A roundness detection device and method for finished aluminum alloy pipe

Abstract

The application discloses a roundness detection device and method for finished aluminum alloy pipes, and relates to the technical field of pipe detection. The method comprises the following steps: constructing a multi-modal perception array and calibrating mapping to the pipe axis, triggering synchronous acquisition of cross-section point cloud and vibration signals by a constant-speed sliding table, generating an attractor feature matrix based on phase space reconstruction and coding dictionary, constructing a weighted directed graph to extract pulse energy distribution for accurate positioning of deviation nodes, combining energy gradient and connectivity change to adaptively correct the decision threshold and dynamically switch the enhanced precision, fast detection and steady tracking modes, continuously optimizing model parameters by online comparison with actual measurement values of a high-precision roundness gauge, and finally generating a quality report containing deviation distribution, grade statistics and unique traceability identification to support batch traceability and remote query. Through the combination of phase space reconstruction attractor coding, pulse energy density analysis and dynamic mode switching, efficient identification, accurate positioning and balance between precision and efficiency of slight deviations are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pipe detection, in particular to a roundness detection device and method for finished aluminum alloy pipe. BACKGROUND

[0002] Aluminum alloy pipes are widely used in key fields such as aerospace, automobile manufacturing, and building structures due to their lightweight and high strength. The roundness precision of the pipe directly affects its assembly performance and service life, and is one of the important indicators for measuring product quality. Currently, the roundness of aluminum alloy pipes is mainly detected by contact measurement and optical scanning in industry, but these traditional methods have many limitations in practical application.

[0003] Contact measurement methods such as micrometer and roundness gauge have high measurement accuracy, but the detection efficiency is low, and point-by-point contact measurement is required, which cannot meet the high-speed detection requirements of modern production lines. At the same time, direct contact between the mechanical probe and the pipe surface may cause scratches on the material surface, which poses a potential risk to the quality of high-precision pipes.

[0004] Optical non-contact measurement methods such as laser scanning or structured light scanning improve the detection speed, but are easily affected by various interference factors in actual industrial environments. The reflective properties of the aluminum alloy pipe surface, residual oil or coolant from processing, and mechanical vibration in the production site can all cause missing or distorted scanning data, affecting the reliability of the detection results. In addition, a single optical detection method can only provide profile information of a static cross-section, and it is difficult to capture the continuous deformation characteristics of the pipe in the axial direction, and the detection sensitivity for defects such as small concave and gradual ovality changes is insufficient.

[0005] Existing detection systems usually use fixed judgment thresholds and sampling parameters, and lack adaptive adjustment capability. In actual production, different batches of pipes may vary in product characteristics due to differences in raw materials, fluctuations in process parameters, and other factors. The detection system with fixed parameters is difficult to adapt to such changes, and is prone to false positives or false negatives, requiring frequent manual intervention and adjustment, which seriously affects the detection efficiency and consistency. SUMMARY

[0006] Based on the above-mentioned shortcomings of the prior art, the present application aims to provide a roundness detection device and method for finished aluminum alloy pipes to solve the above technical problems.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical solution: a roundness detection method for finished aluminum alloy pipes, comprising:

[0008] S1: Construct a multi-modal sensing array composed of laser triangulation, structured light scanning and vibration resonance sensors, calibrate and synchronize, and map the multi-modal sensing array output to the pipe center axis coordinate system;

[0009] S2: Drive the linear slide at a constant speed along the pipe axis, synchronously collect cross-section point clouds and vibration response signals at predetermined interval positions, and generate a time-sequenced profile sequence;

[0010] S3: According to the principle of phase space reconstruction, the profile sequence is mapped to a high-dimensional trajectory, and the trajectory segments are discretely mapped by a preset coding dictionary to generate an attractor feature matrix;

[0011] S4: Set the coding symbols in the attractor feature matrix as nodes, construct a weighted directed graph according to the node co-occurrence frequency and transition probability, extract the pulse energy density distribution in the weighted directed graph, and locate the local topographic deviation;

[0012] S5: Identify the deviation candidate nodes according to the node energy gradient and connectivity change, correct the determination threshold combined with the historical detection results, and output the corresponding deviation nodes and deviation levels according to the determination threshold;

[0013] S6: According to the deviation detection rate and production batch state, switch between the preset detection modes to balance the detection accuracy and efficiency;

[0014] S7: Compare the detection results with the actual measured values of the high-precision roundness gauge, and update the coding dictionary and graph edge weight calculation parameters through online learning;

[0015] S8: Integrate the deviation node position, deviation level and update record to generate a quality report, and generate a unique traceability identification for each pipe.

[0016] The application further provides that step S1 comprises:

[0017] The internal and external parameters of the laser triangulation, structured light scanning and vibration resonance sensors are calibrated by a calibration ruler, a checkerboard calibration board and a rigid cylindrical calibration target;

[0018] The acquisition time sequences of the laser triangulation, structured light scanning and vibration resonance sensors are phase-locked by configuring a hardware trigger signal line, realizing microsecond-level synchronization of multi-modal data;

[0019] The internal parameter correction result and the external parameter registration result of the sensor are combined to generate a coordinate transformation operator, and the collected ranging, point cloud and vibration response data are mapped to the pipe center axis coordinate system.

[0020] The application further provides that step S2 comprises:

[0021] A linear slide is controlled by a closed-loop servo system to move at a constant speed along the axial direction of the pipe;

[0022] A displacement encoder is arranged on the slide rail to generate a trigger signal at a predetermined equidistant position along the axial direction of the pipe;

[0023] The trigger signal is transmitted to the multi-modal sensing array for synchronous acquisition of the multi-modal sensors;

[0024] At each trigger, the laser point cloud, the structured light image and the vibration response are synchronously recorded, a time sequence identifier is generated according to the acquisition time, and a contour sequence with time sequence marking is generated.

[0025] The application further provides that step S3 comprises:

[0026] Based on the phase space reconstruction principle, a multi-dimensional reconstruction vector is constructed from the contour sequence with time sequence marking according to a predetermined delay and embedding dimension;

[0027] The high-dimensional trajectory formed according to the reconstruction vector is divided into a plurality of adjacent trajectory segments;

[0028] Each trajectory segment is mapped to a corresponding code symbol sequence according to a pre-constructed code dictionary;

[0029] An attractor feature matrix is formed according to the code symbol sequence, and is used to represent the dynamic deviation mode of the pipe cross-section contour.

[0030] The application further provides that step S4 comprises:

[0031] The code symbols in the attractor feature matrix are set as nodes of a directed graph;

[0032] The code symbol sequence is scanned in sequence according to the matrix row sequence and column sequence, the number of simultaneous occurrences of any two nodes within a predetermined window range is counted, and the co-occurrence frequency of the node pair is obtained;

[0033] The symbol sequence is traversed, the number of occurrences of a successor node symbol is counted from each node symbol as a starting point, and the transition probability of the node to the successor node is obtained by comparing the total number of occurrences of the starting point symbol;

[0034] Directed edges are added between the nodes according to the co-occurrence frequency or the transition probability, and the corresponding co-occurrence frequency or transition probability value is set as the edge weight, to generate a weighted directed graph;

[0035] A uniform pulse excitation is applied in the weighted directed graph and propagated along the graph edges, and the energy response of each node is recorded and accumulated to generate a pulse energy density distribution;

[0036] The energy density distribution is mapped back to the pipe cross-section space position, and the energy concentration or mutation area is identified to locate the local roundness deviation position.

[0037] The application is further configured that the step S5 comprises:

[0038] extracting the energy response difference and connectivity change information of each node from the weighted directed graph;

[0039] According to the energy response difference and connectivity change, the deviation candidate nodes are marked, and the historical detection records of the deviation candidate nodes are queried to dynamically correct the initial determination threshold;

[0040] According to the corrected determination threshold, the deviation candidate nodes are classified and marked according to the deviation level, and the deviation level includes slight deviation, obvious deviation and serious deviation;

[0041] The marked deviation nodes, spatial positions and deviation levels are output.

[0042] The application is further configured that the step S6 comprises:

[0043] presetting a reinforcement precision mode, a rapid detection mode and a steady tracking mode in the detection system;

[0044] obtaining a deviation detection rate and a production batch state at the end of the detection period;

[0045] When the deviation detection rate is greater than a preset precision standard or the production batch state is a new product first piece, the detection mode is switched to the reinforcement precision mode;

[0046] When the deviation detection rate is less than a preset efficiency standard and the production batch state is a stable batch, the detection mode is switched to the rapid detection mode;

[0047] When the production batch state is a batch switching or needs to be continuously monitored, the detection mode is switched to the steady tracking mode.

[0048] The application is further configured that the step S7 comprises:

[0049] According to a preset sampling period, a pipe section sample of a deviation node that has completed detection and has been determined is sampled, and a high-precision roundness instrument is used to measure the sample;

[0050] The deviation nodes and deviation levels output by the detection are compared with the roundness values measured by the roundness instrument, and the performance index evaluation code entries are evaluated based on the performance indexes, and the performance indexes include symbol classification accuracy, missed detection rate, false detection rate, recognition stability and model contribution;

[0051] The code entries in the code dictionary that do not meet the evaluation performance indexes are merged or removed, and the weight parameters of the corresponding edges in the weighted directed graph are adjusted according to the performance evaluation results;

[0052] The updated code dictionary and edge weight parameters are applied to subsequent roundness detection.

[0053] The application is further configured that the step S8 comprises:

[0054] Collecting the deviation node position, deviation level and model update record, and associating them according to the unique identification of the pipe;

[0055] According to the preset report template, the associated data is formatted into a quality report, and the quality report includes pipe basic information, deviation distribution graph, deviation level statistics and model update history;

[0056] A unique traceable identification code is assigned to each detected pipe, and the traceable identification code and the pipe batch number are embedded in the quality report;

[0057] Based on the traceable identification code, a two-dimensional code or an RFID tag is printed and fixed on the end of the pipe.

[0058] The application also provides a roundness detection device for finished aluminum alloy pipes, which is used to realize the roundness detection method for finished aluminum alloy pipes.

[0059] Array construction module: construct a multi-modal perception array composed of laser triangulation, structured light scanning and vibration resonance sensor, calibrate and synchronize, and map the multi-modal perception array output to the pipe center axis coordinate system;

[0060] Sequence generation module: drive the linear slide at a constant speed along the axial direction of the pipe, synchronously collect cross-section point cloud and vibration response signal at predetermined interval positions, and generate a contour sequence marked with time;

[0061] Matrix generation module: map the contour sequence to a high-dimensional trajectory according to the phase space reconstruction principle, discretize and map the trajectory fragments through a preset coding dictionary, and generate an attractor feature matrix;

[0062] Deviation positioning module: set the coding symbols in the attractor feature matrix as nodes, construct a weighted directed graph according to the node co-occurrence frequency and transition probability, extract the pulse energy density distribution in the weighted directed graph, and locate the local topographic deviation;

[0063] Deviation determination module: identify deviation candidate nodes according to the node energy gradient and connectivity change, correct the determination threshold combined with historical detection results, and output the corresponding deviation node and deviation level according to the determination threshold;

[0064] Mode switching module: switch between preset detection modes according to the deviation detection rate and production batch state, for balancing detection accuracy and efficiency;

[0065] Parameter update module: compare the detection results with the actual measured values of the high-precision roundness gauge, and update the coding dictionary and graph edge weight calculation parameters through online learning;

[0066] Traceability generation module: integrate the deviation node position, deviation level and update record to generate a quality report, and generate a unique traceability identification for each pipe.

[0067] The application provides a roundness detection device and method for finished aluminum alloy pipes, which comprises the following steps: constructing a multi-modal sensing array composed of a laser triangulation, a structured light scanning and a vibration resonance sensor, calibrating and synchronously calibrating, mapping the multi-modal sensing array output to a pipe center axis coordinate system, driving a linear slide along the pipe axial direction at a constant speed, synchronously collecting cross-section point clouds and vibration response signals at predetermined interval positions, generating a time sequence marked contour sequence, mapping the contour sequence to a high-dimensional trajectory according to the phase space reconstruction principle, discretely mapping the trajectory fragments through a preset coding dictionary, generating an attractor feature matrix, setting the coding symbols in the attractor feature matrix as nodes, constructing a weighted directed graph according to the node co-occurrence frequency and transition probability, extracting the pulse energy density distribution in the weighted directed graph, positioning local morphology deviation, identifying deviation candidate nodes according to the node energy gradient and connectivity change, correcting the determination threshold value in combination with historical detection results, outputting the corresponding deviation node and deviation level according to the determination threshold value, switching among preset detection modes according to the deviation detection rate and production batch state, balancing the detection accuracy and efficiency, comparing the detection results with the actual measurement values of a high-precision roundness gauge, updating the coding dictionary and graph edge weight calculation parameters through online learning, integrating the deviation node position, deviation level and update record to generate a quality report, and generating a unique traceability identification for each pipe.

[0068] 1. High-dimensional nonlinear feature extraction: based on the phase space reconstruction and attractor coding technology, the high-dimensional nonlinear feature extraction is carried out on the contour sequence, the micro roundness deviation signal is effectively amplified, and the identification ability for micro morphology abnormalities is improved.

[0069] 2. Pulse energy density positioning: the weighted directed graph and pulse energy density analysis are used to accurately position the local deviation area, and the spatial resolution and interpretability of the detection results are significantly enhanced.

[0070] 3. Adaptive mode switching: an adaptive threshold determination mechanism combining energy gradient and connectivity change is adopted, and the detection mode is dynamically switched according to the deviation detection rate and production batch state, so that the adaptive matching of different pipe specifications and on-site working conditions is realized, and the detection accuracy and efficiency are effectively balanced.

[0071] The above description is only a summary of the technical scheme of the application, in order to more clearly understand the technical means of the application, the application can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described. BRIEF DESCRIPTION OF DRAWINGS

[0072] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor. In the drawings:

[0073] Figure 1 A flow chart of a roundness detection method for finished aluminum alloy pipes is shown for an exemplary embodiment of the present application;

[0074] Figure 2 A structural schematic diagram of a roundness detection device for finished aluminum alloy pipes is shown for an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0075] The embodiments of the present application will be described below with reference to the drawings and preferred embodiments, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. The present application can also be implemented or applied by means of other different specific embodiments, and various modifications or changes can be made to the details in the specification based on different views and applications without departing from the spirit of the present application. It should be understood that the preferred embodiments are only for illustrating the present application, and are not intended to limit the protection scope of the present application.

[0076] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and only the components related to the present application are shown in the diagrams, not the number, shape and size of the components when actually implemented. The actual implementation of each component may be a random change, and the component layout pattern may also be more complex.

[0077] In the following description, a large number of details are discussed to provide a more thorough explanation of the embodiments of the present application, however, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details, and in other embodiments, the known structures and devices are shown in the form of block diagrams rather than in the form of details, to avoid making the embodiments of the present application difficult to understand.

[0078] Embodiment one:

[0079] A roundness detection method for finished aluminum alloy pipes, as shown in Figure 1 , comprises:

[0080] S1: Construct a multi-modal perception array composed of laser triangulation, structured light scanning and vibration resonance sensor, calibrate and synchronize, and map the multi-modal perception array output to the pipe center axis coordinate system;

[0081] S2: driving the linear slide along the pipe axis at a constant speed, synchronously collecting cross-section point clouds and vibration response signals at predetermined interval positions, and generating a contour sequence marked with time;

[0082] S3: mapping the contour sequence to a high-dimensional trajectory according to the phase space reconstruction principle, discretely mapping the trajectory fragments through a preset encoding dictionary, and generating an attractor feature matrix;

[0083] S4: setting the encoding symbols in the attractor feature matrix as nodes, constructing a weighted directed graph according to the node co-occurrence frequency and transition probability, extracting the pulse energy density distribution in the weighted directed graph, and locating the local topographic deviation;

[0084] S5: identifying the deviation candidate nodes according to the node energy gradient and connectivity change, correcting the determination threshold combined with the historical detection results, and outputting the corresponding deviation nodes and deviation levels according to the determination threshold;

[0085] S6: switching between preset detection modes according to the deviation detection rate and the production batch state, for balancing the detection accuracy and efficiency;

[0086] S7: comparing the detection results with the actual measured values of the high-precision roundness gauge, and updating the encoding dictionary and the graph edge weight calculation parameters through online learning;

[0087] S8: integrating the deviation node position, the deviation level and the update record to generate a quality report, and generating a unique traceability identification for each pipe.

[0088] The application is further provided that step S1 comprises:

[0089] The laser triangulation, the structured light scanning and the vibration resonance sensor are calibrated by a scale, a chessboard calibration board and a rigid cylindrical calibration target; specifically, the internal parameter calibration includes: laser triangulation: 10 groups of reflection point data are collected on the scale with a scale interval of 10 mm, a mapping curve of the laser spot and the actual distance is fitted, and a ranging compensation table is generated; the structured light scanning module: a 9*7 chessboard calibration board with each grid being 20 mm*20 mm is used for multi-view shooting, and the camera internal parameters (focal length, principal point position, radial distortion coefficient) and the projector internal parameters are estimated respectively; the vibration resonance module: a rigid cylindrical calibration target with a diameter of 50 mm and a scale ring interval of 5 mm is attached to the surface of the pipe, a known amplitude of 0.1 mm and a frequency of 1 kHz are applied, and the amplitude-frequency response curve of the sensor is recorded; the external parameter registration includes: on the same cylindrical calibration target, the point cloud template is obtained by the structured light, the radial distance is measured by the laser triangulation along the 12 equal angles around the target, and the amplitude timing is recorded by the vibration module at the same position; the rotation matrix R and the displacement vector T between the coordinate systems of the three sensors and the target body are calculated by the rigid registration algorithm, and are saved as external parameters respectively;

[0090] The acquisition timing of the laser triangulation, the structured light scanning and the vibration resonance sensor is phase-locked by configuring a hardware trigger signal line, so that the multi-modal data is synchronized at the microsecond level; specifically, the structured light projector is used as the main clock, and the 5V TTL hardware trigger line is used to trigger the acquisition modules of the three sensors at the same time; a delay adjustment circuit is added to the trigger line, and the laser ranging head and the vibration sampling are delayed to ±100 ns respectively;

[0091] The internal parameter correction result and the external parameter registration result of the sensor are combined to generate a coordinate transformation operator, and the collected ranging, point cloud and vibration response data are mapped to the pipe center axis coordinate system; specifically, the internal parameter matrix K and the external parameter (R, T) are combined to obtain a 4*4 homogeneous transformation matrix from the coordinate system of each sensor to the pipe center axis coordinate system After each acquisition, the original point cloud, the ranging data and the vibration probe coordinates are mapped according to H and fused to the same cross-sectional plane; further, in a feasible embodiment of the present application, the laser triangulation internal parameter calibration: the scale point distance (mm): 100, 200, 300, 400, 500; the measured distance (mm): 100.12, 200.05, 300.08, 400.11, 500.03; the compensation table after fitting is as follows:

[0092]

[0093] The structured light scanning internal parameter calibration: the pixel coordinates of the corner points of the chessboard are measured as follows:

[0094]

[0095] The camera intrinsic matrix is obtained: The projector equivalent camera intrinsic parameters are: ;

[0096] Vibration resonance sensor intrinsic calibration: rigid cylinder diameter: 50.00mm; excitation applied: amplitude 0.10mm, frequency 1kHz; response record as follows:

[0097]

[0098] Extrinsic registration results:

[0099] Laser→ calibration target: rotation matrix , displacement vector ;

[0100] Structured light→ calibration target: rotation matrix , displacement vector ;

[0101] Vibration→ calibration target: rotation matrix , is a third-order unit matrix, displacement vector

[0102] ;

[0103] In the time sequence synchronous calibration, the structured light trigger clock is 100Hz, and the TTL pulse width is 5μs; the laser ranging delay is +200ns, and the vibration sampling delay is-150ns;

[0104] Therefore, the original laser point cloud point , after mapping to the pipe coordinate system: ; the structured light point cloud example pixel, after intrinsic and extrinsic parameter transformation, corresponds to the three-dimensional physical point (10.02, 0.58, 0.47) mm.

[0105] The application is further provided with that step S2 comprises:

[0106] The linear slide is controlled by a closed-loop servo system to move at a constant speed along the pipe axial direction; specifically, a servo driver and a high-resolution rotary encoder are used to form a closed-loop control system, and the linear slide is connected to the servo motor transmission mechanism; the target axial speed is set in the controller, the encoder feedback is collected in real time, and the PID adjustment is used to ensure that the slide running speed is stable at the set value;

[0107] A displacement encoder is arranged on the slide rail to generate a trigger signal at a predetermined equidistant position of the pipe in the axial direction; specifically, a high-precision linear grating ruler or an incremental encoder is uniformly arranged on the linear slide rail, and the resolution thereof meets the level of 0.1 mm; a hardware trigger circuit is configured, and when the displacement feedback of the encoder reaches a predetermined equidistant value (for example, every 2 mm), a TTL trigger pulse is output once;

[0108] The trigger signal is transmitted to a multi-modal sensing array for multi-modal sensor synchronous acquisition; specifically, the trigger pulse is connected to the laser triangulation, the structure light scanning trigger input and the vibration resonance sensor; when the trigger reaches each time, the data acquisition process of the three sets of sensors is started at the same time, and the same trigger sequence number is marked in the acquisition controller;

[0109] At each trigger time, the laser point cloud, the structure light image and the vibration response are recorded synchronously, the time sequence is generated according to the acquisition time, and the contour sequence with the time sequence mark is generated; specifically, in the controller software, a unique time stamp and a sequence number are allocated to each trigger event, and the three-way data under the sequence number are packaged and stored; the laser point cloud, the structure light image and the vibration response stream are aligned according to the trigger sequence number and the time stamp, and sequentially composed into the cross section contour sequence with the time sequence mark; further, the parameters and the description involved in the above embodiment are shown in the following table:

[0110]

[0111] After the servo controller receives the "100 mm / s" set value, the lead screw motor is started, the encoder outputs real-time speed feedback and is adjusted by the PID; when the slide has moved 2 mm (judged by the encoder counting), the central trigger unit outputs a 5V TTL pulse, and the three sets of sensors are synchronously activated to start acquisition; the laser ranging head scans along 360°, and 720 distance points are recorded in turn; the structure light camera takes an image once with a resolution of 1280*720; the vibration sensor collects 100 amplitude / phase data in a 2 ms window at a rate of 50 kHz; the controller writes the trigger sequence number (such as "#127") and the high-precision time stamp (for example, 2025-06-1016:34:12.345678) into the three-way data packet header together, and stores them in the temporary buffer; the slide continues to move to the next 2 mm position, and the above trigger and acquisition are repeated until the scanning of the predetermined length is completed; after the scanning is completed, all "#1" to "#500" data packets in the buffer are exported in order to form a complete contour sequence file with the time sequence mark, which is used for subsequent feature extraction and deviation analysis.

[0112] The application further provides that step S3 comprises:

[0113] Based on the principle of phase space reconstruction, the contour sequence with time sequence mark is constructed into multi-dimensional reconstruction vector according to the predetermined delay and embedding dimension; specifically, the contour sequence with time sequence mark is constructed into multi-dimensional reconstruction vector according to the predetermined delay τ and embedding dimension m; the contour values of sample points from the i th frame to the i + (m-1)·τ frame jointly form an m-dimensional vector, representing the state of the system at that time;

[0114] The high-dimensional trajectory formed according to the reconstruction vector is divided into several adjacent trajectory segments; specifically, the high-dimensional trajectory obtained after reconstruction is divided into several adjacent trajectory segments according to a fixed segment length L; each segment contains L consecutive reconstruction vectors, forming a short-time dynamic manifold;

[0115] Each trajectory segment is mapped to a corresponding code symbol sequence according to a pre-constructed code dictionary; specifically, the code dictionary is constructed based on the typical deviation mode, and the common trajectory segment is corresponded to a unique code symbol (such as A, B, C…); the actual trajectory segment is sequentially queried and mapped to the corresponding symbol;

[0116] An attractor feature matrix is constructed according to the code symbol sequence, which is used to represent the dynamic deviation mode of the pipe cross-section contour; specifically, the symbol sequence corresponding to each trajectory segment is arranged in time sequence to form a matrix row; the matrix column corresponds to the appearance order of each symbol in the whole sequence in the dictionary; further, 10 frame contour time sequence values are processed, and the parameters and example data are shown in the following table:

[0117]

[0118] The code dictionary is shown in the following table:

[0119]

[0120] Wherein, r1...r 10 are respectively the cross-section radius deviation values (unit: mm) of the 1st to 10th frames;

[0121] According to the above description, 8 reconstruction vectors are sequentially constructed for N=10 with τ=1 and m=3: When L=3, the 8 vectors are divided into 6 segments: segment 1: vectors 1-3; segment 2: vectors 2-4; …; segment 6: vectors 6-8; assuming that only four modes A-D are defined in the dictionary, the symbol sequence [A, B, A, C, B, D] is obtained by comparing the real-time measured segment content with the code dictionary table, and the symbol sequence is arranged in time as a 1×6 matrix: If it needs to be expanded to multiple frames in parallel, more symbol sequences of the same batch of pipes can be stacked as multiple rows to form an M×6 feature matrix, which is used for subsequent deviation mode statistics and atlas construction.

[0122] The application is further configured that the step S4 comprises:

[0123] The encoded symbols in the attractor feature matrix are set as nodes of a directed graph; specifically, all different encoded symbols in the attractor feature matrix are extracted as nodes in the graph, and the corresponding cross-section space position or angle mapping relationship of each node is recorded;

[0124] The sequence of encoded symbols is scanned in the order of matrix row and column sequence, and the number of simultaneous occurrence of any two nodes in a predetermined window range is counted to obtain the co-occurrence frequency of the node pair; specifically, a fixed size window is slid in the sequence of symbols of each row (or each column) of the matrix, and the number of simultaneous occurrence of any two symbols in the window is counted to obtain the co-occurrence frequency of the node pair;

[0125] The sequence of symbols is traversed, and the number of occurrence of the successor node symbol is counted with each node symbol as the starting point, and the transition probability from the starting point to the successor node is obtained by comparing with the total number of occurrence of the starting point symbol; specifically, the complete symbol sequence is traversed, and the number of occurrence of the successor (adjacent position) symbol is counted with each symbol as the starting point, and the transition probability from the starting point to each successor symbol is obtained by dividing the number by the total number of occurrence of the starting point symbol;

[0126] A directed edge is added between the nodes according to the co-occurrence frequency or transition probability, and the corresponding co-occurrence frequency or transition probability value is set as the edge weight to generate a weighted directed graph; specifically, the co-occurrence frequency or transition probability is used as the edge weight to add a directed edge between the nodes, and a weighted directed graph structure reflecting the nonlinear interaction between the symbols is generated;

[0127] A uniform pulse excitation is applied in the weighted directed graph and propagated along the edges of the graph, and the energy response of each node is recorded to generate a pulse energy density distribution; specifically, a unit excitation pulse is applied to each node in the constructed directed graph in turn, and single-step or multi-step energy propagation is performed according to the edge weight, and the energy response value of each node in the propagation process is recorded to obtain the pulse energy density distribution;

[0128] The energy density distribution is mapped back to the space position of the pipe cross-section, the energy concentration or mutation area is identified, and the local roundness deviation position is located; specifically, the cumulative energy value of each node in the graph is projected back to the pipe cross-section through its space mapping relationship, and the node position with high energy concentration or mutation is taken as the candidate area of local roundness deviation; further, the attractor feature matrices of two pipe samples A and B are shown in the following table:

[0129]

[0130] The mapping angles of each node on the cross section are: A→0°, B→60°, C→120°, and D→180°; when the window width is 2, the co-occurrence statistics (including cross-sample accumulation) are: (A, B): [A, B], [B, A] in sample A, a total of 2 times; [B, C], [C, A], [A, B] in sample B, a total of 1 time, a total of 3; (B, C): [B, A] in sample A, a total of 0 times; [B, C], [C, A] in sample B, a total of 1 time; transition probability: taking A as an example, the starting point A appears a total of 3 times (twice in sample A and once in sample B), and the subsequent symbols are counted respectively: A→B: 2 times→2 / 3≈0.67; A→C: 0 times→0; A→D: 1 time→1 / 3≈0.33; according to the above statistics, nodes A, B, C and D and the directed edges are constructed, and the edge weight is taken as an example: the weight of A→B is 0.67, the weight of A→D is 0.33, and the weight of B→C is 0.50; a unit excitation (initial energy 1) is applied to each node in turn: excitation A: A retains 1×(1-Σout edge weight), and 0.67 is transmitted along A→B and 0.33 is transmitted along A→D; excitation B: 0.50 is transmitted along B→C and 0.50 is transmitted along B→A;..., the accumulated response energy of each node is taken as an example: A: 0.40, B: 1.20, C: 0.60, and D: 0.8; the energy distribution is mapped back to the angle: 0°(A): 0.40; 60°(B): 1.20; 120°(C): 0.60; and 180°(D): 0.80; the B node with the highest energy corresponds to the 60° position, and the D node with the second highest energy corresponds to the 180° position, that is, the local circularity deviation of the cross section is a significant area.

[0131] The application further provides that step S5 comprises:

[0132] The energy response difference and connectivity change information of each node are extracted from the weighted directed graph; specifically, for each node in the weighted directed graph, the energy response value of this pulse is read and subtracted from the energy value of the last detection period to obtain the energy response difference; the in-degree and out-degree of each node in the current directed graph are counted and compared with the historical graph structure to obtain the connectivity change amount;

[0133] According to the energy response difference and the connectivity change, the deviation candidate nodes are marked, the historical detection records of the deviation candidate nodes are queried, and the initial judgment threshold is dynamically corrected; specifically, the nodes whose energy difference and connectivity change amount simultaneously exceed the respective initial threshold are marked as deviation candidate nodes; for each candidate node, the maximum energy response value and the connectivity change statistics in the historical detection record database are queried; according to the ratio of the historical maximum value to the current average energy level, the initial energy difference threshold is adaptively adjusted up or down to adapt to the batch difference of pipes and the environment difference;

[0134] The deviation candidate nodes are classified and marked according to the modified judgment threshold value, and the deviation levels include slight deviation, obvious deviation and serious deviation; specifically, according to the modified threshold value, the energy difference value of each candidate node is corresponded to the three intervals of slight deviation, obvious deviation and serious deviation, and the connectivity change amount is used as an auxiliary means to confirm the level;

[0135] The marked deviation nodes, spatial positions and deviation levels are output; specifically, all the marked deviation nodes, corresponding spatial positions (such as circumferential angles) and deviation levels are output, which provides data support for subsequent positioning correction and quality reporting; further, the node responses obtained from adjacent two detection periods (periods t-1 and t) of the same pipe are as follows:

[0136]

[0137] The energy difference threshold value is 0.08, and the connectivity change threshold value is ±1 (if the in-degree or out-degree change exceeds 1, it is considered to be significant); in the candidate node marking, node B: the energy difference value 0.15>0.08 and the in-degree change +1, which meets the candidate condition; node C: the energy difference value 0.05<0.08, which is not marked; node D: the energy difference value 0.05<0.08, which is not marked; node A: the energy difference value 0.02<0.08, which is not marked; the maximum energy difference value of node B in the historical library is 0.12, and the current system average energy difference value is about 0.07; the ratio 0.12 / 0.07≈1.71, so the energy difference threshold value is increased from 0.08 to 0.08×1.2≈0.096 to reduce the misjudgment; the modified threshold value is 0.096; the energy difference value of node B is 0.15, which falls into the “serious deviation” interval; the output result is: deviation node: B, spatial position (°): 60, deviation level: serious deviation.

[0138] The application further provides that step S6 comprises:

[0139] The detection system is preset with a reinforced precision mode, a rapid detection mode and a steady tracking mode; specifically, three operation modes are defined in the system controller, the reinforced precision mode: the smallest scanning interval and the largest excitation amplitude; the rapid detection mode: the largest interval and the smallest excitation amplitude; and the steady tracking mode: intermediate parameters between the above two modes;

[0140] The deviation detection rate and the production batch state are obtained at the end of the detection period; specifically, after each detection period ends, the deviation detection rate (the number of marked deviation nodes / the total number of triggers) of the current period is automatically calculated, and the current batch state (first piece, new first piece, stable batch or batch switching) is obtained from the production execution system;

[0141] When the deviation detection rate is greater than the preset precision standard or the production batch state is a new first piece, the detection mode is switched to the intensive precision mode;

[0142] When the deviation detection rate is less than the preset efficiency standard and the production batch state is a stable batch, the detection mode is switched to the rapid detection mode;

[0143] When the production batch state is batch switching or needs to be continuously monitored, the detection mode is switched to the steady-state tracking mode; further, in 5 detection cycles, the detection times, the deviation node number, the detection rate (%), the batch state, the switching condition and the mode after switching are as shown in the following table:

[0144]

[0145] The preset standards are that the precision standard is that the detection rate is greater than 10%, the efficiency standard is that the detection rate is less than 2%, cycle 1: the first piece is first measured, and is directly switched to the intensive precision mode; cycle 2: the detection rate 8.0% is between the standards, and the current mode is kept; cycle 3: the detection rate 1.0% is less than 2% and the batch is stable, and is switched to the rapid detection mode; cycle 4: the detection process encounters batch switching, and is switched to the steady-state tracking mode; and cycle 5: the detection rate 2.4% is greater than 2% and the batch is stable, and is switched back to the intensive precision mode.

[0146] The application further provides that step S7 comprises:

[0147] According to the preset sampling inspection cycle, a pipe section sample that has completed detection and has been judged as a deviation node is sampled, and a high-precision roundness instrument is used to actually measure the sample; specifically, a pipe section that has completed detection and has been marked as a deviation node is randomly selected according to a preset sampling inspection cycle (such as sampling inspection once every 100 pipes); a high-precision roundness instrument is used to actually measure the selected section to obtain a true roundness deviation value;

[0148] The deviation node and the deviation level output by the detection are compared with the roundness instrument actually measured roundness value, and a performance index evaluation coding item is evaluated, and the performance index includes a symbol classification accuracy rate, a missed detection rate, a false detection rate, an identification stability and a model contribution degree; specifically, the deviation node position and the deviation level recorded by the system are one-to-one corresponding to the roundness instrument actually measured value; a performance evaluation data set is constructed, which is used for subsequent index calculation; in the performance index, the symbol classification accuracy rate: the proportion of correctly matching to the actually measured abnormal area in the deviation node corresponding to a certain coding symbol; the missed detection rate: the proportion of the actually measured deviation area that is not marked by the corresponding symbol; the false detection rate: the proportion of the deviation node marked by the symbol that does not appear deviation in the actually measured value; the identification stability: the classification accuracy rate standard deviation of the same symbol in multiple sampling inspections; the model contribution degree: the difference value of the overall detection accuracy before and after excluding the symbol or the edge weight corresponding thereto;

[0149] Merging or eliminating the coding entries that do not meet the performance evaluation indicators in the coding dictionary, and adjusting the weight parameters of the corresponding edges in the weighted directed graph according to the performance evaluation results; specifically, for the symbol entries that do not meet the evaluation results, merging them into the closest high-performance symbol according to the similarity principle, or directly eliminating them; according to the symbol elimination or merging and the evaluation error hotspots, the related edge weights in the weighted directed graph are adjusted by a preset proportion;

[0150] Applying the updated coding dictionary and edge weight parameters to subsequent roundness detection; specifically, the updated coding dictionary and edge weight parameters are distributed to the detection system to replace the original configuration; the subsequent detection automatically loads the new model and continues to perform roundness deviation identification; further, 10 cross sections are sampled every 100 pipes for 10 times of sampling, and a total of 25 deviation nodes are marked by the detection system; the performance indicators are shown in the following table:

[0151]

[0152] The evaluation threshold is: classification accuracy ≥ 80%; false negative rate ≤ 15%; false positive rate ≤ 10%; stability σ ≤ 5%; contribution ≥ 0.5%; symbols C (accuracy 50%, false negative rate 50%) and D (accuracy 66.7%, false negative rate 33.3%) do not meet the standards; C is merged into B and D is merged into A according to similarity; the original edge A→B weight is 0.67, and because D is merged into A, the weight is adjusted down by 5% to 0.64; the edge B→C weight is 0.50, and because C is merged into B, the weight is adjusted up by 10% to 0.55; the new coding dictionary only retains A and B; the updated directed graph edge weight is immediately distributed to the detection system, and automatically takes effect when the subsequent batch of pipes is detected.

[0153] The application further provides that step S8 comprises:

[0154] Collecting the deviation node position, deviation level and model update record, and associating them according to the unique identification of the pipe; specifically, the unique internal identification (such as serial number), batch number and specification information of each pipe are extracted from the database; the deviation node position and deviation level marked by the pipe are combined with the items of the model update and associated with the identification to form a complete data record;

[0155] According to the preset report template, the associated data is formatted into a quality report, and the quality report includes pipe basic information, deviation distribution graph, deviation level statistics and model update history; specifically, a pre-designed report template is loaded, which includes pipe basic information area, deviation distribution graph area, deviation level statistics area and model update history area; the associated data is filled into the corresponding columns, and the deviation distribution heat map and level statistics graph are automatically drawn; the report is exported as a PDF document, and a traceable identification code position is reserved at the document footer;

[0156] Assign a unique traceability code to each detected pipe, embed the correspondence between the traceability code and the pipe batch number in the quality report; Specifically, generate a globally unique traceability code for each pipe; Embed the mapping relationship between the traceability code and the pipe batch number, specification in the report;

[0157] Print a two-dimensional code or RFID tag based on the traceability code and fix it on the end of the pipe; Specifically, encode the traceability code into a two-dimensional code (or generate a corresponding RFID EPC code), and print it into a tear-resistant self-adhesive label through an industrial label printer; Paste the label on the clean and oil-free end of the pipe, or weld and fix the RFID chip to ensure that it can be scanned and identified throughout the production, storage and installation process.

[0158] Embodiment two:

[0159] Please refer to Figure 2 The exemplary roundness detection device for finished aluminum alloy pipes is used to implement the roundness detection method for finished aluminum alloy pipes described above, and the device comprises:

[0160] Array construction module: Construct a multi-modal perception array composed of laser triangulation, structured light scanning and vibration resonance sensors, calibrate and synchronize, and map the multi-modal perception array output to the pipe center axis coordinate system;

[0161] Sequence generation module: drive the linear slide at a constant speed along the axial direction of the pipe, synchronously collect cross-section point clouds and vibration response signals at predetermined intervals, and generate a contour sequence with time sequence markers;

[0162] Matrix generation module: map the contour sequence to a high-dimensional trajectory according to the phase space reconstruction principle, discretize and map the trajectory segments through a pre-set coding dictionary, and generate an attractor feature matrix;

[0163] Deviation positioning module: set the coding symbols in the attractor feature matrix as nodes, construct a weighted directed graph according to the node co-occurrence frequency and transition probability, extract the pulse energy density distribution in the weighted directed graph, and locate the local topographic deviation;

[0164] Deviation judgment module: identify deviation candidate nodes according to the node energy gradient and connectivity change, correct the judgment threshold combined with historical detection results, and output the corresponding deviation nodes and deviation levels according to the judgment threshold;

[0165] Mode switching module: switch between preset detection modes according to the deviation detection rate and production batch state, used to balance the detection accuracy and efficiency;

[0166] Parameter updating module: compare the detection results with the actual measurement values of high-precision roundness gauges, and update the coding dictionary and graph edge weight calculation parameters through online learning.

[0167] Traceability generation module: integrate the deviation node position, deviation level and update record to generate a quality report, and generate a unique traceability identification for each pipe.

[0168] It should be noted that the above embodiment provides a roundness detection device for finished aluminum alloy pipe, and the above embodiment provides a roundness detection method for finished aluminum alloy pipe, which belongs to the same concept. The specific way in which each module and unit performs the operation has been described in detail in the method embodiment, and will not be repeated here. The roundness detection device for finished aluminum alloy pipe provided by the above embodiment can be completed by different functional modules according to the above functions in actual application, that is, the internal structure of the system is divided into different functional modules to complete all or part of the functions described above, and this is not limited here.

[0169] The above embodiments can be realized all or partially by software, hardware, firmware or other any combination. When realized by software, the above embodiments can be realized all or partially in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center and the like containing one or more available medium collections. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0170] It should be understood that the term "and / or" herein only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent three cases of A alone, A and B together, and B alone, where A and B can be singular or plural. In addition, the character " / " herein generally represents that the front and rear associated objects are an "or" relationship, but can also represent an "and / or" relationship, which can be understood according to the context before and after.

[0171] In this application, "at least one" means one or more, "multiple" means two or more. "At least one of the following (one)" or the like means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can mean a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0172] It should be understood that the size of the sequence of the above processes in various embodiments of the present application does not mean the order of execution, and the execution order of the processes should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0173] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0174] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0175] In several embodiments provided in the present application, it should be understood that the disclosed system can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0176] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0177] In addition, each function unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0178] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0179] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for detecting the roundness of finished aluminum alloy tubing, characterized in that, include: S1: Construct a multimodal sensing array consisting of laser triangulation, structured light scanning, and vibration resonance sensors, perform calibration and synchronous calibration, and map the output of the multimodal sensing array to the coordinate system of the pipe's central axis; S2: Drive the linear slide table along the axial direction of the tube at a constant speed, and simultaneously collect the cross-sectional point cloud and vibration response signal at predetermined intervals to generate a time-marked contour sequence. S3: Based on the phase space reconstruction principle, the contour sequence is mapped to a high-dimensional trajectory, and the trajectory segments are discretized and mapped using a preset encoding dictionary to generate an attractor feature matrix; S4: Set the encoded symbols in the attractor feature matrix as nodes, construct a weighted directed graph based on the node co-occurrence frequency and transition probability, extract the pulse energy density distribution in the weighted directed graph, and locate local morphological deviations; S5: Identify candidate nodes for deviation based on node energy gradient and connectivity changes, adjust the judgment threshold by combining historical detection results, and output the corresponding deviation node and deviation level according to the judgment threshold. S6: Based on the deviation detection rate and production batch status, switch between preset detection modes to balance detection accuracy and efficiency; S7: Compare the detection results with the measured values ​​of the high-precision roundness meter, and update the coding dictionary and graph edge weight calculation parameters through online learning; S8: Integrate deviation node locations, deviation levels, and update records to generate a quality report and create a unique traceability identifier for each pipe.

2. The method for roundness detection of finished aluminum alloy tubes according to claim 1, characterized in that, Step S1 includes: The intrinsic and extrinsic parameters of laser triangulation, structured light scanning, and vibration resonance sensors were calibrated using a calibration ruler, a checkerboard calibration plate, and a rigid cylindrical calibration target. By configuring the hardware trigger signal line, the acquisition timing of laser triangulation, structured light scanning and vibration resonance sensor is phase-locked, achieving microsecond-level synchronization of multimodal data; The sensor's intrinsic parameter calibration results and extrinsic parameter registration results are combined to generate a coordinate transformation operator, which maps the collected ranging, point cloud, and vibration response data to the pipe's central axis coordinate system.

3. The method for roundness detection of finished aluminum alloy tubes according to claim 1, characterized in that, Step S2 includes: The linear slide is controlled by a closed-loop servo system to move at a constant speed along the axial direction of the tube. A displacement encoder is installed on the slide rail to generate trigger signals at predetermined equidistant positions along the axial direction of the pipe. The trigger signal is transmitted to the multimodal sensing array for synchronous acquisition by the multimodal sensors; Upon each trigger, the laser point cloud, structured light image, and vibration response are recorded synchronously. A time sequence identifier is generated based on the acquisition time, and a contour sequence of time sequence markers is generated.

4. The method for roundness detection of finished aluminum alloy tubes according to claim 1, characterized in that, Step S3 includes: Based on the phase space reconstruction principle, a multidimensional reconstruction vector is constructed for the contour sequence of time-marked data according to a predetermined delay and embedding dimension; The high-dimensional trajectory formed by the reconstructed vector is divided into several adjacent trajectory segments; Each trajectory segment is mapped to a corresponding sequence of coded symbols based on a pre-built coding dictionary; An attractor feature matrix is ​​constructed based on the sequence of coded symbols, which is used to characterize the dynamic deviation pattern of the pipe cross-sectional profile.

5. The method for roundness detection of finished aluminum alloy tubes according to claim 1, characterized in that, Step S4 includes: Set the encoded symbols in the attractor feature matrix as nodes in the directed graph; Scan the encoded symbol sequence sequentially according to the matrix row order and column order, and count the number of times any two node symbols appear at the same time within a predetermined window to obtain the co-occurrence frequency of node pairs; Traverse the symbol sequence, starting from each node symbol, count the number of times the successor node symbol appears, and compare it with the total number of times the starting symbol appears to obtain the transition probability from the node to the successor node. Add directed edges between nodes based on co-occurrence frequency or transition probability, and set the corresponding co-occurrence frequency or transition probability value as the edge weight to generate a weighted directed graph; A uniform pulse excitation is applied to a weighted directed graph and propagated along the graph edges. The energy response of each node is recorded cumulatively to generate a pulse energy density distribution. The energy density distribution is mapped back to the spatial location of the pipe cross-section to identify areas of energy concentration or abrupt change and locate the location of local roundness deviation.

6. The method for roundness detection of finished aluminum alloy tubes according to claim 1, characterized in that, Step S5 includes: Extract the energy response difference and connectivity change information of each node from the weighted directed graph; Based on the energy response difference and connectivity change, candidate nodes for deviation are marked, and the historical detection records of the candidate nodes for deviation are queried to dynamically adjust the initial judgment threshold. Based on the revised judgment threshold, the candidate nodes of deviation are classified and marked according to the deviation level, which includes slight deviation, obvious deviation and severe deviation. Output the marked deviation nodes, spatial locations, and deviation levels.

7. The method for roundness detection of finished aluminum alloy tubes according to claim 1, characterized in that, Step S6 includes: The detection system is preset with an enhanced accuracy mode, a rapid detection mode, and a steady-state tracking mode. At the end of the testing cycle, obtain the deviation detection rate and production batch status; When the deviation detection rate is greater than the preset accuracy standard or the production batch status is the first piece of a new product, the detection mode will be switched to the enhanced accuracy mode. When the deviation detection rate is less than the preset efficiency standard and the production batch status is a stable batch, the detection mode will be switched to the fast detection mode. When the production batch status is batch switching or requires continuous monitoring, switch the detection mode to steady-state tracking mode.

8. The method for roundness detection of finished aluminum alloy tubes according to claim 1, characterized in that, Step S7 includes: According to the preset sampling cycle, pipe cross-section samples that have been tested and whose deviation nodes have been determined are extracted, and the samples are measured using a high-precision roundness meter. The deviation nodes and deviation levels of the detection output are compared with the roundness values ​​measured by the roundness meter. The coding entries are evaluated based on performance indicators, including symbol classification accuracy, false negative rate, false positive rate, recognition stability, and model contribution. Merge or remove coding entries in the coding dictionary that fail to meet the performance evaluation criteria, and adjust the weight parameters of the corresponding edges in the weighted directed graph based on the performance evaluation results. The updated encoding dictionary and edge weight parameters are then applied to subsequent roundness detection.

9. The method for roundness detection of finished aluminum alloy tubes according to claim 1, characterized in that, Step S8 includes: Collect deviation node locations, deviation levels, and model update records, and associate them with the unique identifier of the pipe material; Based on the preset report template, the associated data is formatted into a quality report, which includes basic information about the pipe, a deviation distribution diagram, deviation level statistics, and model update history. Each tested pipe is assigned a unique traceability code, and the correspondence between the traceability code and the pipe batch number is embedded in the quality report. Based on the traceability identification code, print a QR code or RFID tag and fix it to the end of the pipe.

10. A roundness testing device for finished aluminum alloy tubes, used to implement the roundness testing method for finished aluminum alloy tubes as described in any one of claims 1-9, characterized in that, include: Array construction module: Constructs a multimodal sensing array consisting of laser triangulation, structured light scanning and vibration resonance sensors, performs calibration and synchronous calibration, and maps the output of the multimodal sensing array to the pipe center axis coordinate system; Sequence generation module: Drives a linear slide table at a constant speed along the pipe axis to synchronously acquire cross-sectional point clouds and vibration response signals at predetermined intervals, generating a time-marked contour sequence; Matrix generation module: Based on the phase space reconstruction principle, the contour sequence is mapped to a high-dimensional trajectory. The trajectory segments are discretized and mapped using a preset encoding dictionary to generate an attractor feature matrix. Deviation localization module: Sets the encoded symbols in the attractor feature matrix as nodes, constructs a weighted directed graph based on the node co-occurrence frequency and transition probability, extracts the pulse energy density distribution in the weighted directed graph, and locates local morphological deviations; Deviation determination module: Identifies candidate deviation nodes based on node energy gradient and connectivity changes, corrects the determination threshold by combining historical detection results, and outputs the corresponding deviation node and deviation level according to the determination threshold; Mode switching module: Based on the deviation detection rate and production batch status, it switches between preset detection modes to balance detection accuracy and efficiency; Parameter update module: Compares the detection results with the measured values ​​of the high-precision roundness meter, and updates the coding dictionary and graph edge weight calculation parameters through online learning; Traceability Generation Module: Integrates deviation node locations, deviation levels, and update records to generate a quality report and creates a unique traceability identifier for each pipe.

Images