Machine vision-based pipe deformation adaptive compensation laser cutting device and method

Abstract

The present application relates to the technical field of laser processing, and discloses a pipe deformation adaptive compensation laser cutting equipment and method based on machine vision, which comprises the following steps: controlling a visual detection module to scan a pipe section to extract a geometric center coordinate, combining a feeding position and a rotation angle to construct an initial deformation topology array; extracting a total center deviation vector and separating it into a plastic bending vector and an initial elastic droop vector; analyzing a processing program code to calculate a dynamic section inertia moment to generate a stiffness attenuation coefficient, combining a supporting wheel state to generate a dynamic beam span correction coefficient; and calculating a targeted processing center coordinate according to the above vectors and coefficients to control a machine tool to compensate processing. By decoupling inherent bending and elastic droop, and quantifying the stiffness weakening caused by material removal for feedforward compensation, the error caused by traditional overall compensation is avoided, and the cutting precision and end stability are ensured by preventing the risk of collision of the cutting head when the pipe is close to breaking.

Description

Technical Field

[0001] This invention relates to the field of laser processing technology, specifically to a laser cutting device and method for adaptive compensation of pipe deformation based on machine vision. Background Technology

[0002] In laser cutting of metal tubes, spatial deviation is inevitable during clamping and feeding due to the typically large length-to-diameter ratio of the tubes. This deviation is mainly caused by the inherent plastic bending of the tube during manufacturing and transportation, and the initial elastic sag caused by its own weight within the suspended span after clamping. Existing tube laser cutting machine tools typically compensate for deviations by obtaining the overall center offset of the tube cross-section through mechanical edge finding or basic vision sensors, and inputting this as a single error data point into the control system for position compensation. However, because the direction of plastic bending changes periodically in spatial phase with the rotation of the chuck, while the direction of elastic sag caused by gravity is always fixed downwards, existing overall compensation methods cannot separate these two deformations with distinctly different physical and mechanical properties. This makes it difficult for the control system to accurately establish a three-dimensional deformation model of the tube during multi-axis linkage machining, easily leading to overcompensation or undercompensation during the tube's rotational feed.

[0003] Furthermore, actual laser cutting is a dynamic process of continuous material removal. As various apertures and cut contours on the tube surface are processed through, the effective cross-sectional area of ​​the tube gradually decreases, and the moment of inertia of the section decreases accordingly. This directly leads to a continuous attenuation of the local mechanical stiffness of the tube during processing. Simultaneously, to accommodate the feeding of long tubes, the auxiliary support rollers at the bottom of the machine tool need to continuously adjust their lifting state, causing the actual effective support span of the tube to change in real time. Current follow-up compensation technology is basically based on an initial static physical model of the tube before it is damaged, without considering the attenuation of dynamic mechanical characteristics caused by material removal and changes in support state. When a large area of ​​the tube is cut, its actual downward displacement will far exceed the initial preset data. Especially in the final stage when the tube is about to be completely cut off, the stress at the cross-sectional connection reaches a critical point, easily leading to structural instability, resulting in sudden bending or falling. Existing servo tracking systems lack a mechanism to judge the structural integrity at this time, and will still instinctively control the cutting head to follow this uncontrolled deformation, which not only damages the forming quality of the end cut surface, but also easily causes the cutting head to collide and interfere with the pipe or machine tool. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a machine vision-based adaptive compensation laser cutting device and method for pipe deformation. This solves the problems of existing pipe laser cutting technologies, which struggle to distinguish between the mixed deformation caused by the initial plastic bending of the pipe and the elastic sag due to its own weight. Furthermore, these technologies cannot dynamically compensate for the real-time stiffness attenuation and support span changes caused by material removal during the cutting process, leading to decreased processing accuracy and cutting instability when the pipe is close to fracture.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a machine vision-based adaptive compensation laser cutting device and method for pipe deformation, comprising: The first aspect of this invention provides a machine vision-based adaptive compensation laser cutting method for pipe deformation, comprising: The vision inspection module is controlled to scan the cross-section of the moving pipe, extract the geometric center coordinates, and bind the geometric center coordinates with the real-time feeding position and the real-time rotation angle of the feeding chuck to construct the initial deformation topology array of the pipe. Extract the total center deviation vector from the initial deformation topology array, control the feeding chuck to drive the pipe to rotate in place, and obtain a discrete deviation data sequence. Based on the discrete deviation data sequence, separate the total center deviation vector into a plastic bending vector and an initial elastic sag vector, and combine the two as the initial mechanical reference data of the pipe. The processing program code is parsed to extract the material removal contour information. The dynamic section moment of inertia of the pipe section is calculated based on the material removal contour information to generate the stiffness attenuation coefficient. The running status of the auxiliary support roller is read and combined with the real-time feeding position to generate the dynamic beam span correction coefficient. The coordinates of the target machining center are calculated and output based on the plastic bending vector, initial elastic sag vector, stiffness attenuation coefficient and dynamic beam span correction coefficient. The theoretical coordinates in the machining program code are mapped to the coordinates of the target machining center, and the machine tool body is controlled to perform multi-axis linkage compensation machining actions.

[0006] The process of constructing the initial deformation topology array of the pipe includes: projecting a line laser onto the surface of the pipe using a 2D laser profilometer in the vision inspection module to acquire two-dimensional point cloud data containing the external contour features of the pipe cross-section, and filtering and denoising the two-dimensional point cloud data; using the least squares method to fit the filtered two-dimensional point cloud data to a standard geometric shape, and extracting the actual two-dimensional coordinates of the geometric center of the pipe cross-section in the sensor local coordinate system of the 2D laser profilometer through fitting operations, wherein the actual two-dimensional coordinates of the geometric center are composed of lateral deviation components and longitudinal deviation components; combining the real-time feeding position and real-time rotation angle acquired in the same sampling period with the lateral deviation components and longitudinal deviation components into a joint data group, and combining the continuously acquired joint data groups according to the order of the real-time feeding position to construct a data matrix, forming the initial deformation topology array.

[0007] The process of separating the total center deviation vector into a plastic bending vector and an initial elastic sag vector includes: selecting a reference section from the initial deformation topology array; controlling the feeding mechanism that drives the pipe feed to move the pipe to the absolute feeding position corresponding to the reference section and keeping it axially stationary; driving the feeding chuck to rotate the pipe in situ; obtaining discrete deviation data sequences under different absolute rotation angles; establishing a mathematical relationship equation between the total center deviation vector (the plastic bending vector transformed by a two-dimensional rotation matrix) and the initial elastic sag vector; decomposing the discrete deviation data sequence into a lateral deviation data sequence and a longitudinal deviation data sequence, and respectively... Curve fitting is performed using a sinusoidal fitting function. The amplitude and phase parameters of the fitted periodic curve are mapped to the plastic bending vector of the reference section. The constant bias term obtained from the fitting is mapped to the initial elastic sag vector. The initial elastic sag vector is used as the basic gravity deformation characteristic of the pipe under the current foundation cantilever state. The initial deformation topology array is traversed. The initial elastic sag vector is subtracted from the total center deviation vector corresponding to each real-time feeding position in the initial deformation topology array. The plastic bending vector sequence of each section along the axial direction of the pipe is calculated in reverse. The plastic bending vector sequence is combined with the initial elastic sag vector as the initial mechanical reference data of the pipe.

[0008] The process of calculating the dynamic cross-sectional moment of inertia of the pipe section to generate the stiffness attenuation coefficient includes: establishing a complete geometric model of the current cross-section to be processed; mapping the material removal contour information to the complete geometric model; discretizing the cutting process into multiple processing frames along the processing trajectory; obtaining the dynamic remaining cross-sectional contour corresponding to each processing frame; calculating the initial moment of inertia in the intact state without damage based on the parallel axis theorem; subtracting the moment of inertia distribution of the centroid corresponding to the cut area of ​​the current processing frame from the initial moment of inertia to calculate the dynamic cross-sectional moment of inertia; and defining the ratio of the initial moment of inertia to the dynamic cross-sectional moment of inertia as the stiffness attenuation coefficient.

[0009] The method further includes: setting a structural integrity judgment threshold for monitoring the stiffness attenuation coefficient; when the calculated stiffness attenuation coefficient is less than the structural integrity judgment threshold, determining that the pipe is in a stable stress stage, and continuing to execute dynamic feedforward calculation to update the coordinates of the target machining center; when the calculated stiffness attenuation coefficient is not less than the structural integrity judgment threshold, forcibly stopping the dynamic update of the stiffness attenuation coefficient, and locking the coordinates of the target machining center to the coordinate values ​​of the previous calculation cycle until the machining program code of the current section finishes running.

[0010] The dynamic beam span correction coefficient is generated through the following process: reading the real-time signal status of whether the auxiliary support rollers distributed on the machine tool bed are raised to support the pipe, and determining the most effective support point at the front end of the pipe; calculating the axial distance between the most effective support point at the front end and the detection point as the effective suspension length of the detection point, and the axial distance between the most effective support point at the front end and the machining point as the effective suspension length of the machining point; and defining the cube of the ratio of the effective suspension length of the machining point to the effective suspension length of the detection point as the dynamic beam span correction coefficient.

[0011] The process of calculating and outputting the coordinates of the target machining center includes: performing a two-dimensional rotation transformation matrix operation on the plastic bending vector corresponding to the current absolute rotation angle; multiplying the initial elastic sag vector using the stiffness attenuation coefficient and the dynamic beam span correction coefficient; adding the plastic bending vector after matrix operation to the initial elastic sag vector after multiplication to obtain the comprehensive deviation vector; and performing vector addition operation between the comprehensive deviation vector and the theoretical coordinates of the current section to generate the coordinates of the target machining center.

[0012] The execution process of the multi-axis linkage compensation machining action includes: receiving the reconstructed coordinates of the target machining center; shifting the origin of the overall coordinate system to the coordinates of the target machining center in real time while keeping the relative cutting trajectory size in the machining program code unchanged, and generating dynamic zero-point offset parameters; generating synchronous position drive commands based on the dynamic zero-point offset parameters, and controlling the servo motors in the horizontal and vertical directions to perform translation tracking.

[0013] A second aspect of the present invention provides a machine vision-based adaptive compensation laser cutting device for pipe deformation, comprising: The machine tool body includes a feeding chuck for clamping the pipe, a feeding mechanism for driving the pipe feed, and auxiliary support rollers. A visual inspection module, comprising a 2D laser profilometer installed at the inspection point, the 2D laser profilometer being used to scan the cross-section of the pipe and acquire profilometer scan data; The control system includes a host computer and a numerical control system. The host computer is communicatively connected to the vision inspection module, and the numerical control system is communicatively connected to the machine tool body. The host computer is also communicatively connected to the numerical control system. The host computer is used to calculate the two-dimensional coordinates of the geometric center of the pipe section based on the contour scanning data, bind the two-dimensional coordinates of the geometric center with the real-time feeding position and the real-time rotation angle of the feeding chuck, construct an initial deformation topology array, select a reference section from the initial deformation topology array, and extract the total center deviation vector of the reference section. The CNC system is used to control the feeding mechanism to move the pipe to the real-time feeding position corresponding to the reference section and keep it axially stationary, and to drive the feeding chuck to rotate the pipe in place. The visual inspection module is used to scan the reference section during the in-situ rotation of the pipe and acquire contour scan data at multiple different rotation angles. The host computer is also used to calculate the corresponding two-dimensional coordinates of the geometric center based on the contour scanning data of the reference section at multiple different rotation angles, and synchronously record the feeding chuck angle data corresponding to the two-dimensional coordinates of the geometric center, generate a discrete deviation data sequence composed of multiple rotation angles and their corresponding two-dimensional coordinates of the geometric center, and separate the total center deviation vector of the reference section into a plastic bending vector and an initial elastic sag vector based on the discrete deviation data sequence. The control system is also used to parse the machining program code to extract material removal contour information, calculate the dynamic section moment of inertia of the pipe section based on the material removal contour information to generate a stiffness attenuation coefficient, read the running status of the auxiliary support roller and generate a dynamic beam span correction coefficient in combination with the real-time feeding position, calculate and output the coordinates of the target machining center based on the plastic bending vector, the initial elastic sag vector, the stiffness attenuation coefficient and the dynamic beam span correction coefficient, map the theoretical coordinates in the machining program code to the coordinates of the target machining center, and control the machine tool body to perform compensation machining actions.

[0014] The control system stores a structural integrity judgment threshold. When the stiffness attenuation coefficient is less than the structural integrity judgment threshold, it is in the adaptive compensation zone. When the stiffness attenuation coefficient is not less than the structural integrity judgment threshold, it enters the locking and holding zone, and stops dynamically updating the stiffness attenuation coefficient and locking the coordinates of the target machining center in the locking and holding zone.

[0015] This invention provides a machine vision-based adaptive compensation laser cutting device and method for pipe deformation. It has the following beneficial effects: 1. This invention constructs an initial deformation topology array to accurately separate the extracted total center deviation vector into a plastic bending vector and an initial elastic sag vector. This technical feature can decouple the inherent bending generated during pipe manufacturing from the elastic sag caused by its own weight, and establish independent mechanical reference data for each. This avoids the over-compensation or under-compensation problems caused by the confusion of deformation factors in traditional overall compensation methods, and improves the initial positioning and reference alignment accuracy of the pipe before it enters the cutting stage.

[0016] 2. This invention analyzes the material removal contour information in the processing program code, calculates the dynamic cross-sectional moment of inertia of the pipe section in real time, and generates a stiffness attenuation coefficient. During the continuous cutting of the pipe, the removal of material will change the cross-sectional characteristics of the pipe and cause the structural stress to weaken. This design can quantitatively reflect the degree of weakening of the local stiffness of the pipe by the cutting trajectory and simultaneously convert it into the feedforward compensation amount of the coordinate system. This ensures that when the pipe undergoes large-area hollowing or complex contour cutting, the cutting head can still accurately follow the real-time dynamic deformation trajectory of the pipe.

[0017] 3. This invention generates a dynamic beam span correction coefficient by combining the operating status of the auxiliary support rollers and sets a structural integrity judgment threshold. When the cutting is close to the end or the pipe is about to break, the stiffness attenuation coefficient will trigger the judgment threshold. The system will then enter the locking and holding zone and stop the dynamic update of the coordinates. This effectively prevents the compensation command jump caused by the sudden change in force when the pipe is in a critical fracture state, avoids the collision risk caused by the cutting head tracking abnormal deformation, and ensures the stability of the end cutting process. Attached Figure Description

[0018] Figure 1 This is a perspective view of the present invention; Figure 2 This is a schematic diagram of the cutting structure of the present invention; Figure 3 This is a flowchart of a method according to an embodiment of the present invention.

[0019] Among them, 10. Machine tool body; 11. Laser cutting head; 12. Feeding chuck; 13. Feeding mechanism; 14. Auxiliary support roller; 20. Vision inspection module; 21. 2D laser profilometer; 30. Control system; 31. Host computer; 32. CNC system. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See attached document Figure 1 With appendix Figure 2 The present invention provides a laser cutting device for tube deformation adaptive compensation based on machine vision, comprising: a machine tool body 10, a vision inspection module 20 and a control system 30.

[0022] The machine tool body 10 includes a laser cutting head 11, a feeding chuck 12, a feeding mechanism 13, and auxiliary support rollers 14 mounted on the machine tool bed. The spatial position of the laser cutting head 11 is defined as the processing point. The feeding chuck 12 is used to clamp the tube. The feeding mechanism 13 is mechanically connected to the feeding chuck 12 and drives the tube to feed along the machine tool axis towards the processing point.

[0023] The vision inspection module 20 includes a 2D laser profilometer 21. The 2D laser profilometer 21 is installed upstream of the laser cutting head 11 along the pipe feeding direction. The spatial position of the 2D laser profilometer 21 is defined as the inspection point. There is a fixed physical distance between the inspection point and the processing point.

[0024] The control system 30 includes a host computer 31 and a numerical control system 32. The host computer 31 establishes data communication with the vision inspection module 20 to receive and process contour scanning data. The numerical control system 32 is electrically connected to the servo motors of each motion axis of the machine tool body 10 to control the mechanical movement of the machine tool body 10. The host computer 31 and the numerical control system 32 are connected via an industrial fieldbus to achieve data interaction.

[0025] See attached document Figure 3 This invention provides a machine vision-based adaptive compensation laser cutting method for pipe deformation, comprising the following steps: S100, the control system 30 controls the machine tool body 10 to drive the pipe to move, controls the vision inspection module 20 to continuously scan the pipe cross section at the inspection point, the host computer 31 calculates and extracts the geometric center coordinates of the pipe cross section, and binds the geometric center coordinates with the real-time feeding position and chuck rotation angle fed back by the CNC system 32 to construct the initial deformation topology array of the pipe. S200, the host computer 31 extracts the total center deviation vector in the initial deformation topology array, and separates the total center deviation vector into a plastic bending vector that rotates with the pipe and an initial elastic sag vector with a constant direction through a decoupling algorithm. The plastic bending vector sequence generated by the reverse calculation is combined with the initial elastic sag vector as the initial mechanical reference data of the pipe. S300, the CNC system 32 parses the machining program code to be executed, extracts the material removal contour information, the host computer 31 calculates the dynamic section moment of inertia of the pipe section based on the material removal contour information, generates the stiffness attenuation coefficient corresponding to the change in pipe stiffness, reads the running status of the auxiliary support roller 14 and calculates and generates the dynamic beam span correction coefficient in combination with the feeding position. S400, when the section of the pipe to be processed moves to the processing point, the host computer 31 calculates and outputs the coordinates of the target machining center under the current processing state based on the plastic bending vector, the initial elastic sag vector, the stiffness attenuation coefficient and the dynamic beam span correction coefficient. The CNC system 32 maps the theoretical coordinates in the processing program code to the coordinates of the target machining center and controls the machine tool body 10 to perform multi-axis linkage compensation processing actions.

[0026] See attached document Figure 3 Step S100 involves acquiring high-frequency discrete contour data and establishing a multi-dimensional spatial data mapping relationship before the pipe is subjected to mechanical damage. Since pipes in actual industrial production generally exhibit inherent bending deformation due to manufacturing processes, and also experience sagging deformation under machine tool clamping, the system needs to acquire the actual positional state of each cross-section of the pipe within the machine tool space before the laser cuts and destroys its original structure. This serves as the data basis for subsequent mechanical feedforward compensation. Specifically, this may include the following sub-steps: S110, the control system 30 controls the feeding mechanism 13 of the machine tool body 10 to drive the tube to move towards the processing point along the machine tool feed axis. During the movement of the tube, the 2D laser profilometer 21 in the vision inspection module 20 projects a line laser onto the surface of the tube according to a preset scanning frequency. The photosensitive chip inside the 2D laser profilometer 21 receives the laser beam reflected from the surface of the tube and generates two-dimensional point cloud data containing the external contour features of the tube cross-section. The optical triangulation principle of the 2D laser profilometer 21 in acquiring two-dimensional point cloud data is well known in the art and will not be described in detail here.

[0027] S120, the host computer 31 receives the aforementioned two-dimensional point cloud data in real time through the industrial data interface, and performs filtering and noise reduction processing on the acquired two-dimensional point cloud data to remove wandering noise points. The host computer 31 uses the least squares method to perform standard geometric shape fitting on the filtered two-dimensional point cloud data. For circular cross-section pipes, the least squares circle fitting algorithm is used; for rectangular cross-section pipes, the least squares line fitting algorithm is used, and the intersection point is calculated to extract features. Through fitting operations, the host computer 31 extracts the actual two-dimensional coordinates of the geometric center of the pipe cross-section in the sensor's local coordinate system. These two-dimensional coordinates of the geometric center are specifically represented by lateral deviation components and longitudinal deviation components.

[0028] In step S130, within the same sampling period when the 2D laser profilometer 21 acquires valid scanning data, the CNC system 32 synchronously latches the real-time encoder pulse data of the servo motors of each motion axis of the machine tool via a high-speed fieldbus. The CNC system 32 parses the encoder pulse data and calculates the absolute feeding position of the currently scanned section of the pipe and the absolute rotation angle of the feeding chuck 12. The absolute feeding position is the length coordinate along the axial direction of the pipe. The host computer 31 uses the distributed clock synchronization mechanism of the fieldbus to bind the two-dimensional coordinates of the geometric center calculated and extracted in step S120 with the absolute feeding position and absolute rotation angle fed back by the CNC system 32 at the same moment. This binding process converts the discrete visual measurement results into a joint data set containing the absolute feeding position, absolute rotation angle, and lateral and longitudinal deviation components.

[0029] S140, as the machine tool feeding mechanism 13 continuously drives the pipe feed, the host computer 31 constructs a data matrix from the continuously acquired joint data sets in its internal memory according to the order of absolute feeding positions. This data matrix is ​​the initial deformation topology array of the pipe. In terms of data structure, this initial deformation topology array is represented as a multi-dimensional sequence that maps the actual center coordinates of the pipe cross-section to the chuck angle at the corresponding absolute feeding position. This array data serves as a reference for subsequent mechanical vector decoupling and trajectory compensation calculations by the system.

[0030] Step S200 involves separating the mixed deformation acquired by visual inspection into independent physical components. The pipe is typically in a cantilevered state under machine tool clamping. The deviation data acquired by the visual inspection module 20 at the inspection point includes not only the inherent bending generated during pipe manufacturing but also the downward bending deformation caused by gravity. Conventional compensation methods treat this composite deformation as an indivisible whole, which can easily lead to compensation failure after the internal stiffness of the pipe is cut and damaged. To solve this problem, the system obtains the initial mechanical reference data of the pipe through spatial geometry and statics decoupling, specifically including the following sub-steps: S210, the host computer 31 selects a specific reference section from the initial deformation topology array and extracts the total center deviation vector of the reference section. The total center deviation vector is the two-dimensional position offset detected in the sensor coordinate system. The total center deviation vector is composed of two independent linear superpositions: the plastic bending vector and the initial elastic sag vector. The plastic bending vector represents the inherent permanent deformation of the tube. This component is stationary relative to the tube body, but its direction changes periodically as the tube rotates in the machine tool coordinate system. The initial elastic sag vector represents the cantilever elastic deformation of the tube caused by gravity. This component points downwards constantly in the machine tool coordinate system, and the die length remains constant when the tube is in the same absolute feeding position.

[0031] S220, the control system 30 performs a rotational calibration operation on the reference section to obtain discrete data samples required for separation calculation. The CNC system 32 controls the feeding mechanism 13 to move the pipe to the absolute feeding position corresponding to the reference section and keep it axially stationary. The CNC system 32 drives the feeding chuck 12 to rotate the pipe in place for at least one revolution. During the pipe rotation, the vision inspection module 20 acquires the two-dimensional coordinates of the geometric center of the section at multiple different absolute rotation angles at fixed angular intervals. The host computer 31 synchronously records the corresponding chuck angle data, generating a discrete deviation data sequence consisting of multiple angle indices and corresponding two-dimensional coordinates. This rotational calibration operation is performed only as a one-time operation in the initial stage of processing to avoid frequent machine stoppages for calibration during subsequent continuous cutting.

[0032] S230, the host computer 31 performs analytical calculations on the discrete deviation data sequence of the reference section using a decoupling algorithm. The physical principle of this decoupling algorithm is that the inherent bending of the pipe rotates with it, and its projection onto the horizontal and vertical axes of the stationary machine tool coordinate system exhibits periodic trigonometric function changes, while the sag deformation caused by gravity remains a fixed constant on the axis corresponding to the direction of gravity. The host computer 31 establishes a mathematical equation relating the pipe deformation, where the total center deviation vector is equal to the vector obtained by transforming the plastic bending vector through a two-dimensional rotation matrix, and the vector sum of the initial elastic sag vector. The corresponding matrix equation is expressed as: ; In the formula, Represents the total center deviation vector. Indicates the current absolute rotation angle. Represents the plastic bending vector. This represents the initial elastic droop vector. This represents a two-dimensional rotation transformation matrix constructed based on the current absolute rotation angle. Its specific matrix form is defined by... trigonometric functions and It is composed of vectors that are equivalently mapped from the local coordinate system of the pipe to the global coordinate system of the machine tool.

[0033] S240, using the aforementioned mathematical relationships, the host computer 31 decomposes the discrete deviation data sequence into a lateral deviation data sequence and a longitudinal deviation data sequence, and inputs them respectively into a sine fitting function for curve fitting. In the periodic curve output by the fitting, the amplitude and phase parameters are directly mapped to the magnitudes of the plastic bending vector components on the corresponding coordinate axes, while the constant bias term obtained from the fitting corresponds to the magnitudes of the fixed initial elastic droop vector components. Through calculation, the host computer 31 completes the separation and extraction of the physical field. For the solution process of the sine curve equation fitting, those skilled in the art can use conventional algorithms such as least squares fitting to program and implement it. The specific mathematical iteration process is a well-known technique in this field and will not be elaborated here.

[0034] In step S250, the host computer 31 stores the calculated and extracted plastic bending vector and initial elastic sag vector of the reference section in memory. The host computer 31 uses this initial elastic sag vector as the basic gravity deformation characteristic of the pipe under the current foundation cantilever state. Subsequently, the host computer 31 traverses the initial deformation topology array obtained in step S100, using this basic gravity deformation characteristic to eliminate the gravity sag components of each absolute feeding position in the array that are within the center deviation vector. Specifically, the host computer 31 subtracts the aforementioned initial elastic sag vector from the total center deviation vector corresponding to each absolute feeding position in the initial deformation topology array, and reverse-calculates the inherent plastic bending vector sequence of each section along the axial direction of the pipe. The host computer 31 combines the reverse-calculated plastic bending vector sequence of each section with the initial elastic sag vector of the foundation, using them together as the initial mechanical reference data of the pipe. The initial mechanical reference data reflects the force balance reference of the pipe under the intact cross-section state and is used to support the dynamic mechanical feedforward calculations performed during the subsequent material removal process.

[0035] Step S300 involves converting the physical stiffness loss and boundary support changes caused by subsequent material removal into feedforward control variables. During tube laser cutting, as the laser beam removes the solid material of the tube, the structural integrity of the tube cross-section is compromised, and its stiffness against bending deformation decreases. This decrease in tube stiffness further exacerbates the elastic sag deformation originally caused by gravity. As the machine tool advances, the tube's overhang length continuously changes, and this change in overhang length causes nonlinear changes in sag. Therefore, it is necessary to pre-analyze the machining trajectory and establish a mechanical feedforward prediction model, specifically including the following sub-steps: S310, the CNC system 32 pre-parses the machining program code to be executed. The machining program code is usually in the G-code format of conventional CNC machining. The CNC system 32 extracts the material removal contour information defined in the machining program code on the current cross-section to be machined. The host computer 31 establishes a complete geometric model of the cross-section based on the original outer diameter and wall thickness parameters of the pipe pre-stored in the machine tool control system 30, and maps the extracted material removal contour information to the complete geometric model. The host computer 31 discretizes the cutting process into multiple machining frames along the machining trajectory, and obtains the dynamic remaining cross-sectional contour corresponding to each machining frame during the actual cutting trajectory evolution.

[0036] S320, the host computer 31 calculates the dynamic moment of inertia of the pipe section based on the material removal contour information, and generates a stiffness attenuation coefficient corresponding to the change in pipe stiffness. The moment of inertia is a physical quantity that measures a structure's ability to resist bending deformation. Based on the parallel axis theorem, the host computer 31 calculates the initial moment of inertia of the section in its intact state without damage. Then, it subtracts the moment of inertia distribution of the centroid corresponding to the cut area in the current processing frame from the initial moment of inertia to calculate the dynamic moment of inertia as the cutting trajectory progresses. The host computer 31 defines the ratio of the initial moment of inertia to the dynamic moment of inertia as the stiffness attenuation coefficient. Its corresponding mathematical formula is: ; In the formula, This represents the stiffness attenuation coefficient. Represents the initial moment of inertia. This represents the dynamic moment of inertia. As the cross-sectional material decreases, the dynamic moment of inertia gradually decreases, while the stiffness attenuation coefficient increases accordingly. The stiffness attenuation coefficient characterizes the factor by which the actual sagging deformation of the pipe is proportionally amplified relative to the initial state due to the reduction in its own stiffness. The surface integral solution of the cross-sectional moment of inertia and the algebraic operations of the parallel axis theorem are conventional calculation methods in structural mechanics in this field, and will not be elaborated here.

[0037] S330, to prevent calculation errors caused by the impending breakage of the pipe at the cutting end, the host computer 31 sets a structural integrity judgment threshold for monitoring the stiffness attenuation coefficient. At the moment the cutting is about to complete, the remaining dynamic moment of inertia at the pipe connection approaches zero, causing the stiffness attenuation coefficient calculated by the formula to approach infinity, leading to singularities in the computer's numerical calculations. The structural integrity judgment threshold is pre-calibrated by technicians based on the pipe's yield strength and safety factor and input into the system. Specifically, this structural integrity judgment threshold is taken as the stiffness attenuation state when the remaining effective cross-sectional area of ​​the corresponding pipe accounts for 5% to 10% of the total area. When the calculated stiffness attenuation coefficient is less than this structural integrity judgment threshold, the pipe is determined to be in a stable stress stage, and the control system 30 continues to perform dynamic feedforward calculations. When the calculated stiffness attenuation coefficient is greater than or equal to this structural integrity judgment threshold, the pipe is determined to have entered the critical fracture zone, and the system forcibly stops dynamically updating the stiffness attenuation coefficient to avoid singularities that could cause violent movements of the machine tool's servo motors.

[0038] S340, the host computer 31 reads the operating status of the machine tool auxiliary support roller 14 and calculates and generates a dynamic beam span correction coefficient based on the absolute feeding position. According to the stress deformation principle of cantilever beams, the gravitational sag deformation of the cantilever end of the pipe is proportional to the cube of its suspended span. The host computer 31 reads the current absolute feeding position and the real-time IO signal status of whether the auxiliary support rollers 14 distributed on the machine tool bed are raised to support the pipe. The host computer 31 uses the most effective support point at the foremost end of the pipe as the calculation origin. When the auxiliary support roller 14 is in the lowered state, the most effective support point at the foremost end is the clamping end face of the feeding chuck 12; when the auxiliary support roller 14 is in the raised supporting state, the most effective support point at the foremost end changes to the central axis of the auxiliary support roller 14. Based on the aforementioned mechanical boundary conditions, the host computer 31 calculates the axial distance between the foremost effective support point and the detection point as the effective overhang length of the detection point, and the axial distance between the foremost effective support point and the machining point as the effective overhang length of the machining point. The host computer 31 defines the cube of the ratio of the effective overhang length of the machining point to the effective overhang length of the detection point as the dynamic beam span correction coefficient. Its corresponding mathematical formula is: ; In the formula, This represents the dynamic beam span correction factor. Indicates the effective overhang length of the processing point. This indicates the effective suspended length of the detection point. The dynamic beam span correction factor is used to correct for errors in the amount of sag caused by gravity due to spatial position changes and variations in bottom boundary support conditions.

[0039] Step S400 involves fusing the decoupled mechanical parameters with spatial geometric parameters to guide the underlying servo motors of the machine tool to perform compensating actions. In laser tube cutting, the dynamic deformation of the tube during processing causes the actual machined surface to deviate from the theoretical machining position. The system dynamically reconstructs the machining coordinate origin and performs multi-axis linkage following, specifically including the following sub-steps: S410, when the pipe section to be processed moves to the processing point, the host computer 31, based on the current absolute feeding position, indexes and extracts the plastic bending vector and initial elastic sag vector corresponding to the absolute feeding position from the initial mechanical reference data. Combining the stiffness attenuation coefficient and the dynamic beam span correction coefficient, it calculates and outputs the target machining center coordinates under the current processing state. The host computer 31 obtains the theoretical machining center coordinates of the current section and calculates the comprehensive deviation vector. The physical principle of this calculation is that the inherent plastic bending of the pipe only changes its spatial orientation with the rotation of the pipe itself, therefore only a corresponding angle matrix transformation is needed; while the elastic sag caused by gravity is always vertically downward in the machine tool coordinate system, but its sag amplitude will nonlinearly increase with the decrease in pipe stiffness caused by cutting and the increase in the overhang length caused by pipe feed. Therefore, it is necessary to amplify it by multiplying it with the stiffness attenuation coefficient and the dynamic beam span correction coefficient. The mathematical formula for the comprehensive deviation vector is expressed as: ; In the formula, Represents the overall deviation vector. This indicates the current absolute rotation angle of the pipe at the processing point. This represents a two-dimensional rotation transformation matrix constructed based on the current absolute rotation angle. Represents the plastic bending vector. This represents the dynamic beam span correction factor. This represents the stiffness attenuation coefficient. This represents the initial elastic droop vector.

[0040] Subsequently, the host computer 31 performs a vector addition operation between the comprehensive deviation vector and the theoretical machining center coordinates, and the corresponding mathematical formula is expressed as follows: ; In the formula, Indicates the coordinates of the target machining center. This represents the coordinates of the theoretical machining center. The coordinates of the targeted machining center reflect the actual physical center of the pipe under its current stress state and spatial position.

[0041] S420, the host computer 31 executes segmented coordinate reconstruction logic based on the structural integrity threshold. The host computer 31 divides the entire cutting trajectory execution process into an adaptive compensation zone and a locking and holding zone. When the stiffness attenuation coefficient is less than the structural integrity threshold, machining is in the adaptive compensation zone, and the host computer 31 calculates and updates the target machining center coordinates in real time according to the control cycle. When the stiffness attenuation coefficient is greater than or equal to the structural integrity threshold, machining enters the locking and holding zone. The host computer 31 stops dynamically amplifying the calculation based on the stiffness attenuation coefficient and locks the target machining center coordinates to the coordinate values ​​of the previous calculation cycle before entering the locking and holding zone, maintaining this position until the current section's cutting program ends.

[0042] S430, the CNC system 32 receives the reconstructed coordinates of the target machining center and maps the theoretical coordinates in the machining program code to the target machining center coordinates. While maintaining the relative cutting trajectory dimensions in the machining program code, the CNC system 32 shifts its overall coordinate system origin to the aforementioned target machining center coordinates in real time. Through the translation of the coordinate system origin, the system converts the pipe deformation error into dynamic zero-point offset parameters of the machine tool's working coordinate system.

[0043] S440, the CNC system 32 controls the machine tool body 10 to perform multi-axis linkage compensation machining actions according to the dynamic zero-point offset parameters. Specifically, the CNC system 32 generates synchronous position drive commands to control the X-axis and Y-axis servo motors of the machine tool to perform translation tracking. The X-axis servo motor compensates for the center offset of the tube in the horizontal direction, and the Y-axis servo motor compensates for the center offset of the tube in the vertical direction, so that the central optical axis of the laser cutting head 11 is always aligned with the actual offset center of the tube. For local undulations in the height of the tube surface, the control system 30 assists the Z-axis servo motor to perform follow-up smooth compensation, maintaining a constant focal length between the laser nozzle and the tube surface. For the underlying following control logic of Z-axis height adjustment based on the capacitive sensing principle, those skilled in the art can combine it with a conventional capacitive height adjustment system. Its related hardware structure and control process are well-known technologies in the field and will not be described in detail here. The CNC system 32 synchronously controls the angle offset of the feeding chuck 12 and the movement of the laser cutting head 11 to adaptively correct, completing the compensation machining of the tube under dynamic deformation.

Claims

1. A machine vision-based adaptive compensation laser cutting method for pipe deformation, characterized in that, include: The vision detection module (20) is controlled to scan the cross section of the moving pipe, extract the geometric center coordinates, bind the geometric center coordinates with the real-time feeding position and the real-time rotation angle of the feeding chuck (12) of the machine tool body (10) to construct the initial deformation topology array of the pipe. Extract the total center deviation vector from the initial deformation topology array, control the feeding chuck (12) to drive the pipe to rotate in place to obtain discrete deviation data sequence, and separate the total center deviation vector into plastic bending vector and initial elastic sag vector based on the discrete deviation data sequence, and combine the two as the initial mechanical reference data of the pipe. The processing program code is parsed to extract the material removal contour information. Based on the material removal contour information, the dynamic section moment of inertia of the pipe section is calculated to generate the stiffness attenuation coefficient. The running status of the auxiliary support roller (14) of the machine tool body (10) is read, and the dynamic beam span correction coefficient is generated in combination with the real-time feeding position. The coordinates of the target machining center are calculated and output based on the plastic bending vector, the initial elastic sag vector, the stiffness attenuation coefficient and the dynamic beam span correction coefficient. The theoretical coordinates in the machining program code are mapped to the coordinates of the target machining center, and the machine tool body (10) is controlled to perform multi-axis linkage compensation machining actions.

2. The machine vision-based adaptive compensation laser cutting method for pipe deformation according to claim 1, characterized in that, The process of constructing the initial deformation topology array of the pipe includes: The 2D laser profilometer (21) in the vision detection module (20) projects a line laser onto the surface of the pipe to obtain two-dimensional point cloud data containing the external contour features of the pipe cross section, and performs filtering and noise reduction processing on the two-dimensional point cloud data. The least squares method is used to fit the filtered two-dimensional point cloud data to a standard geometric figure. The actual geometric center two-dimensional coordinates of the pipe section in the local coordinate system of the sensor of the 2D laser profiler (21) are extracted by the fitting operation. The actual geometric center two-dimensional coordinates are composed of lateral deviation components and longitudinal deviation components. The real-time feeding position and real-time rotation angle acquired within the same sampling period are combined with the lateral deviation component and the longitudinal deviation component to form a joint data group. The continuously acquired joint data groups are combined in the order of the real-time feeding position to construct a data matrix to form the initial deformation topology array.

3. The machine vision-based adaptive compensation laser cutting method for pipe deformation according to claim 1, characterized in that, The process of separating the total center deviation vector into a plastic bending vector and an initial elastic sagging vector includes: A reference section is selected from the initial deformation topology array. The feeding mechanism (13) of the machine tool body (10) that drives the pipe feed moves the pipe to the real-time feeding position corresponding to the reference section and keeps it axially stationary. The feeding chuck (12) drives the pipe to rotate in place and obtains discrete deviation data sequences under different real-time rotation angles. A mathematical equation is established to establish the relationship between the total center deviation vector of the reference section and the vector sum of the plastic bending vector of the reference section after transformation by a two-dimensional rotation matrix and the initial elastic sag vector. The discrete deviation data sequence is decomposed into a lateral deviation data sequence and a longitudinal deviation data sequence, and then input into a sine fitting function for curve fitting. The amplitude and phase parameters of the periodic curve output by the fitting are mapped to the plastic bending vector of the reference section, and the constant bias term obtained by fitting is mapped to the initial elastic sagging vector. The initial elastic sag vector is used as the basic gravity deformation feature of the pipe under the current foundation cantilever state. The initial deformation topology array is traversed, and the initial elastic sag vector is subtracted from the total center deviation vector corresponding to each real-time feeding position in the initial deformation topology array. The plastic bending vector sequence of each section of the pipe along the axial direction is calculated in reverse. The plastic bending vector sequence is combined with the initial elastic sag vector as the initial mechanical reference data of the pipe.

4. The machine vision-based adaptive compensation laser cutting method for pipe deformation according to claim 1, characterized in that, The process of calculating the dynamic cross-sectional moment of inertia and the stiffness attenuation coefficient of the pipe cross-section includes: Establish a complete geometric model of the current cross section to be processed, map the material removal contour information to the complete geometric model, discretize the cutting process into multiple processing frames along the processing trajectory, and obtain the dynamic remaining cross section contour corresponding to each processing frame. The initial moment of inertia in the intact state without damage is calculated based on the parallel axis theorem. The moment of inertia distribution of the centroid corresponding to the cut area of ​​the current processing frame is subtracted from the initial moment of inertia to calculate the dynamic cross-sectional moment of inertia. The ratio of the initial moment of inertia to the dynamic cross-sectional moment of inertia is defined as the stiffness attenuation coefficient.

5. The machine vision-based adaptive compensation laser cutting method for pipe deformation according to claim 4, characterized in that, The method further includes: Set a structural integrity judgment threshold for monitoring the stiffness attenuation coefficient; When the calculated stiffness attenuation coefficient is less than the structural integrity judgment threshold, the pipe is determined to be in a stable stress stage, and the coordinates of the target machining center are continuously updated dynamically. When the calculated stiffness attenuation coefficient is not less than the structural integrity judgment threshold, the dynamic update of the stiffness attenuation coefficient is forcibly stopped, and the coordinates of the target machining center are locked to the coordinate values ​​of the previous calculation cycle until the machining program code of the current section finishes running.

6. The machine vision-based adaptive compensation laser cutting method for pipe deformation according to claim 1, characterized in that, The dynamic beam span correction factor is generated through the following process: Read the real-time signal status of whether the auxiliary support rollers (14) distributed on the machine tool bed are raised to support the pipe, and determine the most effective support point at the front end of the pipe; The axial distance between the effective support point at the foremost end and the detection point of the vision inspection module (20) is calculated as the effective suspension length of the detection point, and the axial distance between the effective support point at the foremost end and the processing point of the laser cutting head (11) on the machine tool body (10) is calculated as the effective suspension length of the processing point. The cube of the ratio of the effective suspended length of the processing point to the effective suspended length of the detection point is defined as the dynamic beam span correction coefficient.

7. The machine vision-based adaptive compensation laser cutting method for pipe deformation according to claim 1, characterized in that, The process of calculating and outputting the coordinates of the target machining center includes: Perform a two-dimensional rotation transformation matrix operation on the plastic bending vector corresponding to the current real-time rotation angle, and multiply the initial elastic sag vector using the stiffness attenuation coefficient and the dynamic beam span correction coefficient. The plastic bending vector after matrix operation is added to the initial elastic sagging vector after product operation to obtain the comprehensive deviation vector; The coordinates of the target machining center are generated by performing a vector addition operation between the comprehensive deviation vector and the theoretical coordinates of the current section.

8. The machine vision-based adaptive compensation laser cutting method for pipe deformation according to claim 1, characterized in that, The execution process of the multi-axis linkage compensation machining action includes: Receive the reconstructed coordinates of the target machining center, and while keeping the relative cutting trajectory size in the machining program code unchanged, shift the origin of the overall coordinate system to the coordinates of the target machining center in real time to generate dynamic zero-point offset parameters. Based on the dynamic zero-point offset parameters, a synchronous position drive command is generated to control the horizontal servo motor and the vertical servo motor of the machine tool body (10) to perform translation tracking.

9. A machine vision-based adaptive compensation laser cutting device for pipe deformation, characterized in that, The method for performing the machine vision-based adaptive compensation laser cutting method for pipe deformation according to any one of claims 1 to 8 includes: The machine tool body (10) includes a feeding chuck (12) for clamping the pipe, a feeding mechanism (13) for driving the pipe to feed, and an auxiliary support roller (14). The visual inspection module (20) includes a 2D laser profilometer (21) installed at the inspection point. The 2D laser profilometer (21) is used to scan the cross section of the pipe and obtain profile scanning data. The control system (30) includes a host computer (31) and a numerical control system (32). The host computer (31) is communicatively connected to the vision inspection module (20), and the numerical control system (32) is communicatively connected to the machine tool body (10). The host computer (31) is used to calculate the two-dimensional coordinates of the geometric center of the pipe section based on the contour scanning data, bind the two-dimensional coordinates of the geometric center with the real-time feeding position and the real-time rotation angle of the feeding chuck (12) to construct an initial deformation topology array, select a reference section from the initial deformation topology array, and extract the total center deviation vector of the reference section. The CNC system (32) is used to control the feeding mechanism (13) to move the pipe to the real-time feeding position corresponding to the reference section and keep it axially stationary, and to drive the feeding chuck (12) to drive the pipe to rotate in place; The visual inspection module (20) is used to scan the reference section and acquire contour scanning data at multiple different rotation angles during the in-situ rotation of the pipe. The host computer (31) is also used to calculate the corresponding two-dimensional coordinates of the geometric center based on the contour scanning data of the reference section at multiple different rotation angles, and synchronously record the angle data of the feeding chuck (12) corresponding to the two-dimensional coordinates of the geometric center, generate a discrete deviation data sequence composed of multiple rotation angles and their corresponding two-dimensional coordinates of the geometric center, and separate the total center deviation vector of the reference section into a plastic bending vector and an initial elastic sag vector based on the discrete deviation data sequence. The control system (30) is also used to parse the machining program code to extract material removal contour information, calculate the dynamic section moment of inertia of the pipe section according to the material removal contour information to generate stiffness attenuation coefficient, read the running status of the auxiliary support roller (14) and generate dynamic beam span correction coefficient in combination with the real-time feeding position, calculate and output the target machining center coordinates according to the plastic bending vector, the initial elastic sag vector, the stiffness attenuation coefficient and the dynamic beam span correction coefficient, map the theoretical coordinates in the machining program code to the target machining center coordinates, and control the machine tool body (10) to perform compensation machining actions.

10. The machine vision-based adaptive compensation laser cutting equipment for pipe deformation according to claim 9, characterized in that: The control system (30) stores a structural integrity determination threshold and is in the adaptive compensation zone when the stiffness attenuation coefficient is less than the structural integrity determination threshold. When the stiffness attenuation coefficient is not less than the structural integrity judgment threshold, the system enters the locking and holding zone, and stops dynamically updating the stiffness attenuation coefficient and locking the coordinates of the target machining center in the locking and holding zone.

Images