Rectangular tube eddy current inspection method

Abstract

This invention discloses an eddy current flaw detection method for rectangular cross-section pipes. First, the pipe is straightened multiple times using a precision straightening machine to strictly control longitudinal curvature, flatness, and corner radius deviations. Then, a flexible array probe containing planar and corner modules is driven by an inflatable airbag encased in silicone, with a pressure closed-loop control to maintain a fill coefficient ≥85% and to match differentiated eddy current parameters. Visual positioning and vacuum adsorption conveying ensure coaxiality. Defects are classified using wavelet-based denoising and a CNN neural network, and an encoder achieves ±1mm positioning. Finally, laser marking and pneumatic sorting are performed. This invention improves flaw detection accuracy and efficiency and is applicable to the inspection of pipes of various specifications.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology for pipes, and to an eddy current testing method for rectangular cross-section pipes. In particular, it relates to an eddy current testing method for rectangular cross-section metal pipes (such as high-temperature alloys for aerospace, rectangular structural pipes of titanium alloys, rectangular steel pipes for high-pressure hydraulic systems, etc.), and is especially suitable for rectangular pipes with cross-sectional dimensions of 20×10mm~150×100mm and wall thickness of 1~10mm. Background Technology

[0002] Eddy current testing is a key technology for detecting surface and near-surface defects in metal pipes based on the principle of electromagnetic induction. It is widely used in aerospace, energy, and automotive industries. Existing technologies have relatively mature eddy current testing methods for circular cross-section pipes. For example, existing technologies disclose an eddy current testing method for circular steel pipes, which shortens the blind zone to 100mm through straightening (bending degree ≤ 1.5mm / m), guide sleeve matching (2~3mm larger than the outer diameter), and auxiliary conveying (baffle + electromagnet). However, this method is only applicable to circular cross-section steel pipes and does not address the adaptation problem for irregular rectangular cross-sections.

[0003] In addition, existing technologies have disclosed automated flaw detection methods for thin-walled circular tubes of aluminum and aluminum alloys, which realize the full automation of feeding-inspection-sorting process through PLC control, but have not solved the problems of electromagnetic interference, conveying offset and blind zone optimization for rectangular cross-section tubes.

[0004] Due to their geometric characteristics (electromagnetic differences between the plane and the corners, and ease of rotation during transport), rectangular cross-section pipes face the following technical bottlenecks when using traditional methods:

[0005] Large blind zone in flaw detection: The concentrated eddy current field at the rectangular corners leads to signal disorder at the end. According to GB / T 7735 standard, the undetectable area at the end is 200~300mm. The length of scrap required for each pipe increases, and the yield decreases by 10%~15% (taking a 10m long pipe as an example, the traditional method cuts 400mm, with a yield of 96%; this method can reduce it to 200mm, with a yield of 98%).

[0006] Low detection accuracy: The gap between the probe and the rectangular surface fluctuates (50%~85%), the difference in defect signal amplitude reaches 300%, the detection rate of artificial defects (0.1~0.5mm deep) is only 80%~85%, and the false judgment rate exceeds 10%;

[0007] Poor conveying stability: When conveyed by V-rollers, the lateral offset of the rectangular tube is greater than 1mm, the circumferential rotation angle is greater than 5°, and the standard deviation of signal repeatability is greater than 5%, which cannot meet the "zero defect" inspection requirements of aerospace. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides an eddy current testing method for rectangular cross-section pipes, particularly for rectangular cross-section metal pipes (such as high-temperature alloys for aerospace, titanium alloy rectangular structural pipes, and rectangular steel pipes for high-pressure hydraulic systems). It is especially suitable for rectangular pipes with cross-sectional dimensions of 20×10mm to 150×100mm and wall thicknesses of 1 to 10mm. The method of this invention achieves: ① a blind zone ≤100mm; ② a defect detection rate ≥99% (for defects with a depth ≥0.1mm and a length ≥0.5mm); ③ a transport offset ≤0.2mm, a rotation angle ≤0.5°, and a signal repeatability standard deviation ≤1%. This aims to solve the problems of large blind zones, low accuracy, and unstable transport inherent in traditional eddy current testing methods for rectangular cross-section pipes.

[0009] To achieve the above objectives, this invention provides an eddy current testing method for rectangular cross-section pipes, the specific steps of which are as follows:

[0010] Step 1: Use a precision straightening machine to straighten the rectangular tube to be inspected, and control the longitudinal curvature of the rectangular tube to be inspected to be ≤1.0mm / m, the flatness error to be ≤0.2mm / m, and the radius deviation of the right angle transition zone to be ≤0.1mm;

[0011] Step 2: The inflatable airbag drives the flexible array probe to fit against the outer surface wall of the rectangular pipe to be inspected, with a fill factor ≥85%; wherein, the flexible array probe consists of multiple planar modules and multiple corner modules, the eddy current frequency of the planar modules is 50kHz~100kHz, the eddy current frequency of the corner modules is 1kHz~5kHz, the gain is 40~60dB, and the phase angle is 45°±5°;

[0012] Step 3: The rectangular pipe to be inspected is transported through a visual positioning and vacuum adsorption conveying system, so that it passes through the flexible array probe in Step 2 for inspection. The eddy current meter simultaneously collects dual-frequency signals, and the collected signals are denoised using wavelet transform. Then, a CNN neural network is used to classify the denoised signal data and output the judgment result.

[0013] Step 4: Use laser marking equipment to mark defect information and sort qualified and unqualified products.

[0014] Preferably, in step 1, the upper roller pressure of the straightener is 50kN~70kN, the lower roller pressure is 40kN~60kN, the roller speed is 0.3m / s~0.5m / s, and the number of straightening passes is 3~5.

[0015] Preferably, after the straightening is completed in step 1, the rectangular tube to be inspected is fed into the inflatable airbag through the input guide sleeve, and after the inspection is completed in step 3, it is sent out through the output guide sleeve; the gap between the inner wall of the input guide sleeve and the output guide sleeve and the outer surface wall of the rectangular tube to be inspected is maintained at 1mm~2mm.

[0016] Preferably, in step 2, a substrate is laid on the side of the inflatable airbag facing the outer wall of the rectangular tube to be inspected, and the side of the substrate away from the outer wall of the rectangular tube to be inspected is fixedly connected to the inflatable airbag. The planar module and the corner module are disposed on the side of the substrate facing the inflatable airbag and fixedly connected to it; the substrate is a PVDF piezoelectric film.

[0017] Preferably, in step 2, the planar module is a planar spiral coil with 100 to 200 turns; the corner module is a differential coil with a radius of 5 mm and 150 to 250 turns.

[0018] Preferably, in step 2, the planar module is used to detect surface and near-surface defects of the rectangular tube to be inspected, with a skin depth of 0.1mm to 0.3mm; the corner module is used to detect stress cracks in the right-angle transition zone of the rectangular tube to be inspected, with a skin depth of 0.5mm to 1.0mm.

[0019] Preferably, in step 3, the plane and corner contours of the rectangular tube to be inspected are identified by an edge detection algorithm, and the vacuum suction cup is adjusted in real time to ensure that the coaxiality between the central axis of the rectangular tube to be inspected and the center of the flexible array probe is ≤0.1mm / m.

[0020] Preferably, in step 3, the conveying speed is 0.5 m / s to 1.0 m / s, and the acceleration is ≤0.2 m / s². 2 .

[0021] Preferably, the conveying speed is 0.5 m / s when the wall thickness of the rectangular tube to be inspected is <3 mm; and the conveying speed is 1.0 m / s when the wall thickness of the rectangular tube to be inspected is 3 mm to 10 mm.

[0022] Preferably, in step 3, the wavelet transform is based on the db4 wavelet basis, and the number of decomposition layers is 3; the input data of the CNN neural network is the signal amplitude-phase spectrum, and the output is defect-free / surface crack / corner crack / inclusion.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. This invention addresses the problem that traditional eddy current testing often suffers from blind spots in corner areas or misjudgments of surface defects due to its difficulty in adapting to the irregular contours of rectangular pipes with their "planar + right-angle transition zones." This invention completely solves this problem through a multi-stage collaborative design: Step 1, precise straightening, provides three-dimensional precision control over longitudinal curvature, flatness, and the radius of the right-angle transition zone's fillet, offering a uniform geometric basis for subsequent probe bonding and ensuring the flexible probe's fill factor remains consistently high; Step 2 features a specially designed flexible array probe with a planar module and corner modules, paired with a 50-100kHz high-frequency (suitable for surface and near-surface defects, skin depth 0.1-0.3mm) and a 1-5kHz low-frequency (suitable for stress cracks in the right-angle transition zone, skin depth...) frequency. The dual-frequency parameters (0.5~1.0mm) are used, and adaptive fitting is achieved through inflatable airbags to achieve non-discriminatory coverage detection of the entire contour of the rectangular pipe. The guiding role of the input / output guide sleeve in step 3 ensures that the pipe passes smoothly through the detection area, avoiding eddy current field disturbance caused by offset, and providing a guarantee for stable acquisition of dual-frequency signals. Combined with db4 wavelet basis denoising and CNN neural network classification, the accuracy of defect detection is finally achieved, which greatly improves the defect identification accuracy and completely avoids the detection blind spots and misjudgments caused by the lack of a single link in traditional technology.

[0025] 2. Traditional flaw detection technology lacks full-chain precision control, has poor signal repeatability, and is easily affected by external interference, making it difficult to meet the stringent requirements of high-end manufacturing. This invention constructs a stability assurance system through multi-stage collaboration: In step 1, the pressure parameters of the upper roller (50~70kN) and lower roller (40~60kN) of the straightening machine, along with the precise control of 3~5 straightening passes, ensure the consistency of the pipe's condition upon entering the testing stage, avoiding fluctuations in the testing signal caused by differences in the initial state of the pipe; In step 2, the setting of the PVDF piezoelectric film substrate ensures a stable connection between the planar module, the corner module, and the inflatable airbag. Its excellent flexibility and stability also reduce the impact of airbag pressure changes on the module's testing posture. Combined with parameter optimization of gain 40~60dB and phase angle 45°±5°, electromagnetic interference is further suppressed; In step 3, the synergistic effect of visual positioning and vacuum adsorption conveying system is utilized. The vacuum suction cup is adjusted in real time through an edge detection algorithm to ensure that the coaxiality between the pipe's central axis and the probe center is ≤0.1mm / m. At the same time, the adaptation speed of 0.5~1.0m / s (dynamically adjusted according to wall thickness) and the low acceleration design of ≤0.2m / s² avoid interference from pipe movement and offset during the conveying process. The stability control of each step in the method described in this invention is superimposed on each other, which significantly improves the repeatability of the flaw detection signal and greatly enhances the reliability of the detection results, making it suitable for fields with extremely high requirements for detection accuracy and stability, such as aerospace and high-end automobile manufacturing.

[0026] 3. This invention achieves a dual improvement in testing efficiency and economy through a collaborative design across the entire process: the seamless connection between straightening in step 1 and conveying and testing in step 3, combined with the guiding effect of the input / output guide sleeves, avoids the repetitive operations of pipe transfer and positioning in traditional technologies, significantly improving the continuity of the testing process; the conveying speed in step 3 can be dynamically adjusted according to the pipe wall thickness, maximizing testing efficiency while ensuring testing accuracy; at the same time, precise control of the entire process significantly reduces the blind zone of end-point flaw detection, reducing pipe scrap loss caused by blind zones. Furthermore, the collaboration between laser marking of defect information and the sorting process in step 4 achieves an integrated closed loop of testing and sorting, avoiding the lag and misjudgment costs of manual marking and sorting, further improving production efficiency and reducing labor costs.

[0027] 4. Traditional flaw detection equipment often has a fixed structure, making it difficult to adapt to rectangular pipes of different specifications and wall thicknesses. This results in companies needing to equip themselves with multiple sets of equipment, increasing production costs. The method described in this invention greatly improves its versatility through modular and parameterized collaborative design: In step 1, the straightening machine can be adapted to the straightening requirements of different pipe specifications by adjusting the roller pressure, roller speed, and straightening pass parameters; In step 2, the modular design of the flexible array probe allows for adaptation to pipes of different cross-sectional dimensions by replacing planar modules and corner modules of different sizes, and the adaptive fitting design of the inflatable airbag further improves the compatibility with pipe size deviations; In step 3, the design of dynamically adjusting the conveying speed according to the wall thickness enables the method to cover the inspection needs of rectangular pipes with wall thicknesses of 0.1~10mm.

[0028] 5. The method described in this invention can achieve the following effects: (1) Reduced blind zone: Through straightening control, flexible probe fitting and stable conveying, the blind zone of end flaw detection is reduced from more than 200mm in the traditional method to ≤100mm, reducing waste by 30~50kg per ton of pipe (based on 10m long pipe); (2) Improved accuracy: Dual-frequency signal fusion and CNN classification algorithm make the defect detection rate ≥99% (85%~90% in the traditional method) and the misjudgment rate ≤0.5% (5%~10% in the traditional method); (3) Enhanced stability: Visual positioning and vacuum adsorption conveying make the lateral offset ≤0.2mm, the rotation angle ≤0.5°, and the signal repeatability standard deviation ≤1% (5%~8% in the traditional method), meeting the requirements of the aerospace AS9100 standard. The adaptability design of each step in the method described in this invention works together, so that this invention can be adapted to the flaw detection needs of various specifications of rectangular pipes without major modifications, greatly reducing the equipment investment and adaptation costs of enterprises, and broadening the application scenarios of the method described in this invention. Attached Figure Description

[0029] Figure 1 This is a schematic diagram showing the positional structure of the flexible array probe and the rectangular tube in this invention.

[0030] Figure 2 This is a flowchart of the dual-frequency signal processing in this invention.

[0031] In the diagram: 1 is the rectangular tube to be inspected, 2 is the inflatable airbag, 3 is the corner module, and 4 is the planar module. Detailed Implementation

[0032] The technical solutions of the present invention will be clearly and completely described in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the present invention are within the scope of protection of the present invention.

[0033] Unless otherwise specified in the specific circumstances, the numerical ranges listed herein include upper and lower limits, as well as all integers and fractions within that range, but are not limited to the specific values ​​listed when the range is defined.

[0034] I. An Eddy Current Testing Method for Rectangular Cross-Section Pipes

[0035] This invention addresses the core technical problems of traditional eddy current testing (EDT) for rectangular cross-section pipes, including difficulties in adapting to irregular contours, low detection accuracy, poor stability, and discontinuous processes. It proposes a comprehensive eddy current testing solution that adapts to irregular rectangular pipe structures while balancing accuracy and efficiency, based on a holistic optimization approach. Firstly, addressing the primary issue in traditional techniques where irregular geometric shapes of rectangular pipes lead to poor probe fit and distorted detection signals, this invention establishes a pre-design logic of shaping before testing. Longitudinal bending, planar warping, and uneven fillet radius in right-angle transition zones of rectangular pipes directly disrupt the coupling stability between the probe and the pipe. Therefore, this invention first designs a precision straightening pre-processing stage. By accurately controlling the deviations in longitudinal bending, flatness, and fillet radius in right-angle transition zones, it provides a well-formed and consistent pipe foundation for subsequent testing, mitigating the interference of geometric deviations on detection accuracy from the source. This is a prerequisite for achieving high-precision subsequent testing. Secondly, addressing the issue that traditional single probes cannot adapt to the irregular contours of rectangular pipes with their planar and right-angle transition zones, and are prone to blind spots in edge and corner detection, this invention proposes a modular adaptation + adaptive driving detection probe design. Based on the differences in defect types in different areas of the rectangular pipe (surface defects are mainly surface and near-surface defects, while edge and corner defects are mainly stress cracks), a flexible array probe composed of planar and edge / corner modules is specifically designed, coupled with differentiated dual-frequency parameters (high frequency adapted for planar detection, low frequency adapted for edge / corner detection) to achieve full contour coverage without differences; at the same time, an inflatable airbag driving method is adopted, combined with the flexible characteristics of the PVDF piezoelectric film substrate, to ensure adaptive adhesion between the probe and the pipe surface, ensuring a stable fill factor of ≥85%, and solving the coupling problem of irregular contours. Thirdly, addressing the problem that traditional conveying systems are prone to pipe movement and offset, leading to inaccurate defect positioning, a collaborative conveying approach of visual positioning + vacuum adsorption is designed. The edge detection algorithm of the visual positioning system identifies the pipe contour in real time and dynamically adjusts the vacuum suction cup adsorption position to ensure high-precision coaxiality between the pipe's central axis and the probe center. Simultaneously, the transport parameters are optimized, employing a dynamic transport speed and low acceleration design adapted to the pipe wall thickness to avoid pipe deviation caused by inertia, providing a stable motion reference for synchronous acquisition of dual-frequency signals and precise defect location. Then, addressing the weaknesses of traditional signal processing methods such as weak anti-interference capability and low defect classification accuracy, a signal processing approach combining precise denoising and intelligent classification is established. Considering that eddy current detection signals are easily mixed with electromagnetic interference, vibration interference, and other noise, a db4 wavelet basis is used for three-layer decomposition denoising, effectively preserving defect signal characteristics. Based on the amplitude-phase spectrum of the dual-frequency signals, a CNN neural network is used for classification, accurately outputting four categories of results: no defects, surface cracks, corner cracks, and inclusions, solving the problem of high misclassification rates in traditional classification methods. Finally, to achieve seamless integration between inspection and subsequent processing, a closed-loop finishing strategy of precise marking and efficient sorting is designed.Defect information is marked in real time using laser marking equipment, providing a basis for subsequent traceability; combined with a PLC-controlled pneumatic push rod, qualified and unqualified products are accurately sorted, forming a complete closed loop of detection-marking-sorting, improving the continuity and economy of the detection process. In summary, the design of this invention is not a local optimization of a single link, but rather based on the core logic of mutual support and synergistic effect among each link: straightening preprocessing provides the foundation for probe fitting, probe adaptation and guide sleeve guidance provide the guarantee for signal acquisition, stable conveying provides the premise for accurate positioning, intelligent signal processing provides support for defect identification, and closed-loop sorting provides the guarantee for process implementation. The design of each link closely revolves around the core technical problem of eddy current testing of rectangular pipes, and through the synergistic optimization of the entire process, the goal of high precision, high stability, and high efficiency in flaw detection is ultimately achieved. Therefore, this invention proposes an eddy current testing method for rectangular cross-section pipes, the specific steps of which are as follows:

[0036] Step 1: Use a precision straightening machine to straighten the rectangular tube to be inspected, and control the longitudinal curvature of the rectangular tube to be inspected to be ≤1.0mm / m, the flatness error to be ≤0.2mm / m, and the radius deviation of the right angle transition zone to be ≤0.1mm.

[0037] Step 2: The inflatable airbag 2 drives the flexible array probe to fit against the outer surface wall of the rectangular tube 1 to be inspected, with a filling coefficient ≥85%; wherein, the flexible array probe is composed of multiple planar modules 4 and multiple corner modules 3, the eddy current frequency of the planar modules is 50kHz~100kHz, the eddy current frequency of the corner modules is 1kHz~5kHz, the gain is 40~60dB, and the phase angle is 45°±5°.

[0038] Step 3: The rectangular pipe to be inspected is transported through a visual positioning and vacuum adsorption conveying system, so that it passes through the flexible array probe in Step 2 for inspection. The eddy current meter simultaneously collects dual-frequency signals, and the collected signals are denoised using wavelet transform. Then, a CNN neural network is used to classify the denoised signal data and output the judgment result.

[0039] Step 4: Use laser marking equipment to mark defect information and sort qualified and unqualified products.

[0040] In some embodiments of the present invention, in step 1, the upper roller pressure of the straightener is 50kN~70kN, the lower roller pressure is 40kN~60kN, the roller speed is 0.3m / s~0.5m / s, and the straightening passes are 3~5 times. A 12-roll precision straightener is used to straighten the rectangular tube to be inspected in multiple passes, and the processed parameters are controlled within the following target parameters:

[0041] Longitudinal curvature ≤1.0 mm / m, monitored in real time by a laser bending tester, which is better than 1.5 mm / m for circular pipes, ensuring the overall straightness of the pipe and avoiding the offset of the central axis due to bending during subsequent transportation, thus ensuring the coaxiality control of the pipe and the flexible array probe.

[0042] With a flatness error of ≤0.2mm / m, the gap between the pipe surface and the reference surface is measured by a dial indicator to achieve precise control for rectangular plane warping, which solves the problems of uneven probe coupling gap and eddy current field distortion caused by plane concavity and convexity in traditional flaw detection.

[0043] The radius deviation of the rounded corners in the right-angle transition zone is ≤0.1mm. This is detected using a profilometer, which avoids straightening cracks caused by stress concentration at the corners while ensuring the regularity of the corner contours. This allows the subsequent corner detection module to accurately fit and effectively detect stress cracks. This invention specifically designs the relevant parameters of the straightening machine for irregular structures with a planar surface and a right-angle transition zone in rectangular pipes. Compared to circular pipes, rectangular pipes are more prone to warping in their planar areas and stress concentration at the corners. If single-pass straightening or mismatched parameters are used, it can easily lead to scratches on the pipe surface and cracks at the corners. The multi-pass straightening of 3 to 5 times can gradually release the residual stress generated during pipe rolling and transportation, avoiding straightening cracks caused by stress concentration. The differentiated setting of the upper and lower roller pressure and the low speed matching of 0.3m / s to 0.5m / s can ensure that the straightening force is applied evenly to the plane and corner areas of the pipe, which can not only achieve the flat correction of plane warping, but also avoid deformation of the corners due to excessive force, and finally achieve the regularization of the pipe geometry, providing a consistent workpiece state for subsequent inspection.

[0044] In some embodiments of the present invention, after straightening in step 1, the rectangular tube to be inspected is fed into the inflatable airbag through the input guide sleeve, and after inspection in step 3, it is discharged through the output guide sleeve. The gap between the inner wall of the input and output guide sleeves and the outer surface wall of the rectangular tube to be inspected is maintained at 1mm to 2mm. Both the input and output guide sleeves are made of 99% alumina ceramic material, and the inner hole size is 1 to 2mm larger than the outer diameter of the tube. For example, a 60×40mm tube is fitted with a 62×42mm guide sleeve. The length of the input guide sleeve is controlled at 150mm to ensure stable establishment of the eddy current field. The length of the output guide sleeve is not limited and can be adjusted according to the inspection needs without affecting the stability of the eddy current field.

[0045] In some embodiments of the present invention, in step 2, such as Figure 1As shown, the flexible array probe consists of four planar modules and four corner modules, forming a detection area surrounding the rectangular tube to be inspected. This allows the rectangular tube to pass stably and continuously through the detection area, thereby achieving flaw detection. A substrate is laid on the surface of the inflatable airbag facing the outer wall of the rectangular tube to be inspected. The side of the substrate facing away from the outer wall of the rectangular tube to be inspected is fixedly connected to the inflatable airbag. The planar modules and corner modules are located on the surface of the substrate facing the inflatable airbag and are fixedly connected to it. The substrate is a PVDF piezoelectric film with a thickness of 0.5 mm. The flexible array probe is moved by an inflatable airbag encased in silicone (working pressure 0.3~0.5MPa). As the airbag gradually fills with gas, it propels the flexible array probe toward the outer contour of the rectangular tube, enabling it to adaptively conform to the tube's outer contour. The filling coefficient is controlled in a closed loop by a pressure sensor (prioritizing the detection of the correlation between the airbag's working pressure and the filling coefficient, and then using the pressure sensor to detect the airbag's working pressure to control the filling coefficient), maintaining the filling coefficient at ≥85% (fill coefficient = effective probe coverage area ÷ tube outer surface area × 100%). In this invention, the inflatable airbag is a single unit; separate units cannot be used to move the flexible array probe individually, as this would result in poor conformation. The modular layout of four planar modules and four corner modules enables seamless, full-coverage inspection of the outer contour of rectangular pipes. The four planar modules correspond to the four planar areas of the pipe, while the four corner modules correspond to the four right-angle transition areas. This customized layout fundamentally avoids the problem of insufficient coverage of corner areas by traditional single-probe or asymmetrical probes, preventing defects such as corner stress cracks from being missed due to probe coverage. Furthermore, the independent configuration of each module allows for matching with differentiated eddy current testing parameters (high frequency for planar areas, low frequency for corner areas), laying a hardware foundation for accurate detection of defects in different regions. On the one hand, the substrate serves as the connecting carrier between the planar module, the corner module, and the inflatable airbag, fixing the module to the side facing the airbag. This allows the driving force of the airbag to be evenly transmitted to each module through the substrate, preventing displacement or tilting of the modules due to uneven force distribution and ensuring the relative position stability of the modules during the testing process. On the other hand, the 0.5mm thick PVDF piezoelectric film has excellent flexibility and fatigue resistance, allowing it to adapt to the expansion and contraction of the inflatable airbag, conforming to the minute geometric undulations on the pipe surface. At the same time, its good insulation and wear resistance can prevent electromagnetic interference between the module and the airbag, extending the service life of the probe.The silicone-encased inflatable airbag (working pressure 0.3~0.5MPa) provides the power drive for probe adhesion. The silicone encapsulation buffers the airbag pressure, preventing direct contact with the substrate or tubing and thus avoiding wear and scratches. The optimized working pressure range of 0.3~0.5MPa provides sufficient thrust for probe adhesion to the tubing surface without causing deformation due to excessive pressure. Compared to the fixed adhesion of traditional rigid probes, the adaptive drive of the inflatable airbag adjusts the adhesion force in real time according to the tubing's geometric deviations (such as slight bends or uneven corner radius), ensuring a consistently stable contact between the probe and the tubing surface. The pressure sensor's closed-loop control fill factor is ≥85%, ensuring eddy current coupling stability and detection signal effectiveness. A fill factor that is too low will result in insufficient eddy current field coverage and detection blind spots; excessive fluctuations in the fill factor will cause instability in the amplitude and phase of the eddy current signal, leading to misjudgments of defects. This invention achieves precise and stable control of the filling coefficient by establishing a pressure-filling coefficient correspondence, real-time detection of airbag pressure by a pressure sensor, and closed-loop adjustment of pressure to maintain a filling coefficient ≥85%. This eliminates the need for complex area detection devices; the bonding effect can be indirectly controlled solely through pressure parameters. The control method is simple and efficient, and can adapt to dynamic position changes during pipe transportation, ensuring that the eddy current field always uniformly covers the pipe detection area. This provides a guarantee for the stable acquisition of dual-frequency signals from the planar module and the corner module.

[0046] In some embodiments of the present invention, in step 2, the planar module is a planar spiral coil with 100 to 200 turns; the corner module is a differential coil with a radius of 5 mm and 150 to 250 turns. The planar spiral coil features a large detection area and uniform magnetic field distribution, perfectly matching the planar area of ​​the rectangular tube, enabling full-coverage scanning detection of the planar area and avoiding detection blind spots. The differential coil has advantages such as strong anti-interference capability and high sensitivity to crack-like defects. Furthermore, its 5 mm radius microstructure can be precisely embedded in the right-angle transition area of ​​the rectangular tube, fitting the narrow spatial contour of the corners, solving the problem that traditional coils cannot penetrate deep into corner areas for detection. The planar helical coil is limited to 100-200 turns. This range balances the relationship between detection sensitivity and signal response speed. Too few turns result in insufficient magnetic field strength, failing to effectively excite eddy currents within the pipe and making it difficult to detect minute surface defects. Too many turns increase coil inductance, reducing signal response speed and making it unsuitable for conveyor speeds of 0.5m / s to 1.0m / s, leading to missed defects. The corner module differential coil is limited to 150-250 turns, slightly more than the planar module. This is because stress cracks in corner areas are often deep defects, requiring stronger magnetic field penetration. More turns enhance the coil's magnetic field strength, improving the detection capability for deep defects. Simultaneously, the 5mm miniature structure prevents the coil from becoming too large to fit the corner space due to excessive turns.

[0047] In some embodiments of the present invention, in step 2, the planar module is used to detect surface and near-surface defects of the rectangular tube to be inspected. The eddy current frequency of the planar module is 50kHz~100kHz, and the skin depth is 0.1mm~0.3mm. The corner module is used to detect stress cracks in the right-angle transition zone of the rectangular tube to be inspected. The eddy current frequency of the corner module is 1kHz~5kHz, and the skin depth is 0.5mm~1.0mm. The gain is 40~60dB, the phase angle is 45°±5°, and the corner eddy current interference is suppressed through electromagnetic simulation optimization. The frequency of eddy current detection directly determines the skin depth (the higher the frequency, the shallower the skin depth). This invention achieves precise matching for defect detection through frequency differentiation design: the planar module uses a high frequency of 50kHz~100kHz, corresponding to a skin depth of 0.1mm~0.3mm, which can accurately focus on minute defects (such as surface microcracks and shallow inclusions) on and near the surface of the pipe, avoiding background signal interference caused by excessive penetration of the high-frequency magnetic field; the corner module uses a low frequency of 1kHz~5kHz, corresponding to a skin depth of 0.5mm~1.0mm, which allows the magnetic field to penetrate to the deep layer of the corner area, accurately detecting stress cracks in the right-angle transition zone. These cracks are mostly caused by processing stress and tend to extend to the deep layer, solving the problem that traditional single-frequency methods cannot simultaneously detect surface and deep defects. At the same time, the right-angle transition zone of rectangular pipes is prone to edge effects, leading to disordered eddy current distribution and interference with the detection signal. Therefore, this invention limits the gain to 40dB~60dB, which can amplify weak defect signals to a identifiable range while avoiding excessive gain that would lead to simultaneous amplification of noise signals. The phase angle is limited to 45°±5° (optimized by electromagnetic simulation), which can adjust the phase position of the eddy current signal, allowing the defect signal to be distinguished from the interference signal generated by the edge effect at the corners in phase, thereby effectively suppressing corner eddy current interference, improving the signal-to-noise ratio of the defect signal, and ensuring accurate identification of defects in the corner areas. This invention achieves specialized detection of the entire area of ​​the rectangular pipe from the plane to the corners by clearly defining the detection objects of the two types of modules, avoiding detection overlap or omissions. Combined with the flexible array probe fitting design, it forms a detection mode of full-area coverage + precise matching, further improving the comprehensiveness and accuracy of overall flaw detection.

[0048] In some embodiments of the present invention, in step 3, the plane and corner contours of the rectangular tube to be inspected are identified by an edge detection algorithm, and the vacuum suction cup (adjustable suction force 0.2MPa~0.5MPa) is adjusted in real time to ensure that the coaxiality between the central axis of the rectangular tube to be inspected and the center of the flexible array probe is ≤0.1mm / m. In specific implementations, the conveyor track can be driven by dual linear motors (positioning accuracy ±0.01mm) and equipped with a 20-megapixel CCD vision system (sampling frequency 500fps). However, the present invention is not limited to these two devices; other known prior art devices that can achieve the same effect as these two can be used in the present invention. The vision system, combined with edge detection algorithms, can quickly and accurately identify the contours of the plane and right-angle transition zone of rectangular pipes, and capture the positional offset of the pipes relative to the flexible array probes in real time (such as lateral movement and angular deflection), providing quantitative deviation data for subsequent attitude adjustment; the precise conveying equipment, with a positioning accuracy of ±0.01mm, provides a stable and non-moving motion reference for pipe conveying, avoiding positional fluctuations caused by mechanical vibration and roller wear in traditional conveying methods. Based on deviation data collected by the vision system, the suction force of the vacuum suction cup can be adjusted in real time to flexibly correct minor offsets and tilts of the pipe. This ultimately achieves the control target of a coaxiality of ≤0.1mm / m between the central axis of the rectangular pipe to be inspected and the center of the flexible array probe. This avoids uneven coupling gaps caused by misalignment between the pipe and probe axes. If the coaxiality is too poor, the distance between different areas of the probe and the pipe surface will vary significantly, causing disordered eddy current magnetic field intensity distribution, leading to defect signal distortion and increased positioning deviation. Precise coaxiality control ensures that the coupling gap between each module of the probe and the pipe surface is uniform, keeping the eddy current field stable. This guarantees that the dual-frequency signals from the planar and corner modules can accurately reflect the internal defects of the pipe, significantly improving the defect detection rate and positioning accuracy. In this invention, coaxiality refers to the degree of coincidence between the geometric central axis of the rectangular pipe to be inspected and the detection central axis of the flexible array probe.

[0049] In some embodiments of the present invention, in step 3, the conveying speed is 0.5 m / s to 1.0 m / s, and the acceleration is ≤0.2 m / s². 2Specifically, the conveying speed needs to be dynamically adjusted according to the pipe wall thickness. When the wall thickness of the rectangular pipe to be tested is <3mm, the conveying speed is 0.5m / s; when the wall thickness of the rectangular pipe to be tested is 3mm~10mm, the conveying speed is 1.0m / s. This invention found that the signal acquisition quality of eddy current testing is directly related to the interaction time between the probe and the pipe, and the interaction time is determined by the conveying speed. The interaction time requirements of pipes with different wall thicknesses are fundamentally different: for thin-walled rectangular pipes with a wall thickness <3mm, the defects are mostly concentrated on the surface and near the surface, relying on the high-frequency eddy current detection of 50kHz~100kHz using a planar module. The skin depth of the high-frequency eddy current is only 0.1mm~0.3mm, which requires higher timeliness of signal acquisition. If the conveying speed is too fast, the interaction time between the probe and the pipe is too short, and the eddy current analyzer cannot fully acquire the signal characteristics of the defect area, easily leading to missed detection of defects such as micro-cracks and inclusions. Therefore, the conveying speed is limited to 0.5 m / s. By extending the interaction time, it is ensured that the high-frequency eddy current signal can completely capture the amplitude-phase changes of surface and near-surface defects, thus guaranteeing detection accuracy. For thick-walled rectangular pipes with a wall thickness of 3-10 mm, the focus of detection is on deep stress cracks in the right-angle transition zone. This relies on low-frequency eddy current detection at 1 kHz to 5 kHz using the corner module. The skin depth of low-frequency eddy currents can reach 0.5 mm to 1.0 mm, with stronger signal penetration capabilities. Effective signals of deep defects can be acquired without an excessively long interaction time. Therefore, the conveying speed is limited to 1.0 m / s to maximize detection efficiency without sacrificing the defect detection rate and avoid production waste caused by low-speed conveying. This design of differential speed adjustment based on wall thickness breaks through the limitations of traditional technology that only adapts to a single speed for all pipe materials, achieving the dual goals of precision and efficiency in flaw detection of pipes of different specifications. Meanwhile, the magnitude of acceleration directly determines the inertial impact intensity during pipe startup and speed change: if the acceleration is too high, the pipe will shift or deviate due to inertia, disrupting the coaxiality between the pipe and probe centers established by the visual positioning and vacuum adsorption systems. This causes fluctuations in the coupling gap between the probe and the pipe, leading to eddy current field distortion and resulting in defect signal distortion. The acceleration is limited to ≤0.2 m / s². 2 This invention enables smooth start-up, stopping, and speed changes of the pipe, avoiding interference from inertial impacts on the pipe's posture. It ensures that the coupling gap between each module of the probe and the pipe surface remains uniform throughout the entire transport process, providing reliable motion state assurance for the stable acquisition of dual-frequency eddy current signals. This invention requires no additional hardware; it can adapt to the flaw detection needs of rectangular pipes with different wall thicknesses simply by adjusting parameters at the software level. This avoids the drawbacks of traditional technologies that require changing the transport mechanism or adjusting the probe layout to adapt to different pipe materials. Enterprises can directly set parameters based on existing equipment, significantly reducing the cost of technology implementation and broadening the application scenarios of the solution.

[0050] In some embodiments of the present invention, in step 3, the wavelet transform is a db4 wavelet basis, and the decomposition layer is 3 layers; the input data of the CNN neural network is the signal amplitude-phase spectrum, and the output is four types of results: no defects / surface cracks / corner cracks / inclusions. The pipe is recorded in real time by an encoder to ensure that the defect positioning accuracy reaches ±1mm. Figure 2 As shown, this invention creatively integrates and optimizes three techniques—wavelet transform denoising, CNN neural network classification, and encoder localization—to construct an integrated processing loop for signal purification, defect classification, and location calibration. This overcomes the limitations of traditional eddy current testing, which suffers from the disconnect between signal processing and localization, and the ambiguity in defect classification. It achieves high-precision identification and accurate localization of defects in rectangular pipes. Specifically, addressing the issue that traditional wavelet denoising uses general parameters and cannot adapt to the dual-frequency detection characteristics of high-frequency signals in the plane and low-frequency signals at the corners of rectangular pipes, this invention employs customized parameters of a db4 wavelet basis and a 3-layer decomposition. Utilizing the excellent time-domain and frequency-domain localization characteristics of the db4 wavelet basis, it accurately distinguishes defect signals from interference noise in different frequency bands. The 3-layer decomposition achieves a balance between preserving the characteristics of high-frequency signals and denoising low-frequency signals, avoiding distortion of high-frequency plane micro-crack signals and residual noise in low-frequency corner areas. The processed dual-frequency signals are transformed into an amplitude-phase spectrum integrating both amplitude variation and phase shift information, providing a high-quality data source for subsequent defect classification. To address the limitations of traditional eddy current testing methods, which rely on thresholding or SVM algorithms and can only achieve binary judgment of defects (defective / no defects) while failing to distinguish between the three types of defects unique to rectangular pipes: surface cracks, corner cracks, and inclusions, this invention uses amplitude-phase maps as input to a CNN neural network. Through multi-layer convolution and pooling operations, it automatically extracts deep feature differences from the maps, achieving accurate classification of "no defects / surface cracks / corner cracks / inclusions," thus completely solving the problem of misjudgment inherent in traditional methods. Furthermore, to address the shortcomings of traditional eddy current testing, where signal processing and position recording are independent, resulting in low positioning accuracy and inability to meet the requirements of laser marking and precise sorting, this invention innovatively links the encoder's real-time displacement recording with the CNN defect classification results in a precise time sequence. The encoder simultaneously outputs the pipe's travel position coordinates at the instant the CNN identifies a defect signal. Combined with the stability control and coaxiality assurance of the conveying system, this achieves high-precision positioning of ±1mm, upgrading the detection results from qualitative judgment to a quantitative conclusion of defect type and precise location.

[0051] In some embodiments of the present invention, laser marking equipment is used to mark defect information and sort qualified and unqualified products. During marking, an XYZ three-axis laser marking machine (positioning accuracy ±0.05mm) is used to mark the detected defects on the pipe surface with a QR code indicating "defect type + depth + location" (e.g., "LC-0.2mm-500mm" indicates a 0.2mm deep corner crack 500mm from the end). During sorting, a PLC-controlled pneumatic pusher (response time ≤0.1s) can be used to push qualified and unqualified products into their corresponding baskets, achieving 100% sorting accuracy.

[0052] II. An eddy current testing device for rectangular cross-section pipes

[0053] The eddy current testing equipment for rectangular cross-section pipes includes an inflatable airbag, a flexible array probe, an eddy current meter, a vacuum adsorption conveying system, and a visual positioning device. The inflatable airbag has a receiving cavity for mounting planar and corner modules. Figure 1As shown, the flexible array probe consists of four planar modules and four corner modules, forming a detection area surrounding the rectangular tube to be inspected. This allows the rectangular tube to pass stably and continuously through the detection area, thereby achieving flaw detection. A substrate is laid on the surface of the inflatable airbag facing the outer wall of the rectangular tube to be inspected. The side of the substrate facing away from the outer wall of the rectangular tube to be inspected is fixedly connected to the inflatable airbag. The planar modules and corner modules are located on the surface of the substrate facing the inflatable airbag and are fixedly connected to it. The substrate is a PVDF piezoelectric film with a thickness of 0.5 mm. The flexible array probe is driven by an inflatable airbag encased in silicone (working pressure 0.3~0.5MPa). As the airbag gradually fills with gas, it propels the flexible array probe toward the outer contour of the rectangular tube, enabling it to adaptively conform to the tube's outer contour. The filling coefficient is controlled in a closed loop by a pressure sensor (prioritizing the detection of the correlation between the working pressure of the inflatable airbag and the filling coefficient, and then using the pressure sensor to detect the working pressure of the inflatable airbag to control the filling coefficient), maintaining the filling coefficient at ≥85% (fill coefficient = effective probe coverage area ÷ outer surface area of ​​the tube × 100%). The planar module is a planar helical coil with 100~200 turns; the corner module is a differential coil with a radius of 5mm and 150~250 turns. The planar module is used to detect surface and near-surface defects in the rectangular tube to be inspected. The eddy current frequency of the planar module is 50kHz~100kHz, and the skin depth is 0.1mm~0.3mm. The corner module is used to detect stress cracks in the right-angle transition zone of the rectangular tube to be inspected. The eddy current frequency of the corner module is 1kHz~5kHz, and the skin depth is 0.5mm~1.0mm. The gain is 40~60dB, and the phase angle is 45°±5°. Electromagnetic simulation optimization is used to suppress corner eddy current interference. The vacuum adsorption conveying system includes a vacuum suction cup and a conveying track. The conveying track is driven by dual linear motors (positioning accuracy ±0.01mm). The vacuum suction cup adjusts the position of the rectangular tube to be inspected by suction. The visual positioning device uses a 20-megapixel CCD vision system (sampling frequency 500fps), which can work with the vacuum suction cup to adjust the position of the central axis of the rectangular tube to be inspected, ensuring that its coaxiality with the center of the flexible array probe is ≤0.1mm / m. In practical use, a control system can also be configured. The control system can automatically adjust the suction force of the vacuum suction cup and, in conjunction with the information fed back by the vision positioning device, adjust the position of the rectangular tube to be inspected in a timely manner.This control system is responsible for receiving deviation data from the visual positioning device, calculating correction strategies, and issuing suction adjustment and position adjustment commands. It also collaborates with the flexible array probe, eddy current analyzer, and laser marking equipment to receive signals collected by these devices. The system uses wavelet transform to denoise the signals, and then employs a CNN neural network to classify the denoised signal data and output a judgment result. Subsequently, based on the output judgment result, the control system processes the data and sends it to the laser marking equipment to mark defect information and sort qualified and unqualified products. The control system equipment includes, but is not limited to, programmable logic controllers (PLCs), machine vision motion control devices (such as integrated machine vision motion control machines), and embedded PC-based controllers. The visual positioning device can be an integrated intelligent vision system or a split-type vision system.

[0054] III. Examples and Comparative Examples

[0055] 1. Experimental Materials and Equipment

[0056] (1) Pipe material to be inspected: TC4 titanium alloy rectangular tube (60×40×3mm, length 10m, surface roughness Ra1.6μm), containing artificial defects (0.1~0.5mm deep and 1~5mm long cracks at the corners and edges, and 0.1~0.3mm deep pitting corrosion on the surface).

[0057] (2) Main equipment:

[0058] 12-roll precision straightener (model JZ-12B, roll diameter 80mm, pressure range 0~100kN).

[0059] Flexible array eddy current probe (self-made, planar module coil with 150 turns, corner module with 200 turns);

[0060] Dual-frequency eddy current meter (model ET-3000, frequency 1kHz~100kHz, sampling rate 1MHz).

[0061] Visual positioning system (model VS-2000, resolution 5120×3200, lens focal length 50mm).

[0062] PLC controller (model S7-1214C, with motion control module).

[0063] 2. Examples and Comparative Examples

[0064] (1) Example:

[0065] Step 1: Straightening pretreatment:

[0066] The pipe is fed into the JZ-12B straightening machine, with the upper roller pressure set to 60kN, the lower roller pressure to 50kN, and the roller speed to 0.5m / s. After three passes of straightening, the laser bending tester detects a curvature of 0.8 mm / m, a flatness error of 0.15 mm / m, and a corner radius deviation of 0.08 mm, which meets the requirements of step 1.

[0067] Step 2: Probe and parameter configuration:

[0068] Install a 62×42mm ceramic guide sleeve, inflate the flexible array probe to 0.4MPa, and the pressure sensor feedback fill factor is 88%;

[0069] The planar module was set to 80kHz, the corner module to 3kHz, the gain to 50dB, and the phase angle to 45°. The eddy current meter was calibrated after a 30-minute warm-up period (using a standard test block: Φ2mm through hole, 0.2mm depth).

[0070] Step 3: Transport and Flaw Detection

[0071] The vision system identifies the outline of the pipe, and the vacuum suction cup adsorbs the center of the wide surface of the pipe and pushes it through the probe at a speed of 0.8m / s;

[0072] The eddy current meter synchronously acquires signals, and after wavelet denoising, it is input into the CNN model (the training set contains 1000 sets of defect signals). The defect type and location are output in real time: a 0.15mm deep corner crack is detected at 95mm from the end, and a 0.2mm deep planar crack is detected at 1500mm from the end.

[0073] Step 4: Marking and Sorting

[0074] The laser marking machine generates a QR code at the corresponding position, and the pneumatic pusher pushes the defective product into the waste basket. The sorting time is 0.5 seconds per piece.

[0075] (2) Comparative Example

[0076] 1) Comparative Example 1: Traditional eddy current testing method (refer to GB / T 7735-2016 "Eddy current testing method for steel pipes")

[0077] The operation steps are as follows:

[0078] Pre-treatment: Straightening is performed using a conventional straightening machine, with longitudinal curvature controlled to ≤1.5mm / m, and no requirements for flatness or fillet radius deviation control.

[0079] Probe compatibility: Use a single specification of circular or rectangular eddy current probe. The gap between the probe and the pipe surface is 2~3mm, and the filling factor is usually 50%~70%.

[0080] Detection parameters: A single eddy current frequency is used (conventional techniques often choose 20kHz~50kHz), gain is 30~50dB, and the phase angle is not specifically optimized.

[0081] Conveying method: V-roller or chain conveyor, conveying speed 0.3~0.5m / s, no visual positioning and coaxiality control, lateral offset of pipe >1mm, rotation angle >5°.

[0082] Signal processing: Hardware filtering or simple software filtering is used to determine defects by thresholding, and only "defective / no defect" results can be output.

[0083] Sorting: Defect locations are manually marked, and qualified and unqualified products are manually sorted.

[0084] Technical limitations: The blind zone for end flaw detection in this comparative example reaches 200-300mm, reducing the yield by 10%-15%. The defect detection rate is only 85%-90%, with a false positive rate exceeding 10%, making it unable to effectively identify corner stress cracks. The standard deviation of signal repeatability is >5%, indicating poor detection stability and making it difficult to meet the needs of high-end manufacturing.

[0085] 2) Comparative Example 2: Magnetic particle inspection (refer to JB / T 8290-2011 "Magnetic Particle Inspection Machine")

[0086] The operation steps are as follows:

[0087] Pretreatment: Remove oil and oxide scale from the surface of the pipe, and preheat the pipe with a wall thickness of <3mm before magnetization.

[0088] Magnetization method: Magnetization is carried out by electric current method or rod method, with a magnetic field strength of 1200~1600A / m. For rectangular tubes, the magnetization direction needs to be adjusted to cover the corner areas.

[0089] Magnetic powder application: Use a wet magnetic powder suspension (concentration 10~20g / L) to apply by spraying or soaking, with a residence time of 1~2min.

[0090] Defect observation: Observe the magnetic powder accumulation under a black light (ultraviolet wavelength 365nm) and manually determine the defect type and location.

[0091] Post-processing: demagnetization (residual magnetism < 0.3mT), cleaning, and drying of the pipes.

[0092] Technical limitations: The method in Comparative Example 2 is only applicable to ferromagnetic materials and cannot detect non-ferromagnetic rectangular tubes such as titanium alloys and aluminum alloys. The detection efficiency is low; inspecting a single 10m long tube takes approximately 15-20 minutes, making automated batch inspection difficult. It also has low sensitivity to near-surface defects and is easily affected by the surface roughness of the tube, with defect judgment relying on manual experience.

[0093] (3) Performance verification

[0094] Example:

[0095] Blind zone test: A rectangular sample tube was made according to the YB / T 4083-2020 standard (Φ5mm through holes were set at 50mm, 100mm and 150mm from the end respectively). The flaw detection results showed that the defect signal at 100mm was clear (signal-to-noise ratio 15dB), and the signal at 50mm was weak (signal-to-noise ratio 8dB). The blind zone was confirmed to be 95mm.

[0096] Accuracy test: 100 pipes with artificial defects were inspected, with a defect detection rate of 99.2% (one pipe with a 0.08mm deep crack was missed) and a false positive rate of 0.8% (two pipes with signal interference were falsely identified as defects).

[0097] Stability test: 500 pipes were continuously tested, with a signal amplitude standard deviation of 0.8%, lateral offset of 0.15mm, and rotation angle of 0.3°, meeting the repeatability requirements.

[0098] Comparative example:

[0099] Blind zone test: The same blind zone test as in the example was performed. The results showed that the signal-to-noise ratio of the defect signal was 12dB at 150mm from the end and only 6dB at 100mm from the end, with a blind zone of 180mm.

[0100] Accuracy test: The same accuracy test as in the example was performed. Results: When 100 pipes with artificial defects were inspected, the defect detection rate was only 87% (13 pipes were missed, of which 7 were edge cracks), and the misjudgment rate was 12% (12 pipes were misjudged as defects due to signal interference).

[0101] Stability test: The same stability test as in the example was conducted. Results: After continuous testing of 500 pipes, the standard deviation of signal amplitude was 5.2%, the lateral offset was 1.2 mm, and the rotation angle was 5.5°. The repeatability was poor and could not meet the consistency requirements of batch testing.

[0102] 3. Results Analysis

[0103] As can be seen from the examples and comparative examples:

[0104] (1) In terms of detection accuracy and defect detection capability, the example tested 100 TC4 titanium alloy rectangular tubes with artificial defects, achieving a defect detection rate of 99.2%, with only one 0.08mm deep micro-crack missed; the false judgment rate was as low as 0.8%, with only 2 signal interferences being misjudged; it could accurately identify 0.15mm deep corner cracks and 0.2mm deep planar cracks, and simultaneously output the defect type and precise location. In contrast, Comparative Example 1 (traditional eddy current method) tested the same batch of sample tubes, with a defect detection rate of only 87%, missing 13 tubes (7 of which were corner cracks in the right-angle transition zone); the false judgment rate was as high as 12%, 15 times that of the example, and it could not effectively distinguish the defect type, with extremely poor detection capability for deep stress cracks in the corners. As for Comparative Example 2 (magnetic particle testing method), it could not detect non-ferromagnetic materials such as TC4 titanium alloy at all, had low sensitivity to near-surface defects, and the defect judgment relied entirely on manual experience, without stable quantitative accuracy.

[0105] (2) Regarding the control of the detection blind zone, the test results of the standard sample tube in the example show that the detection blind zone is only 95mm, and the signal-to-noise ratio of the defect signal at 100mm from the end reaches 15dB, which is far superior to the industry standard. In contrast, in the test under the same conditions, the detection blind zone of Comparative Example 1 is as high as 180mm, and the traditional process even has an end blind zone of 200~300mm, which directly leads to a 10%~15% reduction in yield.

[0106] (3) Regarding the stability and batch consistency of the test, the method described in the example can continuously test 500 pipes with a signal amplitude standard deviation of only 0.8%, a pipe lateral offset of 0.15 mm, and a rotation angle of 0.3°. The test repeatability and consistency are excellent, fully meeting the requirements of industrial mass production. However, when the same batch of the comparative example was continuously tested, the signal amplitude standard deviation reached 5.2% (6.5 times that of the example), the lateral offset was 1.2 mm, and the rotation angle was 5.5°. The signal fluctuation was large, the pipe posture was out of control, and the batch test consistency was extremely poor, which could not meet the requirements of stable mass production.

[0107] (4) In terms of detection efficiency and automation level, the embodiment can realize a closed-loop automation of the entire process of detection-marking-sorting, with a conveying speed of up to 0.8 m / s and a sorting time of only 0.5 s for a single pipe. There is no manual intervention throughout the process, and the efficiency is far superior to the traditional solution. In contrast, the conveying speed of Comparative Example 1 is only 0.3~0.5 m / s, which requires manual marking of defect locations and manual sorting, resulting in low efficiency and high labor costs. The detection time of a single 10 m long pipe in Comparative Example 2 is 15~20 min, which requires multiple manual operations and cannot achieve automated batch detection at all.

[0108] (5) Regarding applicability and adaptability, the embodiments can be adapted to various rectangular tubes, including not only non-ferromagnetic rectangular tubes such as TC4 titanium alloy and aluminum alloy, but also ferromagnetic rectangular tubes, taking into account both planar micro-defects and deep corner cracks; the transmission parameters can be dynamically adjusted according to the tube wall thickness, and different specifications of tubes can be adapted without changing the hardware, resulting in low cost of on-site modification. In contrast, Comparative Example 1 uses a single probe and a single frequency, which cannot adapt to the irregular contours of rectangular tubes, has poor adaptability to different specifications of tubes, and suffers from serious missed detection of corner defects. Comparative Example 2 is only applicable to ferromagnetic materials and cannot cover commonly used non-ferromagnetic tubes such as titanium alloy and aluminum alloy, thus its application scenarios are extremely limited.

[0109] (6) As can be seen, the method described in this invention has a high detection rate of corner cracks and full defect coverage capability. The pre-processing of precision straightening controls the longitudinal curvature, flatness, and corner radius deviation of the pipe from the source, eliminating the interference of geometric deviation on probe coupling and laying the foundation for high-precision detection. The flexible array probe design of the planar module and the corner module, combined with the differentiated dual-frequency parameters of high frequency adapting to the planar and low frequency adapting to the corner, perfectly adapts to this irregularly shaped pipe with a transition area between the planar and right angles, fundamentally avoiding the problem of insufficient corner coverage of traditional single probes. At the same time, the use of inflatable airbag drive and the adaptive bonding design of PVDF flexible substrate ensures a stable filling coefficient of ≥85%, solving the problems of unstable coupling and signal distortion of irregular contours, and realizing full contour non-discriminatory coverage detection. The embodiments employing the method described in this invention achieve a high detection rate of 99.2%, an extremely low false positive rate of 0.8%, and accurate classification of micro-defects, demonstrating the feasibility of the signal processing scheme of this invention: customized db4 wavelet-based 3-layer decomposition denoising accurately adapts to the characteristics of dual-frequency detection signals, effectively removes electromagnetic and vibration interference, and completely preserves defect signal features; based on amplitude-phase spectrum CNN neural network intelligent classification, it breaks through the limitation of traditional threshold methods that can only make simple judgments such as "present / no defect", and achieves accurate classification of "no defect / surface crack / corner crack / inclusion", completely solving the problems of high false positive rate and weak defect recognition ability of traditional methods.

[0110] (7) The method described in this invention achieves high detection stability and batch consistency, and can be well adapted to the needs of industrial mass production. The extremely low signal fluctuation and minimal pipe posture deviation of 500 consecutive pipes tested in the example prove that the initial design concept for stability of this invention is reliable: by using a visual positioning and vacuum adsorption collaborative conveying system, high-precision control of the coaxiality between the pipe and the probe center is achieved, which solves the problems of inaccurate defect positioning and uneven coupling gap caused by traditional conveying movement and offset; the dynamic conveying speed and low acceleration design adapted to the wall thickness avoids the interference of inertial impact on the pipe posture and ensures the stability of the coupling state throughout the conveying process; the straightening pretreatment, stable conveying, probe bonding and signal acquisition are mutually supportive, achieving high repeatability and high consistency in the detection process, and solving the core problem of poor batch detection stability in traditional technology. Ultimately, the embodiment's detection blind zone of only 95mm and its detection efficiency far exceeding traditional solutions validate the advantages of the invention's closed-loop design throughout the entire process: the optimized ceramic guide sleeve design and stable conveying control significantly reduce the end detection blind zone, substantially decrease pipe waste, and improve yield; the laser marking real-time defect information combined with the PLC pneumatic push rod design for precise sorting forms a complete closed loop of "detection-marking-sorting," solving the problems of discontinuous processes and heavy reliance on manual labor in traditional technologies. While ensuring high precision, it significantly improves detection efficiency and adapts to the pace of industrialized mass production. The method described in this invention can be adapted to the flaw detection of rectangular pipes of different materials, breaking through the material limitations of traditional methods such as magnetic particle testing. It can cover ferromagnetic and non-ferromagnetic rectangular pipes and is suitable for the detection needs of multiple fields such as aviation, new energy, and high-end equipment.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. An eddy current testing method for rectangular cross-section pipes, characterized in that, The specific steps are as follows: Step 1: Use a precision straightening machine to straighten the rectangular tube to be inspected, and control the longitudinal curvature of the rectangular tube to be inspected to be ≤1.0mm / m, the flatness error to be ≤0.2mm / m, and the radius deviation of the right angle transition zone to be ≤0.1mm; Step 2: The inflatable airbag drives the flexible array probe to fit against the outer surface wall of the rectangular pipe to be inspected, with a fill factor ≥85%; wherein, the flexible array probe consists of multiple planar modules and multiple corner modules, the eddy current frequency of the planar modules is 50kHz~100kHz, the eddy current frequency of the corner modules is 1kHz~5kHz, the gain is 40~60dB, and the phase angle is 45°±5°; Step 3: The rectangular pipe to be inspected is transported through a visual positioning and vacuum adsorption conveying system, so that it passes through the flexible array probe in Step 2 for inspection. The eddy current meter simultaneously collects dual-frequency signals, and the collected signals are denoised using wavelet transform. Then, a CNN neural network is used to classify the denoised signal data and output the judgment result. Step 4: Use laser marking equipment to mark defect information and sort qualified and unqualified products.

2. The method according to claim 1, characterized in that, In step 1, the pressure of the upper roller of the straightener is 50kN~70kN, the pressure of the lower roller is 40kN~60kN, the roller speed is 0.3m / s~0.5m / s, and the number of straightening passes is 3~5.

3. The method according to claim 1, characterized in that, After the straightening is completed in step 1, the rectangular tube to be inspected is fed into the inflatable airbag through the input guide sleeve, and after the inspection is completed in step 3, it is sent out through the output guide sleeve; the gap between the inner wall of the input guide sleeve and the output guide sleeve and the outer surface wall of the rectangular tube to be inspected is maintained at 1mm~2mm.

4. The method according to claim 1, characterized in that, In step 2, a substrate is laid on the side of the inflatable airbag facing the outer wall of the rectangular tube to be inspected. The side of the substrate away from the outer wall of the rectangular tube to be inspected is fixedly connected to the inflatable airbag. The planar module and the corner module are disposed on the side of the substrate facing the inflatable airbag and fixedly connected to it. The substrate is a PVDF piezoelectric film.

5. The method according to claim 1, characterized in that, In step 2, the planar module is a planar spiral coil with 100 to 200 turns; the corner module is a differential coil with a radius of 5 mm and 150 to 250 turns.

6. The method according to claim 1, characterized in that, In step 2, the planar module is used to detect surface and near-surface defects of the rectangular tube to be inspected, with a skin depth of 0.1mm to 0.3mm; the corner module is used to detect stress cracks in the right-angle transition zone of the rectangular tube to be inspected, with a skin depth of 0.5mm to 1.0mm.

7. The method according to claim 1, characterized in that, In step 3, the plane and corner contours of the rectangular tube to be inspected are identified by the edge detection algorithm, and the vacuum suction cup is adjusted in real time to ensure that the coaxiality between the central axis of the rectangular tube to be inspected and the center of the flexible array probe is ≤0.1mm / m.

8. The method according to claim 1, characterized in that, In step 3, the conveying speed is 0.5 m / s to 1.0 m / s, and the acceleration is ≤0.2 m / s². 2 .

9. The method according to claim 8, characterized in that, When the wall thickness of the rectangular tube to be inspected is <3mm, the conveying speed is 0.5m / s; when the wall thickness of the rectangular tube to be inspected is 3mm~10mm, the conveying speed is 1.0m / s.

10. The method according to claim 1, characterized in that, In step 3, the wavelet transform is based on the db4 wavelet basis, and the decomposition layer is 3 layers; the input data of the CNN neural network is the signal amplitude-phase spectrum, and the output is defect-free / surface crack / corner crack / inclusion.

Images