Laser cutting automatic typesetting optimization system

Abstract

The invention discloses a laser cutting automatic typesetting optimization system, and belongs to the technical field of automobile safety air bag production. The system comprises a sensing layer, an edge calculation layer, an execution layer and a collaboration layer which are in communication connection in sequence to form a sensing, calculation, execution and feedback closed loop; the sensing layer acquires cloth characteristic data through double-stage sensing; the edge calculation layer deploys a core innovative immune clustering-NSGA-III fusion algorithm, firstly simplifies and optimizes dimensions according to shape similarity grouping, then dynamically adapts to an optimization direction to realize multi-target equilibrium optimization, and forms a secondary optimization system in cooperation with a CNN proxy model pre-screening module; the execution layer accurately executes cutting and corrects deviation, and the collaboration layer is in butt joint with a production system to achieve data tracing and algorithm self-iteration. The multi-target collaborative optimization of the material utilization rate, the cutting precision, the efficiency and the like is realized, the response speed and the adaptability of the system are improved, and the multi-variety and small-batch production requirements of the safety airbags are met.

Description

Technical Field

[0001] This invention relates to the field of airbag manufacturing technology, and in particular to a laser cutting automatic layout optimization system. Background Technology

[0002] As a core passive safety component that protects the safety of drivers and passengers, the production process of automotive airbags has extremely high requirements for cutting accuracy, material utilization and production efficiency. As the core processing step for airbag fabric, the rationality of the layout plan directly determines the mechanical properties, production cost and production rhythm of the final product.

[0003] Currently, automotive airbag production is developing towards a more diversified and smaller-batch production model. A single production line needs to adapt to the cutting requirements of airbag parts of different models and specifications, which poses a more stringent challenge to laser cutting automatic layout optimization technology.

[0004] In existing laser cutting and layout optimization technologies, multi-objective optimization is the core challenge. However, existing solutions generally suffer from algorithmic limitations. Some technologies use a single multi-objective optimization algorithm to directly perform layout calculations. Due to the diverse shapes and specifications of airbag workpieces, the algorithm needs to process a large number of workpiece parameters, fabric characteristic parameters, and process parameters simultaneously, resulting in complex optimization variables. This not only leads to slow convergence speed, making it difficult to adapt to the production rhythm of small batches and rapid changeovers, but also makes it prone to getting trapped in local optima, failing to achieve a balance between material utilization, cutting efficiency, and texture adaptability. Other technologies attempt to simplify calculations by grouping workpieces using immune clustering algorithms. However, this type of technology can only achieve the single function of grouping by shape similarity and lacks synergy with multi-objective optimization algorithms. After grouping, it still requires manual setting of the optimization direction and cannot dynamically adjust the algorithm strategy according to the grouping complexity, ultimately failing to achieve the desired multi-objective optimization effect.

[0005] Meanwhile, existing layout optimization technologies generally suffer from high computational load and poor real-time performance in global optimization. Since airbag laser cutting requires adjusting the layout scheme based on dynamic characteristics such as fabric texture, tension, and defects, existing systems often directly initiate the global optimization process, necessitating comprehensive calculations of massive amounts of perceived data and process parameters, resulting in lengthy layout output times. Some technologies attempt to improve real-time performance by introducing simple linear models for preliminary screening; however, these models are structurally simple and cannot accurately extract the core information from multi-dimensional feature data. Furthermore, they are not trained on large-scale, multi-scenario historical production data, and the preliminary solutions output often do not conform to actual process constraints. This not only fails to provide an effective initial solution for global optimization but also requires additional computational resources to correct pre-screening deviations, further reducing overall optimization efficiency and making it difficult to adapt to the actual needs of rapid batch switching and timely output of layout schemes in multi-variety, small-batch production.

[0006] Furthermore, existing technologies generally suffer from a disconnect between the pre-screening and global optimization stages, lacking an effective technical collaboration mechanism. On one hand, the preliminary solutions output by the pre-screening module often only contain simple coordinate range information and are not logically related to the parameters of the global optimization algorithm. This necessitates a full-dimensional verification of the pre-screening solution for global optimization, making it impossible to reduce invalid calculations using the pre-screening results. On the other hand, the process adaptation data generated during global optimization cannot be fed back to the pre-screening module, preventing the accuracy of the pre-screening model from continuously improving. This disconnect not only further exacerbates the computational load and reduces optimization efficiency but also makes the system less adaptable to the characteristics of different batches of fabric and different specifications of workpieces. It can easily lead to mismatches between the layout scheme and actual production conditions, ultimately affecting the stability of airbag cutting quality and increasing production costs and the difficulty of quality control.

[0007] Therefore, this invention patent proposes an automatic layout optimization system for laser cutting. Summary of the Invention

[0008] One objective of this invention is to propose an automatic layout optimization system for laser cutting. This invention effectively addresses the core pain points of existing automotive airbag laser cutting layout technology, such as complex multi-objective optimization dimensions, poor convergence effects leading to local optima, high global optimization computational load and poor real-time performance, as well as the disconnect between pre-screening and optimization stages and insufficient system adaptability. It achieves multi-objective balanced optimization of material utilization, texture adaptability, cutting efficiency, tension stability, defect risk control, and cutting accuracy through an immune clustering-NSGA-Ⅲ fusion algorithm. The pre-screening module, based on a CNN-based surrogate model, significantly reduces the global optimization computational load and improves system response speed. Furthermore, a two-level collaborative mechanism between pre-screening and optimization forms a technical linkage, enabling the system to accurately adapt to the multi-variety, small-batch production scenarios of automotive airbags. This ensures stable cutting quality while reducing unnecessary calculations and production costs, providing a high-precision, high-efficiency, and highly adaptable automatic layout solution for airbag laser cutting.

[0009] According to an embodiment of the present invention, an automatic layout optimization system for laser cutting includes a perception layer, an edge computing layer, an execution layer, and a collaboration layer, with each layer interacting with data through an industrial communication module.

[0010] The sensing layer is a flexible material multi-dimensional sensing system that integrates sensors for texture acquisition, tension detection, defect identification, and deformation monitoring. It adopts a dual-stage sensing mode of pre-scanning and real-time monitoring to provide quantitative fabric characteristic data for layout optimization.

[0011] The edge computing layer deploys a multi-objective collaborative optimization engine, the core of which is the immune clustering-NSGA-Ⅲ fusion algorithm. It performs global optimization with multiple objectives, including material utilization and texture adaptability, cutting efficiency and tension stability, and defect risk and cutting accuracy. Its multi-objective optimization objective function set is as follows:

[0012]

[0013] Among them, formula (1) is the comprehensive objective function of material utilization rate and texture adaptation, formula (2) is the comprehensive objective function of cutting efficiency and tension stability, and formula (3) is the comprehensive objective function of defect risk and cutting accuracy. This represents the total area of ​​all workpieces cut in a single operation. This represents the total area of ​​all workpieces cut in a single operation. For texture matching coefficients, This refers to the total number of workpieces cut in a single operation. The time required for a single cut by the CEO The total time consumed by the laser head during idle movement. The tension uniformity coefficient, This represents the defect risk coefficient. D represents the deviation between the actual cutting accuracy and the standard cutting accuracy, and D is the standard accuracy value for laser cutting of automotive airbags; the edge calculation layer simultaneously establishes a dynamic mapping relationship between the layout coordinates and the laser cutting parameters.

[0014] The execution layer controls the laser equipment to complete the cutting and correct the deviation according to the instructions of the edge computing layer. The collaboration layer connects to the production management system to realize order processing, data traceability and algorithm self-iteration, ensuring the collaboration between the system and the production process.

[0015] Furthermore, the quantified fabric characteristic data of the sensing layer includes fiber direction and texture distribution characteristics acquired by the texture acquisition sensor, laying tension uniformity data acquired by the tension detection sensor, coating defects and fiber breakage information acquired by the defect identification sensor, and minute deformation of the fabric during the cutting process captured by the deformation monitoring sensor.

[0016] Furthermore, the multi-objective collaborative optimization engine of the edge computing layer constructs a four-level computing system consisting of data preprocessing, proxy model pre-screening, fusion algorithm optimization, and parameter coupling output. The data preprocessing module uses an improved DS evidence theory to fuse multimodal sensing data and eliminate data conflicts.

[0017] Furthermore, the data preprocessing module normalizes the multi-dimensional perception data to the [0,1] interval by defining the texture matching coefficient K, tension uniformity coefficient T, and defect risk coefficient F, and transforms it into a constraint condition that can be identified by the objective function of the fusion algorithm.

[0018] Where K is the texture matching coefficient, and when K≥0.85 it is determined to be texture adaptation, which meets the texture optimization requirements of formula (1); T is the tension uniformity coefficient, and when T≤0.1 it is determined to be fabric tension stable, which adapts to the efficiency optimization logic of formula (2); F is the defect risk coefficient, and when F=0 it is determined to be no material defect, which ensures that the risk control target of formula (3) is achieved.

[0019] Furthermore, the operation flow of the immune clustering-NSGA-Ⅲ fusion algorithm is as follows:

[0020] First, the workpieces are grouped according to shape similarity using an immune clustering algorithm to simplify the optimization dimension. Then, the improved NSGA-Ⅲ algorithm is substituted into the objective function group described in claim 1 to perform global optimization of the layout position and cutting path of each group of workpieces. The improved NSGA-Ⅲ algorithm dynamically adjusts the optimization direction through an adaptive mutation operator, and the mutation probability is adapted in real time according to the complexity of workpiece grouping in the range of 0.5-0.9 to ensure the balanced convergence of multiple objective functions.

[0021] Furthermore, the proxy model pre-screening module is built on a neural network. After being trained with historical production data, it can quickly output a preliminary layout scheme and laser cutting parameter range that meet the basic constraints, providing an efficient initial solution for global optimization.

[0022] Furthermore, the adaptation logic of the dynamic mapping relationship combines T in the objective function and the workpiece contour curvature feature. When the workpiece contour curvature... or At that time, matching low cutting speed with high energy density laser parameters; when the workpiece contour curvature and At the same time, matching high cutting speed with appropriate energy density laser parameters ensures that the objective function is... .

[0023] Furthermore, the execution layer includes a motion control unit, a laser parameter adjustment unit, and a workpiece sorting unit. The motion control unit drives the laser cutting head to move in a multi-axis coordinated manner, the laser parameter adjustment unit adjusts the laser power and pulse frequency in real time, and the workpiece sorting unit automatically separates finished products from waste materials.

[0024] Furthermore, the motion control unit adopts bus control technology and can correct the motion trajectory based on the feedback data from the deformation monitoring sensor; the workpiece sorting unit, combined with the layout coordinates, completes automated sorting through a pneumatic actuator.

[0025] Furthermore, the collaboration layer uses a standardized communication protocol to communicate with MES and ERP systems. By collecting finished product quality inspection data and feeding it back to the edge computing layer, it enables the agent model parameter update and algorithm self-iteration to adapt to the production needs of different batches.

[0026] The beneficial effects of this invention are:

[0027] 1. This invention effectively addresses the core technical pain points of existing laser cutting and layout technologies, such as complex multi-objective optimization dimensions, poor convergence, and susceptibility to local optima, through the fusion algorithm of immune clustering-NSGA-Ⅲ. In existing technologies, single optimization algorithms either suffer from too many optimization variables and slow convergence due to the complex parameters of various workpieces, or they can only achieve single-objective optimization without balancing material adaptation, cutting accuracy, and efficiency. By extracting workpiece contour feature points and calculating Euclidean distance through the immune clustering algorithm, and grouping workpieces based on shape similarity, the dimensions of subsequent optimization are simplified from the root, reducing the amount of invalid computation. Furthermore, the adaptive mutation operator of the improved NSGA-Ⅲ algorithm is used to dynamically adjust the optimization direction to address the differences in complexity among different groups. This ensures the breadth of optimization when facing complex groups and improves convergence efficiency when facing simple groups. Ultimately, it achieves a balanced optimization of multiple objectives, including material utilization, texture adaptation, cutting efficiency, tension stability, defect risk control, and cutting accuracy, completely avoiding the technical limitations caused by the single nature of traditional algorithms.

[0028] 2. The CNN-based surrogate model pre-screening module in this invention creatively solves the problems of high computational load and poor real-time performance during global optimization in existing layout optimization technologies. It perfectly adapts to the needs of multi-variety, small-batch production scenarios for automotive airbags. The CNN surrogate model, through a complete structure of input layer, convolutional layer, pooling layer, fully connected layer, and output layer, can accurately extract the core information of multi-dimensional feature data after preprocessing. Furthermore, it is trained with a large amount of historical data covering different fabric types, workpiece specifications, perception parameters, and finished product quality, ensuring the accuracy and process adaptability of the pre-screening scheme. At the same time, this module can quickly output a preliminary layout scheme and laser parameter range that meet basic constraints, providing a high-quality initial solution for subsequent global optimization, significantly reducing the computational load of global optimization, effectively improving the overall system response speed, and meeting the actual needs of rapid batch switching and timely output of layout schemes during the production process.

[0029] 3. This invention creatively collaborates with the immune clustering-NSGA-Ⅲ fusion algorithm and the CNN surrogate model pre-screening module to construct a two-level optimization system for efficient pre-screening and accurate optimization. This creates a technical synergy effect that far exceeds the function of a single module, further breaking through the limitation of the disconnect between pre-screening and optimization in existing technologies. The preliminary scheme output by the CNN surrogate model not only provides the initial optimization direction for the fusion algorithm but also eliminates the possibility of layouts that obviously do not conform to constraints such as texture and tension in advance, reducing the invalid optimization range of the fusion algorithm. The parameter logic formed by the fusion algorithm during the optimization process can, in turn, provide data support for the subsequent training of the surrogate model, continuously improving the pre-screening accuracy. This two-way collaborative mechanism not only greatly improves the efficiency and accuracy of the overall layout optimization but also enables the system to better adapt to changes in the characteristics of different batches of fabric and workpiece types, significantly enhancing the practicality and adaptability of the system. Attached Figure Description

[0030] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0031] Figure 1 This is a schematic diagram of the framework structure of a laser cutting automatic layout optimization system proposed in this invention;

[0032] Figure 2 This is a schematic diagram of the operation process of the immune clustering-NSGA-Ⅲ fusion algorithm of the automatic layout optimization system for laser cutting proposed in this invention. Detailed Implementation

[0033] To make the technical means and objectives and effects of the present invention easier to understand, the embodiments of the present invention will be described in detail below with reference to specific illustrations.

[0034] like Figure 1-2 As shown, this invention discloses an automatic layout optimization system for laser cutting, including a perception layer, an edge computing layer, an execution layer, and a collaboration layer. Each layer interacts with data through an industrial communication module, forming a closed-loop control system of perception, calculation, execution, and feedback. This closed-loop system can ensure real-time and efficient data transmission between layers, realizing full-process collaboration from material characteristic acquisition to layout optimization, cutting execution, and data feedback, avoiding problems such as insufficient layout accuracy and low efficiency caused by disconnection between links.

[0035] The industrial communication module can be an EtherCAT bus module or a 5G industrial module, preferably an EtherCAT bus module, with a data transmission latency of ≤200ms, which can meet the real-time control requirements in mass production and ensure the synchronous interaction of instructions and data at each level.

[0036] The sensing layer is a flexible material multi-dimensional sensing system that integrates sensors for texture acquisition, tension detection, defect identification, and deformation monitoring. It adopts a dual-stage sensing mode of pre-scanning and real-time monitoring to provide quantitative fabric characteristic data for layout optimization. Specifically, the texture acquisition sensor uses a high-resolution linear array camera, whose scanning direction is perpendicular to the fabric conveying direction. By continuously capturing images of the fabric surface, the fiber direction and texture distribution features are extracted using an image recognition algorithm. The image sampling frequency is matched with the fabric conveying speed to ensure no texture information is missed. The tension detection sensor uses a distributed strain gauge sensor, which is evenly arranged on the support rollers of the fabric laying platform. It collects the force data of the rollers in real time and calculates the uniformity data of the fabric laying tension through a force-tension conversion model. The defect identification sensor uses a hyperspectral imager with a spectral range covering 400-1000nm. It can accurately identify defects such as silicon coating defects and fiber breaks on the fabric surface. The location and range of defects are determined by analyzing the gray values ​​of pixels. The deformation monitoring sensor uses a laser displacement sensor, which is installed next to the laser cutting head and moves synchronously with the cutting head. It captures the minute displacements of the fabric surface in real time during the cutting process and calculates the deformation data. The aforementioned quantified fabric characteristic data specifically includes fiber direction and texture distribution characteristics acquired by texture acquisition sensors, laying tension uniformity data acquired by tension detection sensors, coating defects and fiber breakage information acquired by defect identification sensors, and minute fabric deformation during the cutting process captured by deformation monitoring sensors. These data provide a comprehensive and accurate basis for subsequent layout optimization.

[0037] The edge computing layer deploys a multi-objective collaborative optimization engine, the core of which is the immune clustering-NSGA-Ⅲ fusion algorithm. It performs global optimization with multiple objectives, including material utilization and texture adaptability, cutting efficiency and tension stability, and defect risk and cutting accuracy. Its multi-objective optimization objective function set is as follows:

[0038]

[0039] Among them, formula (1) is the comprehensive objective function of material utilization rate and texture adaptation, formula (2) is the comprehensive objective function of cutting efficiency and tension stability, and formula (3) is the comprehensive objective function of defect risk and cutting accuracy. This represents the total area of ​​all workpieces cut in a single operation. This represents the total area of ​​all workpieces cut in a single operation. For texture matching coefficients, This refers to the total number of workpieces cut in a single operation. The time required for a single cut by the CEO The total time consumed by the laser head during idle movement. The tension uniformity coefficient, This represents the defect risk coefficient. D represents the deviation between the actual cutting accuracy and the standard cutting accuracy, and D is the standard accuracy value for laser cutting of automotive airbags; the edge calculation layer simultaneously establishes a dynamic mapping relationship between the layout coordinates and the laser cutting parameters.

[0040] Furthermore, the multi-objective collaborative optimization engine of the edge computing layer constructs a four-level computing system: data preprocessing, surrogate model pre-screening, fusion algorithm optimization, and parameter coupling output. The data preprocessing module uses an improved DS evidence theory to fuse multimodal sensing data and eliminate data conflicts. In specific implementation, the data preprocessing module first performs noise reduction processing on the multi-dimensional raw data collected by the sensing layer, using a Gaussian filtering algorithm to remove random noise from texture images and tension data. Then, it uses the improved DS evidence theory to assign trust levels to data sources from different sensors, adjusting the weight coefficients of conflicting data to a reasonable range to avoid optimization deviations caused by errors from a single data source. Simultaneously, the data preprocessing module defines texture matching coefficients... Tension uniformity coefficient and defect risk coefficient The multi-dimensional perceived data is normalized to the [0,1] interval, transforming it into constraints that the objective function of the fusion algorithm can recognize; whereby... The texture matching coefficient is obtained by calculating the cosine of the angle between the preset texture direction of the workpiece and the actual fiber direction of the fabric. When the texture is determined to be compatible, it meets the texture optimization requirements of formula (1); The tension uniformity coefficient is obtained by calculating the ratio of the standard deviation to the mean of the pavement tension data. When the fabric tension is stable, the efficiency optimization logic of formula (2) is applied. This is the defect risk coefficient, when there are no defects. When there are minor defects When there are serious defects Only when When the material is deemed to be defect-free, the risk control objective of formula (3) is achieved.

[0041] The specific operation flow of the immune clustering-NSGA-Ⅲ fusion algorithm is as follows:

[0042] First, the workpieces are grouped according to shape similarity using an immune clustering algorithm to simplify the optimization dimensions. The specific steps are as follows: extract the contour feature points of each workpiece, calculate the Euclidean distance between the feature points, use the distance threshold as the clustering standard, group the workpieces with Euclidean distance less than the threshold into one group, and the shape similarity of each group of workpieces is ≥80%, thereby reducing the number of optimization variables.

[0043] Then, by substituting the above objective function group into the improved NSGA-Ⅲ algorithm, global optimization is performed on the layout position and cutting path of each group of workpieces;

[0044] Among them, the improved NSGA-Ⅲ algorithm dynamically adjusts the optimization direction through an adaptive mutation operator. The mutation probability is adapted in real time to the complexity of workpiece grouping in the range of 0.5-0.9. The more groups there are and the more complex the workpiece shape is, the closer the mutation probability is to 0.9. The fewer groups there are and the simpler the workpiece shape is, the closer the mutation probability is to 0.5. This ensures the balanced convergence of the multi-objective function and avoids getting trapped in local optima.

[0045] The proxy model pre-screening module is built on CNN. The network structure includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The dimension of the input layer is consistent with the dimension of the preprocessed feature data, and the dimension of the output layer corresponds to the key parameters of the preliminary layout scheme.

[0046] The proxy model was trained with over 100,000 sets of historical production data, including different fabric types, workpiece specifications, sensing parameters, layout schemes, and finished product quality data. After training, it can quickly output a preliminary layout scheme and laser cutting parameter range that meet the basic constraints within 50ms, providing an efficient initial solution for global optimization and reducing global optimization time by more than 60%.

[0047] The dynamic mapping relationship established by the parameter coupling output module, and its adaptation logic combined with the objective function The workpiece contour curvature is obtained by calculating the reciprocal of the radius of curvature at each point on the workpiece edge. When the workpiece contour curvature... or At the same time, matching low cutting speed with high energy density laser parameters ensures cutting accuracy under complex contours or unstable tension conditions; when the workpiece contour curvature And all At the same time, by matching high cutting speed with suitable energy density laser parameters, cutting efficiency is improved while ensuring accuracy, ultimately ensuring the objective function is met. It meets the cutting precision requirements for automotive airbags.

[0048] The execution layer controls the laser equipment to complete the cutting and correct the deviation according to the instructions of the edge computing layer. The execution layer includes a motion control unit, a laser parameter adjustment unit and a workpiece sorting unit. The motion control unit drives the laser cutting head to move in a multi-axis coordinated manner. The laser parameter adjustment unit adjusts the laser power and pulse frequency in real time. The workpiece sorting unit automatically separates finished products from waste materials.

[0049] Specifically, the motion control unit adopts EtherCAT bus control technology, connects the X, Y, and Z axis drive motors and rotary motors of the laser cutting head, receives the layout coordinate path data output by the edge computing layer, and controls the cutting head to move along the preset trajectory, with a motion repeatability accuracy of 0.003mm / step.

[0050] Meanwhile, the motion control unit receives deformation data from the deformation monitoring sensor in real time and dynamically corrects the motion trajectory through a PID algorithm. When the deformation amount is ≥0.02mm, trajectory correction is triggered immediately to ensure that the cutting contour deviation is ≤±0.05mm.

[0051] The laser parameter adjustment unit adopts a digital signal control module, which receives real-time parameter commands from the parameter coupling output module, adjusts the power by adjusting the drive current of the laser generator, and adjusts the pulse frequency by adjusting the pulse signal frequency. The response time is ≤10ms, ensuring that the parameter adjustment is synchronized with the cutting motion.

[0052] The workpiece sorting unit includes pneumatic grippers, a conveyor belt, and position sensors. The position sensors acquire the actual coordinates of the finished product and waste material after cutting, compare them with the preset coordinates in the layout scheme, and control the pneumatic grippers to accurately grab the finished product and place it on the finished product conveyor belt. The waste material is transported to the recycling device through the waste material conveyor belt, realizing automated sorting with a sorting accuracy rate of ≥99.8%.

[0053] The collaboration layer connects to the production management system to enable order processing, data traceability, and algorithm self-iteration, ensuring collaboration between the system and the production process.

[0054] The collaboration layer uses the OPCUA standardized communication protocol to communicate with MES and ERP systems. It supports the import of order files in Excel and XML formats, and automatically parses the core requirements in the order, such as workpiece specifications, production quantity, and delivery deadline, and transforms the requirements into control parameters that the system can recognize.

[0055] Meanwhile, the collaboration layer collects key data in the production process in real time, including sensing parameters, layout scheme parameters, cutting parameters, and finished product quality inspection data, and establishes a data association library to form a full life cycle traceability chain for each batch of products, which facilitates the investigation of quality problems.

[0056] In addition, the collaboration layer collects finished product quality inspection data and feeds it back to the edge computing layer, regularly updates the training dataset of the proxy model, retrains the model to optimize parameter weights, realizes algorithm self-iteration, and enables the system to adapt to the production needs of different batches of fabric characteristics and different workpiece types, continuously improving the layout optimization effect.

[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A laser cutting automatic layout optimization system, characterized in that, It includes a perception layer, an edge computing layer, an execution layer, and a collaboration layer, with each layer interacting with data through an industrial communication module; The sensing layer is a flexible material multi-dimensional sensing system that integrates sensors for texture acquisition, tension detection, defect identification, and deformation monitoring. It adopts a dual-stage sensing mode of pre-scanning and real-time monitoring to provide quantitative fabric characteristic data for layout optimization. The edge computing layer deploys a multi-objective collaborative optimization engine, the core of which is the immune clustering-NSGA-Ⅲ fusion algorithm. It performs global optimization with multiple objectives, including material utilization and texture adaptability, cutting efficiency and tension stability, and defect risk and cutting accuracy. Its multi-objective optimization objective function set is as follows: Among them, formula (1) is the comprehensive objective function of material utilization rate and texture adaptation, formula (2) is the comprehensive objective function of cutting efficiency and tension stability, and formula (3) is the comprehensive objective function of defect risk and cutting accuracy. This represents the total area of ​​all workpieces cut in a single operation. This represents the total area of ​​all workpieces cut in a single operation. For texture matching coefficients, This refers to the total number of workpieces cut in a single operation. The time required for a single cut by the CEO The total time consumed by the laser head during idle movement. The tension uniformity coefficient, This represents the defect risk coefficient. D represents the deviation between the actual cutting accuracy and the standard cutting accuracy, and D is the standard accuracy value for laser cutting of automotive airbags; the edge calculation layer simultaneously establishes a dynamic mapping relationship between the layout coordinates and the laser cutting parameters; The execution layer controls the laser equipment to complete the cutting and correct the deviation according to the instructions of the edge computing layer. The collaboration layer connects to the production management system to realize order processing, data traceability and algorithm self-iteration, ensuring the collaboration between the system and the production process.

2. The laser cutting automatic layout optimization system according to claim 1, characterized in that, The quantified fabric characteristic data of the sensing layer includes fiber direction and texture distribution characteristics acquired by the texture acquisition sensor, laying tension uniformity data acquired by the tension detection sensor, coating defects and fiber breakage information acquired by the defect identification sensor, and minute deformation of the fabric during the cutting process captured by the deformation monitoring sensor.

3. The laser cutting automatic layout optimization system according to claim 1, characterized in that, The multi-objective collaborative optimization engine of the edge computing layer constructs a four-level computing system: data preprocessing, proxy model pre-screening, fusion algorithm optimization, and parameter coupling output. The data preprocessing module adopts improved DS evidence theory to fuse multimodal sensing data and eliminate data conflicts.

4. The laser cutting automatic layout optimization system according to claim 1, characterized in that, The data preprocessing module normalizes the multi-dimensional perception data to the [0,1] interval by defining the texture matching coefficient K, tension uniformity coefficient T, and defect risk coefficient F, and transforms it into a constraint condition that can be identified by the objective function of the fusion algorithm. Where K is the texture matching coefficient, and when K≥0.85 it is determined to be texture adaptation, satisfying the texture optimization requirements of formula (1); T is the tension uniformity coefficient, and when T≤0.1 it is determined to be fabric tension stable, adapting to the efficiency optimization logic of formula (2); F is the defect risk coefficient. When F=0, it is determined that there are no material defects, thus ensuring that the risk control target of formula (3) is achieved.

5. The laser cutting automatic layout optimization system according to claim 1, characterized in that, The operation flow of the immune clustering-NSGA-Ⅲ fusion algorithm is as follows: First, the workpieces are grouped according to shape similarity using an immune clustering algorithm to simplify the optimization dimension. Then, the improved NSGA-Ⅲ algorithm is substituted into the objective function group described in claim 1 to perform global optimization of the layout position and cutting path of each group of workpieces. The improved NSGA-Ⅲ algorithm dynamically adjusts the optimization direction through an adaptive mutation operator, and the mutation probability is adapted in real time according to the complexity of workpiece grouping in the range of 0.5-0.9 to ensure the balanced convergence of multiple objective functions.

6. The laser cutting automatic layout optimization system according to claim 3, characterized in that, The proxy model pre-screening module is built on a neural network. After being trained with historical production data, it can quickly output a preliminary layout scheme and laser cutting parameter range that meet the basic constraints, providing an efficient initial solution for global optimization.

7. The laser cutting automatic layout optimization system according to claim 1, characterized in that, The adaptation logic of the dynamic mapping relationship combines the objective function T and the workpiece contour curvature feature. When the workpiece contour curvature... or At that time, matching low cutting speed with high energy density laser parameters; when the workpiece contour curvature and At the same time, matching high cutting speed with appropriate energy density laser parameters ensures that the objective function is... .

8. The laser cutting automatic layout optimization system according to claim 1, characterized in that, The execution layer includes a motion control unit, a laser parameter adjustment unit, and a workpiece sorting unit. The motion control unit drives the laser cutting head to move in a multi-axis coordinated manner, the laser parameter adjustment unit adjusts the laser power and pulse frequency in real time, and the workpiece sorting unit automatically separates finished products from waste materials.

9. The laser cutting automatic layout optimization system according to claim 8, characterized in that, The motion control unit adopts bus control technology and can correct the motion trajectory based on the feedback data from the deformation monitoring sensor; the workpiece sorting unit combines the layout coordinates and completes automated sorting through a pneumatic actuator.

10. The laser cutting automatic layout optimization system according to claim 1, characterized in that, The collaboration layer uses a standardized communication protocol to communicate with MES and ERP systems. By collecting finished product quality inspection data and feeding it back to the edge computing layer, it enables the agent model parameters to be updated and the algorithm to be iterated, thus adapting to the production needs of different batches.

Images