Laser cutting method for cutting complex shapes

Abstract

The invention discloses a laser cutting method for cutting complex shapes, and particularly relates to the field of laser cutting. A three-dimensional curved surface model of a to-be-cut workpiece is obtained, and a surface normal tensor field is constructed; extracting a driving point set based on the curvature derivative and generating a sliding path network; constructing a laser power density mapping model in combination with the thermal diffusion response model and the thermophysical parameters of the material; thermal feedback images are collected in the cutting process, an energy offset function is constructed, laser focus displacement and the scanning rate are dynamically adjusted through fuzzy control, and closed-loop control over power, paths and thermal feedback is achieved; according to the method, self-adaptive adjustment of laser power distribution, curved surface adaptation of a cutting path and thermal field real-time compensation in the machining process can be achieved under a complex geometric structure, the cutting precision, stability and energy efficiency are effectively improved, and the method is suitable for laser machining application of high-requirement complex structures such as aerospace, mold manufacturing and medical instruments.

Description

Technical Field

[0001] The present invention relates to the technical field of laser cutting, in particular to a laser cutting method for cutting complex shapes. Background Art

[0002] Traditional laser cutting methods generally rely on static path presets and fixed power control when processing workpieces with complex boundary shapes or multi-curvature continuous surfaces. This type of method often uses a linear expansion path planning strategy, combined with constant or step-by-step adjustment of laser parameters, and is difficult to dynamically respond to changes in the micro-geometry of the workpiece and the thermal feedback characteristics of the material. When processing hollow thin-walled structures, anisotropic curved surfaces, and heat-sensitive materials, problems such as cutting path drift, rough cross-sections, or thermal melt collapse often occur. In addition, existing methods generally lack a reverse coupling perception mechanism for the cutting environment, which makes it difficult to form closed-loop control of the laser processing process. The cutting effect relies on human experience adjustment, and the process stability is poor.

[0003] Some studies have introduced methods such as machine vision or posture compensation to enhance path control accuracy. However, the path model is generally disconnected from the laser energy field control, preventing a collaborative feedback chain from forming. This limits the system's ability to adaptively process complex spatial structures. Therefore, a laser cutting method with a bidirectional path-energy coupling control mechanism that can drive dynamic deformation of the laser field through curvature changes is urgently needed to improve the processing accuracy and process intelligence of complex structures. Summary of the Invention

[0004] The purpose of the present invention is to provide a laser cutting method for cutting complex shapes to solve the shortcomings of the background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a laser cutting method for cutting complex shapes, comprising: Obtain a three-dimensional surface model of the workpiece to be cut and construct a surface normal tensor field; Extracting a differential curvature driving point set based on the normal tensor field and defining a laser path seed point array; Perform tensor projection transformation on the seed point array to generate an adaptive surface sliding path network; Constructing a laser field parameter mapping model coupled with the path network, and calculating the corresponding laser power density distribution according to the instantaneous curvature derivative of each sliding path point; Start the laser processing controller to perform laser cutting operations along the sliding path network and collect thermal feedback images in real time during the process; An energy distribution offset function is constructed based on the thermal feedback image, and is reversely coupled with a laser power density mapping model to dynamically adjust the laser focus displacement and scanning rate.

[0006] Preferably, constructing the surface normal tensor field comprises: Use structured light 3D reconstruction method to obtain high-precision point cloud data of the workpiece to be cut, and remove abnormal boundary points; Based on the point cloud local tangent plane fitting algorithm, the normal vector and the main direction of curvature at each sampling point are calculated; Encode each normal vector and its local curvature information into a symmetric second-order tensor and construct a tensor mapping matrix; Gaussian filtering regularization is performed on the tensor mapping matrix to form a normal tensor field.

[0007] Preferably, defining the laser path seed point array comprises: Based on the normal tensor field, a continuous differential curvature function field is constructed on the workpiece surface to calculate the gradient change rate of each point in the main curvature direction; Perform multi-scale decomposition on the curvature gradient field and extract the set of main direction curvature mutation points as candidate path excitation points; Using Gaussian curvature contour lines to generate a guided flow field on the workpiece surface, and projecting the excitation point onto a region with dense contour lines; A stable seed point array is constructed in the projection area according to the local normal distortion rate as the initial control point set of the laser path.

[0008] Preferably, generating an adaptive surface sliding path network includes: Extract the principal direction eigenvalues and eigenvector groups of the corresponding normal tensor at the seed point to form the principal axis basis set of the local surface; Taking each seed point as the starting point, a directed cutting trajectory segment is constructed along the main direction of the tensor, and its continuous deformation trend in the curvature tensor field is recorded; The path connection weight is established using the offset angle between the main directions of the local tensor, and the sliding path is generated by connecting adjacent seed points based on the minimum energy deformation criterion; All trajectory segments are spliced together to form a sliding path network covering the target area, and the path adaptively adjusts its direction and density under the guidance of the main direction of the tensor.

[0009] Preferably, constructing a laser field parameter mapping model coupled with the path network includes: Collect the local principal curvature value K and its first-order derivative in the path direction of each path point in the sliding path network , forming a continuous curvature derivative field; K represents the principal curvature value at a discrete point on the sliding path, that is, the curvature in the main direction, dK represents the slight change in the curvature value, that is, the curvature difference between adjacent path points. ds represents the slight displacement or path spacing along the sliding path direction, Represents the curvature derivative value per unit path length; The curvature derivative value is introduced as an input parameter into the thermal diffusion response model ; where α is the basic power coefficient, β∈[1.2,1.8] is the nonlinear response factor; Based on the material thermal conductivity κ, melting point and reflectivity R, construct the material property coupling function ;in, is the ambient temperature, which is used as the power correction coefficient to adjust the initial thermal response power value; The final calculated laser power density value is bound to the path point to generate a set of laser power density mapping data that automatically adjusts as the path point moves. The data is input into the laser field parameter mapping model, and the expression is: Where, Indicates a point The laser power density at .

[0010] Preferably, dynamically adjusting the laser focus displacement and scanning rate includes: The thermal feedback image is mapped into a temperature field, and the position (xc, yc) of the maximum temperature point Tmax in each frame image is extracted as the actual thermal center; Compare the path points corresponding to the theoretical laser energy center (xp, yp) and construct a two-dimensional space offset vector ; At the same time record its module length ; and calculate its rate of change through sliding window difference ; Modulate the offset vector and rate of change Input fuzzy controller; calculate laser focus displacement correction based on empirical rule set , and scan speed correction ; Specifically including: Laser focus displacement correction value: ; Scanning speed correction value: ; represents the fuzzy control function for laser focus displacement correction, A fuzzy control function representing the scanning speed; The output results are defuzzified and sent to the laser processing controller in real time to perform the following compensation control: ; ; Indicates the initial reference focus position. Indicates the position of the laser focus at the current moment; Indicates the initial set reference scanning speed; Indicates the laser scanning speed at the current moment; The corrected laser focus displacement and scanning speed are written back to the laser control system in real time and executed in conjunction with the power density mapping model.

[0011] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. This invention overcomes the limitations of traditional regular path planning on complex three-dimensional surfaces by introducing a path generation mechanism that combines a normal tensor field with curvature derivatives, enabling sliding, adaptive laser cutting path construction. Furthermore, by coupling a thermal diffusion response model with material properties to control power density, laser energy can be precisely distributed across complex geometric areas, effectively avoiding overburning, undercutting, and energy waste, significantly improving cutting quality and efficiency.

[0012] 2. This invention, for the first time, incorporates thermal feedback images into cutting process control. By constructing an energy distribution offset function and driving fuzzy adjustment of focus position and scanning speed, closed-loop control of the cutting process is achieved. This mechanism enhances the system's dynamic response to local workpiece deformation, heat accumulation, and path errors, improving the stability, precision, and intelligence of the overall machining system. It is particularly suitable for high-precision machining of high-curvature, hollow, thin-walled, and heat-sensitive materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0014] Figure 1 This is a mind map of the method of the present invention. DETAILED DESCRIPTION

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0016] For examples, see Figure 1 As shown, the laser cutting method for complex shape cutting described in this embodiment includes: Obtain a three-dimensional surface model of the workpiece to be cut and construct a surface normal tensor field; Extracting a differential curvature driving point set based on the normal tensor field and defining a laser path seed point array; Perform tensor projection transformation on the seed point array to generate an adaptive surface sliding path network; Constructing a laser field parameter mapping model coupled with the path network, and calculating the corresponding laser power density distribution according to the instantaneous curvature derivative of each sliding path point; Start the laser processing controller to perform laser cutting operations along the sliding path network and collect thermal feedback images in real time during the process; An energy distribution offset function is constructed based on the thermal feedback image, and is reversely coupled with a laser power density mapping model to dynamically adjust the laser focus displacement and scanning rate.

[0017] In this embodiment, to achieve high-precision laser cutting of complex structure workpieces, it is first necessary to perform 3D modeling of the surface of the workpiece to be cut and construct a continuous and smooth surface normal tensor field. The specific steps are as follows: A structured light 3D scanner (such as the Gocator 3506) scans the workpiece from multiple angles to obtain complete point cloud data of the workpiece's exterior surface. During the scanning process, a multi-frame stitching and fusion algorithm is used to align the data from multiple viewpoints. The point cloud resolution is set to 0.1 mm to ensure that even subtle changes in curvature are captured.

[0018] To eliminate scanning errors, the Statistical Outlier Removal (SOR) algorithm is used to clean up the noise of the original point cloud. Valid points with a distance from the center within twice the standard deviation are retained to improve the quality of the point cloud data.

[0019] For the cleaned point cloud, a local tangent plane least squares fitting algorithm is used to construct a spherical neighborhood with a radius of r (r = 3 mm) centered at each point. Covariance analysis is performed on the neighborhood points to determine the normal vector corresponding to the minimum eigenvalue, which is used as the unit normal direction for that point.

[0020] At the same time, the maximum and minimum principal curvature directions (principal axis directions) are extracted as directional components of the constructed tensor. This normal-principal curvature combination reflects the morphological trend of the workpiece surface at the microscopic scale.

[0021] The unit normal vector and the principal curvature direction of each point 、 Encoded as a symmetric second-order tensor ,in 、 is the corresponding curvature coefficient, symbol The tensor set of all points constitutes the tensor mapping matrix T(x,y,z), where (x,y,z) represents the three-dimensional coordinate position of a point in the workpiece surface space; this matrix describes the normal distribution and deformation trend of the entire surface space.

[0022] To avoid normal field jumps caused by scanning errors, a Gaussian filter regularization operation is applied to the tensor matrix. The convolution kernel radius is set to 2 mm, and all tensor components are weighted averaged to smooth the normal field changes between points.

[0023] The regularized tensor field meets the continuity requirements and serves as the basic input data of the laser sliding path network to guide the direction adjustment and energy distribution control of the subsequent cutting path.

[0024] In this embodiment, in order to achieve adaptive planning of the laser path for complex curved surfaces, it is necessary to extract a set of points with processing guidance from the surface normal tensor field constructed above as the initial control seed points of the laser cutting path. The specific method is as follows: Based on the regularized normal tensor field, the differential curvature function field on the workpiece surface is constructed by calculating local derivatives along the principal directions of the tensor. The principal curvature direction at each point is determined by the eigenvectors of the tensor, and the curvature derivative is calculated from the rate of change of the tensor field's neighborhood. Finite differences are used to approximate the gradient change along the principal curvature direction, forming a continuous curvature rate map.

[0025] To enhance responsiveness to complex geometries, the curvature rate of change map was subjected to multi-scale Gaussian filtering (σ = 1mm, 2mm, and 5mm) to extract curvature mutation points at different spatial frequencies. Using the Laplace-of-Gaussian (LoG) operator response, a set of stable points that maintain high responsiveness at multiple scales was identified. These points are defined as principal direction curvature mutation points, representing regions of drastic local geometric change.

[0026] Using Gaussian curvature as a scalar function, a continuous contour network is constructed on the workpiece surface. Ridge-Valley tracing is used to track the distribution of curvature contours along the principal curvature directions. The aforementioned discontinuity points are projected onto areas of dense contour density to enhance machining guidance and geometric resolution at these points. This step ensures that the selected points are concentrated in areas with dense features rather than isolated local extreme points.

[0027] To avoid path distortion caused by minor curvature perturbations, the normal distortion rate of candidate points within the projection area is analyzed. A "distortion rate response factor" is constructed using the derivative of the normal vector along the principal curvature direction. This filter eliminates areas of high variability and retains only points with stable normal field variations and continuous curvature variations as laser path seed points. Ultimately, an array of seed points covering the entire processing area is constructed, serving as the starting point input for the subsequent sliding path generation algorithm.

[0028] In this embodiment, to achieve a laser sliding cutting path with directional consistency and structural matching on complex curved surfaces, a tensor projection transformation is performed based on the extracted seed point array and normal tensor field to generate an adaptive sliding path network covering the target area. The specific steps are as follows: For each seed point , extract its corresponding normal tensor By tensor eigendecomposition, the maximum eigenvalue λ1 and the corresponding eigenvector corresponding to the main direction are obtained. This eigenvector represents the most significant deformation direction of the local surface and serves as the initial guiding direction for the sliding path at that point. The eigenvector is normalized and stored in the path initial direction array.

[0029] Each seed point is the starting point of the path, along the characteristic vector The trajectory segment is constructed by extending the fixed step length s = 1.5 mm. The extension point is recorded as . In the new point Repeat tensor feature extraction at each segment, update the guidance direction, and continue to build the path until it reaches a boundary or a direction jump occurs (angle greater than 45°), at which point expansion stops. Each segment of the path retains the tensor's main direction, local curvature value, and connection angle information for subsequent weight calculation.

[0030] For each pair of adjacent seed points, the local tensor main direction offset angle θ is calculated, and the connection weight function is calculated. ,in =0.2 is the directional sensitivity coefficient. By minimizing the directional deformation energy in the entire path network ,The path segments with the minimum connection cost are selected for splicing to construct a sliding path with good global continuity and smooth direction.

[0031] The above sliding trajectory generation process is performed across the entire target area, and all trajectories are merged to form a preliminary sliding path network. To avoid redundant cutting caused by overly dense paths, the paths are thinned based on the "inverse curvature density principle": paths are dense in high-curvature areas and sparse in low-curvature areas. The final output is a sliding path network with consistent direction, self-adjustable density, and adaptive curvature.

[0032] To achieve intelligent control of power density during laser cutting and improve processing consistency and cross-section quality in complex geometric structures, this embodiment provides a method for constructing a laser field parameter mapping model coupled with a sliding path network, specifically comprising the following steps: For each discrete path point in the sliding path network , first extract its principal curvature value , which is obtained by calculating the curvature characteristics of the main direction of the tensor field mentioned above. At the same time, using the orderliness of the path, the first-order derivative is approximately calculated by the finite difference method: Where Δs represents the spacing between path points (set to 1 mm), and K represents the principal curvature at a discrete point on the sliding path. This refers to the curvature of the point in the principal direction, reflecting the degree of curvature of the workpiece surface at that point. dK represents the small change in curvature, or the difference in curvature between adjacent path points. ds represents the small displacement along the sliding path, or path spacing, which is the length interval between discrete path points. Represents the curvature derivative value per unit path length; this formula is used to construct a continuous curvature derivative field, which reflects the geometric change rate of the path in space.

[0033] The curvature derivative value at each path point As the driving variable of the laser energy response, it is input into the following nonlinear thermal diffusion power model: ;in: represents the initial thermal response power density (unit: W / mm²); α = 0.6 represents the base power coefficient, set empirically based on the laser type (e.g., pulsed fiber laser); β∈[1.2,1.8] represents the nonlinear response exponent, reflecting the increased power sensitivity in regions of high curvature variation (default is 1.5). This model automatically increases the laser power density in regions of high curvature variation to compensate for possible insufficient thermal diffusion and path deviation.

[0034] In order to further match the thermophysical properties of different materials, a material property correction coefficient is introduced based on the initial power density. The following material coupling function is constructed: ; Where: κ is the thermal conductivity of the material (W / m·K), is the melting point temperature of the material (K), R is the laser reflectivity of the material (0–1), is the ambient temperature (default 298K).

[0035] Laser power density correction value It is given by: .

[0036] The final corrected power density value With waypoints One-to-one binding generates path position power density pairs. This mapping is used as the input of the laser control module to achieve continuous adjustment of laser power along the path points, completing the coupled modeling of the laser parameter field and the path field. The final laser field parameter mapping model expression is: Where, Indicates a point The laser power density at .

[0037] In this embodiment, to ensure that the laser cutting operation can be accurately executed along the previously generated sliding path network and to effectively monitor and adjust heat accumulation, ablation anomalies, or energy offsets that occur during the actual processing process, a synchronous laser control and infrared thermal feedback image acquisition system are used for closed-loop processing. The specific steps are as follows:

[0038] It uses an industrial-grade laser cutting system (such as the IPG YLS-2000 fiber laser), paired with a six-axis CNC platform and an adaptive optical head. The control system is implemented by a combination of a PLC master control and a motion control card.

[0039] Import the above sliding path network and corresponding power density map into the controller: The cutting path is converted into NC interpolation instructions (such as G code or custom intermediate layer path commands) in vector form; Each path segment corresponds to a dynamic power adjustment instruction, which couples the laser power output with the curvature derivative and material parameters; The processing head is equipped with an electric focusing module, which supports automatic adjustment of focal length and working distance according to the position of the path point.

[0040] After initialization, the system starts the laser emission module and automatically cuts along the sliding path network. The laser parameters are adjusted in real time with the path position, including: The laser power output is adjusted according to the mapping model; Focus depth is automatically compensated according to the path normal; Cutting speed can be linked with power (when power increases, speed decreases relatively to ensure thermal coupling time); After each path segment is cut, a 0.2s delay buffer is performed to avoid continuous superposition of heat accumulation.

[0041] To monitor the thermal field during the machining process, a high-sensitivity infrared imager (such as the FLIR A655sc, with a thermal resolution of <0.05K and a frame rate of 30Hz) is installed next to the cutting head. It is arranged at an angle to maintain a 20° viewing angle with the laser path to avoid laser interference.

[0042] The image acquisition module captures real-time thermal distribution images of the cutting area, with an image size of 640 × 480 pixels and a spatial accuracy of approximately 0.2 mm / pixel. The images are stored in a time series format, with each image labeled with the path point corresponding to the time ti for subsequent closed-loop energy regulation.

[0043] During the image acquisition process, the system has a built-in thermal anomaly warning module: If a certain area continuously has high-temperature patches higher than the set threshold Tmax = 850°C for more than 100ms, it will be marked as a heat accumulation area; If the thermal distribution of a path segment in the image is discontinuous and there is a thermal hole at the boundary (temperature difference greater than 50°C), it is judged as insufficient cutting; Abnormal information is transmitted back to the main controller in real time, triggering laser power adjustment or path local rescanning commands.

[0044] In order to further improve the thermal control accuracy and energy utilization efficiency in complex morphology areas during laser cutting, this embodiment introduces an energy distribution offset modeling method based on infrared thermal feedback images, and realizes dynamic linkage adjustment of focus depth and scanning speed through fuzzy control strategy.

[0045] An infrared thermal imager (FLIR A655sc) is used to monitor the laser cutting process in real time at a frame rate of 30Hz. The system integrates an image processing module and uses the grayscale-temperature mapping relationship: Where: I(x,y,t): thermal image pixel grayscale value; T(x,y,t): equivalent temperature value (unit: °C); a=0.25, b=20: linear transformation coefficients obtained through system calibration. In each image frame, the current actual thermal center is identified by searching for the maximum temperature value Tmax and its coordinate location (xc,yc).

[0046] Compare the thermal center (xc, yc) with the path theory laser energy action center (xp, yp) to construct a two-dimensional offset vector: ; At the same time record its module length ; and calculate its rate of change through sliding window difference ;The above two variables and As the two input variables of the fuzzy controller, they are divided into three linguistic variable levels (such as: small / medium / large, decreasing / stable / increasing), and the following empirical rule set is constructed: If the offset is large and continues to increase, "significant focus down" and "significant deceleration" are output; If the offset is small and stable, output "maintain current focus" and "normal speed"; If the offset is moderate and the rate of change is large, "slight focus adjustment" and "slow deceleration" are output.

[0047] The fuzzy controller uses the above rule set to output two control variables: the laser focus displacement correction value: ; Scanning speed correction value: ; The fuzzy control function representing the correction of laser focus displacement outputs an adjustment value of focus displacement according to the thermal center offset and its changing trend. If the thermal offset is large and increasing, adjust the laser focus slightly lower (more focused). If the thermal offset is small or stable, maintain or slightly adjust the focus. A fuzzy control function represents the scanning speed. It outputs a correction value Δv for the scanning speed, enabling dynamic acceleration and deceleration. If the thermal offset is increasing, it indicates that the laser heating is deviating from the center, and the speed should be reduced. If the offset is decreasing, the speed can be gradually restored or slightly increased.

[0048] The output results are defuzzified (e.g., by the center of gravity method), and the numerical results are sent to the laser processing controller in real time to perform the following compensation control: ; ; Indicates the initial reference focus position. Indicates the position of the laser focus at the current moment; Indicates the initial set reference scanning speed; Indicates the laser scanning speed at the current moment; The controller updates the focus and speed parameters in real time in each cycle (e.g., 33 ms), and executes them in conjunction with the power density mapping model to achieve three-dimensional closed-loop control of the laser parameter field based on thermal response.

[0049] This invention provides a laser cutting method for complex shapes. Based on three-dimensional surface modeling, normal tensor field construction, curvature derivative-driven path generation, and adaptive power density control, this method generates a sliding path network through tensor guidance. Combined with a thermal diffusion response model and material property coupling mechanism, it achieves coordinated control of laser power, focal depth, and scanning speed. An energy distribution offset function is constructed using infrared thermal feedback images, forming a closed-loop adjustment system for laser parameters. This method effectively improves cutting accuracy, energy utilization, and process stability in complex structures with multiple curvatures. It is suitable for demanding manufacturing scenarios such as aviation, precision molds, and medical devices.

[0050] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0051] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0052] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0053] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. Laser cutting method for cutting complex shapes, characterized by: include: Obtain a three-dimensional surface model of the workpiece to be cut and construct a surface normal tensor field; Extracting a differential curvature driving point set based on the normal tensor field and defining a laser path seed point array; Perform tensor projection transformation on the seed point array to generate an adaptive surface sliding path network; Constructing a laser field parameter mapping model coupled with the path network, and calculating the corresponding laser power density distribution according to the instantaneous curvature derivative of each sliding path point; Start the laser processing controller to perform laser cutting operations along the sliding path network and collect thermal feedback images in real time during the process; An energy distribution offset function is constructed based on the thermal feedback image, and is reversely coupled with a laser power density mapping model to dynamically adjust the laser focus displacement and scanning rate.

2. The laser cutting method for complex shape cutting according to claim 1, characterized in that: The constructing of the surface normal tensor field comprises: Use structured light 3D reconstruction method to obtain high-precision point cloud data of the workpiece to be cut, and remove abnormal boundary points; Based on the point cloud local tangent plane fitting algorithm, the normal vector and the main direction of curvature at each sampling point are calculated; Encode each normal vector and its local curvature information into a symmetric second-order tensor and construct a tensor mapping matrix; Gaussian filtering regularization is performed on the tensor mapping matrix to form a normal tensor field.

3. The laser cutting method for complex shape cutting according to claim 1, characterized in that: The definition of the laser path seed point array includes: Based on the normal tensor field, a continuous differential curvature function field is constructed on the workpiece surface to calculate the gradient change rate of each point in the main curvature direction; Perform multi-scale decomposition on the curvature gradient field and extract the set of main direction curvature mutation points as candidate path excitation points; Using Gaussian curvature contour lines to generate a guided flow field on the workpiece surface, and projecting the excitation point onto a region with dense contour lines; A stable seed point array is constructed in the projection area according to the local normal distortion rate as the initial control point set of the laser path.

4. The laser cutting method for complex shape cutting according to claim 1, characterized in that: The generation of adaptive surface sliding path network includes: Extract the principal direction eigenvalues and eigenvector groups of the corresponding normal tensor at the seed point to form the principal axis basis set of the local surface; Taking each seed point as the starting point, a directed cutting trajectory segment is constructed along the main direction of the tensor, and its continuous deformation trend in the curvature tensor field is recorded; The path connection weight is established using the offset angle between the main directions of the local tensor, and the sliding path is generated by connecting adjacent seed points based on the minimum energy deformation criterion; All trajectory segments are spliced together to form a sliding path network covering the target area, and the path adaptively adjusts its direction and density under the guidance of the main direction of the tensor.

5. The laser cutting method for complex shape cutting according to claim 1, characterized in that: The construction of a laser field parameter mapping model coupled with the path network includes: Collect the local principal curvature value K and its first-order derivative in the path direction of each path point in the sliding path network , forming a continuous curvature derivative field; K represents the principal curvature value at a discrete point on the sliding path, that is, the curvature in the main direction, dK represents the slight change in the curvature value, that is, the curvature difference between adjacent path points. ds represents the slight displacement or path spacing along the sliding path direction, Represents the curvature derivative value per unit path length; The curvature derivative value is introduced as an input parameter into the thermal diffusion response model ; where α is the basic power coefficient, β∈[1.2,1.8] is the nonlinear response factor; Based on the material thermal conductivity κ, melting point and reflectivity R, construct the material property coupling function ;in, is the ambient temperature, which is used as the power correction coefficient to adjust the initial thermal response power value; The final calculated laser power density value is bound to the path point to generate a set of laser power density mapping data that automatically adjusts as the path point moves. The data is input into the laser field parameter mapping model, and the expression is: Where, Indicates a point The laser power density at .

6. The laser cutting method for complex shape cutting according to claim 1, characterized in that: The dynamic adjustment of laser focus displacement and scanning rate includes: The thermal feedback image is mapped into a temperature field, and the position (xc, yc) of the maximum temperature point Tmax in each frame image is extracted as the actual thermal center; Compare the path points corresponding to the theoretical laser energy center (xp, yp) and construct a two-dimensional space offset vector ; At the same time record its module length ; and calculate its rate of change through sliding window difference ; Modulate the offset vector and rate of change Input fuzzy controller; calculate laser focus displacement correction based on empirical rule set , and scan speed correction ; Specifically including: Laser focus displacement correction value: ; Scanning speed correction value: ; represents the fuzzy control function for laser focus displacement correction, A fuzzy control function representing the scanning speed; The output results are defuzzified and sent to the laser processing controller in real time to perform the following compensation control: ; ; Indicates the initial reference focus position. Indicates the position of the laser focus at the current moment; Indicates the initial set reference scanning speed; Indicates the laser scanning speed at the current moment; The corrected laser focus displacement and scanning speed are written back to the laser control system in real time and executed in conjunction with the power density mapping model.

Images