Laser pipe cutting machine electric chuck automatic positioning method based on dynamic learning algorithm

Abstract

The invention discloses an automatic positioning method for an electric chuck of a laser pipe cutting machine based on a dynamic learning algorithm. The automatic positioning method comprises the following steps: S1, constructing a pipe shape and mechanical data set; s2, an adaptive non-Euclidean manifold optimization method and a Hodge-Laplacian constraint reconstruction method are adopted, and a clamping topology mapping matrix is generated; s3, mechanical distribution of chuck clamping points is optimized in combination with a graph neural geometric network, and a variable stiffness clamping strategy is formed by using a Riemannian manifold constraint variable stiffness control method; s4, analyzing stress evolution characteristics of each area of the pipe in the clamping process, and calculating potential error influence factors; s5, calculating a nonlinear evolution path of the clamping error of the chuck, and optimizing an automatic compensation strategy of the chuck in combination with a non-Euclidean error coding technology; and S6, dynamically adjusting the clamping torque, the position compensation parameter and the error correction mechanism of the chuck in real time, and generating an automatic positioning scheme of the chuck. Intelligent positioning of the electric chuck of the laser pipe cutting machine is achieved by combining a dynamic learning algorithm, an error compensation technology and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic positioning of laser pipe cutting machines, and particularly to an automatic positioning method for an electric chuck of a laser pipe cutting machine based on a dynamic learning algorithm. Background Art

[0002] In the field of laser pipe cutting, the electric chuck, as a key clamping component, is responsible for the fixation and precise positioning of pipes. Its clamping accuracy directly determines the cutting quality and processing stability. Existing laser pipe cutting machines usually adopt a rigid fixed electric chuck, which clamps the pipe by presetting the distribution of clamping points and a fixed torque to ensure the stability of the pipe position during the cutting process. However, there are multiple problems in the clamping process of traditional electric chucks, including fixed clamping stiffness, inefficient error compensation, insufficient deformation prediction, and poor adaptability. These defects are particularly prominent when processing complex pipes (such as thin-walled pipes, elliptical pipes, polygonal pipes, and variable cross-section pipes).

[0003] In traditional electric chuck positioning methods, the distribution of clamping points is preset and not intelligently optimized according to the geometric shape and mechanical properties of the pipe, resulting in problems such as uneven clamping, stress concentration, and excessive deformation when clamping special-shaped pipes. Especially for thin-walled materials or high-strength alloy pipes, the fixed clamping torque may cause local overstress, leading to plastic deformation or clamping slip, thus affecting the cutting accuracy. Most of the rigid chucks in the prior art are designed based on a fixed mechanical model and lack the ability to adaptively adjust according to different materials and shapes, resulting in unreasonable stiffness adjustment during the clamping process and affecting the cutting accuracy and processing stability. In addition, during the clamping process, traditional methods fail to fully consider the local deformation and overall stress evolution of the pipe and cannot dynamically adjust the clamping stiffness, resulting in cumulative errors due to local deformation during the cutting process and ultimately affecting the overall processing accuracy.

[0004] The insufficient error compensation ability is also one of the main challenges faced by traditional electric chucks. Existing error compensation methods mainly rely on linear correction models or simple force feedback control, and cannot fully consider the non-linear deformation and stress distribution changes of the pipe during the clamping process, resulting in limited error compensation effects. In addition, existing error compensation algorithms are usually based on the linear assumption of Euclidean space and are difficult to handle the error propagation of complex-shaped pipes in a non-uniform mechanical environment, making the compensation strategy ineffective in practical applications. Especially in high-precision processing scenarios, the cumulative effect of errors will cause the cutting deviation to gradually increase, affecting the final processing quality.

[0005] The adaptability of existing electric chucks is relatively poor. When processing pipes of different shapes and materials, manual adjustment of clamping parameters is required, resulting in a decrease in processing efficiency. At the same time, the lack of intelligent clamping optimization strategies makes the adaptability to different pipes during the cutting process relatively poor. Due to significant differences in the stiffness, deformation characteristics, and force response of pipes, a single fixed clamping strategy is difficult to meet the diverse processing requirements. Existing technologies are difficult to optimize the distribution of clamping points in real time through data-driven or intelligent calculation methods, making it difficult to ensure processing accuracy and stability during the cutting process. In addition, traditional methods lack an effective adaptive adjustment mechanism in complex mechanical environments and cannot adjust the chuck positioning scheme according to real-time clamping torque, error feedback, and pipe shape, resulting in uneven stress during the processing and affecting the final cutting effect.

[0006] To address the above problems, some studies have used intelligent optimization algorithms to improve the clamping method of electric chucks, such as the clamping point distribution adjustment method based on topology optimization, the clamping force optimization strategy based on force feedback, and the error prediction model based on finite element analysis. However, these methods still have limitations in practical applications. First, the topology optimization method is usually based on static modeling and lacks the ability of dynamic adjustment, unable to respond to the force changes during the processing in real time. Second, the clamping optimization method based on force feedback is limited by the sensor accuracy and response speed and is difficult to achieve precise control in a high-dynamic environment. In addition, although the error prediction model can optimize the error compensation strategy to a certain extent, it is limited by the computational complexity and algorithm efficiency and is difficult to meet the high-precision requirements in a real-time processing environment.

[0007] Therefore, how to provide an automatic positioning method for the electric chuck of a laser pipe cutting machine based on a dynamic learning algorithm is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0008] An object of the present invention is to propose an automatic positioning method for the electric chuck of a laser pipe cutting machine based on a dynamic learning algorithm. The present invention combines a dynamic learning algorithm, non-Euclidean optimization, Riemannian flow deformation stiffness control, and multi-modal error compensation technology to construct an intelligent clamping point optimization, stiffness adaptive regulation, and real-time error compensation mechanism for the automatic positioning problem of the electric chuck of a laser pipe cutting machine, ensuring that the chuck can dynamically adjust the clamping strategy according to the pipe shape and force characteristics. Compared with the traditional rigid clamping method, the present invention can accurately optimize the distribution of clamping points, reduce error accumulation, improve the adaptability to special-shaped pipes, effectively improve the positioning accuracy, cutting stability, and intelligent level of the chuck, and meet the high-precision pipe processing requirements.

[0009] The automatic positioning method for the electric chuck of a laser pipe cutting machine based on a dynamic learning algorithm according to an embodiment of the present invention includes the following steps:

[0010] S1. Measure the geometric shape, wall thickness distribution, and internal stress state of the pipe to be processed using multi-frequency laser coherent sensing, and construct a pipe shape and mechanics dataset;

[0011] S2. Based on the pipe shape and mechanics dataset, use the adaptive non-Euclidean manifold optimization method to calculate the optimal clamping point distribution of the chuck, and combine the Hodge-Laplacian constraint reconstruction method to adjust the interaction relationship between the clamping points and generate a clamping topology mapping matrix;

[0012] S3. Based on the clamping topology mapping matrix, combine the graph neural geometric network to optimize the mechanical distribution of the chuck clamping points, and use the Riemannian manifold constraint variable stiffness control method to adjust the clamping stiffness according to the pipe type to form a variable stiffness clamping strategy;

[0013] S4. Based on the variable stiffness clamping strategy, analyze the stress evolution characteristics of each region of the pipe during the clamping process, calculate the potential error influencing factors, and generate an error analysis result;

[0014] S5. Use the multi-modal physical embedding error compensation network to calculate the non-linear evolution path of the chuck clamping error, and combine the non-Euclidean error coding technology to optimize the automatic compensation strategy of the chuck to form an error compensation scheme;

[0015] S6. Integrate the clamping topology mapping matrix, variable stiffness clamping strategy, and error compensation scheme, and dynamically adjust the chuck clamping torque, position compensation parameters, and error correction mechanism in real time to finally generate a chuck automatic positioning scheme.

[0016] Optionally, the S2 specifically includes:

[0017] S21. Establish a clamping point set based on the pipe shape and mechanics dataset, and the clamping point set is expressed as P = {p 夹持1 , p 夹持2 ,..., p 夹持n}, where n represents the total number of clamping points, and map the surface curvature of the pipe to the non-Euclidean manifold A to construct an initial topological model of the clamping area;

[0018] S22. Use the adaptive non-Euclidean manifold optimization method to define an optimization objective function for the clamping point distribution:

[0019]

[0020] Among them, L 优化 represents the optimization objective function, λ 刚度i represents the stiffness adjustment parameter of the i-th clamping point p 夹持i , represents the non-Euclidean gradient operator, μ 相互ij represents the i-th clamping point p 夹持iand the interaction coefficient d between the j-th clamping point p 夹持j ; d(p A (p 夹持i , p 夹持j ) represents the geometric distance between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j ;

[0021] S23. Based on the Hodge-Laplacian constraint reconstruction method, establish the clamping point topological constraint matrix:

[0022]

[0023] where H 拓扑ij represents the topological connection weight between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j ; F(p 夹持i ) represents the pipe shape feature vector at the i-th clamping point p 夹持i ; F(p 夹持j ) represents the pipe shape feature vector at the j-th clamping point p 夹持j ; F(p 夹持k ) represents the pipe shape feature vector at the k-th clamping point p 夹持k ; ∥·∥ A represents the non-Euclidean measure between the clamping point shape feature vectors; exp represents the exponential function;

[0024] S24. Calculate the gradient update amount of each clamping point:

[0025]

[0026] where Δ Hodge p 夹持i represents the gradient update amount of the i-th clamping point p 夹持i ; N(i) represents the neighborhood set of the i-th clamping point p 夹持i ; p 夹持j -p 夹持i represents the relative displacement of the clamping point position;

[0027] S25. Based on the gradient update amount of each clamping point, construct the clamping topology mapping matrix:

[0028]

[0029] where M 映射ij represents the element of the clamping topology mapping matrix; H 拓扑ik represents the topological connection weight between the i-th clamping point p 夹持i and the k-th clamping point p 夹持k ; H 拓扑kjDenote the $k$-th clamping point $p$ 夹持k and the $j$-th clamping point $p$ 夹持j The topological connection weight $\delta$ 映射ij Denote the Kronecker delta matrix $\alpha$ for constraining the topological stability of the clamping points 梯度 Denote the gradient adjustment coefficient;

[0030] S26. The optimization goal is to adjust the positions of the clamping points through gradient update and ensure the stability of the distribution of the clamping points on the non-Euclidean manifold $A$.

[0031] Optionally, the specific steps of S3 are as follows:

[0032] S31. Initialize the mechanical feature matrix $F$ based on the clamping topology mapping matrix and combined with the pipe material category information 0 $=\{F$ 0 $(p$ 夹持1 $),\cdots,F$ 0 $(p$ 夹持n )$\}$;

[0033] S32. Use the graph neural geometry network to perform multi-layer mechanical aggregation on the mechanical feature matrix $F$ 0 to construct the mechanical distribution iterative equation:

[0034] $F(p$ 夹持i )$ = F$ 0 $(p$ 夹持i )$+\sum$ j∈N(i) $H$ 拓扑ij $(F$ 0 $(p$ 夹持j )$-F$ 0 $(p$ 夹持i ));

[0035] where $F(p$ 夹持i ) represents the comprehensive mechanical feature at the $i$-th clamping point $p$ 夹持i , $N(i)$ represents the neighborhood set of the $i$-th clamping point $p$ 夹持i , and $H$ 拓扑ij represents the topological connection weight between the $i$-th clamping point $p$ 夹持i and the $j$-th clamping point $p$ 夹持j ;

[0036] S33. Use the Riemannian manifold constraint variable stiffness control method to construct the variable stiffness regulation functional $\Psi$ 刚度 for describing the optimal distribution of the stiffness of the chuck clamping points:

[0037]

[0038] where $n$ represents the total number of clamping points, $\alpha$ 局部i represents the local stiffness weight, $K$ 刚度iDenote the stiffness value at the $i$-th clamping point $p$ 夹持i ; Denote the gradient operator on the Riemannian manifold $B$, $\beta$ 交互ij Denote the regulation factor, $K$ 刚度j Denote the stiffness value at the $j$-th clamping point $p$ 夹持j ; $\zeta$ 力学i Denote the mechanical influence coefficient, $\vert\vert F(p$ 夹持i )\vert\vert$ represents the comprehensive mechanical characteristic modulus at the $i$-th clamping point $p$ 夹持i ;

[0039] S34. Set the pipe constraint function $\Theta$ for different pipe categories 管材 :

[0040]

[0041] where, $\max$ represents the maximum value function represents the stiffness threshold corresponding to the pipe category represents the pipe category, $\Theta$ 管材 constrain the variable stiffness value not to exceed the material adaptation range, and meet the differential requirements of brittleness and pipes;

[0042] S35. Construct the stiffness iteration equation:

[0043]

[0044] where, represents the stiffness value at the $i$-th clamping point $p$ at the $(t + 1)$-th iteration 夹持i ; represents the stiffness value at the $i$-th clamping point $p$ at the $t$-th iteration 夹持i ; $\delta$ 阈值 represents the threshold penalty coefficient;

[0045] S36. Through multiple rounds of iterative optimization, make the stiffness of the clamping points gradually converge to the optimal distribution on the Riemannian manifold, and form a variable stiffness clamping strategy.

[0046] Optionally, the specific steps of S4 include:

[0047] S41. Based on the variable stiffness clamping strategy, obtain the instantaneous stress state of each clamping point of the pipe during the clamping process, construct a pipe stress distribution model, and characterize the stress conditions at different positions during the clamping process;

[0048] S42. Analyze the change trend of stress over time during the clamping process, calculate the stress transfer path according to the force relationship between the clamping points, and determine the force balance state and stress concentration characteristics of the pipe during the clamping process;

[0049] S43. Based on the stress evolution analysis, calculate the stress gradients in each area of the pipe and identify the areas where the stress gradient changes are greater than the set threshold;

[0050] S44. Evaluate the influence of the variable stiffness adjustment on the stress distribution in combination with the stress gradient change situation, identify the key influencing factors causing errors during the clamping process, and generate an error analysis result.

[0051] Optionally, the specific steps of S5 are as follows:

[0052] S51. Based on the error analysis result, construct a multi-modal physical embedding error compensation network, define the clamping error state set E = {E 夹持1 , E 夹持2 ,..., E 夹持n}, where E 夹持i represents the instantaneous error at the i-th clamping point p 夹持i , and set the error adjustment parameter ξ 误差i in combination with the pipe type, and n represents the total number of clamping points;

[0053] S52. In the error compensation network, introduce a non-linear error evolution path modeling method and define the error evolution equation:

[0054]

[0055] Among them, represents the error value at the i-th clamping point p 夹持i at the (t + 1)-th iteration, represents the error value at the i-th clamping point p 夹持i at the t-th iteration, N(i) represents the domain set of the i-th clamping point p 夹持i , γ 误差传递ij represents the error transfer coefficient, represents the error value at the j-th clamping point p 夹持j at the t-th iteration, Φ 非线性i represents the non-linear correction term of the error, which is used to compensate the dynamic evolution trend of the clamping error;

[0056] S53. Based on the error evolution path, adopt a non-Euclidean error coding technology and define the error compensation optimization objective:

[0057]

[0058] Among them, Ψ 补偿 represents the error compensation optimization objective function, ρ 局部误差i represents the local error adjustment parameter, represents the non-Euclidean gradient operator, E 夹持j represents the instantaneous error at the j-th clamping point p 夹持j τ误差交互ij Represents the error interaction weight;

[0059] S54. Based on the error compensation optimization objective function and combined with the dynamic change characteristics of the chuck clamping error, an error compensation scheme is finally formed.

[0060] Optionally, the S6 specifically includes:

[0061] S61. Based on the clamping topology mapping matrix, extract the spatial distribution characteristics of the chuck clamping points, and combine the variable stiffness clamping strategy and the error compensation scheme to initialize the automatic positioning optimization variables of the chuck, including the clamping point position, clamping stiffness, and error correction parameters;

[0062] S62. Real-time monitor the torque change during the chuck clamping process, analyze the stress state of the clamping points, and adjust the clamping torque of the chuck in combination with the clamping topology mapping matrix to meet the stress requirements of different pipe materials;

[0063] S63. Combine the variable stiffness clamping strategy to dynamically adjust the position compensation parameters of the chuck clamping points, optimize the distribution of the clamping points according to the stress conditions of the pipe materials, and enable the chuck to maintain the optimal positioning state under different pipe material forms;

[0064] S64. Based on the error compensation scheme, adjust the error correction parameters in real time to make the error correction process respond to the changes in the chuck clamping state in real time;

[0065] S65. Based on the clamping torque, position compensation parameters, and error correction mechanism of the chuck, construct an automatic positioning optimization stop criterion:

[0066]

[0067] Where max represents the maximum value function, Represents the position vector of the i-th clamping point p at the (t + 1)-th iteration 夹持i , Represents the position vector of the i-th clamping point p at the t-th iteration 夹持i , ∈ 收敛 Represents the convergence threshold of position optimization, Represents the clamping torque of the i-th clamping point p at the (t + 1)-th iteration 夹持i , Represents the clamping torque of the i-th clamping point p at the t-th iteration 夹持i , δ 稳定 Represents the stability threshold of clamping torque optimization;

[0068] That is, when the maximum change amount of the optimization variable is less than the set threshold, the automatic positioning optimization process terminates;

[0069] S66. Finally, an automatic positioning solution for the chuck is generated, enabling the chuck to achieve real-time dynamic adjustment under different pipe categories and clamping states.

[0070] The beneficial effects of the present invention are as follows:

[0071] Firstly, by introducing a dynamic learning algorithm, the present invention optimizes the automatic positioning process of the electric chuck of the laser pipe cutting machine, overcomes the limitations of the traditional rigid clamping method, and improves the machining accuracy, stability, and adaptability. Compared with the prior art, the chuck clamping solution of the present invention can adapt to the geometric shape and mechanical properties of the pipe, optimizes the distribution of clamping points using the adaptive non-Euclidean manifold optimization method, and combines the Hodge-Laplacian constraint reconstruction method to make the interaction relationship between clamping points more reasonable, constructs a high-precision clamping topology mapping matrix, thereby improving the clamping stability and reducing the deformation error caused by uneven clamping during the cutting process.

[0072] Secondly, the variable stiffness clamping strategy proposed by the present invention uses Riemannian manifold constraints to optimize the stiffness of clamping points under different pipe categories, enabling the stiffness to be dynamically adjusted during the machining process, thereby effectively reducing the stress concentration or clamping slip phenomenon caused by fixed stiffness. Based on the graph neural geometry network, the mechanical distribution of clamping points is optimized to make the clamping process more intelligent, and to ensure that the optimal stiffness control scheme is maintained during the cutting of thin-walled, special-shaped, and high-strength pipes, avoiding plastic deformation caused by excessive clamping, while improving the adaptive ability of clamping stiffness and ensuring the stability of the cutting process.

[0073] Thirdly, in response to the problem of insufficient traditional error compensation, the present invention proposes a non-linear error correction method based on a multi-modal physical embedding error compensation network. By calculating the non-linear evolution path of the clamping error and combining non-Euclidean error coding technology, the error propagation mode can be accurately identified, the error compensation strategy can be optimized, and the compensation scheme can adapt to different pipe categories, shapes, and force characteristics. Compared with the traditional error compensation method based on linear correction, the method of the present invention can correct the clamping error more accurately in a complex mechanical environment, reduce error accumulation, and improve the final cutting accuracy.

[0074] Fourthly, the automatic positioning solution of the chuck of the present invention realizes the real-time optimization of the clamping torque, position compensation parameters, and error correction mechanism of the chuck by integrating the clamping topology mapping matrix, variable stiffness clamping strategy, and error compensation scheme, enabling the chuck to perform dynamic adjustment for different pipe categories, and improving the machining accuracy and efficiency. Description of the Drawings

[0075] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0076] Figure 1 This is the overall flowchart of the automatic positioning method for the electric chuck of a laser pipe cutting machine based on a dynamic learning algorithm proposed by the present invention. Specific embodiments

[0077] Now, the present invention will be further described in detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0078] Reference Figure 1 , the automatic positioning method for the electric chuck of a laser pipe cutting machine based on a dynamic learning algorithm includes the following steps:

[0079] S1. Use multi-frequency laser coherence sensing to measure the geometric shape, wall thickness distribution, and internal stress state of the pipe to be processed, and construct a pipe shape and mechanics data set;

[0080] S2. Based on the pipe shape and mechanics data set, adopt an adaptive non-Euclidean manifold optimization method to calculate the optimal clamping point distribution of the chuck, and combine the Hodge-Laplacian constraint reconstruction method to adjust the interaction relationship between the clamping points to generate a clamping topology mapping matrix;

[0081] S3. Based on the clamping topology mapping matrix, combine the graph neural geometry network to optimize the mechanical distribution of the chuck clamping points, and use the Riemannian manifold constraint variable stiffness control method to adjust the clamping stiffness according to the pipe type to form a variable stiffness clamping strategy;

[0082] S4. Based on the variable stiffness clamping strategy, analyze the stress evolution characteristics of each region of the pipe during the clamping process, calculate the potential error influencing factors, and generate an error analysis result;

[0083] S5. Use a multi-modal physical embedding error compensation network to calculate the non-linear evolution path of the chuck clamping error, and combine the non-Euclidean error coding technology to optimize the automatic compensation strategy of the chuck to form an error compensation scheme;

[0084] S6. Integrate the clamping topology mapping matrix, the variable stiffness clamping strategy, and the error compensation scheme, and dynamically adjust the chuck clamping torque, position compensation parameters, and error correction mechanism in real time, and finally generate a chuck automatic positioning scheme.

[0085] In this embodiment, the specific content of S2 includes:

[0086] S21. Establish a clamping point set based on the pipe shape and mechanics data set, and the clamping point set is expressed as P = {p 夹持1 , p 夹持2 ,..., p 夹持n}, where n represents the total number of clamping points, maps the surface curvature of the pipe to the non-Euclidean manifold A, and constructs the initial topological model of the clamping area;

[0087] S22. Adopt an adaptive non-Euclidean manifold optimization method to define the optimization objective function for the distribution of clamping points:

[0088]

[0089] where L 优化 represents the optimization objective function, λ 刚度i represents the stiffness adjustment parameter of the i-th clamping point p 夹持i . represents the non-Euclidean gradient operator, μ 相互ij represents the interaction coefficient between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j , d A (p 夹持i , p 夹持j ) represents the geometric distance between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j ;

[0090] S23. Based on the Hodge-Laplacian constraint reconstruction method, establish the topological constraint matrix of the clamping points:

[0091]

[0092] where H 拓扑ij represents the topological connection weight between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j , F(p 夹持i ) represents the pipe shape feature vector at the i-th clamping point p 夹持i , F(p 夹持j ) represents the pipe shape feature vector at the j-th clamping point p 夹持j , F(p 夹持k ) represents the pipe shape feature vector at the k-th clamping point p 夹持k , ∥·∥ A represents the non-Euclidean measure between the clamping point shape feature vectors, and exp represents the exponential function;

[0093] S24. Calculate the gradient update amount of each clamping point:

[0094]

[0095] where Δ Hodge p 夹持i represents the i-th clamping point p 夹持iThe gradient update amount, and N(i) represents the i-th clamping point p 夹持i 's neighborhood set, p 夹持j -p 夹持i represents the relative displacement of the clamping point position;

[0096] S25. Based on the gradient update amount of each clamping point, construct a clamping topology mapping matrix:

[0097]

[0098] where M 映射ij represents the element of the clamping topology mapping matrix, and H 拓扑ik represents the i-th clamping point p 夹持i and the k-th clamping point p 夹持k 's topological connection weight, and H 拓扑kj represents the k-th clamping point p 夹持k and the j-th clamping point p 夹持j 's topological connection weight, and δ 映射ij represents the Kronecker delta matrix, which is used to constrain the topological stability of the clamping points, and α 梯度 represents the gradient adjustment coefficient;

[0099] S26. The optimization goal is to adjust the clamping point position through gradient update and ensure the stability of the clamping point distribution on the non-Euclidean manifold A.

[0100] In this embodiment, the specific steps of S3 are as follows:

[0101] S31. Based on the clamping topology mapping matrix and combined with the pipe material category information, initialize the mechanical feature matrix F 0 ={F 0 (p 夹持1 ),…,F 0 (p 夹持n )};

[0102] S32. Use the graph neural geometric network to perform multi-layer mechanical aggregation on the mechanical feature matrix F 0 and construct a mechanical distribution iteration equation:

[0103] F(p 夹持i ) = F 0 (p 夹持i ) + ∑ j∈N(i) H 拓扑ij (F 0 (p 夹持j ) - F 0 (p 夹持i ));

[0104] where F(p 夹持i ) represents the i-th clamping point p夹持i The comprehensive mechanical characteristics at [location], and N(i) represents the i-th clamping point p 夹持i The domain set of, H 拓扑ij Represents the i-th clamping point p 夹持i And the j-th clamping point p 夹持j The topological connection weight between them;

[0105] S33. Using the Riemannian manifold constraint variable stiffness control method, construct the variable stiffness regulation functional Ψ 刚度 , which is used to describe the optimal distribution of the stiffness of the chuck clamping points:

[0106]

[0107] Among them, n represents the total number of clamping points, α 局部i Represents the local stiffness weight, K 刚度i Represents the stiffness value at the i-th clamping point p 夹持i At the location, Represents the gradient operator on the Riemannian manifold B, β 交互ij Represents the regulation factor, K 刚度j Represents the j-th clamping point p 夹持j At the location, ζ 力学i Represents the mechanical influence coefficient, ||F(p 夹持i )|| Represents the comprehensive mechanical characteristic modulus at the i-th clamping point p 夹持i At the location;

[0108] S34. Set the pipe constraint function Θ for different pipe categories 管材 :

[0109]

[0110] Among them, max represents the maximum value function, Represents the stiffness threshold corresponding to the pipe category, Represents the pipe category, Θ 管材 Constrains the variable stiffness value not to exceed the material adaptation range, meeting the differential requirements of brittleness and pipes;

[0111] S35. Construct the stiffness iteration equation:

[0112]

[0113] Among them, Represents the stiffness value at the i-th clamping point p at the (t + 1)-th iteration 夹持i At the location, Represents the stiffness value at the i-th clamping point p at the t-th iteration 夹持i At the location, δ 阈值 Represents the threshold penalty coefficient;

[0114] S36. Through multiple rounds of iterative optimization, the stiffness of the clamping points gradually converges to the optimal distribution on the Riemannian manifold, forming a variable stiffness clamping strategy.

[0115] In this embodiment, the specific steps of S4 are as follows:

[0116] S41. Based on the variable stiffness clamping strategy, obtain the instantaneous stress state of each clamping point of the pipe during the clamping process, construct a pipe stress distribution model to characterize the force conditions at different positions during the clamping process;

[0117] S42. Analyze the change trend of stress over time during the clamping process, calculate the stress transfer path according to the force relationship between the clamping points, and determine the force balance state and stress concentration characteristics of the pipe during the clamping process;

[0118] S43. On the basis of stress evolution analysis, calculate the stress gradient of each region of the pipe and identify the regions where the stress gradient change is greater than the set threshold;

[0119] S44. Evaluate the influence of variable stiffness adjustment on the stress distribution in combination with the stress gradient change situation, identify the key influencing factors causing errors during the clamping process, and generate an error analysis result.

[0120] In this embodiment, the specific steps of S5 are as follows:

[0121] S51. Based on the error analysis result, construct a multi-modal physical embedding error compensation network, define the clamping error state set E = {E 夹持1 , E 夹持2 ,..., E 夹持n}, where E 夹持i represents the instantaneous error at the i-th clamping point p 夹持i , and combine with the pipe type to set the error adjustment parameter ξ 误差i , and n represents the total number of clamping points;

[0122] S52. In the error compensation network, introduce a non-linear error evolution path modeling method and define the error evolution equation:

[0123]

[0124] Where, represents the error value at the i-th clamping point p 夹持i at the (t + 1)-th iteration, represents the error value at the i-th clamping point p 夹持i at the t-th iteration, N(i) represents the domain set of the i-th clamping point p 夹持i , γ 误差传递ij represents the error transfer coefficient, represents the error value at the j-th clamping point p 夹持jThe error value at Φ 非线性i represents the non-linear correction term of the error, which is used to compensate for the dynamic evolution trend of the clamping error;

[0125] S53. Based on the error evolution path, adopt the non-Euclidean error coding technology to define the error compensation optimization objective:

[0126]

[0127] Among them, Ψ 补偿 represents the error compensation optimization objective function, ρ 局部误差i represents the local error adjustment parameter, represents the non-Euclidean gradient operator, E 夹持j represents the instantaneous error at the j-th clamping point p 夹持j at, τ 误差交互ij represents the error interaction weight;

[0128] S54. Based on the error compensation optimization objective function, combined with the dynamic change characteristics of the chuck clamping error, finally form an error compensation scheme.

[0129] In this embodiment, the S6 specifically includes:

[0130] S61. Based on the clamping topology mapping matrix, extract the spatial distribution characteristics of the chuck clamping points, and combine the variable stiffness clamping strategy and the error compensation scheme to initialize the chuck automatic positioning optimization variables, including the clamping point position, clamping stiffness, and error correction parameters;

[0131] S62. Real-time monitor the torque change during the chuck clamping process, analyze the stress state of the clamping points, and adjust the clamping torque of the chuck in combination with the clamping topology mapping matrix to meet the stress requirements of different pipes;

[0132] S63. Combine the variable stiffness clamping strategy to dynamically adjust the position compensation parameters of the chuck clamping points, optimize the distribution of the clamping points according to the pipe stress situation, and make the chuck maintain the optimal positioning state under different pipe shapes;

[0133] S64. Based on the error compensation scheme, adjust the error correction parameters in real time to make the error correction process respond to the change of the chuck clamping state in real time;

[0134] S65. Based on the clamping torque, position compensation parameters, and error correction mechanism of the chuck, construct an automatic positioning optimization stop criterion:

[0135]

[0136] Among them, max represents the maximum value function, represents the position vector of the i-th clamping point p 夹持i at the (t + 1)-th iteration, Denote the position vector of the \(i\)-th clamping point \(p\) at the \(t\)-th iteration, \(\in\) 夹持i 收敛 Denote the convergence threshold for position optimization, Denote the clamping moment of the \(i\)-th clamping point \(p\) at the \((t + 1)\)-th iteration, 夹持i Denote the clamping moment of the \(i\)-th clamping point \(p\) at the \(t\)-th iteration, \(\delta\) 夹持i 稳定 Denote the stability threshold for clamping moment optimization;

[0137] That is, when the maximum change of the optimization variable is less than the set threshold, the automatic positioning optimization process terminates;

[0138] S66. Finally, generate an automatic positioning scheme for the chuck, enabling the chuck to achieve real-time dynamic adjustment under different pipe categories and clamping states.

[0139] Example 1:

[0140] To verify the feasibility of the present invention in implementation, the present invention is applied to the laser pipe cutting workshop of an aviation manufacturing enterprise. This workshop is responsible for processing aircraft fuel pipes, hydraulic pipes, and structural support pipe fittings, with extremely high requirements for the processing accuracy of pipes. Due to the complex pipe materials, including high-strength titanium alloy, stainless steel, and carbon fiber composite materials, and some pipes having elliptical cross-sections, polygonal cross-sections, or even non-uniform wall thicknesses, traditional electric chucks have problems such as fixed clamping stiffness, inefficient error compensation, and insufficient deformation prediction during the clamping process, resulting in a decrease in cutting accuracy and seriously affecting the subsequent assembly quality.

[0141] Traditional electric chucks adopt a fixed clamping moment method. When facing different pipes, it is necessary to manually adjust the clamping parameters. The selection of clamping points and the mechanical distribution lack intelligent optimization, resulting in plastic deformation of some thin-walled pipes due to excessive local stress during the clamping process, thereby affecting the cutting quality. In addition, the stress distribution cannot be analyzed in real time during the clamping process, resulting in small displacements of high-strength pipes due to the accumulation of thermal stress during the cutting process, making the final cutting error exceed the design tolerance range. Especially when the pipe length exceeds 2 meters, the cumulative error can reach 0.8 mm, far exceeding the quality standard of the enterprise for the processing error of fuel pipes not exceeding 0.3 mm. In addition, for non-circular cross-section pipes, there is a high risk of slippage during the clamping process of traditional chucks, resulting in a repeat processing rate as high as 12.5%, seriously affecting production efficiency. Therefore, the intelligent electric chuck automatic positioning method of the present invention is applied to the laser pipe cutting machine in this workshop.

[0142] ​​​First, a multi-frequency laser coherent sensing system is used to scan the geometric shape, wall thickness distribution, and internal stress state of the pipe, and a dataset of pipe shape and mechanics is constructed. This dataset is input into an adaptive non-Euclidean manifold optimization model to calculate the optimal clamping point distribution, and the Hodge-Laplacian constraint method is combined to optimize the interaction relationship between clamping points, finally generating a clamping topology mapping matrix. Compared with the traditional fixed clamping method, this method can adaptively adjust the clamping point distribution according to the pipe shape, increasing the clamping stability by about 27.3%.

[0143] Based on the clamping topology mapping matrix, the present invention further uses a graph neural geometry network to optimize the mechanical distribution of clamping points, and combines a Riemannian manifold constraint variable stiffness control strategy to achieve intelligent stiffness regulation. In actual tests, when cutting a titanium alloy pipe with a thickness of 0.5 mm and a length of 1.5 m, the variable stiffness control strategy of the present invention improves the mechanical balance degree between clamping points by 31.2% and reduces the local deformation amount by 48.7%. Compared with the traditional fixed stiffness clamping method, this method can effectively reduce the incidence of plastic deformation during the processing of high-strength materials, reducing the processing error from the original 0.6 mm to 0.22 mm, meeting the requirements of aviation manufacturing for high-precision pipe fitting processing.

[0144] In addition, the present invention monitors the stress evolution characteristics of clamping points in real time during the processing, calculates potential error influencing factors, and uses a multi-modal physical embedding error compensation network to predict the non-linear evolution path of errors. Compared with the traditional error compensation method, the non-Euclidean error coding technology of the present invention optimizes the error compensation strategy, adjusts the error compensation parameters of the chuck during the dynamic clamping process, and improves the error compensation accuracy by about 36.5%. When processing a batch of stainless steel pipes with complex cross-sections and a length of 2.2 m, this technology reduces the processing error from 0.8 mm to 0.18 mm, and the error compensation accuracy is significantly improved.

[0145] Finally, the present invention synthesizes the clamping topology mapping matrix, variable stiffness clamping strategy, and error compensation scheme, dynamically adjusts the clamping torque, position compensation parameters, and error correction mechanism of the chuck, and generates an automatic positioning scheme. Experimental data shows that this method improves the average clamping stability to 98.6% during the processing, reduces the cutting precision error to 0.15 mm, and reduces the repeated processing rate to 3.2%, which is 74.4% less than the traditional method. In addition, since the position and stiffness of the clamping points can be dynamically adjusted, workers do not need to manually adjust the clamping parameters frequently, and the processing efficiency is increased by 19.8%. Table 1 below summarizes the experimental data, demonstrating the advantages of the present invention compared with the traditional method.

[0146] Table 1 Performance comparison table between the traditional method and the present invention in the automatic positioning of the electric chuck of the laser pipe cutting machine

[0147]

[0148]

[0149] In this embodiment, by applying the present invention to the automatic positioning of the electric chuck of a laser pipe cutting machine, its intelligent optimization ability in complex pipe processing is verified. Through multi-frequency laser coherent sensing measurement, non-Euclidean optimization, Riemannian flow variable stiffness control and error compensation strategy, this method realizes the intelligent distribution of chuck clamping points, dynamic stiffness adjustment and high-precision error compensation, effectively solving the problems of fixed clamping stiffness, low-efficiency error compensation and insufficient deformation prediction in high-precision cutting of traditional fixed clamping methods. The experimental results show that the present invention significantly improves the chuck clamping stability, processing accuracy and production efficiency, reduces the cutting error and repeated processing rate, and provides reliable technical support for the intelligent and precise processing of laser pipe cutting machines.

[0150] As mentioned above, the above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. An automatic positioning method for an electric chuck of a laser pipe cutting machine based on a dynamic learning algorithm, characterized in that, It includes the following steps: S1. Measure the geometric shape, wall thickness distribution, and internal stress state of the pipe to be processed using multi-frequency laser coherent sensing, and construct a pipe shape and mechanics dataset; S2. Based on the pipe shape and mechanics dataset, adopt an adaptive non-Euclidean manifold optimization method to calculate the optimal clamping point distribution of the chuck, and combine the Hodge-Laplacian constraint reconstruction method to adjust the interaction relationship between the clamping points to generate a clamping topology mapping matrix; S3. Based on the clamping topology mapping matrix, combine the graph neural geometry network to optimize the mechanical distribution of the chuck clamping points, and use the Riemannian manifold constraint variable stiffness control method to adjust the clamping stiffness according to the pipe type to form a variable stiffness clamping strategy; S4. Based on the variable stiffness clamping strategy, analyze the stress evolution characteristics of each area of the pipe during the clamping process, calculate the potential error influencing factors, and generate an error analysis result; S5. Use a multi-modal physical embedding error compensation network to calculate the non-linear evolution path of the chuck clamping error, and combine the non-Euclidean error coding technology to optimize the automatic compensation strategy of the chuck to form an error compensation scheme; S6. Synthesize the clamping topology mapping matrix, variable stiffness clamping strategy, and error compensation scheme, and dynamically adjust the chuck clamping torque, position compensation parameters, and error correction mechanism in real time, and finally generate a chuck automatic positioning scheme.

2. The automatic positioning method of the electric chuck of the laser pipe cutting machine based on the dynamic learning algorithm according to claim 1, characterized in that, The specific content of S2 includes: S21. Establish a set of clamping points based on the pipe shape and mechanical data set. The set of clamping points is represented as P = {p 夹持1 , p 夹持2 ,..., p 夹持n}, where n represents the total number of clamping points, and map the surface curvature of the pipe to the non-Euclidean manifold A to construct an initial topological model of the clamping area; S22. Adopt an adaptive non-Euclidean manifold optimization method to define an optimization objective function for the clamping point distribution; Among them, L 优化 represents the optimization objective function, and λ 刚度i represents the stiffness adjustment parameter of the i-th clamping point p 夹持i . represents the non-Euclidean gradient operator, and μ 相互ij represents the interaction coefficient between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j , and d A (p 夹持i , p 夹持j ) represents the geometric distance between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j . S23. Based on the Hodge-Laplacian constraint reconstruction method, establish a clamping point topology constraint matrix; Among them, H 拓扑ij represents the topological connection weight between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j . F(p 夹持i ) represents the pipe shape feature vector at the i-th clamping point p 夹持i . F(p 夹持j ) represents the pipe shape feature vector at the j-th clamping point p 夹持j . F(p 夹持k ) represents the pipe shape feature vector at the k-th clamping point p 夹持k . ||·|| A represents the non-Euclidean measure between the clamping point shape feature vectors, and exp represents the exponential function; S24. Calculate the gradient update amount of each clamping point; Among them, Δ Hodge p 夹持i represents the gradient update amount of the i-th clamping point p 夹持i , N(i) represents the neighborhood set of the i-th clamping point p 夹持i , p 夹持j -p 夹持i represents the relative displacement of the clamping point position; S25. Based on the gradient update amount of each clamping point, construct a clamping topology mapping matrix; Among them, M 映射ij represents the element of the clamping topological mapping matrix, H 拓扑ik represents the topological connection weight between the i-th clamping point p 夹持i and the k-th clamping point p 夹持k , H 拓扑kj represents the topological connection weight between the k-th clamping point p 夹持k and the j-th clamping point p 夹持j , δ 映射ij represents the Kronecker delta matrix, which is used to constrain the topological stability of the clamping points, and α 梯度 represents the gradient adjustment coefficient; S26. The optimization objective is to adjust the clamping point position through gradient update and ensure the stability of the clamping point distribution on the non-Euclidean manifold A.

3. The automatic positioning method of the electric chuck of the laser pipe cutting machine based on the dynamic learning algorithm according to claim 1, characterized in that The specific content of S3 includes: S31. Initialize the mechanical feature matrix F based on the clamping topological mapping matrix and combined with the pipe material category information 0 ={F 0 (p 夹持1 ),…,F 0 (p 夹持n )}; S32. Use the graph neural geometry network to perform multi-layer mechanical aggregation on the mechanical feature matrix F 0 to construct a mechanical distribution iterative equation: F(p 夹持i ) = F 0 (p 夹持i ) + ∑ j∈N(i) H 拓扑ij (F 0 (p 夹持j ) - F 0 (p 夹持i )); Among them, F(p 夹持i ) represents the comprehensive mechanical characteristics at the i-th clamping point p 夹持i . N(i) represents the domain set of the i-th clamping point p 夹持i . H 拓扑ij represents the topological connection weight between the i-th clamping point p 夹持i and the j-th clamping point p 夹持j . S33. Using the Riemannian manifold-constrained variable stiffness control method, construct a variable stiffness regulation functional Ψ 刚度 , which is used to describe the optimal distribution of the stiffness at the chuck clamping points: Among them, n represents the total number of clamping points, α 局部i represents the local stiffness weight, K 刚度i represents the stiffness value at the i-th clamping point p 夹持i . represents the gradient operator on the Riemannian manifold B, β 交互ij represents the regulation factor, K 刚度j represents the stiffness value at the j-th clamping point p 夹持j . 力学i represents the mechanical influence coefficient, ||F(p 夹持i )|| represents the comprehensive mechanical characteristic modulus at the i-th clamping point p 夹持i ; S34. Set the pipe constraint function Θ for different pipe categories 管材 : where max represents the maximum value function, represents the stiffness threshold corresponding to the pipe material category, represents the pipe material category, Θ 管材 constrains the variable stiffness value not to exceed the material adaptation range, meeting the differential requirements of brittleness and pipes; S35. Construct a stiffness iteration equation; Among them, represents the stiffness value at the i-th clamping point p 夹持i at the (t + 1)-th iteration, represents the stiffness value at the i-th clamping point p 夹持i at the t-th iteration, and δ 阈值 represents the threshold penalty coefficient; S36. Through multiple rounds of iterative optimization, make the stiffness of the clamping points gradually converge to the optimal distribution on the Riemannian manifold to form a variable stiffness clamping strategy.

4. The automatic positioning method of the electric chuck of the laser pipe cutting machine based on the dynamic learning algorithm according to claim 1, characterized in that, The specific content of S4 includes: S41. Based on the variable stiffness clamping strategy, obtain the instantaneous stress state of each clamping point of the pipe during the clamping process, construct a pipe stress distribution model, and characterize the force conditions at different positions during the clamping process; S42. Analyze the change trend of stress over time during the clamping process, calculate the stress transfer path according to the force relationship between the clamping points, and determine the force balance state and stress concentration characteristics of the pipe during the clamping process; S43. On the basis of the stress evolution analysis, calculate the stress gradient of each area of the pipe and identify the areas where the stress gradient change is greater than the set threshold; S44. Combine the stress gradient change situation to evaluate the influence of the variable stiffness adjustment on the stress distribution, identify the key influencing factors causing errors during the clamping process, and generate an error analysis result.

5. The automatic positioning method of the electric chuck of the laser pipe cutting machine based on the dynamic learning algorithm according to claim 1, wherein The specific content of S5 includes: S51. Based on the error analysis results, construct a multi-modal physical embedding error compensation network, and define the clamping error state set E = {E 夹持1 , E 夹持2 ,..., E 夹持n}, where E 夹持i represents the instantaneous error at the i-th clamping point p 夹持i , and set the error adjustment parameter ξ 误差i in combination with the pipe material category, and n represents the total number of clamping points; S52. In the error compensation network, introduce a non-linear error evolution path modeling method and define an error evolution equation: Among them, represents the error value at the i-th clamping point p 夹持i at the (t + 1)-th iteration, represents the error value at the i-th clamping point p 夹持i at the t-th iteration, N(i) represents the domain set of the i-th clamping point p 夹持i , γ 误差传递ij represents the error transfer coefficient, represents the error value at the j-th clamping point p 夹持j at the t-th iteration, Φ 非线性i represents the non-linear correction term of the error, which is used to compensate the dynamic evolution trend of the clamping error; S53. Based on the error evolution path, adopt non-Euclidean error coding technology and define the error compensation optimization objective: Among them, Ψ 补偿 represents the error compensation optimization objective function, ρ 局部误差i represents the local error adjustment parameter, represents the non-Euclidean gradient operator, E 夹持j represents the instantaneous error at the j-th clamping point p 夹持j , τ 误差交互ij represents the error interaction weight; S54. Based on the error compensation optimization objective function and combined with the dynamic change characteristics of the chuck clamping error, finally form an error compensation scheme.

6. The automatic positioning method of the electric chuck of the laser pipe cutting machine based on the dynamic learning algorithm according to claim 1, characterized in that, The specific content of S6 is as follows: S61. Based on the clamping topology mapping matrix, extract the spatial distribution characteristics of the chuck clamping points, and combined with the variable stiffness clamping strategy and the error compensation scheme, initialize the automatic positioning optimization variables of the chuck, including the clamping point position, clamping stiffness, and error correction parameters; S62. Real-time monitor the torque change during the chuck clamping process, analyze the stress state of the clamping points, and adjust the clamping torque of the chuck in combination with the clamping topology mapping matrix to meet the stress requirements of different pipes; S63. Combined with the variable stiffness clamping strategy, dynamically adjust the position compensation parameters of the chuck clamping points, optimize the distribution of the clamping points according to the pipe stress conditions, and make the chuck maintain the optimal positioning state under different pipe forms; S64. Based on the error compensation scheme, adjust the error correction parameters in real time to make the error correction process respond to the changes in the chuck clamping state in real time; S65. Based on the clamping torque, position compensation parameters, and error correction mechanism of the chuck, construct an automatic positioning optimization stop criterion: where max represents the maximum value function, represents the position vector of the i-th clamping point p 夹持i at the (t + 1)-th iteration, represents the position vector of the i-th clamping point p 夹持i at the t-th iteration, ∈ 收敛 represents the convergence threshold for position optimization, represents the clamping moment of the i-th clamping point p 夹持i at the (t + 1)-th iteration, represents the clamping moment of the i-th clamping point p 夹持i at the t-th iteration, δ 稳定 represents the stability threshold for clamping moment optimization; That is, when the maximum change amount of the optimization variable is less than the set threshold, the automatic positioning optimization process terminates; S66. Finally, generate an automatic positioning scheme for the chuck to enable the chuck to achieve real-time dynamic adjustment under different pipe categories and clamping states.

Images