Automatic detection device for round-to-square pipe with zinc layer defect identification

Abstract

This invention relates to the field of non-destructive testing technology for pipes, and discloses an automated inspection device for round-to-square galvanized pipes with zinc layer defect identification. The device includes: a conveying and rotating module for driving the pipe to be inspected and outputting a spatial coordinate reference; a laser-ultrasonic excitation module for emitting pulsed laser light onto the pipe surface to excite surface waves and volume waves; a multi-physics synchronous detection module for acquiring time-series thermal wave signals and visual images characterizing surface texture and spectral properties; a multi-source data fusion module for spatially aligning first-type and second-type defect candidate areas; a defect qualitative identification module for identifying defective areas as substrate defects or zinc layer defects; and a sorting execution module for sorting the pipes to different collection areas. This invention achieves differentiated sorting of pipes with substrate defects and those with zinc layer defects, providing a reliable solution for fully automated online quality control of round-to-square galvanized pipes.

Description

Technical Field

[0001] This invention relates to the field of non-destructive testing technology for pipes, and in particular to an automated testing device for round-to-square pipes with zinc layer defect identification. Background Technology

[0002] Round-to-square tubing is produced by roll bending round billets into square or rectangular cross-sections. To improve corrosion resistance, the surface typically undergoes hot-dip galvanizing. During the round-to-square forming process, the radius (R) of the tube cross-section undergoes severe plastic deformation, potentially leading to substrate cracks. The hot-dip galvanizing process can introduce zinc layer defects such as incomplete galvanizing, zinc nodules, scratches, and zinc layer peeling. Substrate cracks are structural safety defects and must be scrapped if present. Zinc layer defects, occurring only on the coating surface or interface, leave the substrate intact and can be repaired for continued use. The handling methods for these two types of defects are drastically different; therefore, inspection requires not only identifying the defects but also distinguishing their root causes.

[0003] Existing pipe surface defect detection technologies primarily rely on machine vision, which uses industrial cameras to capture surface images and image processing algorithms to identify abnormal areas. However, machine vision can only determine the presence of anomalies on the surface; it cannot distinguish whether the anomaly is a substrate crack or a zinc layer defect. When a substrate crack is covered by an intact zinc layer, visual inspection cannot detect it; conversely, when scratches appear on the zinc layer surface, visual inspection may misjudge them as substrate cracks. This limitation, relying solely on surface images and failing to address the root cause of defects at a mechanistic level, leads to a large number of good pipes being rejected or inferior pipes being released, severely impacting the accuracy and cost-effectiveness of product quality control.

[0004] Therefore, this invention proposes an automated inspection device for round-to-square pipes with zinc layer defect identification. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of the inability to distinguish between substrate defects and zinc layer defects, frequent misjudgments, and crude sorting in the prior art. To this end, an automated detection device for round-to-square pipes with zinc layer defect identification is proposed.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an automated inspection device for round-to-square pipes with zinc layer defect identification, comprising:

[0007] The conveying rotation module is used to drive the pipe to be inspected to move at a constant speed along the axial direction and rotate in a controlled manner around its own axis, and outputs the axial displacement and circumferential rotation angle of the pipe in real time as a spatial coordinate reference.

[0008] The laser ultrasonic excitation module is set at the inspection station and is used to emit pulsed lasers to the surface of the pipe with a preset energy density and pulse width. In the double-layer structure formed by the zinc layer and the substrate of the pipe, surface waves propagating along the surface of the zinc layer and volume waves propagating inside the substrate are simultaneously excited.

[0009] The multi-physics synchronous detection module is used to synchronously acquire time-series thermal wave signals and visual images characterizing surface texture and spectral properties of the pipe surface after laser excitation under the same trigger signal, based on a spatial coordinate reference, so that the two are synchronized in time and co-located in space. The time-series thermal wave signals reflect the transient temperature rise of the surface caused by the propagation of surface waves and volume waves.

[0010] The multi-source data fusion module is used to spatially align the first type of defect candidate region generated from visual image segmentation with the second type of defect candidate region extracted from time-series thermal wave signals using a spatial coordinate reference.

[0011] The defect qualitative identification module is used to cross-determine the aligned candidate areas according to the pre-established correspondence rules between acoustic mode energy distribution and defect type, and identify the defect area as a substrate defect or a zinc layer defect.

[0012] The sorting execution module is used to sort pipes identified as having base material defects and pipes identified as having zinc layer defects into different collection areas based on the identification results of the defect qualitative identification module.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0014] This invention uses a laser ultrasonic excitation module to simultaneously excite surface waves and volume waves in the double-layer structure of the zinc layer and the substrate of the pipe. A multi-physics synchronous detection module simultaneously acquires time-series thermal wave signals reflecting the transient temperature rise of the surface caused by the propagation of surface waves and volume waves, as well as visual images characterizing surface texture and color features. This allows for the acquisition of differentiated acoustic wave feature information that distinguishes substrate defects from zinc layer defects from a physical mechanism perspective.

[0015] This invention uses a multi-source data fusion module to spatially align the first type of defect candidate region generated by visual image segmentation with the second type of defect candidate region extracted from time-series thermal wave signals using a spatial coordinate reference. The defect qualitative identification module then performs cross-judgment on the aligned candidate regions according to the correspondence rules between acoustic modal energy distribution and defect type. This enables the defect region to be accurately identified as a substrate defect or a zinc layer defect, solving the technical problem that existing visual inspection technologies cannot perform root cause qualitative identification of defects.

[0016] This invention uses a sorting execution module to sort pipes identified as having base material defects and pipes identified as having zinc layer defects into different collection areas based on the qualitative identification results of defects. This enables refined automatic sorting based on the root cause type of defects, avoiding the mistaken rejection of good materials and the misplacement of inferior materials. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the device configuration provided in an embodiment of the present invention;

[0019] Figure 2 The flowchart illustrates the cross-determination process of the defect qualitative identification module provided in this embodiment of the invention. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an automated inspection device for round-to-square pipes with zinc layer defect identification proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] The following examples are for illustrative purposes and are not intended to limit the scope of the invention.

[0023] The following description, in conjunction with the accompanying drawings, details the specific scheme of an automated inspection device for identifying defects in round-to-square pipes with zinc coating provided by this invention.

[0024] Please see Figure 1 The diagram illustrates the device configuration of an automated inspection device for round-to-square pipes with zinc layer defect identification, according to an embodiment of the present invention, comprising:

[0025] The conveying rotation module is used to drive the pipe to be inspected to move at a constant speed along the axial direction and rotate in a controlled manner around its own axis, and outputs the axial displacement and circumferential rotation angle of the pipe in real time as a spatial coordinate reference.

[0026] The laser ultrasonic excitation module is set at the inspection station and is used to emit pulsed lasers to the surface of the pipe with a preset energy density and pulse width. In the double-layer structure formed by the zinc layer and the substrate of the pipe, surface waves propagating along the surface of the zinc layer and volume waves propagating inside the substrate are simultaneously excited.

[0027] The multi-physics synchronous detection module is used to synchronously acquire time-series thermal wave signals and visual images characterizing surface texture and spectral properties of the pipe surface after laser excitation under the same trigger signal, based on a spatial coordinate reference. This makes the two synchronous in time and co-located in space. The time-series thermal wave signal reflects the transient temperature rise of the surface caused by the propagation of surface waves and volume waves.

[0028] The multi-source data fusion module is used to spatially align the first type of defect candidate region generated from visual image segmentation with the second type of defect candidate region extracted from time-series thermal wave signals using a spatial coordinate reference.

[0029] The defect qualitative identification module is used to cross-determine the aligned candidate areas according to the pre-established correspondence rules between acoustic mode energy distribution and defect type, and identify the defect area as a substrate defect or a zinc layer defect.

[0030] The sorting execution module is used to sort pipes identified as having base material defects and pipes identified as having zinc layer defects into different collection areas based on the identification results of the defect qualitative identification module.

[0031] It should be noted that round-to-square tubes refer to special-shaped tubes that are gradually rolled into square or rectangular cross-sections from round tube blanks through a roll bending process. The cross-section consists of four alternating planar segments and four rounded corner transition segments. The rounded corner areas undergo severe plastic deformation during the forming process, making them high-risk areas for substrate cracking.

[0032] The zinc coating is a protective layer formed on the surface of the pipe substrate through a hot-dip galvanizing process, and its thickness is typically tens to hundreds of micrometers. The zinc coating and the substrate form a double-layer structure. The two materials have different acoustic impedances, and ultrasonic waves will be reflected, refracted, and undergo mode conversion at the interface of this double-layer structure.

[0033] A double-layer structure refers to a composite structure in which a round-to-square tube is made of two different materials: an outer zinc layer and an inner substrate layer. When laser ultrasound is excited in this double-layer structure, surface waves mainly propagate on the surface of the zinc layer, while volume waves mainly propagate inside the substrate. The two acoustic modes have different response characteristics to different types of defects.

[0034] The thermoelastic effect refers to the physical mechanism by which a material absorbs laser energy and undergoes transient thermal expansion when a pulsed laser irradiates its surface, thereby generating ultrasonic waves. Under the dominance of the thermoelastic effect, laser energy is converted into heat and mechanical energy, generating broadband ultrasonic waves without causing ablation damage to the material surface.

[0035] Mode conversion refers to the physical phenomenon where, during propagation, ultrasonic waves encounter a defect interface and, due to differences in acoustic impedance, some of their energy is converted from one mode to another. For example, a bulk wave at a substrate crack interface will partially convert to a surface wave, and a surface wave at a zinc layer peeling interface will partially convert to a bulk wave. The mode conversion process is accompanied by local energy dissipation, resulting in a transient temperature rise.

[0036] Transient temperature rise refers to the instantaneous localized temperature increase at the defect interface caused by the dissipation of some of the acoustic mechanical energy into heat energy during mode switching of ultrasonic waves at the defect interface. The duration of transient temperature rise is typically on the order of microseconds to milliseconds, and the temperature rise amplitude is usually 0.1 to 2 Kelvin, which can be captured by high frame rate infrared thermal imaging units.

[0037] Acoustic modal energy distribution refers to the distribution of energy with frequency in the frequency domain after a time-series thermal wave signal undergoes a short-time Fourier transform. This distribution includes energy information from both surface wave and volume wave bands. Surface waves and volume waves occupy different frequency ranges in the spectrum, and by analyzing the energy distribution of each frequency band, the propagation characteristics of different acoustic modes can be deduced.

[0038] Multiple echoes of bulk waves refer to a series of equally timed echo signals that appear in a time-series thermal wave signal when a bulk wave propagates within a substrate and encounters multiple reflections at interfaces with discontinuous acoustic impedance, such as cracks in the substrate. The number of echoes and the rate of amplitude decay are related to the size and orientation of the crack.

[0039] Dominant frequency shift refers to the phenomenon where the center frequency of the received bulk wave signal shifts towards a lower frequency relative to the center frequency of the excitation ultrasonic wave. When ultrasonic waves pass through a cracked area of ​​a substrate, the selective attenuation of high-frequency components by the crack leads to a weakening of high-frequency components and a relative enhancement of low-frequency components in the received signal, manifested as a shift of the main peak of the spectral energy distribution towards lower frequencies.

[0040] Surface wave scattering noise refers to the irregular noise signal generated in the spectral energy distribution when surface waves propagate along the zinc layer and encounter scattering and interference at zinc layer peeling or virtual bonding interfaces. This noise signal exhibits diffuse characteristics in spatial distribution, with blurred boundaries, and is significantly different in morphology from concentrated hot spots generated by substrate cracks.

[0041] A sparse bond refers to an interface defect where, although there is physical contact between the zinc layer and the substrate, the bond strength is far lower than that of a normal metallurgical bond. In the sparse bond region, the zinc layer and the substrate are not completely separated, but there are tiny gaps or weak bonding layers. Surface waves will scatter and undergo mode conversion in this region, but volume waves can still partially penetrate, which differs from the acoustic response characteristics of a completely delaminated bond.

[0042] Spatial alignment refers to the process of registering the positions of first-type defect candidate regions and second-type defect candidate regions under a unified spatial coordinate reference. Since visual images and thermal wave signals come from different sensors, their pixel coordinate systems and spatial resolutions are different. Spatial alignment transforms the two types of candidate regions to the same coordinate system through a coordinate mapping matrix, so that surface anomalies and subsurface anomalies at the same physical location can correspond one-to-one.

[0043] The short-time Fourier transform (SFT) is a mathematical transformation method for time-frequency analysis of time-series signals. This method, by windowing the signal and performing Fourier transforms in segments, can simultaneously extract the signal's characteristics in both the time and frequency domains. It is suitable for analyzing the time-varying spectral characteristics of ultrasonic surface waves and volume waves at defect interfaces.

[0044] Weighted Mahalanobis distance is a distance metric that considers the correlation between features, assigning different weights to different feature components based on standard Mahalanobis distance. In the improved Otsu algorithm, weighted Mahalanobis distance is used as a regularization constraint term in the fitness function to enhance the algorithm's sensitivity to defect edge pixels while suppressing pseudo-edge interference caused by zinc layer crystal patterns.

[0045] The R-corner segment is the arc-shaped transition area connecting two adjacent planar segments in the cross-section of a round-to-square tube. Due to the characteristics of the round-to-square forming process, the radius of curvature of the R-corner segment is usually small, and the surface orientation changes drastically. It is a high-incidence area for substrate cracks and zinc layer defects, and also a technically challenging area for optical and ultrasonic testing.

[0046] The planar segments are the four planar regions in the cross-section of a round-to-square tube that form a square or rectangular outline. The surface normals of the planar segments are in the same direction, and the optical imaging conditions are relatively stable. However, zinc layer defects such as incomplete galvanizing, zinc nodules, scratches, and zinc ash may occur during the hot-dip galvanizing process.

[0047] In one specific embodiment, the pipe to be inspected is a round-to-rectangular pipe with dimensions of 200 mm x 300 mm, a wall thickness of 9.2 mm, a material of Q235 steel, a zinc coating thickness of 60 micrometers, and a length of 6 meters. The radius of curvature of the R-angle transition section in the cross-section of this pipe is 23 mm.

[0048] The contouring clamping unit for the conveying rotation module includes six sets of clamping wheels evenly distributed along the circumference of the pipe. The contour of the clamping surface of each set of clamping wheels is specially contoured according to the curvature change of the planar section and the R-angle transition section in the cross-section of the round-to-square pipe of the above specifications, so that the clamping surface makes surface contact with each section of the pipe surface. The radial adjustment unit dynamically adjusts the radial position of each clamping wheel through a servo hydraulic cylinder based on the circumferential rotation angle collected in real time by the incremental encoding unit. When the pipe rotates from the planar section to the R-angle section, the clamping wheels extend outward radially to compensate for the reduction in the distance from the pipe surface to the rotation axis. When the pipe rotates back from the R-angle section to the planar section, the clamping wheels retract inward radially. The coaxiality of the pipe axis and the rotation axis is always maintained when the pipe alternates between the planar section and the R-angle section. The servo drive unit drives the pipe to rotate around its own axis, while the axial conveying unit drives the pipe to move at a constant axial speed of 0.5 meters per second along the axial direction. The axial movement speed is coordinated with the circumferential rotation speed to form equally spaced spiral scanning tracks on the pipe surface with an overlap rate of approximately 35% between adjacent spiral tracks. The incremental encoding unit collects the circumferential rotation angle and axial displacement of the pipe in real time and outputs the above position data as a unified spatial coordinate reference at a frequency of 1000 times per second.

[0049] The pulsed laser of the laser ultrasonic excitation module is a 1064 nm Nd:YAG laser, located at the detection station. The pulse width is set to 10 nanoseconds, the single pulse energy to 50 millijoules, and the repetition frequency to 100 Hz. A beam shaping unit is positioned in the output path of the pulsed laser, using a cylindrical lens group to shape the pulsed laser beam into a line source spot with a width of 1.0 mm and a length of 15 mm. The long axis of the line source spot is parallel to the pipe axis, forming an axially extending linear excitation region on the pipe surface. This linear excitation region covers a 23 mm radius of curvature transition section and a 2 mm extension connecting it to an adjacent planar section in the circumferential direction of the pipe. The laser control unit is electrically connected to the pulsed laser and the beam shaping unit. It receives the current circumferential rotation angle of the tube in real time. When the tube rotates to the planar section, it maintains the standard parameters of 50 millijoules for single pulse energy and 10 nanoseconds for pulse width. When the tube rotates to the R-angle section, it finely adjusts the single pulse energy to 48 millijoules and the line source spot width to 0.8 mm to compensate for the signal attenuation caused by acoustic wave divergence at the R-angle curved surface, so that the ultrasonic signal intensity generated by the line source spot is consistent in the planar section and the R-angle section.

[0050] The infrared thermal imaging unit of the multi-physics synchronous detection module uses a cooled indium antimonide focal plane array detector with a pixel count of 640 x 512, a thermal sensitivity of 15 millikrvin, and a frame rate of 1000 Hz. This frame rate matches the 100 Hz repetition frequency of the pulsed laser, acquiring 10 frames of thermal wave signals within a 10-millisecond interval between two laser pulses. This continuous acquisition of time-series thermal wave signals from the pipe surface after laser excitation completely captures the round-trip propagation and mode conversion process of the ultrasonic bulk wave along the 9.2 mm wall thickness. The color vision imaging unit employs a multi-angle ring-shaped LED light source and a CMOS area array camera. The ring-shaped light source comprises eight independently controlled LED groups evenly distributed along the circumference, with adjacent groups spaced 45 degrees apart. The illumination axes of four groups are angled at 15 degrees to the area array camera's optical axis, corresponding to the normal directions of the four planar segments of the pipe, providing low-angle uniform illumination. The other four groups are angled at 45 degrees to the area array camera's optical axis, corresponding to the normal directions of the four rounded corner segments of the pipe, providing high-angle illumination to suppress specular reflection. The color temperature of each LED group is independently adjustable within the range of 4500 Kelvin to 6500 Kelvin. Each group is equipped with a polarizing filter at its front end, and the area array camera lens has a rotatable polarizer at its front end. By adjusting the rotation angle of the polarizer, the polarization directions of the two polarizing elements are made orthogonal, further suppressing specular reflection interference from the galvanized layer surface and obtaining a visual image of the pipe surface with uniform brightness. The area array camera has a resolution of 2048 x 1024 pixels and a frame rate of 1000 Hz, strictly matching the frame rate of the infrared thermal imaging unit. A synchronous trigger unit generates a square wave rising edge trigger signal with a frequency of 1000 Hz and a duty cycle of 50%, simultaneously connecting to the external trigger interfaces of both the infrared thermal imaging unit and the area array camera, ensuring strict alignment of the acquisition times for each frame of the two signals. The spatial stamping unit receives the spatial coordinate reference output from the rotation module via a gigabit Ethernet interface, encoding the axial displacement and circumferential rotation angle of the pipe into 32-bit digitized spatial coordinate stamps, which are embedded into each frame of the time-series thermal wave signal and each frame of the visual image data, respectively, enabling the two heterogeneous signals to achieve frame-level precision commonality under the spatial coordinate reference.

[0051] The visual defect segmentation unit of the multi-source data fusion module first converts the color visual image from the RGB color space to the HSV color space, extracts the luminance channel for analyzing surface brightness features, and extracts the saturation and hue channels for analyzing color features. It then uses a gray-level co-occurrence matrix to extract four texture descriptors: energy, contrast, correlation, and uniformity. The window size is set to 16 x 16 pixels. The segmentation algorithm employs a modified Otsu's algorithm. Based on the conventional inter-class variance maximization criterion, a second-order oscillating particle swarm optimization algorithm is introduced for iterative optimization of the gray-level threshold. The particle swarm size is set to 30 particles, with each particle's position encoded as a candidate gray-level threshold value. The initial positions are randomly distributed within the gray-level range of 0 to 255, the maximum number of iterations is set to 100, the inertia weight decreases linearly from 0.9 to 0.4, and the learning factor is set to 2.0. The fitness function consists of two parts: an inter-class variance term and a weighted Mahalanobis distance regularization term. In the weight matrix, the weights for luminance features are set to 0.5, texture features to 0.3, and color features to 0.2. After segmentation, the generated candidate regions are filtered by area. Isolated noise points with an area smaller than 0.5 square millimeters are filtered out, and adjacent candidate regions with similar features are merged to generate a first-class defect candidate region carrying information on defect location, area, average brightness, main texture direction, and main color component attributes. The thermal wave feature extraction unit performs a short-time Fourier transform on the time-series data of each pixel location in the time-series thermal wave signal. A Hamming window is selected as the window function, with a window length of 64 sampling points and an overlap rate of 75%. After the transform, the time-spectrum map of each pixel location is obtained, and the acoustic energy mode conversion features generated by ultrasonic surface waves and volume waves at the defect interface are extracted to generate a second-class defect candidate region reflecting subsurface structural anomalies. The spatial alignment unit establishes an affine transformation matrix between the pixel coordinates of the visual image and the pixel coordinates of the heat wave signal based on the spatial coordinate reference. The pixel resolution of the visual image is 0.15 mm per pixel, and the pixel resolution of the heat wave signal is 0.3 mm per pixel. The affine transformation matrix includes a scaling factor of 2.0, a rotation angle of zero, and a coordinate translation amount. The first type of defect candidate area and the second type of defect candidate area are spatially mapped and aligned with pixel-level precision in the same coordinate system.

[0052] The acoustic mode identification unit of the defect qualitative identification module performs a short-time Fourier transform on the time-series data of each pixel position in the time-series thermal wave signal. A Hamming window is used as the window function, with a window length of 64 sampling points and an overlap rate of 75%. After the transform, a time-spectrum image of each pixel position is obtained. The frequency resolution is determined by the sampling rate and the window length. The spectral energy distribution, which simultaneously includes the ultrasonic surface wave band and the volume wave band, is extracted. The volume wave feature analysis unit detects multiple echoes in the volume wave band. The detection method involves searching for a sequence of energy peaks on the time axis. When the time interval between two adjacent energy peaks is within a reasonable range of the round-trip propagation time of the ultrasound in the pipe wall thickness, and the peak amplitude shows a gradually decreasing trend, it is determined to be a multiple echo of the volume wave. The number of echo occurrences and the amplitude ratio of the first to the last echo are recorded. Simultaneously, the frequency corresponding to the energy peak in the volume wave band is compared with the center frequency of the pulsed laser-excited ultrasound to calculate the degree of dominant frequency shift. The surface wave characteristic analysis unit analyzes the distribution characteristics of scattered noise in the surface wave frequency band. It quantifies the noise by calculating the coefficient of variation of spectral energy in the spatial domain within the surface wave frequency band. The coefficient of variation for normal zinc layer regions is typically less than 0.2, for zinc layer peeling regions it is greater than 0.5, and for virtual bonding regions it is between 0.3 and 0.5. The cross-judgment unit spatially correlates candidate regions of the first and second types of defects. When an aligned candidate region exhibits multiple echoes of bulk waves and a shift in the dominant frequency in its spectral energy distribution, the region is identified as a substrate defect. When only surface wave scattering noise is present without any abnormal multiple echoes of bulk waves, it is identified as a zinc layer defect. When both multiple echoes of bulk waves and a shift in the dominant frequency, along with surface wave scattering noise, are present, it is identified as a substrate defect. When neither multiple echoes of bulk waves nor surface wave scattering noise is present, the region is identified as a qualified area.

[0053] The defect location calculation unit of the sorting execution module binds the type label of each defect identified by the defect qualitative identification module with the physical coordinates of the defect along the pipe length direction based on the spatial coordinate reference, generating defect location information carrying the defect type label. The delay compensation unit reads the real-time pipe conveying speed of 0.5 meters per second and the pre-calibrated response delay of the sorting execution mechanism of 50 milliseconds, and calculates the compensated sorting trigger delay. The delay is equal to the physical distance between the defect location and the sorting air port divided by the real-time pipe conveying speed and then subtracting the response delay. The sorting execution unit includes four sets of sorting air ports and corresponding collection bins arranged sequentially along the pipe conveying direction. The spacing between adjacent air ports is 1.5 meters. The first bin is used to collect pipes with base material defects, and the second bin is used to collect pipes with zinc coating defects. According to the defect type label, the sorting control unit prioritizes base material defects. When a pipe has both base material defects and zinc coating defects, it prioritizes sorting the pipe to the first bin. When the compensated sorting trigger delay arrives, it sends a trigger signal to the corresponding sorting air port, and the high-pressure airflow blows the pipe into the corresponding collection bin. Continuous testing of 200 round-to-square galvanized rectangular tubes of the above specifications showed that the detection rate of substrate cracks was 98.5%, the identification accuracy of zinc layer defects was 96.7%, the detection coverage of the R-corner area was 100%, and the overall sorting accuracy was 95.2%.

[0054] 1. Conveying Rotary Module

[0055] The conveying rotation module includes a contour clamping unit, a radial adjustment unit, a servo drive unit, an axial conveying unit, and an incremental encoding unit;

[0056] The contour-following clamping unit includes multiple sets of clamping wheels distributed along the circumference of the pipe. The clamping surface contour of each set of clamping wheels is adapted to the curvature change of the plane section and the R-angle transition section in the cross-section of the round-to-square pipe.

[0057] The radial adjustment unit is used to dynamically adjust the radial position of each clamping wheel according to the circumferential rotation angle of the pipe, so as to maintain the coaxiality of the pipe axis and the rotation axis when the flat section and the R-angle section of the pipe alternately pass through.

[0058] The servo drive unit is used to drive the pipe to rotate around its own axis, and outputs the first rotation speed when the pipe rotates to the planar segment and the second rotation speed when the pipe rotates to the R-angle segment, based on the circumferential rotation angle fed back by the incremental encoding unit.

[0059] The axial conveying unit is used to drive the pipe to move at a constant axial speed along the axial direction. The axial speed is matched with the circumferential rotation speed driven by the servo drive unit to form a spiral scanning trajectory with equal spacing on the surface of the pipe.

[0060] Incremental encoding units are used to collect the circumferential rotation angle and axial displacement of the pipe in real time to output a spatial coordinate reference.

[0061] It should be noted that contour clamping refers to a mechanical clamping method in which the clamping surface contour of the clamping component is machined according to the actual cross-sectional shape of the object being clamped, so that the clamping surface and the surface of the object being clamped can achieve surface contact rather than point contact or line contact at multiple angles and positions. In the conveying and rotating module, contour clamping enables the clamping wheels to provide uniform clamping force on both the flat section and the radius (R-angle) section of the tube.

[0062] The clamping surface profile refers to the geometric shape of the working surface on the clamping wheel that directly contacts the pipe surface. This profile is CNC machined based on the planar geometry of the flat segment and the arc geometry of the R-angle transition segment in the cross-section of the round-to-square pipe, so that the clamping surface is a plane at the corresponding flat segment position and an arc surface matching the radius of curvature of the R-angle at the corresponding R-angle segment position.

[0063] The R-angle transition section is the arc-shaped transition area connecting two adjacent planar segments in a round-to-square tube cross-section. In the round-to-square forming process, the circular tube blank is gradually deformed into a square or rectangular cross-section through multiple roll passes. Ideal right angles cannot be formed at the four corners; instead, arc-shaped transitions with a certain radius of curvature are retained. The radius of curvature of the R-angle transition section is determined by the die's profile design and is typically within the range of two to three times the tube wall thickness.

[0064] Curvature variation refers to the circumferential variation of the degree of curvature at various points on the cross-sectional profile of a pipe transitioning from round to square. In the planar section, the curvature is zero, and the surface is flat; in the R-angle transition section, the curvature gradually increases from zero to its maximum value and then gradually decreases back to zero, resulting in a continuously changing arc surface. The clamping surface profile needs to adapt to this curvature variation to maintain a good fit during pipe rotation.

[0065] Radial position refers to the distance of the clamping wheels relative to the axis of rotation along the radial direction of the pipe cross-section. During the rotation of the round-to-square pipe, the distance from each point on the pipe surface to the axis of rotation changes periodically, with larger radial distances corresponding to flat sections and smaller radial distances to angular sections. The radial adjustment unit compensates for these distance changes by altering the radial position of the clamping wheels.

[0066] Dynamic adjustment refers to the working method in which the radial adjustment unit continuously and in real time adjusts the radial position of the clamping wheels based on the real-time feedback of the rotation angle signal during the rotation of the pipe. Unlike static fixing, dynamic adjustment requires the adjustment mechanism to have sufficient response speed to complete the adjustment of the radial position within a short period of time when the pipe rotates from the planar section to the R-angle section. The response time is usually required to be within tens of milliseconds.

[0067] The circumferential rotation angle refers to the angle value that the pipe rotates around its own axis, starting from a certain reference position. This angle is acquired in real time by the incremental encoding unit and is used to determine whether the pipe is currently facing the plane segment or the R-angle segment of the detection station. It serves as the reference signal for the radial adjustment unit to perform dynamic adjustment and the servo drive unit to switch speeds.

[0068] The first rotational speed is the drive speed output by the servo drive unit when the pipe rotates to the planar section. Since the distance from the surface of the planar section of the pipe to the axis of rotation is large, the surface linear velocity is high at the same angular velocity. The first rotational speed is determined by reverse calculation based on the distance from the planar section to the axis of rotation and the preset detection linear velocity requirement, so that the surface linear velocity of the planar section when it passes through the detection station is consistent with the preset value.

[0069] The second rotational speed is the drive speed output by the servo drive unit when the pipe rotates to the R-angle segment. The distance from the surface of the R-angle segment to the axis of rotation is smaller than that of the planar segment, resulting in a lower surface linear velocity at the same angular velocity. To compensate for this difference, the second rotational speed is higher than the first rotational speed, ensuring that the surface linear velocity of the R-angle segment when it passes the inspection station remains the same as that of the planar segment.

[0070] Surface linear velocity refers to the instantaneous linear velocity of a point on the pipe surface as it passes the inspection station during rotation. Its value is equal to the rotational angular velocity multiplied by the distance from that point to the axis of rotation. Since the cross-section of the pipe converted from round to square is non-circular, the distances from the plane segment and the radius (R-angle) segment to the axis of rotation are different. Under a constant angular velocity, the surface linear velocities of the two segments differ, requiring compensation through variable speed drive.

[0071] Axial speed refers to the linear velocity of the pipe as it is conveyed forward along its axis. The ratio of axial speed to circumferential speed determines the pitch of the helical scanning trajectory, and the relationship between the two directly affects the integrity of the detection coverage and the overlap rate between adjacent scanning trajectories.

[0072] Pitch refers to the distance the pipe travels along its axis during one revolution. Pitch equals the axial speed divided by the circumferential rotation speed. The pitch setting must ensure that there are no uncovered blank areas between adjacent helical scanning tracks, nor excessive overlap that would reduce detection efficiency. Typically, the overlap rate between adjacent tracks is set to 30% to 40%.

[0073] Coaxiality refers to the degree of coincidence between the pipe's own axis and the rotation axis of the conveying rotating module. Maintaining coaxiality is a crucial prerequisite for achieving a constant laser ultrasonic excitation path in the inspection of round-to-square pipes. If the pipe's axis deviates from the rotation axis, the distance from various points on the pipe surface to the laser ultrasonic excitation module will change with the rotation angle, leading to fluctuations in excitation energy density and instability in the ultrasonic signal.

[0074] The acoustic path refers to the distance a pulsed laser travels from the laser ultrasonic excitation module to the excitation point on the pipe surface. Whether the acoustic path is constant or not directly affects the energy density of the laser on the pipe surface and the ultrasonic excitation efficiency. Fluctuations in the acoustic path will lead to inconsistent ultrasonic signal intensity, affecting the accuracy of subsequent feature extraction and defect identification.

[0075] Incremental encoding is a position measurement method that converts the angular or linear displacement of a rotating shaft into pulse signals. The encoder outputs a pulse for each certain angle of rotation or distance traveled, and the cumulative value of the angular or linear displacement is obtained by counting and determining the direction of the pulses. Incremental encoding does not output absolute position values; a reference point zeroing operation is required when the system is powered on.

[0076] The conveying rotation module of this application solves the technical problem of maintaining coaxiality and constant surface linear velocity during the rotation of round-to-square pipes due to the non-circular cross-section through the coordinated design of contour clamping, radial dynamic adjustment and variable speed drive.

[0077] Specifically, firstly, the multiple clamping wheels of the contour clamping unit are designed to conform to the curvature changes of the plane section and the R-angle transition section in the cross-section of the round-to-square pipe, so that the contour of the clamping surface matches the actual cross-sectional shape of the pipe, and can provide uniform clamping force throughout the entire rotation cycle of the pipe.

[0078] Next, the radial adjustment unit dynamically adjusts the radial position of each clamping wheel as the pipe alternates between the planar section and the radius-angle section, based on the real-time feedback of the pipe's circumferential rotation angle from the incremental coding unit. When the pipe rotates from the planar section to the radius-angle section, the distance from the pipe surface to the rotation axis decreases, and the radial adjustment unit drives the clamping wheels to extend radially outward to compensate for the aforementioned distance change; when the pipe rotates back from the radius-angle section to the planar section, the clamping wheels retract radially inward. Through this dynamic adjustment, the pipe axis and the rotation axis always remain aligned.

[0079] Then, based on the circumferential rotation angle fed back by the incremental encoding unit, the servo drive unit outputs a first rotational speed when the pipe rotates to the planar segment and a second rotational speed when the pipe rotates to the radius (R-angle segment). Because the distance from the pipe surface to the rotation axis is larger in the planar segment, the surface linear velocity is higher; while the distance from the pipe surface to the rotation axis is smaller in the radius (R-angle segment), the surface linear velocity is lower at the same angular velocity. By using variable rotational speed drive, the linear velocity of each point on the pipe surface remains uniform as it passes through the laser ultrasonic excitation station.

[0080] Meanwhile, the axial conveying unit drives the pipe to move along the axial direction at a constant axial speed. This axial speed is coordinated with the circumferential rotation speed of the servo drive unit to form equally spaced spiral scanning tracks on the surface of the pipe, ensuring that laser ultrasonic excitation and multi-physics field signal acquisition can cover the entire outer surface of the pipe without omission.

[0081] Finally, the incremental coding unit collects the circumferential rotation angle and axial displacement of the pipe in real time, and outputs the above position data as a unified spatial coordinate reference for use by the spatial stamping unit of the subsequent multi-physics synchronous detection module and the spatial alignment unit of the multi-source data fusion module.

[0082] II. Laser-Ultrasonic Excitation Module

[0083] The laser ultrasonic excitation module includes a pulsed laser, a beam shaping unit, and a laser control unit;

[0084] A pulsed laser, installed at the inspection station, is used to emit pulsed laser beams onto the surface of the pipe.

[0085] The beam shaping unit is located on the output optical path of the pulsed laser and is used to shape the pulsed laser beam into a line source spot. The long axis of the line source spot is parallel to the axial direction of the tube, so as to form a linear excitation region extending along the axial direction on the surface of the tube. The width of the linear excitation region in the circumferential direction of the tube is configured to cover the R-corner segment of the round-to-square tube cross-section.

[0086] The laser control unit is used to dynamically adjust the energy density, pulse width, and size of the pulsed laser beam according to the cross-sectional shape of the pipe and the testing requirements, so as to ensure that the ultrasonic signal intensity generated by the line source spot is consistent on the planar section and the R-angle section of the pipe.

[0087] Furthermore, the pulsed laser has a wavelength of 1064 nm, a pulse width of 8 to 15 ns, a single pulse energy of 30 to 80 mJ, and a repetition frequency of 50 to 200 Hz; the beam shaping unit shapes the laser beam into a line source spot with a width of 0.5 to 1.5 mm and a length of 8 to 20 mm.

[0088] It should be noted that a pulsed laser is a light source device that outputs a high-energy laser beam in pulse form. It is positioned at the inspection station and emits a pulsed laser beam with a specific wavelength, energy density, pulse width, and repetition frequency onto the surface of the pipe. After the laser beam irradiates the pipe surface, it excites broadband ultrasonic waves in the surface and subsurface layers of the material through the thermoelastic effect.

[0089] A pulsed laser beam is a high-energy-density laser beam emitted from a pulsed laser with extremely short time intervals. The energy density and pulse width of the pulsed laser beam together determine the efficiency and mode of ultrasonic excitation. If the energy density is too low, the signal-to-noise ratio of the excited ultrasonic signal will be insufficient, while if the energy density is too high, it may cause thermal damage to the surface of the galvanized layer.

[0090] Energy density refers to the laser energy deposited per unit area by a pulsed laser beam, measured in millijoules per square millimeter. The magnitude of energy density directly affects the intensity of the thermoelastic effect, which in turn determines the amplitude and spectral width of the excited ultrasound. Therefore, it needs to be optimized based on the zinc layer thickness and substrate material of the pipe.

[0091] Pulse width refers to the duration of a single laser pulse, measured in nanoseconds. Shorter pulse widths are advantageous for exciting broadband ultrasonic waves under the thermoelastic effect-dominated mechanism, avoiding surface damage caused by heat conduction. Simultaneously, the generated ultrasonic waves have a richer frequency composition, which is beneficial for detecting defects at different depths.

[0092] The repetition rate refers to the number of laser pulses emitted per second by a pulsed laser, measured in Hertz (Hz). The repetition rate needs to be matched with the linear speed of the pipe spiral scan and the frame rate of the infrared thermal imaging unit to ensure that the spacing between adjacent excitation points is small enough to cover all areas without omissions, while also allowing sufficient time for thermal signal acquisition between two consecutive excitations.

[0093] A line source spot is a thin, elongated strip of light formed on the surface of a pipe by modulating a pulsed laser beam through a beam shaping unit. The long axis of the line source spot is parallel to the axial direction of the pipe, creating a linear excitation region extending along the axial direction on the pipe surface. Compared to a circular spot, the line source spot can generate wide-angle ultrasonic coverage in curved surface regions with rounded corners, improving the detection sensitivity for defects in curved areas. The width and length of the line source spot are adjusted according to the pipe specifications and inspection requirements.

[0094] The linear excitation region is a slender, elongated ultrasonic excitation area extending along the pipe's axial direction, formed by the linear laser spot irradiating the pipe surface. Within this region, the pulsed laser energy is absorbed by the pipe's surface material, simultaneously exciting surface waves propagating along the zinc layer and volume waves propagating within the substrate through the thermoelastic effect. The circumferential width of the linear excitation region must at least cover the radius (R-angle) transition section and its junction with the planar section to ensure complete coverage of the R-angle region.

[0095] Wavelength refers to the electromagnetic wave wavelength of the laser output from a pulsed laser, measured in nanometers. A laser with a wavelength of 1064 nanometers belongs to the near-infrared band and exhibits good absorption and thermoelastic excitation efficiency on galvanized surfaces, making it a commonly used wavelength in laser ultrasonic testing. This wavelength of laser can generate sufficiently strong ultrasonic signals while avoiding surface damage.

[0096] The laser ultrasonic excitation module of this application, through the coordinated work of a pulsed laser, a beam shaping unit, and a laser control unit, simultaneously excites surface waves and volume waves in the double-layer structure composed of the zinc layer and the substrate of the pipe, providing differentiated characteristic information of two different acoustic modes for subsequent qualitative identification of defects.

[0097] Specifically, firstly, a pulsed laser emits a pulsed laser beam with a preset energy density and pulse width onto the surface of the pipe. The energy density and pulse width are configured according to parameters such as the thickness of the zinc layer, the material of the substrate, and the wall thickness of the pipe. This allows the laser energy to simultaneously excite surface waves propagating along the surface of the zinc layer and volume waves propagating inside the substrate within the double-layer structure formed by the zinc layer and the substrate, under the mechanism dominated by the thermoelastic effect, without causing thermal damage to the surface of the galvanized layer.

[0098] Next, a beam shaping unit is positioned in the output optical path of the pulsed laser, and optically modulates the circular cross-section pulsed laser beam into a line source spot. The long axis of the line source spot is parallel to the axial direction of the pipe, forming an axially extending linear excitation region on the pipe surface. The circumferential width of the linear excitation region covers the R-corner transition section of the round-to-square pipe cross-section and its connection area with the planar section, thus fully incorporating the R-corner, a high-incidence area of ​​defects, into the ultrasonic excitation range.

[0099] Then, the laser control unit dynamically adjusts the energy density, pulse width, and line source spot size of the pulsed laser beam according to the cross-sectional shape of the pipe and the detection requirements. When the pipe rotates to the planar section, the laser control unit outputs a set of excitation parameters based on the curvature characteristics of the planar section; when the pipe rotates to the R-angle section, the laser control unit adjusts the excitation parameters based on the curvature characteristics of the R-angle section, so that the ultrasonic signal intensity generated by the line source spot is consistent in the planar section and the R-angle section, ensuring the consistency of detection sensitivity in different areas.

[0100] III. Multiphysics Field Synchronous Detection Module

[0101] The multi-physics field synchronous detection module includes an infrared thermal imaging unit, a color visual imaging unit, a synchronous triggering unit, and a spatial stamping unit;

[0102] The infrared thermal imaging unit is used to continuously acquire time-series thermal wave signals from the surface of the pipe after laser excitation. Its frame rate is configured to capture the transient temperature rise signals generated by the mode switching of ultrasonic surface waves and volume waves at the defect interface.

[0103] The color vision imaging unit includes a multi-angle ring LED light source and an area array camera. The illumination angle of each LED group in the multi-angle ring LED light source corresponds to the normal direction of the plane segment and the R-corner segment in the cross-section of the round-to-square pipe. This is used to suppress specular reflection caused by the curvature change of the R-corner segment in order to obtain a visual image of the pipe surface with uniform brightness. The frame rate of the area array camera is matched with the frame rate of the infrared thermal imaging unit.

[0104] The synchronous triggering unit is used to generate the same trigger signal so that the acquisition of the time-series thermal wave signal and the visual image is synchronized in time.

[0105] The spatial stamping unit is used to receive the spatial coordinate reference and embed the spatial coordinate reference into each frame of the time series thermal wave signal and each frame of the visual image, so that the two are spatially coherent.

[0106] It should be noted that the time-series thermal wave signal is a sequence of thermal images continuously acquired by an infrared thermal imaging unit at a fixed frame rate, reflecting the temperature change of the pipe surface over time after laser excitation. Each frame of the thermal image records the temperature distribution at various points on the pipe surface at a given moment, and multiple consecutive frames constitute a complete record of the temperature evolution over time. When ultrasonic surface waves and volume waves encounter a defect interface during propagation, mode conversion occurs, and some of the acoustic mechanical energy is dissipated into heat energy, generating a transient temperature rise at the defect interface. This temperature rise is recorded in the corresponding frame of the time-series thermal wave signal. By performing time-frequency analysis on the time-series thermal wave signal, acoustic characteristic parameters such as the propagation speed, attenuation characteristics, and spectral changes of surface waves and volume waves can be extracted.

[0107] The visual image is a two-dimensional color image of the pipe surface acquired by the color vision imaging unit. This image records the intensity distribution of reflected light on the pipe surface in the visible light band, containing three types of feature information: surface brightness, texture, and color. Zinc layer defects such as incomplete plating, zinc nodules, scratches, and zinc ash appear in the visual image as localized brightness abrupt changes, abnormal texture direction, or color deviation. The visual image provides a data foundation for the preliminary identification and localization of surface defects.

[0108] Synchronous acquisition of thermal signals and visual images refers to the simultaneous exposure of the infrared thermal imaging unit and the area array camera at the same frame rate under the control of the same trigger signal, ensuring that the acquisition time of each frame of thermal signal is strictly aligned with the acquisition time of the corresponding frame of visual image. Time synchronization is a prerequisite for subsequent multi-source data fusion. Only when the two signals are synchronized in time can they be cross-determined for the same physical location after spatial alignment.

[0109] Spatial coordinate stamp embedding refers to the spatial stamping unit encoding the axial displacement and circumferential rotation angle output by the conveying rotation module into digital tags, which are then written into the data structures of the thermal wave signal frame and visual image frame acquired at the corresponding time. After embedding the spatial coordinate stamp, each frame of data not only contains the physical signal acquired by the sensor, but also carries the precise spatial location information of the pipe when the signal was acquired, providing coordinate basis for the spatial alignment of the multi-source data fusion module.

[0110] Transient temperature rise capture refers to the process by which an infrared thermal imaging unit, with a sufficiently high frame rate and thermal sensitivity, distinguishes and records the subtle, instantaneous temperature changes caused by ultrasonic mode switching from background temperature noise. Since the amplitude of transient temperature rises is typically 0.1 to 2 Kelvin, and the duration is only on the order of microseconds to milliseconds, the infrared thermal imaging unit needs both a high frame rate and high thermal sensitivity to effectively capture them. The captured transient temperature rise signal appears as a localized temperature peak in the time-series thermal wave signal, and its amplitude, pulse width, and spatial distribution are related to the defect type and size.

[0111] The illumination angle configuration of a multi-angle ring LED light source refers to setting the angle between the illumination axis of each LED group and the optical axis of the area array camera based on the different normal directions of the planar segment and the R-angle segment in the cross-section of the round-to-square tube. For the planar segment where the normal direction points towards the camera, a smaller illumination angle is used to obtain uniform front illumination; for the R-angle segment where the normal direction deviates from the camera, a larger illumination angle is used to make the illumination light nearly parallel to the normal direction of the R-angle segment surface, thereby reducing specular reflection and ensuring that the R-angle segment and the planar segment have similar brightness levels in the visual image.

[0112] Polarization-suppressed specular reflection refers to placing a polarizing filter at the front end of the LED light assembly and a rotatable polarizer at the front end of the area array camera lens. By adjusting the rotation angle of the polarizer, the polarization directions of the two polarizing elements are made orthogonal. According to Fresnel's law of reflection, the polarization state of specularly reflected light is the same as that of the incident light, while the polarization state of diffusely reflected light is randomly distributed. When the polarization directions of the two polarizing elements are orthogonal, the specularly reflected light is significantly attenuated, while some diffusely reflected light still passes through, thus effectively suppressing specular reflection interference on the surface of the galvanized layer in the R-angle section and improving the visual detectability of defects in this area.

[0113] Frame rate matching refers to ensuring that the frame rate of the area scan camera is the same as that of the infrared thermal imaging unit. Both imaging units are controlled by the same trigger signal at the same frame rate, ensuring that visual and thermal information of the same area on the pipe surface is acquired within the same exposure time window. Frame rate mismatch will cause the two signals to be misaligned on the time axis. After subsequent spatial alignment, the same frame may not correspond to the pipe state at the same physical moment, affecting the accuracy of cross-judgment.

[0114] Spatiotemporal co-location is the core technical objective of the multiphysics synchronous detection module, referring to the simultaneous one-to-one correspondence between time-series thermal wave signals and visual images in both time and spatial dimensions. In the time dimension, a synchronous triggering unit ensures strict synchronization of the acquisition times of the two signals; in the spatial dimension, a spatial stamping unit embeds the same spatial coordinate reference into each frame of data from both signals. Once spatiotemporal co-location is achieved, the thermal wave characteristics and visual characteristics of any physical location on the pipe surface at any given time can be accurately correlated, providing a data foundation for the subsequent cross-judgment by the defect qualitative identification module.

[0115] The multiphysics synchronous detection module of this application synchronously acquires time-series thermal wave signals and visual images under a unified spatial coordinate reference, so that the two heterogeneous signals are synchronized in time and co-located in space, providing accurate spatiotemporal alignment data for subsequent multi-source data fusion and defect cross-judgment.

[0116] Specifically, firstly, the infrared thermal imaging unit continuously acquires time-series thermal wave signals from the pipe surface after laser excitation at a high frame rate. These thermal wave signals reflect the transient temperature rise caused by surface wave and volume wave propagation, and include mode conversion information for both types of acoustic waves at the defect interface. The frame rate of the infrared thermal imaging unit is configured to capture these transient temperature rise signals, completely recording the propagation and attenuation process of ultrasonic waves along the pipe wall thickness direction.

[0117] Simultaneously, the color vision imaging unit acquires visual images of the pipe surface. The illumination angle of each LED group in the multi-angle ring LED light source is independently configured according to the normal direction of the planar segment and the R-angle segment in the round-to-square pipe cross-section. For the planar segment where the normal direction points towards the area array camera, low-angle illumination is used; for the R-angle segment where the normal direction deviates from the area array camera, high-angle illumination is used. This ensures that the illumination light in each area is nearly parallel to the normal direction of that area, suppressing specular reflection caused by the abrupt curvature change in the R-angle segment and obtaining a visual image of the pipe surface with uniform brightness. The frame rate of the area array camera is kept the same as that of the infrared thermal imaging unit.

[0118] Next, the synchronous triggering unit generates the same trigger signal and sends it to the infrared thermal imaging unit and the area array camera simultaneously, so that the acquisition of each frame of the two is strictly synchronized in time, ensuring that the thermal wave signal and visual image corresponding to the same timestamp reflect the pipe status at the same physical moment when the data is fused in the future.

[0119] Then, the spatial stamping unit receives the spatial coordinate reference output by the transmission rotation module, encodes the axial displacement and circumferential rotation angle of the pipe into spatial coordinate stamps, and embeds them into each frame of the time series thermal wave signal and each frame of the visual image data, so that the two heterogeneous signals coexist under the spatial coordinate reference, providing a unified coordinate reference for the spatial alignment of the multi-source data fusion module.

[0120] IV. Multi-source data fusion module

[0121] The multi-source data fusion module includes a visual defect segmentation unit, a thermal wave feature extraction unit, and a spatial alignment unit;

[0122] The visual defect segmentation unit is used to perform image segmentation on the visual image. By analyzing the surface brightness, texture and color features, it extracts abnormal regions that match the zinc layer defect features and generates the first type of defect candidate region.

[0123] The thermal wave feature extraction unit is used to perform spatiotemporal domain transformation based on frequency domain or time-frequency domain on the time series thermal wave signal, extract the acoustic energy mode conversion features generated by ultrasonic surface waves and volume waves at the defect interface, and generate a second type of defect candidate region reflecting subsurface structural anomalies.

[0124] The spatial alignment unit is used to spatially map and align the first type of defect candidate area and the second type of defect candidate area according to the spatial coordinate reference, so that the defect qualitative identification module can jointly determine the surface anomaly and the subsurface structure anomaly at the same physical location.

[0125] Furthermore, the visual defect segmentation unit uses an improved Otsu algorithm for image segmentation. The improved Otsu algorithm introduces a second-order oscillating particle swarm optimization algorithm to perform grayscale threshold iterative optimization based on the conventional inter-class variance maximization criterion, and uses the weighted Mahalanobis distance of defect edge pixels as the regularization constraint of the fitness function.

[0126] It should be noted that visual defect segmentation refers to the process of processing the visual image of the pipe surface acquired by the color vision imaging unit to separate pixel areas with abnormal surface brightness, abnormal texture direction, or color deviation from the normal background. Zinc layer defects in visual images manifest as significant differences in local grayscale values ​​compared to the surrounding normal areas. Simultaneously, the texture direction and color distribution of the defect edges also exhibit characteristic changes. The visual defect segmentation algorithm automatically extracts candidate regions for zinc layer defects by comprehensively analyzing these multi-dimensional features.

[0127] Surface brightness characteristics refer to the grayscale values ​​and spatial distribution of each pixel in a visual image of a pipe surface. Zinc layer defects, such as uncoated areas, expose a darker substrate due to the lack of zinc coating, appearing as low-grayscale areas in the image; zinc nodules, due to localized thickening of the zinc layer, form protrusions and appear as high-grayscale areas in the image. Surface brightness characteristics are the fundamental features for distinguishing defective areas from normal areas.

[0128] Surface texture features refer to the spatial arrangement and directionality of pixel grayscale values ​​in a visual image of a pipe surface. A normal galvanized layer surface exhibits a uniform crystalline texture with randomly distributed texture directions; while scratches appear as elongated textures along a specific direction, and zinc ash deposits appear as irregular dotted or sheet-like textures. Surface texture features are quantified using texture descriptors such as grayscale co-occurrence matrix or local binary mode to distinguish between zinc layer defects and normal crystalline textures.

[0129] Color characteristics refer to the numerical distribution of pixels in the red, green, and blue color channels of a visual image of a pipe surface. Different types of zinc coating defects exhibit different color appearances: uncoated areas appear as dark gray as the substrate, zinc nodules appear as bright white, and zinc gray areas appear as grayish-black. Color characteristics are quantified using descriptors such as color histograms or color moments, providing a basis for preliminary identification of defect types.

[0130] Zinc coating defect characteristics refer to the collective brightness, texture, and color features of various zinc coating defects on the pipe surface as observed in visual images. Uncoated areas are characterized by low brightness, a dark gray tone, and a smooth texture; zinc nodules are characterized by high brightness, a bright white tone, and locally raised textures; scratches are characterized by thin, low-brightness stripes along a specific direction; and zinc ash is characterized by irregular gray-black dotted or patchy areas. These characteristics are pre-established as standard feature templates. The visual defect segmentation unit extracts abnormal areas that match the zinc coating defect characteristics by matching the features of the area to be detected with the standard templates.

[0131] The first type of defect candidate region is a set of pixel regions initially identified as having zinc layer defects, output by the visual defect segmentation unit after segmenting the visual image. Each candidate region contains attribute information such as pixel coordinates, area, average brightness, texture direction, and color distribution in the visual image. Since the first type of defect candidate region is generated solely based on surface visual features, it cannot distinguish whether the anomaly is a simple zinc layer defect or a reflection of a substrate crack on the surface. Therefore, it needs to be spatially aligned with the second type of defect candidate region for joint determination.

[0132] Spatiotemporal domain transformation refers to the process of converting the time-series data of each pixel location in a time-series thermal wave signal into the frequency domain or time-frequency domain using mathematical methods such as Fourier transform, short-time Fourier transform, or wavelet transform. After transformation, the temperature change curve over time in the time domain yields the energy distribution in the frequency domain. Different frequency components correspond to different ultrasonic modes and propagation paths, providing a basis for extracting acoustic energy mode conversion characteristics.

[0133] The acoustic energy mode conversion characteristics refer to the characteristic changes in the spectral energy distribution caused by reflection, refraction, and mode conversion at the defect interface when ultrasonic waves propagate in a double-layer structure consisting of a zinc layer and a substrate. These characteristics include energy enhancement or reduction in specific frequency bands, the appearance of new frequency components, and frequency shifts in energy peak positions. For bulk waves at the substrate crack interface, mode conversion manifests as multiple echoes in the bulk wave frequency band and a shift of the dominant frequency to lower frequencies. For surface waves at the zinc layer peeling interface, mode conversion manifests as increased scattering noise in the surface wave frequency band. These characteristics are extracted by comparing the spectral differences between the defective and normal regions.

[0134] Subsurface structural anomalies refer to structural discontinuities or damage existing below the zinc layer on the pipe surface and within the substrate. These mainly include microscopic cracks, through-cracks, inclusions, and pores within the substrate. Subsurface structural anomalies cannot be directly observed through visual images, but they cause changes in volume wave propagation characteristics. The presence and severity of subsurface structural anomalies can be inferred by analyzing multiple echoes and dominant frequency shifts in volume waves.

[0135] The second type of defect candidate region is a set of pixel regions initially identified as having subsurface structural anomalies, output by the thermal wave feature extraction unit after performing spatiotemporal transformation on the time-series thermal wave signal. Each candidate region includes attribute information such as the pixel coordinates of that region in the thermal wave image, the number of occurrences of multiple volume wave echoes, the dominant frequency offset, and the surface wave scattering noise intensity. The second type of defect candidate region reflects structural anomalies in the depth direction of the pipe material, and may or may not spatially overlap with the surface anomalies reflected by the first type of defect candidate region.

[0136] Joint determination refers to the process by which the defect qualitative identification module, within the same spatial coordinate system, comprehensively considers surface anomaly information provided by the first type of defect candidate region and subsurface anomaly information provided by the second type of defect candidate region, and makes a final conclusion on the defect type according to pre-established determination rules. The core of joint determination lies in using the spatial correspondence between surface anomalies and subsurface anomalies for cross-validation to distinguish the root causes of defects.

[0137] The improved Otsu algorithm is an image segmentation method derived from the conventional maximum inter-class variance thresholding algorithm by introducing an optimization algorithm to improve the threshold search strategy and fitness function. The conventional Otsu algorithm searches for the grayscale threshold that maximizes the variance between the foreground and background classes by traversing grayscale levels, which is ineffective for segmenting low-contrast defects on zinc surface. The improved Otsu algorithm, while maintaining the maximization of inter-class variance criterion, introduces a second-order oscillating particle swarm optimization algorithm to accelerate threshold optimization and uses the weighted Mahalanobis distance of defect edge pixels as a regularization constraint term in the fitness function, enhancing sensitivity to low-contrast defect edges while suppressing false edge interference from zinc layer crystal patterns.

[0138] The inter-class variance maximization criterion is the core principle of the Otsu algorithm for determining the optimal segmentation threshold. This criterion divides image pixels into foreground and background classes based on a grayscale threshold, calculates the grayscale variance between the two classes, and selects the grayscale value that maximizes the inter-class variance as the optimal threshold. A larger inter-class variance indicates a more significant grayscale difference between the foreground and background classes, resulting in better segmentation.

[0139] The second-order oscillating particle swarm optimization (PSO) algorithm is an improved swarm intelligence search algorithm that introduces a second-order oscillation term into the standard PSO algorithm. In the standard PSO algorithm, particles update their velocity and position only based on their own historical best position and the swarm's historical best position, which easily leads to getting trapped in local optima. The second-order oscillating PSO algorithm adds an oscillation term to the velocity update formula, allowing particles to maintain a certain degree of search diversity in the later stages of the search and improving their ability to escape local optima. In the improved Otsu algorithm, this optimization algorithm is used to accelerate the search for the optimal segmentation threshold in grayscale space, with a particle swarm size of 30, a maximum number of iterations of 100, and an inertia weight that decreases linearly from 0.9 to 0.4.

[0140] The fitness function is an objective function used in particle swarm optimization (PSO) to evaluate the merits of each particle's search position. In the improved Otsu algorithm, the fitness function consists of two parts: an inter-class variance term and a weighted Mahalanobis distance regularization term. The inter-class variance term drives the particle search for the threshold that maximizes the separation between the foreground and background, while the regularization term suppresses the interference of zinc layer grain texture on the segmentation results. A larger fitness function value indicates better segmentation performance at the corresponding grayscale threshold.

[0141] Regularization constraints are introduced into the fitness function to prevent optimization algorithms from overfitting or producing unexpected solutions. In the improved Otsu algorithm, the weighted Mahalanobis distance of defect edge pixels is used as a regularization constraint, enabling the optimization algorithm to maximize inter-class variance while also ensuring consistency between the segmentation boundary and the actual defect edge, thus avoiding interference from pseudo-edges of zinc layer crystal patterns in the segmentation results.

[0142] The multi-source data fusion module of this application spatially aligns the first type of defect candidate region generated by visual defect segmentation with the second type of defect candidate region generated by thermal wave feature extraction under a unified spatial coordinate reference. This enables the joint analysis of surface anomaly information and subsurface structure anomaly information at the same physical location, providing a data basis for accurate registration for the cross-determination of the defect qualitative identification module.

[0143] Specifically, firstly, the visual defect segmentation unit performs image segmentation on the visual image. The segmentation algorithm analyzes surface brightness, texture, and color features to separate abnormal regions in the image that match known zinc layer defect feature templates from the normal background, generating first-class defect candidate regions. Each candidate region contains attribute information such as the region's pixel coordinates, area, average brightness, texture direction, and color distribution.

[0144] Simultaneously, the thermal wave feature extraction unit performs a spatiotemporal transformation based on the frequency domain or time-frequency domain on the time-series data of each pixel location in the time-series thermal wave signal. After the transformation, the energy distribution in the frequency domain is obtained, and the acoustic energy mode conversion features generated by ultrasonic surface waves and volume waves at the defect interface are extracted, including multiple echoes in the volume wave frequency band, the dominant frequency offset, and the scattering noise intensity in the surface wave frequency band, generating a second type of defect candidate region reflecting subsurface structural anomalies.

[0145] Then, the spatial alignment unit establishes a mapping transformation matrix between the pixel coordinates of the visual image and the pixel coordinates of the thermal wave signal based on the spatial coordinate reference, transforming the first type of defect candidate area and the second type of defect candidate area to the same spatial coordinate system. After alignment, if both visual anomalies and thermal wave anomalies exist at the same physical location on the pipe surface, the two candidate areas at that location will overlap spatially; if only visual anomalies or only thermal wave anomalies exist, there will be only one candidate area.

[0146] Finally, the two types of candidate regions after spatial alignment are provided to the defect qualitative identification module, which jointly determines the surface anomalies and subsurface structural anomalies at the same physical location.

[0147] Please see Figure 2 The flowchart illustrates the cross-determination process of the defect qualitative identification module provided in this embodiment of the invention.

[0148] V. Defect Qualitative Identification Module

[0149] The defect qualitative identification module includes an acoustic mode identification unit, a bulk wave feature analysis unit, a surface wave feature analysis unit, and a cross-determination unit;

[0150] The acoustic mode recognition unit is used to perform short-time Fourier transform on the time-series thermal wave signal to extract the spectral energy distribution that simultaneously includes the ultrasonic surface wave frequency band and the volume wave frequency band.

[0151] The bulk wave characteristic analysis unit is used to analyze the occurrence of multiple echoes of bulk waves and the degree of main frequency shift in the spectral energy distribution. Multiple echoes of bulk waves and main frequency shift are used to identify crack defects in the substrate.

[0152] The surface wave feature analysis unit is used to analyze the distribution characteristics of surface wave scattering noise in the spectral energy distribution. Surface wave scattering noise is used to identify zinc layer peeling or virtual bonding defects.

[0153] The cross-determination unit is used to spatially associate the first type of defect candidate area with the second type of defect candidate area, and perform cross-determination based on the occurrence of multiple echoes of the bulk wave, the degree of main frequency shift and the distribution characteristics of surface wave scattering noise, so as to identify the defect area as a substrate defect or a zinc layer defect.

[0154] Furthermore, the cross-determination logic executed by the cross-determination unit is as follows:

[0155] When a certain aligned candidate region exhibits multiple echoes of bulk waves and a shift in the dominant frequency in the spectral energy distribution, the aligned candidate region is identified as a substrate defect.

[0156] When an aligned candidate region exhibits only surface wave scattering noise and no volume wave multiple echo anomaly in the spectral energy distribution, the aligned candidate region is identified as a zinc layer defect.

[0157] When an aligned candidate region exhibits multiple echoes of bulk waves with a shift in dominant frequency and surface wave scattering noise in its spectral energy distribution, the aligned candidate region is identified as a substrate defect.

[0158] When an aligned candidate region does not exhibit multiple echo anomalies of body waves in its spectral energy distribution and does not show surface wave scattering noise, the aligned candidate region is identified as a qualified region.

[0159] It should be noted that acoustic mode identification refers to the process of performing a short-time Fourier transform on a time-series thermal wave signal to separate and identify two different acoustic modes: ultrasonic surface waves and volume waves, from the spectral energy distribution. Surface waves propagate along the zinc layer surface, and their energy is mainly concentrated in the lower frequency band of the spectrum; volume waves propagate inside the substrate, and their energy is distributed over a wider frequency band. By analyzing the energy distribution characteristics of different frequency bands in the spectrum, the signal components of the two types of acoustic waves can be separated, providing a basis for subsequently extracting the characteristic parameters of surface waves and volume waves respectively.

[0160] Spectral energy distribution refers to the distribution curve of signal energy as a function of frequency on the frequency axis after a short-time Fourier transform of a time-series thermal wave signal. The energy magnitude of different frequency components reflects the intensity of the corresponding acoustic wave modes. The energy of surface waves is mainly concentrated in the 200 kHz to 800 kHz frequency band, while the energy distribution of volume waves is spread across a wider frequency band from 200 kHz to 5 MHz. By analyzing the energy peak positions, peak amplitudes, and bandwidths of different frequency bands in the spectral energy distribution, propagation characteristic parameters of surface waves and volume waves can be extracted.

[0161] The ultrasonic surface wave frequency band refers to the frequency range corresponding to ultrasonic surface wave signals in the spectral energy distribution. Surface waves propagate along the zinc layer surface, and their frequency range mainly depends on the excitation characteristics of the pulsed laser and the acoustic parameters of the zinc layer material, typically concentrated between 200 kHz and 800 kHz. Within this frequency band, the energy of the surface wave signal is relatively concentrated. When the surface wave encounters a zinc layer peeling or virtual bonding interface, characteristic scattering noise will be generated in this frequency band.

[0162] The bulk wave frequency band refers to the frequency range corresponding to ultrasonic bulk wave signals in the spectral energy distribution. Bulk waves propagate within the substrate and include both longitudinal and transverse wave modes, typically ranging from 200 kHz to 5 MHz. The bulk wave frequency band covers a relatively wide frequency range. When a bulk wave encounters a crack in the substrate, multiple echoes will occur within this band, and the frequency position corresponding to the energy peak will shift towards lower frequencies.

[0163] Zinc plating refers to a zinc plating defect in which the zinc layer completely separates from the substrate. In the peeled area, an air gap exists between the zinc layer and the substrate. When surface waves propagate to the boundary of the peeled area, due to the significant difference in acoustic impedance between the zinc-air interface and the zinc-substrate interface, the surface waves undergo strong reflection and scattering at this boundary, generating significant scattered noise in the surface wave frequency band. The substrate itself in the peeled area remains intact and poses no structural safety hazard.

[0164] Spatial location association refers to the process by which the cross-determination unit establishes a one-to-one correspondence between first-type defect candidate areas and second-type defect candidate areas at the same physical location based on spatially aligned coordinate information. If a physical location contains both first-type and second-type defect candidate areas, that location is marked as an associated region; if only one type of candidate area exists, that location is marked as a non-associated region. Spatial location association is a prerequisite step for cross-determination.

[0165] Cross-judgment refers to the process of comprehensively judging candidate areas after spatial location correlation based on the occurrence of multiple body wave echoes, the degree of dominant frequency shift, and the distribution characteristics of surface wave scattering noise. The core logic of cross-judgment is as follows: multiple body wave echoes and dominant frequency shift are sufficient criteria for substrate defects. As long as multiple body wave echoes and dominant frequency shift are present, regardless of whether surface wave scattering noise is also present, it is identified as a substrate defect; only when no abnormality of multiple body wave echoes is present and only surface wave scattering noise is present is it identified as a zinc layer defect; when neither of these abnormalities is present, it is identified as a qualified area.

[0166] Substrate defects refer to damage occurring within the base metal of the pipe, primarily including micro-cracks and through-cracks generated in the R-corner area during the round-to-square forming process. Substrate defects are structural safety defects; their presence substantially impairs the pipe's load-bearing capacity and safety performance, necessitating product scrapping. In this invention, the acoustic criterion for substrate defects is the appearance of multiple volume wave echoes with a shift in dominant frequency to lower frequencies. Regardless of whether the surface visual image shows abnormalities, as long as the volume wave abnormality criterion is met, it is determined to be a substrate defect.

[0167] Zinc layer defects refer to damage occurring on the surface of the galvanized layer or at the interface between the zinc layer and the substrate. These mainly include incomplete plating, zinc nodules, scratches, zinc ash adhesion, zinc layer peeling, and weak bonding. Zinc layer defects only affect the integrity and appearance of the plating layer; the substrate itself remains intact, and the product can still meet usage requirements after rework, repair, or regalvanizing. In this invention, the criterion for zinc layer defects is the presence of only surface wave scattering noise without abnormal volume wave echoes, indicating that only the zinc layer is damaged while the substrate structure remains intact.

[0168] A qualified area refers to the region on the pipe surface that exhibits neither abnormal volume wave echoes nor surface wave scattering noise. Within this area, both the zinc layer and the substrate are intact, without any defects requiring treatment. Pipe sections corresponding to these qualified areas can be shipped directly as finished products.

[0169] The defect qualitative identification module of this application performs acoustic mode identification and feature analysis on time-series thermal wave signals, combines the spatially aligned candidate regions provided by the multi-source data fusion module, and performs cross-judgment according to the pre-established correspondence rules between acoustic mode features and defect types, thereby achieving the root cause differentiation between substrate defects and zinc layer defects from a physical mechanism perspective.

[0170] Specifically, firstly, the acoustic mode identification unit performs a short-time Fourier transform on the time-series thermal wave signal to extract the spectral energy distribution that simultaneously contains both ultrasonic surface wave and volume wave frequency bands. By analyzing the energy distribution of different frequency bands in the spectrum, the signal components of the two acoustic modes, surface wave and volume wave, are separated.

[0171] Next, the bulk wave characteristic analysis unit analyzes the occurrence of multiple bulk wave echoes and the degree of dominant frequency shift in the spectral energy distribution. The frequency and attenuation rate of multiple bulk wave echoes reflect the size and orientation of the internal cracks in the substrate, while the degree of dominant frequency shift reflects the strength of the attenuation effect of the crack on the high-frequency components of the ultrasonic waves. These two features are used together to identify substrate crack defects.

[0172] Simultaneously, the surface wave characteristic analysis unit analyzes the distribution characteristics of surface wave scattering noise in the spectral energy distribution. The intensity and spatial distribution of surface wave scattering noise reflect the extent and severity of zinc layer peeling or virtual bonding. The presence of surface wave scattering noise indicates interfacial damage between the zinc layer and the substrate.

[0173] Then, the cross-determination unit spatially correlates the candidate areas for the first type of defect with those for the second type of defect, and performs cross-determination based on the occurrence of multiple body echoes, the degree of dominant frequency shift, and the distribution characteristics of surface wave scattering noise. When a candidate area with aligned multiple body echoes and a dominant frequency shift appears, it indicates that there is a crack inside the substrate and is identified as a substrate defect; when only surface wave scattering noise appears without abnormal multiple body echoes, it indicates that only the zinc layer is damaged and the substrate is intact and is identified as a zinc layer defect; when multiple body echoes, dominant frequency shift, and surface wave scattering noise appear simultaneously, it indicates that the substrate crack has extended to the zinc layer or that both types of defects exist simultaneously, and is identified as a substrate defect according to the principle of prioritizing substrate defects; when there are no abnormal multiple body echoes and no surface wave scattering noise, it is identified as a qualified area.

[0174] VI. Sorting Execution Module

[0175] The sorting execution module includes a defect location calculation unit, a delay compensation unit, a sorting execution unit, and a sorting control unit;

[0176] The defect location calculation unit is used to bind the defect type identified by the defect qualitative identification module with the physical coordinates of the defect in the pipe length direction according to the spatial coordinate reference, and generate defect location information carrying defect type label;

[0177] The delay compensation unit is used to calculate the compensated sorting trigger delay based on the real-time conveying speed of the pipe and the pre-calibrated response delay of the sorting actuator, so that the sorting trigger delay matches the defect location information.

[0178] The sorting execution unit includes multiple sorting air ports and corresponding collection bins arranged sequentially along the pipe conveying direction. The collection bins include a first bin for collecting pipes identified as having base material defects and a second bin for collecting pipes identified as having zinc layer defects.

[0179] The sorting control unit is used to sort the pipe to the first bin according to the defect type label and the preset substrate defect priority principle. When the pipe has both substrate defects and zinc layer defects, it will be sorted first. When the compensated sorting trigger delay arrives, the sorting execution unit will be controlled to perform the sorting action.

[0180] It should be noted that defect location calculation refers to the process of binding the type label of each defect identified by the defect qualitative identification module with the physical coordinates of the defect along the pipe length direction, based on a spatial coordinate reference. The defect qualitative identification module outputs the pixel coordinates of the defect in the visual image or thermal image, as well as the defect type. The defect location calculation unit converts the pixel coordinates into the actual physical distance along the pipe length direction using the spatial coordinate reference, generating defect location information carrying the defect type label and physical location coordinates.

[0181] The defect type label is a classification marker assigned by the defect qualitative identification module to each identified defect area. The label content includes two types: substrate defect or zinc layer defect. The defect type label is transmitted to the sorting control unit along with the defect location information, serving as the core basis for sorting decisions.

[0182] Defect location information is structured data generated by the defect location calculation unit, containing defect type labels and the physical coordinates of the defect along the pipe length. Each defect location information corresponds to an identified defect area, recording the type of the defect and its actual distance from the pipe end face, providing the sorting control unit with the triggering basis and triggering position for the sorting action.

[0183] Delay compensation refers to calculating the compensated sorting trigger delay based on the real-time conveying speed of the pipe and the pre-calibrated response delay time of the sorting actuator, so that the sorting action is triggered exactly when the defect area reaches the sorting air inlet. Because there is a fixed response delay from receiving the control command to completing the mechanical action, during which the pipe continues to move forward at a certain speed, without delay compensation, the sorting action will only be performed after the defect area has already passed the sorting air inlet, leading to sorting errors.

[0184] The response delay of the sorting actuator refers to the time required for the solenoid valve at the sorting port to receive an electrical signal and for the high-pressure airflow to be fully ejected. This includes the electromechanical response time of the solenoid valve, the pressure build-up time of the air path, and the airflow acceleration time. This delay time is obtained through pre-calibration experiments. For a specific model of solenoid valve and air path system, its value is a fixed constant, stored in the delay compensation unit for calculation purposes.

[0185] The compensated sorting trigger delay is the waiting time calculated and output by the delay compensation unit from the current moment until the sorting control unit should issue a sorting trigger signal. This delay is equal to the physical distance between the defect location and the sorting air port divided by the real-time conveying speed of the pipe, minus the response delay time of the sorting actuator.

[0186] The sorting air ports are high-pressure air nozzles arranged sequentially along the pipe conveying direction in the sorting execution unit, with each port corresponding to a collection bin. When the sorting control unit sends a trigger signal, the corresponding sorting air port ejects high-pressure air, blowing the pipe passing in front of the air port away from its original conveying path and causing it to fall into the corresponding collection bin.

[0187] The collection bins are containers used to collect pipes with different sorting results, and are arranged sequentially at corresponding positions of each sorting air inlet along the pipe conveying direction. The collection bins include at least a first bin for collecting pipes with base material defects and a second bin for collecting pipes with zinc coating defects. Pipes collected in the first bin will be treated as scrap, while pipes collected in the second bin will undergo rework and repair.

[0188] The first silo is a collection container used to collect pipes identified as having base material defects. Base material defects are structural safety defects, and the pipes must be scrapped. The pipes collected in the first silo will be sent to the waste disposal area.

[0189] The second hopper is a collection container used to collect pipes identified as having zinc coating defects. Zinc coating defects only affect the integrity of the coating; the substrate is intact. Pipes collected in the second hopper will be sent to the rework and repair area, where they can be regalvanized or partially repaired and still be used normally.

[0190] The substrate defect priority principle is the sorting priority rule followed by the sorting control unit when processing a pipe that has both substrate and zinc layer defects. Since substrate cracks are considered structural safety defects and must be scrapped if present, regardless of whether zinc layer defects are also present, the pipe should be prioritized as a substrate defect and sorted to the first hopper. This principle ensures that any pipe with structural safety hazards is not mistakenly sorted to the reworkable collection hopper.

[0191] The sorting trigger signal is an electrical control signal sent by the sorting control unit to the solenoid valve of the corresponding sorting air port of the sorting execution unit when the compensated sorting trigger delay arrives. This signal causes the solenoid valve to open, and high-pressure airflow is ejected from the sorting air port, blowing the pipe into the corresponding collection bin.

[0192] The sorting execution module of this application binds the qualitative identification results of defects with the physical location of the pipe, and combines a delay compensation algorithm and a sorting strategy that prioritizes base material defects to achieve refined automatic sorting of pipes, separating base material defect pipes from zinc layer defect pipes into different collection bins.

[0193] Specifically, firstly, the defect location calculation unit binds the type label of each defect identified by the defect qualitative identification module with the physical coordinates of the defect in the pipe length direction according to the spatial coordinate reference, and generates defect location information carrying the defect type label.

[0194] Next, the delay compensation unit calculates the compensated sorting trigger delay based on the real-time conveying speed of the pipe and the pre-calibrated response delay time of the sorting actuator. This delay ensures that when the high-pressure airflow is ejected from the sorting air port, the defect area on the pipe is exactly in front of the sorting air port, achieving accurate sorting.

[0195] Then, the sorting control unit makes sorting decisions based on the defect type label in the defect location information and according to the preset substrate defect priority principle. When only substrate defects exist on a pipe, the sorting control unit controls the pipe to be sorted to the first hopper; when only zinc layer defects exist on a pipe, the sorting control unit controls the pipe to be sorted to the second hopper; when both substrate defects and zinc layer defects exist on a pipe, the sorting control unit prioritizes the sorting to the first hopper according to the substrate defect priority principle.

[0196] When the compensated sorting trigger delay arrives, the sorting control unit sends a sorting trigger signal to the corresponding sorting air port, and the sorting execution unit sprays high-pressure airflow to blow the pipe into the corresponding collection bin.

[0197] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An automated inspection device for round-to-square pipes with zinc layer defect identification, characterized in that, include: The conveying rotation module is used to drive the pipe to be inspected to move at a constant speed along the axial direction and rotate in a controlled manner around its own axis, and outputs the axial displacement and circumferential rotation angle of the pipe in real time as a spatial coordinate reference. The laser ultrasonic excitation module is set at the inspection station and is used to emit pulsed lasers to the surface of the pipe with a preset energy density and pulse width. In the double-layer structure formed by the zinc layer and the substrate of the pipe, surface waves propagating along the surface of the zinc layer and volume waves propagating inside the substrate are simultaneously excited. The multi-physics synchronous detection module is used to synchronously acquire time-series thermal wave signals and visual images characterizing surface texture and spectral properties of the pipe surface after laser excitation under the same trigger signal, based on a spatial coordinate reference, so that the two are synchronized in time and co-located in space. The time-series thermal wave signals reflect the transient temperature rise of the surface caused by the propagation of surface waves and volume waves. The multi-source data fusion module is used to spatially align the first type of defect candidate region generated from visual image segmentation with the second type of defect candidate region extracted from time-series thermal wave signals using a spatial coordinate reference. The defect qualitative identification module is used to cross-determine the aligned candidate areas according to the pre-established correspondence rules between acoustic mode energy distribution and defect type, and identify the defect area as a substrate defect or a zinc layer defect. The sorting execution module is used to sort pipes identified as having base material defects and pipes identified as having zinc layer defects into different collection areas based on the identification results of the defect qualitative identification module.

2. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 1, characterized in that, The conveying rotation module includes: The contour-following clamping unit includes multiple sets of clamping wheels distributed along the circumference of the pipe. The clamping surface contour of each set of clamping wheels is adapted to the curvature change of the plane section and the R-angle transition section in the cross section of the round-to-square pipe. The radial adjustment unit is used to dynamically adjust the radial position of each clamping wheel according to the circumferential rotation angle of the pipe, so as to maintain the coaxiality of the pipe axis and the rotation axis when the flat section and the R-angle section of the pipe alternately pass through. The servo drive unit is used to drive the pipe to rotate around its own axis, and outputs the first rotation speed when the pipe rotates to the planar segment and the second rotation speed when the pipe rotates to the R-angle segment, based on the circumferential rotation angle fed back by the incremental encoding unit. An axial conveying unit is used to drive the pipe to move at a constant axial speed along the axial direction. The axial speed is coordinated with the circumferential rotation speed driven by the servo drive unit to form a spiral scanning trajectory with equal spacing on the surface of the pipe. Incremental encoding units are used to collect the circumferential rotation angle and axial displacement of the pipe in real time to output a spatial coordinate reference.

3. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 1, characterized in that, The laser ultrasonic excitation module includes: A pulsed laser, installed at the inspection station, is used to emit pulsed laser beams onto the surface of the pipe. A beam shaping unit is disposed on the output optical path of a pulsed laser to shape the pulsed laser beam into a line source spot. The long axis of the line source spot is parallel to the axial direction of the tube to form a linear excitation region extending along the axial direction on the surface of the tube. The width of the linear excitation region in the circumferential direction of the tube is configured to cover the R-corner segment of the round-to-square tube cross-section. The laser control unit is used to dynamically adjust the energy density, pulse width, and size of the pulsed laser beam according to the cross-sectional shape of the pipe and the testing requirements, so as to ensure that the ultrasonic signal intensity generated by the line source spot is consistent on the planar section and the R-angle section of the pipe.

4. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 3, characterized in that, The pulsed laser has a wavelength of 1064nm, a pulse width of 8 to 15ns, a single pulse energy of 30 to 80mJ, and a repetition frequency of 50 to 200Hz; the beam shaping unit shapes the laser beam into a line source spot with a width of 0.5 to 1.5mm and a length of 8 to 20mm.

5. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 1, characterized in that, The multiphysics field synchronous detection module includes: An infrared thermal imaging unit is used to continuously acquire the time-series thermal wave signal of the pipe surface after laser excitation. Its frame rate is configured to capture the transient temperature rise signal generated by the mode switching of ultrasonic surface waves and volume waves at the defect interface. The color vision imaging unit includes a multi-angle ring LED light source and an area array camera. The illumination angle of each LED group in the multi-angle ring LED light source corresponds to the normal direction of the plane segment and the R-corner segment in the cross-section of the round-to-square pipe. This is used to suppress specular reflection caused by the curvature change of the R-corner segment in order to obtain a visual image of the pipe surface with uniform brightness. The frame rate of the area array camera is matched with the frame rate of the infrared thermal imaging unit. The synchronous triggering unit is used to generate the same trigger signal so that the acquisition of the time-series thermal wave signal and the visual image is synchronized in time. The spatial stamping unit is used to receive the spatial coordinate reference and embed the spatial coordinate reference into each frame of the time series thermal wave signal and each frame of the visual image, so that the two are spatially coherent.

6. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 1, characterized in that, The multi-source data fusion module includes: The visual defect segmentation unit is used to perform image segmentation on the visual image. By analyzing the surface brightness, texture and color features, it extracts abnormal regions that match the zinc layer defect features and generates the first type of defect candidate region. The thermal wave feature extraction unit is used to perform spatiotemporal domain transformation based on frequency domain or time-frequency domain on the time series thermal wave signal, extract the acoustic energy mode conversion features generated by ultrasonic surface waves and volume waves at the defect interface, and generate a second type of defect candidate region reflecting subsurface structural anomalies. The spatial alignment unit is used to spatially map and align the first type of defect candidate area and the second type of defect candidate area according to the spatial coordinate reference, so that the defect qualitative identification module can jointly determine the surface anomaly and the subsurface structure anomaly at the same physical location.

7. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 6, characterized in that, The visual defect segmentation unit uses an improved Otsu algorithm for image segmentation. The improved Otsu algorithm introduces a second-order oscillating particle swarm optimization algorithm to perform grayscale threshold iterative optimization based on the conventional inter-class variance maximization criterion, and uses the weighted Mahalanobis distance of defect edge pixels as the regularization constraint of the fitness function.

8. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 1, characterized in that, The defect qualitative identification module includes: The acoustic mode recognition unit is used to perform short-time Fourier transform on the time-series thermal wave signal to extract the spectral energy distribution that simultaneously includes the ultrasonic surface wave frequency band and the volume wave frequency band. The bulk wave characteristic analysis unit is used to analyze the occurrence of multiple echoes of bulk waves and the degree of main frequency shift in the spectral energy distribution. The multiple echoes of bulk waves and the main frequency shift are used to identify crack defects in the substrate. The surface wave feature analysis unit is used to analyze the distribution characteristics of surface wave scattering noise in the spectral energy distribution. The surface wave scattering noise is used to identify zinc layer peeling or virtual bonding defects. The cross-determination unit is used to spatially associate the first type of defect candidate area with the second type of defect candidate area, and perform cross-determination based on the occurrence of multiple echoes of the bulk wave, the degree of main frequency shift and the distribution characteristics of surface wave scattering noise, so as to identify the defect area as a substrate defect or a zinc layer defect.

9. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 8, characterized in that, The cross-determination logic executed by the cross-determination unit is as follows: When a candidate region after alignment exhibits multiple echoes of bulk waves and a shift in the dominant frequency in the spectral energy distribution, the candidate region after alignment is identified as a substrate defect. When an aligned candidate region exhibits only surface wave scattering noise and no volume wave multiple echo anomaly in the spectral energy distribution, the aligned candidate region is identified as a zinc layer defect. When an aligned candidate region exhibits multiple echoes of bulk waves with a shift in dominant frequency and surface wave scattering noise in its spectral energy distribution, the aligned candidate region is identified as a substrate defect. When an aligned candidate region does not exhibit multiple echo anomalies of body waves in its spectral energy distribution and does not show surface wave scattering noise, the aligned candidate region is identified as a qualified region.

10. The automated inspection device for round-to-square pipes with zinc layer defect identification according to claim 1, characterized in that, The sorting execution module includes: The defect location calculation unit is used to bind the defect type identified by the defect qualitative identification module with the physical coordinates of the defect in the pipe length direction according to the spatial coordinate reference, and generate defect location information carrying defect type label; The delay compensation unit is used to calculate the compensated sorting trigger delay based on the real-time conveying speed of the pipe and the pre-calibrated response delay of the sorting actuator, so that the sorting trigger delay matches the defect location information. The sorting execution unit includes multiple sorting air ports and corresponding collection bins arranged sequentially along the pipe conveying direction. The collection bins include a first bin for collecting pipes identified as having base material defects and a second bin for collecting pipes identified as having zinc layer defects. The sorting control unit is used to sort the pipe to the first bin according to the defect type label and the preset substrate defect priority principle. When the pipe has both substrate defects and zinc layer defects, it will be sorted first. When the compensated sorting trigger delay arrives, the sorting execution unit will be controlled to perform the sorting action.

Images