adaptive segmentation on the current frame of full-focus ultrasound image, and extract background noise feature value N and bottom surface echo intensity feature value R; Step S22: Generate signal quality parameter Q = (R- N) / R0 based on background noise characteristic value N and bottom echo intensity characteristic value R; Where R0 is the preset bottom surface echo reference intensity; Step S23: Compare the signal quality parameter Q with a preset signal failure threshold Q. fail and signal recovery threshold Q rec Compare: If Q<Q fail If so, the signal is deemed invalid; If Q≥Q rec If so, the signal is considered valid; Step S3, based on the signal validity determination result, implements scanning control linkage, specifically including: Before the scan begins, the system needs to be adjusted until Q ≥ Q. rec Only after this can the scan begin, during which Q is allowed to drop to Q. fail ≤Q<Q rec When Q is greater than or equal to Q fail At this point, scanning can continue, but when Q < Q fail If the signal is deemed invalid, the weld seam scanning motion is paused, and the probe posture is adjusted. The probe is only allowed to adjust its posture in the direction perpendicular to the workpiece surface and in the rotational direction around the probe axis until Q ≥ Q. rec , When the signal is determined to be valid, the scanning continues from the paused position along the preset trajectory until the scanning is completed.

2. The automatic detection method for thin-walled welds according to claim 1, characterized in that: The fully focused ultrasonic image is generated by performing time-delay superposition within at least two pixel angle sub-regions; the pixel angle sub-region is an imaging sector defined in the workpiece cross-section coordinate system, which is determined by the preset ultrasonic propagation path type and the nominal incident angle of the wedge; the first pixel angle sub-region, with α1° as the center line, covers the lower and root regions of the weld, and this mode is the first fully focused imaging mode; the second pixel angle sub-region, with α2° as the center line, covers the upper and near-surface regions of the weld.

3. The automatic detection method for thin-walled welds according to claim 2, characterized in that: The ultrasonic propagation path corresponding to the first pixel angle sub-region is transverse wave-bottom surface reflection-defect-bottom surface reflection-transverse wave path.

4. The automatic detection method for thin-walled welds according to claim 2, characterized in that: The ultrasonic propagation path corresponding to the second pixel angle sub-region is a transverse wave-defect-transverse wave path.

5. The automatic detection method for thin-walled welds according to claim 1, characterized in that: The depth interval mask and adaptive segmentation described in step S21 specifically include: Step S211, preset background noise ROI bg Its depth ranges from 0.3T to 0.7T of the workpiece thickness T; preset bottom surface echo ROI bot Its depth range is T±1mm; Step S212: Apply Otsu adaptive threshold segmentation to the entire image I to obtain a binary image B, where regions with a connected component area greater than 50 pixels are considered as "weld or defect" regions. Step S213: Compare the binary image B with the ROI respectively. bg and ROI bot Perform a logical AND operation to obtain the background region mask M containing defect interference. bg_defect and bottom area mask M bot_defect ; Step S214, for M bg_defect and M bot_defect Invert each one and compare it with the original ROI. bg ROI bot AND, to obtain a pure background noise mask M bg and pure bottom surface echo mask M bot , used for subsequent calculations.

6. The automatic detection method for thin-walled welds according to claim 5, characterized in that: The background noise feature value N is a pure background noise mask M. bg The root mean square of the inner pixel magnitude is calculated using the following formula: , Where: M bg (i,j) is a binary mask for the pure background noise region, with a value of 1 or 0; I(i,j) is the pixel amplitude of the full-focus ultrasound image at position (i,j).

7. The automatic detection method for thin-walled welds according to claim 5, characterized in that: The characteristic value R of the bottom surface echo intensity is the pure bottom surface echo mask M. bot The following statistics for the amplitude of an inner pixel are used: (1) the arithmetic mean of the top 1% of pixels by amplitude; (2) the average of all pixels with amplitude greater than the -6 dB threshold; (3) the root mean square value of all pixels; where the defective region has been masked by M. bot Excluded, not included in the statistics.

8. The automatic detection method for thin-walled welds according to claim 5, characterized in that: The background noise ROI bg Avoiding the heat-affected zone, bottom reflection zone, and surface blind zone; the bottom echo ROI bot Avoid the center of the weld, defect areas, and areas with geometric reflectors.

9. The automatic detection method for thin-walled welds according to claim 8, characterized in that: The background noise ROI bg It is located in the middle layer of the base material body, away from the heat-affected zone of the weld, with a depth ranging from 0.3T to 0.7T.

10. The automatic detection method for thin-walled welds according to claim 5, characterized in that: The bottom surface echo ROI bot It is placed on the bottom surface of the base material near the weld, with a depth range of T±1 mm, and is located horizontally outside the heat-affected zone.

11. The automatic detection method for thin-walled welds according to claim 5, characterized in that: The background noise ROI bg With bottom echo ROI bot The position is fixed according to the workpiece geometry parameters during the initialization phase of the detection program and is dynamically updated with the probe position during the scanning process.

12. The automatic detection method for thin-walled welds according to claim 1, characterized in that: The signal failure threshold and signal recovery threshold are respectively set as follows:   ， N0 is the background noise baseline value.

13. The automatic detection method for thin-walled welds according to claim 12, characterized in that: The methods for obtaining the bottom surface echo reference intensity R0 and the background noise reference value N0 include: Step S231: Select a known calibration area in the workpiece base material area that is defect-free, has a smooth surface, and is well coupled; Step S232: Using the same full-focusing parameters as the formal scan, the probe incident angle is changed in 0.5° steps within a ±5° range by an electronically controlled rotation mechanism. The A-scan waveform at each angle is collected and the bottom surface echo peak value is recorded. Step S233: Select the angle with the highest peak value of the bottom echo as the representative angle, and adjust the bottom echo amplitude at this angle to 80% of the full scale of the display, and lock the system gain; Step S234: Acquire a sector-scan full-focus image under the stated gain condition, and extract the pure bottom-echo mask M from it. bot The average value of the pixel intensity is denoted as R0; Step S235, pure background noise mask M in the same image bg The root mean square of the pixel magnitude within the region is calculated and denoted as N0.

14. The automatic detection method for thin-walled welds according to claim 13, characterized in that: The probe attitude adjustment includes: The probe is raised and lowered along a direction perpendicular to the workpiece surface to fine-tune the coupling distance; The probe holder is rotated in small-angle steps around the probe axis, and a fully focused image is acquired and the Q value is calculated at each step; If after adjustment Q < Q fail Then repeat the lifting and rotating operations to find the optimal value until Q ≥ Q. rec ; When Q≥Q rec When the time comes, resume the weld seam scanning motion and continue executing the preset trajectory from the coordinates recorded during the pause.

Images