Laser pipe cutting machine machining optimization method for small pipes

Abstract

The invention relates to the technical field of laser pipe cutting machining, and discloses a laser pipe cutting machine machining optimization method for small pipes. The method comprises the following steps: acquiring three-dimensional point cloud data of a to-be-processed pipe, and extracting a pipe axis direction vector and pipe wall thickness distribution data through geometric topology analysis; a dynamic machining coordinate system is established based on the axis direction vector, and the theoretical projection position of the laser focus in the coordinate system is calculated; material absorption characteristic parameters are matched according to the pipe wall thickness distribution data, and a laser power density gradient adjusting instruction is generated; monitoring the radial run-out amount in the rotating process of the pipe in real time, and generating a laser focus track compensation parameter in combination with a theoretical projection position; and finally, the laser power density gradient adjusting instruction and the focus track compensation parameters are fused to generate a cutting path control signal, and a laser cutting head is driven to complete machining. According to the method, the problems of insufficient precision and unstable quality in the machining process of small pipes are effectively solved, and the machining adaptability is improved.

Description

Technical Field

[0001] This invention relates to the field of laser tube cutting technology, specifically an optimized method for laser tube cutting machine processing of small tubes. Background Technology

[0002] Small tubular parts are widely used in fields such as machinery manufacturing, aerospace, and medical devices. These tubular materials are typically characterized by small diameter, long length, thin and unevenly distributed wall thickness, requiring extremely high processing precision and quality. Laser tube cutting technology, with its advantages of high cutting speed and small heat-affected zone, has become the main method for processing small tubular parts, but it still faces many technical challenges in practical applications. Traditional laser tube cutting machines typically use a fixed coordinate system for programming and cutting. This coordinate system references the machine's mechanical standards and does not account for the geometric deviations of the tube itself. Because thin tubes are prone to bending and skewing during rolling, transportation, or storage, their actual axis deviates from the theoretical axis, causing the laser focus to be unable to accurately align with the cutting position, resulting in cutting dimensional errors. Furthermore, current technologies rely heavily on preset parameters or localized sampling measurements to detect tube wall thickness, making it difficult to obtain comprehensive dynamic distribution data on wall thickness. When the tube has uneven wall thickness, if the laser power remains constant, problems such as overheating and burning through thin-walled areas and incomplete cutting through thick-walled areas can easily occur, severely affecting the cut quality. Radial runout is another key factor affecting processing accuracy during rotary cutting of pipes. Thin pipes have poor rigidity and, during rotation, are affected by their own gravity and the clamping force of the fixture, resulting in periodic radial displacement. While traditional processing equipment is equipped with simple runout detection devices, the detection response is slow, compensation is delayed, and it cannot correct the laser focus trajectory in real time, leading to wavy deviations in the cutting trajectory. Furthermore, existing laser power adjustments are mostly based on empirical values ​​or single parameters set in segments, without establishing a dynamic relationship with the real-time geometric parameters of the pipe. This makes it difficult to adapt to the cutting needs of different positions and wall thicknesses, resulting in poor processing adaptability. These problems collectively lead to a low pass rate in laser cutting of thin pipes, hindering the improvement of processing efficiency and restricting the development of related industries. Summary of the Invention

[0003] The purpose of this invention is to provide an optimized method for laser tube cutting for thin tubes, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides an optimized method for laser tube cutting machine processing of thin tubes, the method comprising: Acquire the 3D point cloud data of the pipe to be processed, and extract the pipe axis direction vector and pipe wall thickness distribution data through geometric topology analysis; A dynamic machining coordinate system is established based on the axis direction vector, and the theoretical projection position of the laser focus in the dynamic machining coordinate system is calculated. Based on the tube wall thickness distribution data, match the material absorption characteristic parameters to generate a laser power density gradient adjustment command; The radial runout during the pipe rotation process is monitored in real time, and the laser focus trajectory compensation parameters are generated by combining the theoretical projection position. The laser power density gradient adjustment command and the focus trajectory compensation parameters are combined to generate a cutting path control signal, which drives the laser cutting head to complete the pipe processing.

[0005] Preferably, acquiring the three-dimensional point cloud data of the pipe to be processed includes: A structured light scanning device was used to collect point clouds on the surface of the pipe, and a non-rigid registration algorithm was used to align the point cloud data from multiple perspectives. Gaussian curvature filtering is applied to the registered point cloud to remove noise points and then the continuous topological mesh of the pipe surface is reconstructed. The minimum circumscribed ellipse of the cross section is calculated based on a continuous topological mesh, and the spatial axis of the pipe is fitted with the sequence of the ellipse's center points.

[0006] Preferably, the calculation of the theoretical projection position of the laser focus in the dynamic processing coordinate system includes: Discretize the spatial axis of the pipe into equidistant nodes, and establish a local orthogonal coordinate system at each node; Based on the projection coordinates of the preset cutting path in the local orthogonal coordinate system, the normal distance of the laser focus relative to the pipe surface is mapped. By superimposing the pipe wall thickness distribution data, material refraction compensation is performed on the normal distance to generate a set of theoretical spatial coordinates for the laser focus.

[0007] Preferably, the real-time monitoring of radial runout during pipe rotation includes: A high-speed linear array sensor is installed along the rotation axis of the pipe to collect time-series changes in the pipe surface profile. Fourier envelope analysis was performed on the time-series variation data to extract the fundamental frequency vibration component and the random disturbance component; The fundamental frequency vibration component is mapped to a rigid offset of the dynamic machining coordinate system, and the random disturbance component is mapped to an elastic deformation.

[0008] Preferably, the generated laser focus trajectory compensation parameters include: Convert the rigid offset into a translation transformation matrix for the dynamic machining coordinate system; Calculate the surface curvature gradient of the pipe based on the elastic deformation, and generate a thermal deformation correction matrix; By combining the translation transformation matrix and the thermal deformation correction matrix, an affine transformation is performed on the theoretical spatial coordinate set.

[0009] Preferably, the absorption characteristic parameters of the matching material include: Construct a database of spectral reflectance of pipe materials and establish a mapping relationship between reflectance and laser wavelength; The material density distribution is inverted based on the pipe wall thickness distribution data, and the equivalent absorption depth coefficient is calculated. By integrating the equivalent absorption depth coefficient with the mapping relationship, a power attenuation ratio factor is generated for different cutting positions.

[0010] Preferably, the laser power density gradient adjustment command includes: Establish the axial power distribution curve based on the power attenuation ratio factor; Edge effect compensation is performed on the power distribution curve by superimposing the curvature radius of the cutting path; The compensated power distribution curve is converted into a duty cycle timing sequence of the laser modulation pulse.

[0011] Preferably, the generation of the cutting path control signal includes: The set of spatial coordinates after affine transformation is converted into a sequence of points representing the motion trajectory of the cutting head. The motion trajectory point sequence is smoothed by acceleration to generate an impact-free interpolation path; The spatiotemporal correspondence between the synchronous duty cycle timing sequence and the interpolation path is used to generate a multi-axis collaborative control instruction set.

[0012] Preferably, the method further includes: The temperature field distribution in the cutting molten pool area is collected in real time using an infrared thermal imager. Extract the isotherm gradient variation features in the temperature field distribution to identify regions with abnormal cut widths; The local parameters of the duty cycle timing sequence are adjusted based on the location feedback of the abnormal kerf width area.

[0013] Preferably, the identification of abnormal kerf width areas includes: Construct a temperature gradient direction histogram to detect regions of discontinuous abrupt changes in isotherms; The convolutional neural network is used to classify the slit morphology of the abrupt region and output the type of overheating or incomplete melting defect. Defect types are mapped to power compensation amounts, and incremental corrections are performed on local duty cycle time series.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This optimized laser tube cutting machine processing method for thin tubes effectively improves many problems in the processing of thin tubes through multi-stage collaborative optimization. By acquiring the three-dimensional point cloud data of the tube to be processed and performing geometric topology analysis, it can comprehensively capture the actual geometric characteristics of the tube, including subtle changes in axial direction and wall thickness. This avoids the incomplete information problems caused by traditional reliance on preset parameters or local measurements, making the processing process more closely resemble the actual state of the tube. A dynamic machining coordinate system is established based on the axis direction vector, overcoming the limitations of traditional fixed coordinate systems that cannot adapt to pipe geometric deviations. The dynamic coordinate system can adjust in real time according to the actual axis of the pipe, ensuring that the calculation of the theoretical projection position of the laser focus is always based on the actual shape of the pipe, reducing positioning errors caused by pipe bending and skew, and laying the foundation for subsequent precise cutting. By matching material absorption characteristic parameters with pipe wall thickness distribution data and generating laser power density gradient adjustment commands, dynamic adaptation of laser energy is achieved. Different wall thickness regions have different laser energy requirements. Through this dynamic adjustment, thin-walled regions can obtain moderate energy to avoid burn-through, while thick-walled regions can obtain sufficient energy to ensure thorough cutting. This improves the unstable cutting quality under fixed power and enhances the consistency of the cut. Real-time monitoring of radial runout during pipe rotation and generation of focal trajectory compensation parameters enables timely responses to dynamic deviations during rotation. Radial runout during the rotation of small pipes can cause cutting trajectory deviations. Real-time compensation ensures that the laser focus always tracks the ideal cutting path, reducing wavy errors caused by runout and resulting in a smoother and more regular cutting trajectory. By integrating laser power density gradient adjustment commands with focal trajectory compensation parameters to generate cutting path control signals, coordinated energy adjustment and trajectory correction are achieved. This fusion mechanism ensures the synchronization of laser power changes and focal position adjustments, avoiding the coordination deficiencies that may result from single parameter optimization. This makes the entire processing process more coordinated and stable, improving the processing effect of small tubular parts. Attached Figure Description

[0015] Figure 1 This is a timing diagram of an optimized laser tube cutting machine processing method for small tubes as described in this invention. Figure 2 A flowchart for acquiring 3D point cloud data; Figure 3 A flowchart for real-time monitoring of radial runout; Figure 4 This is a flowchart for matching the absorption characteristic parameters of materials. Detailed Implementation

[0016] The technical solutions of 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.

[0017] Please see Figure 1 This invention provides an optimized method for laser tube cutting machine processing of thin tubes, the method comprising: By acquiring 3D point cloud data of the pipe to be processed, the pipe's axial direction vector and wall thickness distribution data are extracted based on geometric topology analysis. A dynamic processing coordinate system is established based on the axial direction vector to calculate the theoretical projection position of the laser focus within the dynamic processing coordinate system. After matching the wall thickness distribution data with material absorption characteristic parameters, a laser power density gradient adjustment command is generated. The radial runout during pipe rotation is acquired through real-time monitoring and combined with the theoretical projection position to generate laser focus trajectory compensation parameters. Finally, the laser power density gradient adjustment command and focus trajectory compensation parameters are fused to generate a cutting path control signal, driving the laser cutting head to complete high-precision pipe processing.

[0018] Example 1: See Figure 2 The acquisition of 3D point cloud data utilizes a structured light scanning device, consisting of a high-precision projection module and an industrial camera. The projection module projects an coded grating pattern onto the pipe surface, while the industrial camera simultaneously acquires deformed grating images of the pipe surface from multiple angles. The coded grating is generated using Gray code combined with a phase-shifting method. The absolute phase value is calculated using a phase unwrapping algorithm, converting the 2D image coordinates into 3D point cloud coordinates. During multi-view scanning, the pipe is fixed to a rotating platform, and scanning is triggered after each fixed-angle rotation, ensuring full circumferential coverage of the pipe surface. Ambient light interference is suppressed using a narrow-band filter during scanning, and the projected light intensity is adaptively adjusted according to the pipe surface reflectivity to avoid overexposure or underexposure leading to missing point cloud data.

[0019] The alignment of multi-view point cloud data employs a non-rigid registration algorithm, which is an improvement upon the iterative nearest point principle and incorporates curvature feature constraint matching. During initial registration, curvature extrema points of each viewpoint are extracted as keypoints, and a correspondence between keypoints is established using a fast point feature histogram descriptor. After coarse registration, a bidirectional distance metric is used to optimize the transformation matrix, and point pair weights are dynamically adjusted based on local curvature similarity. In the fine registration stage, a thin-plate spline deformation model is incorporated to compensate for nonlinear displacements caused by pipe clamping deformation. Registration errors are evaluated using reprojection errors, and outliers exceeding a threshold are removed using a random sampling consensus algorithm, ultimately achieving spatial alignment of multi-view point clouds with sub-pixel accuracy.

[0020] The registered point cloud undergoes Gaussian curvature filtering to calculate the maximum and minimum principal curvature of each point, constructing a curvature anisotropy index. Curvature filtering is implemented using a sliding window, with the window size adaptively selected based on the pipe diameter, filtering out noise points with significant curvature abrupt changes. The retained point cloud data is input into a Poisson surface reconstruction algorithm to generate a closed, continuous topological mesh. During mesh generation, the octree depth is automatically optimized based on the point cloud density, and surface normal consistency is ensured through global optimization. The reconstructed mesh model undergoes non-manifold edge detection and repair to eliminate self-overlapping and void defects, ultimately outputting a triangular mesh surface with good watertightness.

[0021] The axis extraction based on a continuous topological mesh employs a cross-sectional analysis method. A series of parallel planes are generated along a preset direction and their intersection with the mesh are determined to obtain the cross-sectional profile of the pipe. Each cross-sectional profile is fitted using a least-squares ellipse, eliminating abnormal ellipses caused by local deformation. The sequence of ellipse center points is smoothed using Savitzky-Golay filtering and then fitted with a cubic B-spline curve to generate the spatial axis. Curvature constraints are introduced during axis fitting to avoid axis oscillations caused by cross-sectional eccentricity. Fitting errors are statistically evaluated using radial deviation, and out-of-tolerance sections trigger local resampling and fitting. The final generated spatial axis satisfies second-order continuous differentiability, providing a benchmark for establishing a dynamic machining coordinate system.

[0022] The structured light scanning device is calibrated using a multi-layer checkerboard target, and the intrinsic and extrinsic parameters of the camera and projector are solved using the beam adjustment method. Online temperature compensation is performed before scanning, dynamically correcting the lens distortion coefficient based on equipment temperature changes. During point cloud acquisition, point cloud density and coverage are monitored in real time, with automatic rescanning triggered for missing areas. The convergence criterion for non-rigid registration comprehensively considers registration error and deformation energy to avoid overfitting leading to local distortion. The threshold of Gaussian curvature filtering is dynamically adjusted based on the pipe surface roughness, preserving effective features while suppressing high-frequency noise. The gradient weights of Poisson reconstruction are correlated with the point cloud normal accuracy, ensuring that the mesh surface closely matches the actual geometry.

[0023] The calculation of the minimum bounding ellipse is accelerated using the convex hull algorithm. First, the vertices of the convex hull of the cross-sectional profile are extracted, and then an iterative minimum area ellipse fitting is applied. The ellipse fitting uses an improved Khachiyan algorithm, incorporating a regularization term to enhance numerical stability under ill-conditioned conditions. The window length for axis smoothing is matched to the pipe bending radius, preserving macroscopic bending characteristics while filtering out high-frequency fluctuations. The establishment of the dynamic machining coordinate system relies on the continuity and consistency of the axis tangent vectors; the Frenet frame theory ensures a smooth transition between the coordinate systems of each node. The entire data processing workflow adopts a parallel computing architecture. The point cloud preprocessing, registration, reconstruction, and axis extraction modules are executed collaboratively in a pipelined manner to improve computational efficiency.

[0024] The detection of pipe clamping status is integrated into the scanning process. By comparing the deviation distribution between the theoretical model and the measured point cloud, the clamping deformation area is identified. An alarm is triggered when deformation exceeds the limit, prompting a readjustment of the clamping posture. The multi-view scanning perspective planning is optimized based on visibility analysis to ensure complete coverage of complex structural areas. The initial transformation matrix for point cloud registration can be initialized using robotic arm pose data, reducing the number of iterations. Topology simplification after mesh reconstruction employs an edge-folding algorithm, reducing computational load while maintaining geometric accuracy. The node spacing for axis fitting is adjustable according to subsequent processing accuracy requirements, and sampling points are automatically densified in high-curvature areas.

[0025] The 3D point cloud data is stored using a hierarchical octree structure, supporting rapid spatial queries and local updates. Quality logs record key parameters at each stage of the scanning process for subsequent process traceability and analysis. The deformation field from non-rigid registration can be exported as a displacement map to aid in assessing the initial stress distribution of the pipe. The point cloud after Gaussian curvature filtering supports interactive editing, allowing manual correction of automatically handled outliers. The subdivision level of the continuous topological mesh matches the density of the original point cloud, avoiding redundant calculations due to over-subdivision. The direction of cross-sectional analysis can be set as the initial axis estimation direction, forming an iterative optimization closed-loop processing flow.

[0026] The exposure time for structured light scanning is set in stages according to the reflectivity of the material, and polarization filtering is used to suppress specular reflection in highly reflective areas. Feature extraction for multi-view registration supports manual annotation assistance, addressing special cases where automatic matching fails. Gradient field calculations for Poisson reconstruction are accelerated using GPUs, significantly improving the processing speed of large-scale point clouds. The number of B-spline control points for axis fitting is automatically optimized using the Akaike Information Criterion, balancing fitting accuracy and model complexity. The entire system achieves flexible configuration through modular design, and the algorithms and parameters of each processing stage can be specifically adjusted for different pipe types.

[0027] The coordinates of point cloud data are unified using a fixture reference sphere. Before scanning, the center coordinates of the reference sphere are collected to establish a global coordinate system. The non-rigid registration deformation model supports anisotropic stiffness settings to adapt to the deformation characteristics of different materials. Curvature analysis of the mesh surface employs discrete differential geometric operators to accurately extract geometric feature lines. The fitting quality of the cross-sectional ellipse is evaluated through residual analysis; low-quality fitting triggers local point cloud resampling. The boundary conditions for axis smoothing are set to natural boundaries to avoid abrupt curvature changes at endpoints. Intermediate results of the data processing workflow support visual review, facilitating interactive verification of the output quality at each stage by process engineers.

[0028] Example 2: See Figure 3The discretization of the pipe's spatial axis begins with the parametric representation of the axis curve. The spatial axis is uniformly divided into a sequence of nodes with fixed spacing using an arc length parameter. The node spacing is dynamically set according to the pipe diameter and processing accuracy requirements. A local orthogonal coordinate system is established at each node location, and the coordinate directions are determined using Frenet frame theory: the tangent vector serves as the Z-axis, the principal normal vector as the X-axis, and the secondary normal vector, generated through a cross product, serves as the Y-axis. The origin of the local coordinate system is fixed at the axis node location, and a rotation matrix is ​​used to achieve a smooth transition in direction between the node coordinate systems, avoiding coordinate jumps.

[0029] The projection of the preset cutting path into the local orthogonal coordinate system employs orthogonal projection transformation. First, the cutting path is discretized into a set of spatial points, and the perpendicular coordinates from each point to the current node section are calculated. The perpendicular point is parametrically represented in the XY plane based on the local coordinate system, and its XY coordinate components are directly mapped to the projected coordinates. The theoretical projection position of the laser focus is generated by superimposing the projected coordinates with a defocusing amount, which is pre-calibrated based on the characteristics of the focusing lens. When mapping the normal distance, the focus position is converted into a distance value relative to the normal direction of the pipe surface through coordinate transformation. This distance value includes the combined effects of the lens focal length and the material's refractive index.

[0030] The pipe wall thickness distribution data is used for normal distance correction. The thickness data is converted into a radial thickness distribution map in the local coordinate system, and the laser propagation path inside the pipe wall is calculated according to Fresnel's law. The refractive compensation model considers the dispersion characteristics of the material's refractive index with wavelength, and establishes corresponding optical path compensation parameters for different laser wavelengths. The theoretical spatial coordinate set generated after compensation contains both three-dimensional spatial position and normal projection angle information, forming the original trajectory reference for the laser focus motion.

[0031] Monitoring of the radial runout of the rotating pipe is achieved using a high-speed linear array sensor. The sensor employs the principle of laser triangulation, configuring paired laser emitters and linear CCD receivers, with three measurement stations spaced apart along the pipe's axial direction. Each station acquires the contour timing data of the rotating pipe at a sampling frequency of 1 kHz, with the sampling timing strictly synchronized with the rotary encoder pulses. The raw contour data undergoes median filtering preprocessing to eliminate random electrical noise interference. After spatiotemporal synchronization, the data from multiple stations forms the spatial motion trajectory of the pipe surface.

[0032] Fourier envelope analysis is used to process time-series contour data. A short-time Fourier transform is performed on the radial displacement sequences of each axial section, and the baseband frequency is selected as the rotating fundamental frequency component. The amplitude envelope curve of the fundamental frequency component is obtained through Hilbert transform demodulation, and the phase information is locked by the synchronous encoder signal. Random disturbance components are separated by band-stop filtering, with the stopband range covering the fundamental frequency and its harmonic frequencies. The fundamental frequency vibration component is mapped to a global rigid displacement, and the displacement vector includes radial and tangential components. The frequency domain energy distribution of the random disturbance component is analyzed, and it is decoupled into elastic vibration modes of different orders using a modal parameter identification algorithm.

[0033] The rigid offset is transformed in the dynamic machining coordinate system using homogeneous coordinate transformation. The rigid displacement at the current moment is represented by a pose transformation matrix, which consists of a displacement vector and a rotation matrix. The rotation matrix is ​​calculated based on the actual curvature direction of the axis at the measurement point, maintaining consistency with the spatial orientation of the theoretical trajectory. The elastic deformation is processed by interpolation to establish a continuous deformation field between discrete nodes. The deformation gradient field is constructed based on thin-shell theory to calculate the instantaneous position correction of the current laser projection point. Finally, the theoretical spatial coordinates are superimposed with rigid offset compensation in the position dimension and incorporate angular deviation correction caused by elastic deformation in the direction dimension, generating a compensated coordinate set with four-dimensional spatial and temporal information.

[0034] The sensor calibration process includes geometric calibration and temperature compensation. Geometric calibration is performed using a standard cylinder with equally spaced annular markings on the surface of the calibration piece. By comparing the systematic deviations between the measured and theoretical profiles, an error mapping table for the sensor's spatial position is established. The temperature compensation module monitors the equipment's operating temperature in real time and automatically corrects for sensor installation position offsets based on the coefficient of thermal expansion. The data synchronization system uses the PTP precision time protocol to ensure time alignment accuracy of data acquisition from different stations at the microsecond level.

[0035] The offset component processing algorithm classifies time-varying signals. The fundamental frequency vibration component is phase-locked through a tracking filter, outputting the real-time trajectory of the dynamic rotation center. Higher-order vibration components employ a random subspace identification method to extract modal parameters, and modal coordinates are estimated using a Kalman filter. The offset prediction model incorporates time series analysis, establishing an ARMA prediction model based on historical data to compensate for the inherent delay response of the control system. When abnormal vibration modes are detected, the system automatically increases the sampling frequency and expands the analysis bandwidth to ensure the accuracy of transient deformation capture.

[0036] The generation and storage of compensation parameters employ a time-domain processing mechanism. The time base is divided into discrete time intervals at the microsecond level, with a corresponding coordinate compensation value for each time interval. The compensation parameters express spatial attitude changes in quaternion form, while simultaneously recording the translational components in the XYZ directions. The parameter table is stored in a circular buffer and retrieved in real time based on the processing progress. When the pipe rotation speed changes beyond a set threshold, dynamic recalculation of the parameter table is triggered to adapt to the nonlinear deformation characteristics during acceleration and deceleration.

[0037] The interpolation model for elastic deformation is based on physical constraints. Between adjacent measurement stations, displacement field interpolation is constructed using finite element shape functions. The shape functions are selected considering the pipe material properties and cross-sectional geometry, satisfying equilibrium equations and boundary continuity conditions. For complex bending regions, denser virtual nodes are set, and the node density gradient is determined by the strain energy density distribution. The displacement of each laser projection point is solved using the natural adjacency interpolation method, with its weight determined by the Vinograph partitioning to ensure the rigid body motion invariance of the interpolation results.

[0038] Offset compensation is achieved through coordinate transformation. The spatial pose of the dynamic machining coordinate system at the current moment is described by a 4×4 transformation matrix. This matrix consists of rotation blocks and translation vectors, where the rotation blocks are determined by the real-time orientation of the dynamic coordinate system, and the translation vectors originate from the rigid displacement compensation. The theoretical coordinate points are transformed using homogeneous coordinates, first by applying rigid displacement, and then by superimposing local positional offsets from elastic deformation. The compensated set of coordinate points forms the actual motion trajectory command source of the laser focus in physical space. The entire data processing flow adopts a pipelined parallel architecture, with modules such as raw signal acquisition, feature extraction, offset calculation, and coordinate compensation working collaboratively in real time.

[0039] Example 3: See Figure 4 The transformation from rigid offset to the dynamic machining coordinate system employs transformation matrix operations, where the translation transformation matrix is ​​constructed based on the displacement vector acquired at the current moment. Let the rigid displacement vector detected by the measurement system be... ,in Indicates the displacement component in the X-axis direction. Indicates the displacement component in the Y-axis direction. Let Z represent the displacement component along the Z-axis, then the corresponding translation transformation matrix is... It can be represented as:

[0040] When this matrix is ​​applied to the theoretical coordinate system, it achieves rigid displacement compensation of the entire coordinate system. The handling of elastic deformation is more complex, requiring gradient analysis based on the curvature variation characteristics of the pipe surface. The principal curvature value at the current measurement point is calculated using differential geometry methods, establishing a mapping relationship between curvature variation and local deformation. The curvature gradient field is calculated using the central difference method, establishing a curvature rate tensor in the neighborhood of the measurement point. The eigenvalues ​​of this tensor are decomposed to obtain the direction of maximum deformation and its intensity.

[0041] The generation of the thermal deformation correction matrix is ​​based on the material's thermodynamic parameters and real-time temperature field distribution. During the pipe's rotation, an infrared thermometer continuously collects surface temperature data to construct a three-dimensional temperature field model. The relationship between temperature gradient and thermal strain is established using the material's thermal expansion coefficient matrix, which includes the thermal expansion characteristics in the axial, circumferential, and radial directions. The calculation of the correction matrix considers the time-delay effect of heat conduction, using the heat conduction equation to predict the temperature field and converting the predicted temperature distribution into an equivalent deformation field.

[0042] The database of spectral reflectance of pipe materials was constructed using spectrophotometric measurement methods. Spectral reflectance measurements were performed on typical pipe samples, covering the operating wavelength range of commonly used industrial lasers. Measurement data were stored categorized by material grade, surface treatment state, and thickness, and a multi-dimensional parametric query interface was established. The mapping relationship between reflectance and laser wavelength was achieved through an interpolation algorithm; for any input wavelength, the corresponding reflectance value and its characteristic curve as a function of the incident angle can be output.

[0043] The inversion of pipe wall thickness distribution data into material density distribution employs an iterative optimization algorithm. Based on the principle of X-ray transmission, thickness measurement data contains information on the material's absorption of photons of specific energies. Deconvolution operations are used to separate the influencing factors of thickness and density. The calculation of the equivalent absorption depth coefficient introduces the concept of effective penetration depth, considering the multiple reflections and scattering effects of the laser within the material. This coefficient, together with the incident laser power, determines the energy deposition distribution within the material, thus affecting the molten pool formation characteristics during the cutting process.

[0044] The generation process of the power attenuation scaling factor integrates multiple source parameters. The equivalent absorption depth coefficient and reflectivity mapping relationship are combined through a weighted superposition method, with the weighting coefficients dynamically adjusted according to the material type. For high-reflectivity materials, the influence of reflection characteristics on power attenuation is given greater weight; for high-absorption materials, the adjustment effect of penetration depth is emphasized. The scaling factor is spatially discretely distributed along the cutting path, and the power adjustment parameter corresponding to each discrete point forms a power density distribution curve that continuously varies along the cutting path.

[0045] A real-time update mechanism for the dynamic machining coordinate system ensures compensation accuracy. The execution frequency of coordinate transformation is strictly synchronized with the working cycle of the laser control system, completing the entire process of coordinate reading, matrix operation, and position correction within each control cycle. The interpolation algorithm for the transformation matrix adopts quaternion spherical linear interpolation to avoid the gimbal lock problem that may occur with Euler angle interpolation. For high-speed rotating pipes, additional deformation caused by Coriolis forces is also considered, and an inertial force compensation term under the rotating coordinate system is introduced into the transformation matrix.

[0046] Adaptive adjustment of material parameters enhances system robustness. When a deviation between the actual cutting effect and the expected result is detected, a parameter self-learning mechanism is activated. By comparing the theoretical molten pool morphology with the molten pool characteristics acquired by the actual infrared thermal imager, the equivalent absorption depth coefficient and reflectivity correction factor are optimized in reverse. The optimization process employs the gradient descent method, gradually adjusting parameters while ensuring processing continuity until the cutting quality reaches a stable state. The learning results are updated to the material database, accumulating empirical data for subsequent processing.

[0047] Dynamic compensation of the laser focus trajectory achieves multi-physics coupling. The compensation system simultaneously considers rigid displacement caused by mechanical vibration, elastic deformation caused by thermal deformation, and power adjustment requirements resulting from changes in material properties. All compensation quantities are coordinated and unified in the spatiotemporal domain, achieving parallel computation and synchronous execution through a distributed control system. The compensation parameters are verified using a motion capture system, simulating actual working conditions in a non-processing state to verify the spatial accuracy and dynamic response characteristics of the compensated trajectory.

[0048] The system's operational status monitoring and safety protection mechanisms comprehensively cover all aspects. Real-time monitoring includes key indicators such as coordinate transformation calculation load, compensation parameter update delay, and sensor signal quality. An abnormal state triggers a tiered alarm mechanism, forming multi-level protection from automatic parameter adjustment to emergency shutdown. The data recording system completely saves all compensation parameters and status variables during the processing, supporting post-process analysis and process optimization. The system maintenance module periodically executes self-test procedures to calibrate sensor zero points and standardize transformation parameters, maintaining long-term operational stability.

[0049] The dynamic optimization algorithm for the processing path runs continuously in the background. Based on real-time collected processing status data, the path optimization algorithm analyzes the balance between cutting efficiency and quality. The optimization objectives include multiple factors such as cutting speed, cut quality, and equipment wear, and a Pareto optimal solution is sought through a multi-objective optimization algorithm. Optimization suggestions are presented to the operator in a non-interrupted manner, and after confirmation, the system smoothly transitions to the new parameters. The entire system adopts a modular architecture design, with each functional module interacting through standardized interfaces, supporting flexible configuration and expansion according to specific processing needs.

[0050] Environmental compensation mechanisms enhance system adaptability. To address environmental disturbances such as workshop temperature fluctuations and foundation vibrations, the system integrates data from multiple environmental sensors. By establishing a correlation model between environmental parameters and processing errors, feedforward compensation control is achieved. Optical path drift caused by changes in air temperature and humidity is corrected in real time using active optical elements, and mechanical waves transmitted to the equipment from foundation vibrations are detected and compensated for by an inertial measurement unit. The environmental compensation module works in conjunction with the core compensation system to ensure the stability of processing accuracy under complex operating conditions.

[0051] The human-machine interface provides comprehensive process visualization capabilities. 3D virtual reality technology displays the pipe processing status in real time, including the theoretical path, compensated trajectory, and material removal process. Operators can view processing details of any cross-section and adjust the viewing angle and display parameters through the interactive device. Alarm information and system status are displayed with hierarchical color coding, supporting rapid location of anomaly sources. The historical processing data playback and analysis function helps process engineers study the formation process of specific phenomena, providing intuitive evidence for continuous improvement.

[0052] Example 4: Taking the precision grooving of a medical stainless steel thin tube (2mm outer diameter, 0.2mm wall thickness) as an example, the implementation process of power density gradient adjustment and cutting path control is described in detail. The tube needs to be machined with axial array grooves with a width of 0.1mm and a spacing of 0.5mm. The material is 316L stainless steel, and the laser used is a fiber laser source with adjustable pulse width.

[0053] The power attenuation ratio factor is generated by the material property analysis module, and its parameter settings are shown in Table 1.

[0054] Table 1: Parameter settings table for power attenuation ratio factor.

[0055] Axial position (mm) Pipe wall thickness (μm) Reflectance (%) Absorption coefficient (mm) Power Proportion Factor 0.0 205 62.3 8.7 1.00 2.5 198 61.8 8.9 1.02 5.0 210 63.1 8.5 0.98 7.5 193 60.5 9.2 1.05 10.0 207 62.7 8.6 1.01 The axial power distribution curve is generated based on tabular data. First, cubic spline interpolation is performed on discrete points to form a continuously varying curve. Curve smoothing uses a moving average filter, with the window width synchronized to the pipe rotation cycle. The edge effect compensation module detects curvature changes along the cutting path and automatically adds power compensation values ​​at the corner positions of the groove. The compensation amount is set in stages according to the sharpness of the corner. For the right-angle corner in the example, the compensation coefficient is set to 1.3 times the baseline value, and the transition region uses a sinusoidal gradient.

[0056] The duty cycle timing sequence conversion is achieved through a laser modulator. The power distribution curve value is mapped to a pulse width modulation signal, and a single rotation cycle is divided into 512 equally divided time slots. The duty cycle value corresponding to each time slot is output through a 12-bit DAC to control the Q-switching drive circuit of the laser. The update rate of the timing sequence is synchronized with the tube rotary encoder signal, triggering the data buffer switch at the zero position of each revolution. The rise and fall times of the modulation pulse are strictly controlled within 50ns to avoid optical pulse distortion during power transitions.

[0057] The processing of the motion trajectory point sequence begins with the compensated set of spatial coordinates. The coordinate data is first converted to a unified unit, transforming millimeter-level coordinate values ​​into the number of position command pulses for the servo motor. Interpolation preprocessing detects the distance difference between adjacent points; if the difference exceeds the maximum allowable step size, an intermediate point is automatically inserted. The acceleration smoothing algorithm uses an S-curve acceleration / deceleration model, setting different jerk parameters based on the inertia characteristics of each axis. For the small pipe in the example, the X / Y axis jerk is set to 5000 mm / s³, the Z-axis to 3000 mm / s³, and the rotation axis to 8000 rad / s³.

[0058] The generation of the multi-axis collaborative control command set follows a specific protocol format. Each control cycle (1ms) sends a data packet containing five-axis position commands. The data packet structure includes a frame header checksum, axis address code, absolute position value, and tracking error tolerance. The position commands use incremental encoding; the difference between the position of the previous cycle and the current cycle is rounded and converted into an integer pulse number. The coupling relationship between the rotary and linear axes is set through electronic gear ratio parameters, ensuring a strict match between the axial feed rate and the groove spacing every 360° rotation of the pipe.

[0059] The synchronous control system relies on a real-time Ethernet communication network. The spatiotemporal correspondence between the duty cycle timing sequence and the interpolation path is ensured through a timestamp mechanism; the main controller broadcasts a global time synchronization signal at the beginning of each millisecond period. Upon receiving the time synchronization signal, the laser modulator starts its internal high-precision timer, triggering pulse output at a specified time. The motion controller simultaneously receives the synchronization signal and triggers position command updates via hardware interrupts. Network transmission delay is eliminated through a time compensation algorithm; the compensation value is automatically measured and stored during system initialization.

[0060] The system's monitoring interface displays the real-time status of key parameters during operation. The power adjustment module displays the power ratio factor, actual output power, and deviation value of the set power at the current operating position. The motion control module displays the actual position, following error, and load rate curves of each axis. The synchronization status monitoring interface uses concentric circles to represent the spatiotemporal alignment; a yellow warning is triggered when the concentricity deviation exceeds 5%, and a red alarm is triggered and processing is suspended when it exceeds 10%.

[0061] The anomaly handling mechanism includes a multi-level recovery strategy. When an abnormal power output is detected, the system first attempts to reload the duty cycle sequence data; if the anomaly persists, it switches to the backup modulation channel. When the motion axis experiences an excessive following error, the feed rate is automatically reduced to 80% of the set value; if the error still cannot be eliminated, an emergency stop is executed and the current position is saved. All abnormal events generate detailed logs, recording a snapshot of the system state at the time of the anomaly and the contents of the historical data buffer for the most recent 10ms.

[0062] The specific execution process of the processing case demonstrates typical parameter interactions. After the operator inputs the pipe parameters (outer diameter 2mm, wall thickness 0.2mm) and processing requirements (groove width 0.1mm, depth 0.15mm) into the interface, the system automatically generates an initial process plan. The power allocation module recommends a base power of 80W based on the material database, with an edge compensation coefficient set to 1.3. The motion parameter module calculates the optimal rotational speed as 300rpm and the axial feed rate as 0.25mm / s. After the operator confirms the parameters, processing begins, and the system sequentially performs preparatory steps such as reference point reset, safety space check, and auxiliary gas pre-activation.

[0063] The real-time adjustment function allows operators to intervene in the operation process. During machining, the power reference value can be modified by ±10% via handwheel fine-tuning, and the adjustment effect is immediately reflected in the actual machining. The axial feed speed can be dynamically adjusted within the range of 50%-150% of the initial value, and the system automatically adjusts the rotational speed synchronously to maintain a constant groove spacing. All adjustment operations are recorded in the process log, including information such as adjustment time, modified parameters, and operator ID.

[0064] The quality control module evaluates the processing effect through multi-sensor data fusion. A coaxial vision system captures the groove edge morphology and measures the deviation between the actual groove width and the set value. An acoustic emission sensor monitors the vibration spectrum characteristics during the cutting process, identifying potential incomplete cuts or overheating. A distance sensor scans the outer diameter of the processed pipe to detect possible thermal deformation. All detection data is stored in association with process parameters, forming a complete processing quality traceability chain.

[0065] The equipment maintenance subsystem tracks the lifespan of key components. Laser operating hours and cumulative output energy are displayed in real time, with early warnings issued when maintenance cycles are approaching. Wear conditions of the motion axis lead screw are assessed through current ripple analysis, and abnormal wear patterns are automatically calibrated for focused inspection. The degree of contamination of optical lenses is determined by transmittance monitoring; cleaning and maintenance are recommended when transmittance drops to 90% of its initial value. Maintenance records are correlated with processing data for analysis, optimizing preventative maintenance plans.

[0066] The system architecture is designed to support modular expansion. The power regulation module is compatible with various laser types and achieves device independence through a driver adapter layer. The motion control module supports up to 8-axis linkage configuration, with independent parameter settings for each axis. The sensor interface provides a standardized data access protocol for easy integration of new detection devices. All functional modules communicate via a message bus, and key data is shared using a publish / subscribe model to reduce system coupling.

[0067] Example 5: The infrared thermal imager configuration adopts a coaxial integrated optical path design. The thermal imager is coupled to the laser optical path through a dichroic mirror, with the center of the field of view aligned with the laser's point of action, achieving a spatial resolution of no less than 640×512 pixels. The imaging wavelength range is selected in the 3.9-4.1 μm mid-wave infrared band, which has the best response characteristics to the thermal radiation of the molten pool. The sampling frequency is fixed at 200 Hz, and the exposure time for each frame is dynamically adjusted, automatically controlled within the range of 50-500 μs based on the brightness of the molten pool. The thermal imager incorporates a Stirling cooler to maintain the detector at an operating temperature of 77 K, reducing the impact of thermal noise. The optical system is equipped with an automatic cleaning device that periodically blows protective gas to prevent smoke and dust from contaminating the lenses.

[0068] The temperature field processing workflow includes geometric calibration and radiometric correction. Upon system startup, thermal reference calibration is performed, establishing a grayscale-temperature mapping table within the 800–1500℃ range using a standard blackbody radiation source. During processing, lens contamination is monitored in real-time, and radiometric coefficient compensation is automatically triggered based on temperature deviations in the background area. Geometric distortion of the temperature field is corrected using a nonlinear mapping method, with a pre-stored distortion correction parameter table for 128 feature points. Molten pool region segmentation employs an adaptive threshold algorithm, determining the boundary of the action area based on temperature gradient changes.

[0069] A multi-level analysis method is employed for isotherm gradient feature extraction. After Gaussian smoothing preprocessing, the temperature gradient components are calculated along the X and Y directions. The gradient magnitudes are synthesized using the Euclidean norm to form a gradient mode field image. The isotherm tracking algorithm is based on connected component analysis, starting from a set threshold temperature point and connecting isotherms along the gradient direction. The measured distances between adjacent isotherms constitute a gradient change feature sequence; when the distance fluctuation exceeds a set threshold, it is marked as an anomaly candidate region.

[0070] The temperature gradient direction histogram is constructed using an eight-part quantization strategy. Within each anomaly candidate region, the 0–360° directional space is divided into eight sectors with 45° intervals. The gradient direction angles of all pixels within each sector are calculated, and the sum of the amplitudes within each sector is statistically analyzed. The peak-to-valley ratio of the histogram is used as an indicator of discontinuity; a ratio of major to minor peaks below 1.5 is considered a valid abrupt change. The distribution width of the bimodal structure is calculated using the full width at half maximum (FWHM) method, with narrow peaks corresponding to abrupt changes in the thermal boundary.

[0071] The convolutional neural network model employs a residual network architecture. The input data consists of 128×128 pixel temperature field slices, containing three channels: raw temperature data, gradient magnitude image, and orientation histogram feature map. The network output layer contains four nodes, representing the probabilities of four categories: normal, overheated, incompletely melted, and edge defect. Training data is derived from labeled samples of historical processing cases, with sample augmentation using mirror rotation and local elastic deformation techniques. During online inference, a sliding ensemble strategy is employed, with weighted voting of classification results from five consecutive frames to determine the final type.

[0072] The mapping from defect type to power compensation is based on an expert rule base. Overburning defects correspond to positive power compensation, with the initial compensation amount set at 5% of the current power, while simultaneously triggering a 15% increase in auxiliary gas flow. Incomplete melting defects employ a negative compensation mechanism, reducing power by 8% and accompanied by a positive adjustment of the decoking amount. The compensation operation incorporates a gradual change mechanism, with the actual adjustment amount transitioning to the target value exponentially, and the transition time set to 100ms. The boundary defect handling strategy includes special rules; when the heat-affected zone at the groove edge is detected to exceed the limit, both power and cutting speed are adjusted simultaneously.

[0073] The incremental correction of the duty cycle timing sequence employs an online editing mode. Within the power modulator's circular buffer, the timing data segment corresponding to the region to be corrected is located. The compensation operation modifies the pulse width value in memory without interrupting the current output waveform. The correction range covers the defect location and the preceding and following buffer areas, with a trapezoidal gradient contour set for the transition region. The correction operation implements a double-buffering mechanism; the editing area is switched to the active state only after background preparation is complete, ensuring waveform continuity.

[0074] The closed-loop feedback system enables dynamic adjustment of the circuit. Thermal imager data is transmitted to the analysis module with a processing delay controlled within 8ms. The total response time from defect identification to power compensation taking effect does not exceed 30ms. The system incorporates a compensation effectiveness detection mechanism, monitoring temperature trends in abnormal areas after correction. If no improvement is observed for three consecutive frames, parameter learning is triggered, increasing the compensation amount by 20% increments until the expected effect is achieved.

[0075] The system's self-protection function monitors multiple abnormal states. When the thermal image saturation exceeds 90%, the gain is automatically reduced, and an event log is recorded. When the temperature gradient in the molten pool region exceeds a safety threshold, power limiting mode is activated regardless of the defect status. Network communication latency exceeding 5ms triggers synchronization recalibration, resetting the controller clock reference. The equipment's cumulative operating status is displayed through health indicators, comprehensively considering parameters such as optical system cleanliness, detector noise level, and cooling efficiency.

[0076] The process database stores a complete history of adjustments. Each compensation operation saves the timestamp, location coordinates, defect type, original power, correction magnitude, and effect evaluation data. Operators can trace the adjustment process of a specific processing section and observe the correspondence between temperature field change sequences and adjustment actions. The database supports feature search functionality, allowing users to find reference solutions for similar historical cases through similarity matching.

[0077] System integration testing covers multiple scenario verification projects. Typical tests include a rapid start-up molten pool state tracking test: a full-power laser is applied instantaneously to cold pipes, requiring the system to establish a stable temperature field analysis capability within 300ms. A sudden change response test simulates an abnormal heat accumulation process, artificially creating localized hot spots to verify the system's identification and compensation timeliness. A long-term stability test executes continuous 24-hour processing, monitoring the drift of key system parameters and triggering automatic calibration to maintain stable accuracy.

[0078] The human-computer interface provides visual decision support. The 3D temperature field is rendered in real-time using pseudo-color, overlaid with isotherms and gradient vectors. Defect areas are marked with flashing outlines, accompanied by type identifiers and confidence percentages. The power correction process is visualized, displaying the editing status of the buffer, with transition areas represented by a semi-transparent overlay effect. Historical adjustment curves are displayed alongside current data, supporting rapid comparative analysis.

[0079] The system maintenance function incorporates a periodic self-test program. Upon first startup each day, it automatically performs an optical window transmittance test and verifies temperature measurement accuracy using a standard blackbody radiation source. A forced neural network calibration program runs weekly, using a built-in test sample set to verify classification accuracy. Monthly optical path collimation calibration is recommended, using a crosshair calibration target to confirm the alignment of the imaging center with the laser focus. All maintenance records are compiled into an electronic log, and maintenance schedule reminders are automatically generated based on equipment usage time.

[0080] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0081] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method of laser tube machining optimization for small tube classes, characterized by, The application relates to a laser cutting method and device for pipe machining. The application comprises the following steps: acquiring three-dimensional point cloud data of a pipe to be machined, and extracting a pipe axis direction vector and pipe wall thickness distribution data through geometric topology analysis; establishing a dynamic machining coordinate system based on the pipe axis direction vector, and calculating a theoretical projection position of a laser focal point in the dynamic machining coordinate system; matching material absorption characteristic parameters according to the pipe wall thickness distribution data, and generating laser power density gradient adjustment instructions; real-time monitoring of a radial runout amount in the pipe rotation process, and combining the theoretical projection position to generate laser focal point trajectory compensation parameters; 2. The method of laser tube machining optimization for tube-like structures according to claim 1, characterized in that, fusing the laser power density gradient adjustment instructions and the focal point trajectory compensation parameters to generate cutting path control signals, and driving a laser cutting head to complete pipe machining. The acquisition of the three-dimensional point cloud data of the pipe to be machined comprises the following steps: collecting pipe surface point cloud by using a structured light scanning device, and aligning multi-view point cloud data through a non-rigid registration algorithm; performing Gaussian curvature filtering on the registered point cloud, and reconstructing a continuous topology grid of the pipe surface after removing noise points; 3. A method of laser tube machining optimization for tube classes according to claim 2, characterized in that, calculating a minimum circumscribed ellipse of a cross section based on the continuous topology grid, and fitting a pipe space axis with an elliptic center point sequence. The calculation of the theoretical projection position of the laser focal point in the dynamic machining coordinate system comprises the following steps: discretizing the pipe space axis into equidistant nodes, and establishing a local orthogonal coordinate system at each node; mapping the normal distance of the laser focal point relative to the pipe surface according to the projection coordinates of the preset cutting path in the local orthogonal coordinate system; 4. The method of laser tube machining optimization for tube-like structures according to claim 3, characterized in that, superimposing the pipe wall thickness distribution data to perform material refraction compensation on the normal distance, and generating a theoretical spatial coordinate set of the laser focal point. The real-time monitoring of the radial runout amount in the pipe rotation process comprises the following steps: arranging a high-speed linear array sensor in the axial direction of the pipe to collect pipe surface profile time sequence change data; performing Fourier envelope analysis on the time sequence change data, and extracting a fundamental frequency vibration component and a random disturbance component; 5. A method of laser tube machining optimization for tube-like structures according to claim 4, characterized in that, mapping the fundamental frequency vibration component into a rigid offset amount of the dynamic machining coordinate system, and mapping the random disturbance component into an elastic deformation amount. The generation of the laser focal point trajectory compensation parameters comprises the following steps: converting the rigid offset amount into a translation transformation matrix of the dynamic machining coordinate system; calculating a pipe surface curvature change gradient according to the elastic deformation amount, and generating a thermal deformation correction matrix; 6. The method of laser tube machining optimization for tube-like structures according to claim 1, characterized in that, combining the translation transformation matrix and the thermal deformation correction matrix to perform affine transformation on the theoretical spatial coordinate set. The matching of the material absorption characteristic parameters comprises the following steps: constructing a pipe material spectral reflectivity database, and establishing a mapping relationship between reflectivity and laser wavelength; inverting material density distribution according to the pipe wall thickness distribution data, and calculating an equivalent absorption depth coefficient; 7. A method of laser tube machining optimization for tube classes according to claim 6, characterized in that, fusing the equivalent absorption depth coefficient and the mapping relationship to generate power attenuation proportion factors at different cutting positions. The generation of the laser power density gradient adjustment instructions comprises the following steps: establishing an axial power distribution curve according to the power attenuation proportion factors; superimposing a cutting path curvature radius to perform edge effect compensation on the power distribution curve; 8. The method of laser tube machining optimization for tube-like structures according to claim 5, characterized in that, converting the compensated power distribution curve into a duty cycle time sequence sequence of laser modulated pulses. The generation of the cutting path control signals comprises the following steps: converting the affine transformed spatial coordinate set into a cutting head motion trajectory point sequence; performing acceleration smoothing processing on the motion trajectory point sequence to generate a non-impact interpolation path; The space-time correspondence between the synchronization duty cycle time sequence and the interpolation path is generated to generate a multi-axis cooperative control instruction set.

9. A method of laser tube machining optimization for tube classes according to claim 8, characterized in that, Also include: Real-time acquisition of cutting molten pool area temperature field distribution by infrared thermal imager; Extract the isotherm gradient change characteristics in the temperature field distribution to identify the abnormal area of the cutting seam width; According to the position feedback of the abnormal area of the cutting seam width, adjust the local parameters of the duty cycle time sequence.

10. The method of laser tube machining optimization for tube-like structures according to claim 9, characterized in that, The identification of the abnormal area of the cutting seam width includes: Construct a temperature gradient direction histogram to detect the non-continuous mutation area of the isotherm; Classify the cutting seam morphology of the mutation area through a convolutional neural network to output the overburning or incomplete penetration defect type; Map the defect type to the power compensation amount to incrementally correct the local duty cycle time sequence.

Images