Positioning control method and system for metal cutting

Abstract

This invention discloses a positioning control method and system for metal cutting, relating to the field of metal cutting technology. The method includes: simultaneously acquiring a cutting monitoring dataset while a laser cutting machine cuts a metal workpiece according to a cutting control scheme; performing multi-dimensional cutting accident simulation based on the cutting monitoring dataset to obtain a cutting accident simulation map; adjusting the cutting control scheme with multiple degrees of freedom to obtain a first cutting adjustment space; performing cutting evaluation optimization on the first cutting adjustment space to obtain a second cutting adjustment space; performing cutting assurance degree analysis optimization on the second cutting adjustment space to obtain a first cutting guide population that meets a predetermined cutting assurance degree; performing multiple rounds of breeding optimization to determine the cutting adjustment optimization result, and optimizing the positioning control of the metal workpiece. This solves the technical problems of insufficient positioning accuracy and poor stability in metal cutting in existing technologies, achieving the technical effect of improving the positioning accuracy and stability of metal workpiece cutting.

Description

Technical Field

[0001] This invention relates to the field of metal cutting technology, and more specifically to a positioning control method and system for metal cutting. Background Technology

[0002] With the increasing demand for high precision and efficiency in the metal processing industry, laser cutting technology has been widely applied in fields such as automobile manufacturing, aerospace, shipbuilding, and precision machining. Traditional metal cutting methods mainly rely on preset cutting control schemes during the cutting process, lacking multi-dimensional monitoring and dynamic adjustment of the real-time status. Due to uncertainties such as differences in workpiece material, complexity of the cutting path, and external environmental interference, the cutting positioning accuracy is often insufficient, and the cutting process is unstable, thus affecting cutting quality and production efficiency. Summary of the Invention

[0003] This application provides a positioning control method and system for metal cutting, which solves the technical problems of insufficient positioning accuracy and poor stability of the cutting process in the prior art.

[0004] The first aspect of this application provides a positioning control method for metal cutting, the method comprising:

[0005] When a laser cutting machine cuts a metal workpiece according to a cutting control scheme, it simultaneously acquires a cutting monitoring dataset, which includes cutting head positioning data, cutting machine status data, and workpiece status data. Based on the cutting monitoring dataset, a multi-dimensional cutting accident simulation is performed to obtain a cutting accident simulation map. Based on the workpiece cutting requirements, the cutting control scheme is adjusted with multiple degrees of freedom according to the cutting accident simulation map to obtain a first cutting adjustment space. A multi-dimensional cutting evaluation model is introduced to optimize the first cutting adjustment space, resulting in a second cutting adjustment space. A cutting assurance degree calculation model is introduced to analyze and optimize the second cutting adjustment space, obtaining a first cutting guide population that meets a predetermined cutting assurance degree. Based on the first cutting guide population, the second cutting adjustment space undergoes multiple rounds of breeding optimization to determine the cutting adjustment optimization result, and the metal workpiece is optimized and positioned according to the cutting adjustment optimization result.

[0006] A second aspect of this application provides a positioning control system for metal cutting, the system comprising:

[0007] Data Acquisition Module: When the laser cutting machine cuts a metal workpiece according to the cutting control scheme, it simultaneously acquires a cutting monitoring dataset, which includes cutting head positioning data, cutting machine status data, and workpiece status data. Deduction Module: Based on the cutting monitoring dataset, it performs multi-dimensional cutting accident deduction to obtain a cutting accident deduction map. Scheme Adjustment Module: Based on the workpiece cutting requirements of the metal workpiece, it adjusts the cutting control scheme with multiple degrees of freedom according to the cutting accident deduction map to obtain a first cutting adjustment space. First Optimization Module: It introduces a multi-dimensional cutting evaluation model to optimize the first cutting adjustment space, obtaining a second cutting adjustment space. Second Optimization Module: It introduces a cutting assurance degree calculation model to perform analytical optimization of the cutting assurance degree in the second cutting adjustment space, obtaining a first cutting guide population that meets the predetermined cutting assurance degree. Positioning Control Module: Based on the first cutting guide population, it performs multiple rounds of breeding optimization in the second cutting adjustment space to determine the cutting adjustment optimization result, and optimizes the positioning control of the metal workpiece based on the cutting adjustment optimization result.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] When a laser cutting machine cuts a metal workpiece according to a cutting control scheme, it simultaneously acquires a cutting monitoring dataset, which includes cutting head positioning data, cutting machine status data, and workpiece status data. Based on this dataset, a multi-dimensional cutting accident simulation is performed to obtain a cutting accident simulation map. Further, based on the workpiece's cutting requirements and the cutting accident simulation map, the cutting control scheme is adjusted with multiple degrees of freedom to obtain a first cutting adjustment space. Subsequently, a multi-dimensional cutting evaluation model is introduced to optimize the first cutting adjustment space, resulting in a second cutting adjustment space. A cutting assurance calculation model is then introduced to analyze and optimize the cutting assurance of the second cutting adjustment space, thus obtaining a first cutting guide population that meets the predetermined cutting assurance. Finally, the second cutting adjustment space is subjected to multiple rounds of breeding optimization based on the first cutting guide population to determine the cutting adjustment optimization result. Based on this result, the metal workpiece is optimized for positioning control. This solves the technical problems of insufficient positioning accuracy and poor stability in metal cutting in existing technologies, achieving the technical effect of improving the positioning accuracy and stability of metal workpiece cutting. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A schematic flowchart of a positioning control method for metal cutting provided in an embodiment of this application;

[0012] Figure 2 This is a schematic diagram of the positioning control system for metal cutting provided in an embodiment of this application.

[0013] Explanation of reference numerals in the attached diagram: Data acquisition module 11, deduction module 12, scheme adjustment module 13, first optimization module 14, second optimization module 15, positioning control module 16. Detailed Implementation

[0014] This application provides a positioning control method and system for metal cutting, which solves the technical problems of insufficient positioning accuracy and poor stability of the cutting process in the prior art.

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0016] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0017] Example 1, as Figure 1 As shown, this application provides a positioning control method for metal cutting, wherein the method includes:

[0018] When the laser cutting machine cuts a metal workpiece according to the cutting control scheme, it simultaneously acquires a cutting monitoring dataset, which includes cutting head positioning data, cutting machine status data, and workpiece status data.

[0019] When the laser cutting machine performs cutting operations on a metal workpiece according to the preset cutting control scheme, the system simultaneously activates the data acquisition module to monitor and acquire various types of status data in real time during the cutting process, forming a cutting monitoring dataset. This dataset includes: cutting head positioning data (current position, displacement trajectory, speed, and attitude angle of the cutting head in the cutting path, reflecting the real-time spatial position of the cutting head relative to the workpiece); cutting machine status data (laser output power, cutting speed, nozzle air pressure, machine tool operating current, and vibration characteristics, characterizing the operating conditions of the cutting machine during the cutting task); and workpiece status data (workpiece surface temperature distribution, material thickness, stress deformation, and kerf width, reflecting the dynamic response state of the workpiece during the cutting process).

[0020] Based on the cutting monitoring dataset, a multi-dimensional cutting accident simulation was performed to obtain a cutting accident simulation map.

[0021] Furthermore, based on the aforementioned cutting monitoring dataset, a multi-dimensional cutting accident simulation is performed to obtain a cutting accident simulation map, including:

[0022] Anomaly detection is performed based on the cutting monitoring dataset to construct a cutting anomaly detection vector; cutting quality accidents are extrapolated from the cutting anomaly detection vector based on the cutting quality accident event set to obtain a first accident extrapolation result; cutting machine health accidents are extrapolated from the cutting machine health accident event set to obtain a second accident extrapolation result; cutting safety accidents are extrapolated from the cutting anomaly detection vector based on the cutting safety accident event set to obtain a third accident extrapolation result; and a cutting accident extrapolation map is constructed based on the first accident extrapolation result, the second accident extrapolation result, and the third accident extrapolation result.

[0023] The positioning data of the cutting head is subjected to trajectory fitting analysis to identify the deviation between the cutting head's movement trajectory and the preset cutting path, thus obtaining the positioning anomaly detection result. The cutting machine status data is compared with parameters and thresholds are determined to identify conditions such as laser power fluctuations, abnormal nozzle air pressure, and operating current overload, thus obtaining the cutting machine anomaly detection result. The workpiece status data is analyzed for thermal response and deformation to identify conditions such as excessive temperature, stress concentration, or abnormal kerf extension, thus obtaining the workpiece anomaly detection result. The positioning anomaly detection results, cutting machine anomaly detection results, and workpiece anomaly detection results are fused to construct a cutting anomaly detection vector. Based on this, preset accident event sets are invoked for extrapolation analysis for different types of accident risks. Specifically, based on the cutting quality accident event set, the cutting anomaly detection vector is extrapolated for the quality dimension, resulting in the first accident extrapolation result; based on the cutting machine health accident event set, the cutting anomaly detection vector is extrapolated for the equipment health dimension, resulting in the second accident extrapolation result; and based on the cutting safety accident event set, the cutting anomaly detection vector is extrapolated for the operational safety dimension, resulting in the third accident extrapolation result. Finally, the simulation results of the first, second, and third accidents are fused together to establish a multi-dimensional segmented accident simulation map. This map uses nodes and edges to represent accident characteristics, triggering factors, and their causal relationships, intuitively reflecting the evolution paths and interactions of different accident types.

[0024] Furthermore, based on the aforementioned cutting monitoring dataset, anomaly detection is performed to construct a cutting anomaly detection vector, including:

[0025] According to the cutting control scheme, the cutting head positioning is fitted to construct a reference positioning dynamic mesh; the cutting head positioning data is meshed to obtain a monitoring positioning dynamic mesh; anomaly detection is performed on the monitoring positioning dynamic mesh based on the reference positioning dynamic mesh to obtain a positioning anomaly detection result; anomaly detection is performed on the cutting machine status data based on the cutting control scheme to obtain a cutting machine anomaly detection result; anomaly detection is performed on the workpiece status data based on the cutting control scheme to obtain a workpiece anomaly detection result; and the cutting anomaly detection vector is generated by combining the positioning anomaly detection result and the cutting machine anomaly detection result.

[0026] Specifically, based on the cutting control scheme, the theoretical motion trajectory of the cutting head is fitted and calculated to obtain a set of reference positioning dynamic grids. These reference positioning dynamic grids can describe the spatial position change law of the cutting head under ideal cutting conditions. Subsequently, the collected cutting head positioning data is discretized and gridded to obtain the corresponding monitoring positioning dynamic grid. By comparing the monitoring positioning dynamic grid with the reference positioning dynamic grid point by point, the deviation between the actual running trajectory of the cutting head and the preset trajectory is identified. If the deviation exceeds a preset threshold, it is judged as a positioning anomaly, and the positioning anomaly detection result is output.

[0027] The system performs mesh fitting and anomaly detection on the cutting machine's status data based on the cutting control scheme. Specifically, the system fits the cutting machine's operating parameters (including laser power, nozzle air pressure, operating current, vibration frequency, etc.) with the target operating parameters in the control scheme to generate a corresponding baseline status mesh. Then, the real-time collected cutting machine status data is mapped to the monitoring status mesh and compared with the baseline status mesh. If anomalies such as power fluctuations, air pressure imbalances, or current overruns are detected, the system outputs the cutting machine anomaly detection result.

[0028] Similarly, the system performs fitted mesh anomaly detection on the workpiece status data. By constructing the ideal cutting state (such as target temperature distribution, kerf width range, deformation limit value, etc.) set in the workpiece control scheme as the workpiece reference mesh, and comparing it item by item with the monitoring workpiece mesh formed by monitoring the workpiece status data, if there are abnormal temperatures, excessive kerf width, or excessive material deformation, the workpiece anomaly detection result is output.

[0029] The system integrates the results of location anomaly detection, cutting machine anomaly detection, and workpiece anomaly detection to construct a unified cutting anomaly detection vector.

[0030] Furthermore, based on the set of cutting quality incidents, the cutting anomaly detection vector is used to perform cutting quality incident deduction to obtain a first incident deduction result, including:

[0031] Based on the set of cutting quality incidents, feature identification of incident results is performed to determine the top-level features of each incident; the causes of each incident are traced back to the top-level features of each incident based on the set of cutting quality incidents to determine the triggering factors of each incident; logical association is performed between the top-level features of each incident and the triggering factors of each incident to construct a set of cutting quality incident paths; perturbation injection is performed on the set of cutting quality incident paths using an adversarial sample generator to obtain a perturbation domain for cutting quality incidents; adversarial training is performed on the perturbation domain for cutting quality incidents based on a Bayesian network to obtain a cutting quality incident inference network; the cutting anomaly detection vector is input into the cutting quality incident inference network to generate the first incident inference result.

[0032] Specifically, the system accesses a historical database of cutting quality accident events and performs feature analysis on the recorded historical accident samples to extract top-level feature parameters for various types of accidents, such as excessively wide kerf, excessive surface roughness, abnormal slag adhesion, and expansion of the heat-affected zone, thus forming top-level features for each accident. Based on these top-level features, the system traces the causes of each accident feature to identify potential triggering factors, such as insufficient laser power, abnormal nozzle pressure, mismatched cutting speed, and uneven workpiece material. By logically mapping and causal modeling between the top-level features and corresponding triggering factors, the system constructs a set of cutting quality accident paths covering multiple accident chains to characterize possible accident evolution paths. Furthermore, an adversarial example generator is introduced to inject perturbations into the cutting quality accident path set, generating a perturbation domain for cutting quality accidents. Perturbation injection methods include numerical perturbation of accident feature parameters and random perturbation of the probability of triggering factors, thereby constructing more boundary samples and expanding the coverage of accident paths. Next, adversarial training was performed on the perturbation domain of the cutting quality accident based on a Bayesian network. This enabled the network to maintain high inference accuracy and robustness even when facing uncertainties and interference samples, thus training a cutting quality accident inference network. Finally, the cutting anomaly detection vector was input into the cutting quality accident inference network. By comparing the matching degree between the anomaly detection vector and the accident path features, the network output the corresponding accident risk probability and accident evolution trend, thereby generating the first accident inference result.

[0033] Furthermore, based on the cutting machine health accident event set, the cutting anomaly detection vector is used to perform cutting machine health accident inference to obtain a second accident inference result. This includes: calling the cutting machine health accident event set, analyzing historical accident samples, identifying top-level accident features at the equipment level, such as laser power attenuation, nozzle blockage, cooling system failure, and transmission component wear; subsequently, tracing the causes of the above accident features to determine the corresponding triggering factors, such as laser lifespan degradation, insufficient coolant, motor overload, and excessive vibration. Based on the causal mapping relationship between accident features and triggering factors, the system constructs a cutting machine health accident path set, and injects perturbations into the path set through an adversarial example generator to form a health accident perturbation domain. Then, a Bayesian network is used to perform adversarial training on this perturbation domain to obtain a cutting machine health accident inference network. The cutting anomaly detection vector is input into the inference network, and the corresponding equipment health risk probability and evolution trend are output, which is the second accident inference result.

[0034] Furthermore, based on the cutting safety accident event set, the cutting anomaly detection vector is used to perform cutting safety accident inference to obtain the third accident inference result. This includes: calling the cutting safety accident event set, analyzing historical accident samples, identifying top-level accident characteristics at the safety level, such as sparks, gas leaks, local overheating, and excessive dust in the operating area during the cutting process; then tracing the causes of the accidents to identify corresponding triggering factors, such as abnormal gas pipeline sealing, excessive laser energy concentration, insufficient ventilation, and unstable workpiece clamping. Through causal modeling of accident characteristics and triggering factors, the system constructs a cutting safety accident path set and injects perturbations based on an adversarial example generator to form a safety accident perturbation domain. Then, the perturbation domain is adversarially trained using a Bayesian network to obtain the cutting safety accident inference network. Finally, the cutting anomaly detection vector is input into this inference network to obtain the corresponding safety accident risk probability and potential evolution path, which serves as the third accident inference result.

[0035] Based on the workpiece cutting requirements of the metal workpiece, the cutting control scheme is adjusted in multiple degrees of freedom according to the cutting accident projection diagram to obtain the first cutting adjustment space.

[0036] Based on the workpiece cutting requirements of metal workpieces, the system extracts the target cutting parameters, including the workpiece's geometric dimensions, material type, thickness distribution, cutting accuracy level, and surface quality requirements. The system then compares and analyzes these cutting requirements with cutting accident projection maps to identify potential problems in the current cutting control scheme that could lead to substandard cutting quality, increased equipment health risks, or safety hazards. To address these problems, the system dynamically adjusts multiple degrees of freedom parameters in the cutting control scheme. These parameters include at least the cutting head movement path, cutting speed, laser power, nozzle air pressure, pulse frequency, and cutting focal point position. By combining and adjusting these multiple degrees of freedom parameters, a set of candidate parameters covering various possible control configurations is generated, thus constructing the first cutting adjustment space.

[0037] A multidimensional model for cutting evaluation is introduced to perform cutting evaluation and optimization on the first cutting adjustment space to obtain the second cutting adjustment space.

[0038] Furthermore, a multi-dimensional cutting evaluation model is introduced to optimize the first cutting adjustment space, thereby obtaining a second cutting adjustment space, including:

[0039] Based on each cutting adjustment scheme within the first cutting adjustment space, the metal workpiece is virtually cut to obtain cutting fitting data for each scheme; the multi-dimensional cutting evaluation model is activated, which includes a cutting quality evaluation model, a cutting machine health evaluation model, and a cutting safety evaluation model; the cutting fitting data for each scheme is input into the multi-dimensional cutting evaluation model to generate multiple cutting evaluation results; based on the multiple cutting evaluation results, the first cutting adjustment space is optimized and filtered according to the multi-dimensional constraints of the cutting evaluation to generate the second cutting adjustment space.

[0040] Furthermore, the multidimensional constraints of the cutting evaluation include cutting quality evaluation constraints, cutting machine health evaluation constraints, and cutting safety evaluation constraints.

[0041] For each cutting adjustment scheme within the first cutting adjustment space, the system invokes the virtual cutting module to simulate cutting the target metal workpiece. Based on inputs such as cutting path parameters, laser power, air pressure conditions, and workpiece material characteristics, corresponding cutting fitting data is generated. This cutting fitting data reflects characteristics such as the kerf width, heat-affected zone size, residual stress distribution, and equipment operating status that may occur under the corresponding adjustment scheme. Subsequently, the multi-dimensional cutting evaluation model is activated to analyze the cutting fitting data for each scheme. The multi-dimensional cutting evaluation model consists of a cutting quality evaluation model, a cutting machine health evaluation model, and a cutting safety evaluation model. All three are constructed using machine learning methods, enabling quantitative evaluation of the effects of the adjustment schemes across different dimensions. Specifically, the cutting quality evaluation model assesses the accuracy and surface quality of the cutting results; the cutting machine health evaluation model assesses energy consumption, power fluctuations, and component wear during equipment operation; and the cutting safety evaluation model assesses the environmental safety, spark splashing, and potential hazards during the cutting process.

[0042] The cutting quality evaluation model constructs a training set by collecting a large amount of historical cutting sample data. This sample data includes cutting result images and corresponding quality indicators (such as kerf width, surface roughness, slag residue, and heat-affected zone size) under different workpiece materials, thicknesses, cutting speeds, and laser power conditions. During the training phase, algorithms such as Convolutional Neural Networks (CNN), Support Vector Machines (SVM), or Gradient Boosting Trees (GBDT) are used to extract features and perform regression prediction on the input cutting fitting data, thereby outputting quality evaluation coefficients that reflect the degree of cutting quality.

[0043] The health assessment model for the cutting machine is constructed based on equipment operating status data. Training samples include laser output power fluctuation curves, nozzle air pressure changes, operating current loads, vibration spectra, and cooling system temperature curves. Using time-series modeling methods such as Long Short-Term Memory (LSTM), Random Forest, or Bayesian networks, a mapping relationship between equipment operating status and health level is established to predict the health risk level of the cutting machine under the current adjustment scheme, and outputs a health assessment coefficient.

[0044] The cutting safety evaluation model is built by learning from safety-related data during the cutting process. Training samples include spark splash range, workpiece clamping stability, gas leak detection values, ambient dust concentration, and abnormal temperature alarms. Methods such as multilayer perceptron (MLP), ensemble learning, or graph neural network (GNN) are used to extract safety hazard characteristics of the cutting environment and working conditions, and output safety risk coefficients.

[0045] The system sequentially inputs the cutting fitting data corresponding to each adjustment scheme into the multi-dimensional cutting evaluation model, generating multiple cutting evaluation results. Each cutting evaluation result consists of coefficients output by three types of evaluation models, corresponding to the cutting quality evaluation coefficient, the cutting machine health evaluation coefficient, and the cutting safety evaluation coefficient, thus forming a multi-dimensional quantitative index of the adjustment scheme. Further, the system filters these multiple cutting evaluation results based on preset multi-dimensional cutting evaluation constraints. These constraints include cutting quality evaluation constraints, cutting machine health evaluation constraints, and cutting safety evaluation constraints. Each type of constraint has corresponding threshold conditions, such as the kerf width not exceeding a preset upper limit, equipment power fluctuation not exceeding the allowable range, and spark spatter level not exceeding the safety level threshold. By comparing each scheme in the first cutting adjustment space, the system eliminates adjustment schemes that do not meet any of the constraints, retaining only the set of schemes that simultaneously meet all three types of constraints, ultimately generating the second cutting adjustment space.

[0046] A cutting guarantee degree calculation model is introduced to perform cutting guarantee degree analysis and optimization on the second cutting adjustment space to obtain a first cutting guide population that meets the predetermined cutting guarantee degree.

[0047] Furthermore, the cutting assurance calculation model is obtained by weighting the multidimensional indicators of the cutting evaluation multidimensional model. The multidimensional indicators of the cutting evaluation include cutting quality evaluation indicators, cutting machine health evaluation indicators, and cutting safety evaluation indicators.

[0048] Specifically, the cutting assurance calculation model is a function constructed based on the output indicators of the multi-dimensional cutting evaluation model. After completing the scheme evaluation, the multi-dimensional cutting evaluation model outputs three-dimensional indicators for each adjustment scheme: cutting quality evaluation indicator Q (such as the comprehensive score of kerf width, cutting surface roughness, etc.), cutting machine health evaluation indicator H (such as the comprehensive score of equipment power stability, operating load, and component wear), and cutting safety evaluation indicator S (such as the comprehensive score of spark splash level, gas leakage risk, and probability of safety hazards).

[0049] Cutting assurance calculation model: G=αQ+βH+γS, where G represents the cutting assurance value, Q is the cutting quality evaluation index, H is the cutting machine health evaluation index, S is the cutting safety evaluation index, and α, β, and γ are the corresponding weight coefficients, which can be set according to the workpiece cutting requirements, production environment requirements, or industry standards.

[0050] By calculating the cut guarantee degree of all candidate schemes in the second cut adjustment space and comparing it with a predetermined guarantee degree threshold, the system selects a set of schemes with a guarantee degree greater than or equal to the threshold as the first cut guiding population.

[0051] The first cutting guide population performs multiple rounds of breeding and optimization in the second cutting adjustment space to determine the cutting adjustment optimization result, and the metal workpiece is optimized and positioned based on the cutting adjustment optimization result.

[0052] After obtaining the first cutting guide population, the system uses this population as the initial solution set to perform multiple rounds of breeding and optimization in the second cutting adjustment space. Specifically, difference detection is performed within the first cutting guide population to identify key differences between candidate schemes, and a new set of candidate schemes is generated through random mutation and crossover operations to expand the search range. Subsequently, the generated candidate schemes are input into the multi-dimensional cutting evaluation model and the cutting guarantee degree calculation model for joint evaluation, and excellent schemes that meet the predetermined guarantee degree conditions are selected and updated to the new guide population. Through the above-mentioned breeding, evaluation and selection cycle, the system continuously optimizes the distribution of candidate schemes in multiple iterations, so that the population gradually converges towards better cutting quality, stronger equipment health and lower safety risks.

[0053] When the number of iterations reaches a preset threshold or the cutting accuracy converges to near the optimal solution, the system outputs the final cutting adjustment optimization result. This result includes optimized configurations of multiple degrees of freedom parameters such as the cutting path, laser power, cutting speed, nozzle pressure, and focusing position. Based on this result, the system optimizes the positioning control of the metal workpiece, enabling the laser cutting machine to execute according to the optimized parameters during actual cutting. This improves cutting positioning accuracy, reduces the incidence of cutting accidents, and ensures the overall stability and safety of the cutting process.

[0054] Furthermore, based on the first cutting-guided population, multiple rounds of reproductive optimization are conducted in the second cutting-regulation space to determine the cutting-regulation optimization results, including:

[0055] Based on the multidimensional model of cutting evaluation, the second cutting adjustment space is guided to perform reproductive optimization according to the first cutting guidance population to obtain a first cutting reproductive optimization set; the first cutting reproductive optimization set is analyzed and optimized according to the cutting guarantee degree calculation model to obtain a second cutting guidance population that satisfies the predetermined cutting guarantee degree; based on the multidimensional model of cutting evaluation, the first cutting reproductive optimization set is further guided to perform reproductive optimization according to the second cutting guidance population until multiple cutting reproductive optimization sets that satisfy the predetermined number of reproductive optimizations are obtained; the cutting guarantee degree is maximized according to the first cutting guidance population and the multiple cutting reproductive optimization sets to generate the cutting adjustment optimization result.

[0056] The system invokes a multi-dimensional cutting evaluation model, using the first cutting guide population as the initial solution set, and performs reproductive operations on the second cutting adjustment space. Specifically, it detects key differences between different candidate solutions, performs random mutation and crossover operations based on these differences, generates a new set of candidate solutions, and conducts a three-dimensional evaluation of quality, health, and safety using the multi-dimensional cutting evaluation model to form the first cutting reproductive optimization set. Subsequently, the system invokes a cutting guarantee degree calculation model to analyze the cutting guarantee degree of the first cutting reproductive optimization set, calculates the cutting guarantee degree value of each candidate solution, compares it with a preset guarantee degree threshold, and selects the set of solutions with a guarantee degree greater than or equal to the threshold to form the second cutting guide population.

[0057] Based on this, the system again invokes the multidimensional model for cut evaluation, using the second cut-guided population as the new population baseline, to continue guiding the first cut reproductive optimization set through multiple rounds of reproductive iteration. In each iteration, differential feature detection, random mutation, crossover, and multidimensional model evaluation are performed to obtain a new reproductive optimization set. The guarantee degree screening process is repeated until the predetermined number of reproductive optimization attempts or guarantee degree convergence conditions are reached, thereby obtaining multiple cut reproductive optimization sets.

[0058] The system integrates the first cutting guide population with the multiple cutting reproduction optimization sets, calls the cutting guarantee degree calculation model to comprehensively compare all candidate solutions, performs cutting guarantee degree maximization optimization, and selects the optimal solution as the cutting adjustment optimization result.

[0059] Furthermore, based on the aforementioned multidimensional model of cut evaluation, and according to the first cut-guided population, the second cut adjustment space is guided to perform reproductive optimization to obtain the first cut reproductive optimization set, including:

[0060] Based on the first cutting-guided population, difference detection is performed on the second cutting regulation space to obtain a cutting regulation difference feature set; random mutation is performed on the cutting regulation difference feature set to obtain a regulation mutation vector set; crossover mutation is performed on the second cutting regulation space based on the regulation mutation vector set to obtain a first cutting breeding set; cutting evaluation optimization is performed on the first cutting breeding set based on the cutting evaluation multidimensional model to obtain the first cutting breeding optimization set.

[0061] The system uses the first cutting guide population as a reference to perform difference detection on the second cutting adjustment space. Specifically, it calculates the differences in multi-dimensional parameters between each candidate scheme in the first cutting guide population and other candidate schemes in the second cutting adjustment space, such as cutting head trajectory deviation, laser power configuration differences, and nozzle air pressure setting differences, to form a cutting adjustment difference feature set.

[0062] Random mutation operations are performed based on the segmented adjustment difference feature set. Specifically, some parameters in the difference features are perturbed, such as small random adjustments to parameters like power, speed, and air pressure within a preset range, thereby generating an adjustment mutation vector set.

[0063] The second cut regulation space is subjected to cross mutation based on the regulation mutation vector set. Specifically, through parameter recombination and cross operation, the feature vectors of different regulation schemes are partially exchanged and combined to obtain a new set of candidate solutions, namely the first cut breeding set.

[0064] The system invokes a multi-dimensional cutting evaluation model to comprehensively evaluate candidate solutions within the first cutting breeding set. Specifically, based on cutting quality evaluation, cutting machine health evaluation, and cutting safety evaluation, and combined with preset multi-dimensional constraints, candidate solutions that do not meet any of the constraints are eliminated, and candidate solutions that meet the requirements are retained, ultimately forming the first cutting breeding optimization set.

[0065] In summary, the embodiments of this application have at least the following technical effects:

[0066] When a laser cutting machine cuts a metal workpiece according to a cutting control scheme, it simultaneously acquires a cutting monitoring dataset, which includes cutting head positioning data, cutting machine status data, and workpiece status data. Based on this dataset, a multi-dimensional cutting accident simulation is performed to obtain a cutting accident simulation map. Further, based on the workpiece's cutting requirements and the cutting accident simulation map, the cutting control scheme is adjusted with multiple degrees of freedom to obtain a first cutting adjustment space. Subsequently, a multi-dimensional cutting evaluation model is introduced to optimize the first cutting adjustment space, resulting in a second cutting adjustment space. A cutting assurance calculation model is then introduced to analyze and optimize the cutting assurance of the second cutting adjustment space, thus obtaining a first cutting guide population that meets the predetermined cutting assurance. Finally, the second cutting adjustment space is subjected to multiple rounds of breeding optimization based on the first cutting guide population to determine the cutting adjustment optimization result. Based on this result, the metal workpiece is optimized for positioning control. This solves the technical problems of insufficient positioning accuracy and poor stability in metal cutting in existing technologies, achieving the technical effect of improving the positioning accuracy and stability of metal workpiece cutting.

[0067] Example 2, based on the same inventive concept as the positioning control method for metal cutting in the foregoing examples, such as... Figure 2 As shown, this application provides a positioning control system for metal cutting, wherein the system includes:

[0068] Data acquisition module 11: When the laser cutting machine cuts a metal workpiece according to the cutting control scheme, it synchronously acquires a cutting monitoring dataset, which includes cutting head positioning data, cutting machine status data, and workpiece status data; Deduction module 12: Based on the cutting monitoring dataset, it performs multi-dimensional cutting accident deduction to obtain a cutting accident deduction map; Scheme adjustment module 13: Based on the workpiece cutting requirements of the metal workpiece, it adjusts the cutting control scheme with multiple degrees of freedom according to the cutting accident deduction map to obtain a first cutting adjustment space; First optimization module 14: It introduces a multi-dimensional cutting evaluation model to perform cutting evaluation optimization on the first cutting adjustment space to obtain a second cutting adjustment space; Second optimization module 15: It introduces a cutting assurance degree calculation model to perform cutting assurance degree analytical optimization on the second cutting adjustment space to obtain a first cutting guide population that meets the predetermined cutting assurance degree; Positioning control module 16: Based on the first cutting guide population, it performs multiple rounds of breeding optimization on the second cutting adjustment space to determine the cutting adjustment optimization result, and optimizes the positioning control of the metal workpiece according to the cutting adjustment optimization result.

[0069] Furthermore, the deduction module 12 is used to perform the following methods:

[0070] Anomaly detection is performed based on the cutting monitoring dataset to construct a cutting anomaly detection vector; cutting quality accidents are extrapolated from the cutting anomaly detection vector based on the cutting quality accident event set to obtain a first accident extrapolation result; cutting machine health accidents are extrapolated from the cutting machine health accident event set to obtain a second accident extrapolation result; cutting safety accidents are extrapolated from the cutting anomaly detection vector based on the cutting safety accident event set to obtain a third accident extrapolation result; and a cutting accident extrapolation map is constructed based on the first accident extrapolation result, the second accident extrapolation result, and the third accident extrapolation result.

[0071] Furthermore, the deduction module 12 is used to perform the following methods:

[0072] According to the cutting control scheme, the cutting head positioning is fitted to construct a reference positioning dynamic mesh; the cutting head positioning data is meshed to obtain a monitoring positioning dynamic mesh; anomaly detection is performed on the monitoring positioning dynamic mesh based on the reference positioning dynamic mesh to obtain a positioning anomaly detection result; anomaly detection is performed on the cutting machine status data based on the cutting control scheme to obtain a cutting machine anomaly detection result; anomaly detection is performed on the workpiece status data based on the cutting control scheme to obtain a workpiece anomaly detection result; and the cutting anomaly detection vector is generated by combining the positioning anomaly detection result and the cutting machine anomaly detection result.

[0073] Furthermore, the deduction module 12 is used to perform the following methods:

[0074] Based on the set of cutting quality incidents, feature identification of incident results is performed to determine the top-level features of each incident; the causes of each incident are traced back to the top-level features of each incident based on the set of cutting quality incidents to determine the triggering factors of each incident; logical association is performed between the top-level features of each incident and the triggering factors of each incident to construct a set of cutting quality incident paths; perturbation injection is performed on the set of cutting quality incident paths using an adversarial sample generator to obtain a perturbation domain for cutting quality incidents; adversarial training is performed on the perturbation domain for cutting quality incidents based on a Bayesian network to obtain a cutting quality incident inference network; the cutting anomaly detection vector is input into the cutting quality incident inference network to generate the first incident inference result.

[0075] Furthermore, the first optimization module 14 is used to perform the following method:

[0076] Based on each cutting adjustment scheme within the first cutting adjustment space, the metal workpiece is virtually cut to obtain cutting fitting data for each scheme; the multi-dimensional cutting evaluation model is activated, which includes a cutting quality evaluation model, a cutting machine health evaluation model, and a cutting safety evaluation model; the cutting fitting data for each scheme is input into the multi-dimensional cutting evaluation model to generate multiple cutting evaluation results; based on the multiple cutting evaluation results, the first cutting adjustment space is optimized and filtered according to the multi-dimensional constraints of the cutting evaluation to generate the second cutting adjustment space.

[0077] Furthermore, the positioning control module 16 is used to perform the following methods:

[0078] Based on the multidimensional model of cutting evaluation, the second cutting adjustment space is guided to perform reproductive optimization according to the first cutting guidance population to obtain a first cutting reproductive optimization set; the first cutting reproductive optimization set is analyzed and optimized according to the cutting guarantee degree calculation model to obtain a second cutting guidance population that satisfies the predetermined cutting guarantee degree; based on the multidimensional model of cutting evaluation, the first cutting reproductive optimization set is further guided to perform reproductive optimization according to the second cutting guidance population until multiple cutting reproductive optimization sets that satisfy the predetermined number of reproductive optimizations are obtained; the cutting guarantee degree is maximized according to the first cutting guidance population and the multiple cutting reproductive optimization sets to generate the cutting adjustment optimization result.

[0079] Furthermore, the positioning control module 16 is used to perform the following methods:

[0080] Based on the first cutting-guided population, difference detection is performed on the second cutting regulation space to obtain a cutting regulation difference feature set; random mutation is performed on the cutting regulation difference feature set to obtain a regulation mutation vector set; crossover mutation is performed on the second cutting regulation space based on the regulation mutation vector set to obtain a first cutting breeding set; cutting evaluation optimization is performed on the first cutting breeding set based on the cutting evaluation multidimensional model to obtain the first cutting breeding optimization set.

[0081] Furthermore, the first optimization module 14 is used to perform the following method:

[0082] The multidimensional constraints for cutting evaluation include cutting quality evaluation constraints, cutting machine health evaluation constraints, and cutting safety evaluation constraints.

[0083] Furthermore, the second optimization module 15 is used to perform the following method:

[0084] The cutting assurance calculation model is obtained by weighting the multidimensional indicators of the cutting evaluation multidimensional model. The multidimensional indicators of the cutting evaluation include cutting quality evaluation indicators, cutting machine health evaluation indicators, and cutting safety evaluation indicators.

[0085] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0086] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0087] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A positioning control method for metal cutting, characterized by, The method comprises: When the laser cutting machine cuts the metal workpiece according to the cutting control scheme, a cutting monitoring data set is synchronously acquired, the cutting monitoring data set comprising cutting head positioning data, cutting machine state data and workpiece state data; Multi-dimensional cutting accident deduction is performed according to the cutting monitoring data set, and a cutting accident deduction graph is obtained; Based on the workpiece cutting requirements of the metal workpiece, the cutting control scheme is adjusted in multiple degrees of freedom according to the cutting accident deduction graph, and a first cutting adjustment space is obtained; A cutting evaluation multi-dimensional model is introduced to perform cutting evaluation optimization on the first cutting adjustment space, and a second cutting adjustment space is obtained; A cutting guarantee degree calculation model is introduced to perform cutting guarantee degree analysis optimization on the second cutting adjustment space, and a first cutting guide population satisfying a predetermined cutting guarantee degree is obtained; According to the first cutting guide population, the second cutting adjustment space is subjected to multi-round reproduction optimization, a cutting adjustment optimization result is determined, and the metal workpiece is subjected to optimized positioning control according to the cutting adjustment optimization result; A cutting evaluation multi-dimensional model is introduced to perform cutting evaluation optimization on the first cutting adjustment space, and a second cutting adjustment space is obtained, comprising: According to each cutting adjustment scheme in the first cutting adjustment space, the metal workpiece is subjected to virtual cutting respectively, and cutting fitting data of each scheme is obtained; The cutting evaluation multi-dimensional model is activated, and the cutting evaluation multi-dimensional model comprises a cutting quality evaluation model, a cutting machine health evaluation model and a cutting safety evaluation model; The cutting fitting data of each scheme is input into the cutting evaluation multi-dimensional model, and a plurality of cutting evaluation results are generated; Based on the plurality of cutting evaluation results, the first cutting adjustment space is screened according to cutting evaluation multi-dimensional constraint optimization, and the second cutting adjustment space is generated; According to the first cutting guide population, the second cutting adjustment space is subjected to multi-round reproduction optimization, a cutting adjustment optimization result is determined, comprising: Based on the cutting evaluation multi-dimensional model, the second cutting adjustment space is guided to perform reproduction optimization according to the first cutting guide population, and a first cutting reproduction optimization set is obtained; According to the cutting guarantee degree calculation model, the first cutting reproduction optimization set is subjected to cutting guarantee degree analysis optimization, and a second cutting guide population satisfying the predetermined cutting guarantee degree is obtained; Based on the cutting evaluation multi-dimensional model, the first cutting reproduction optimization set is continuously guided to perform reproduction optimization according to the second cutting guide population, until a plurality of cutting reproduction optimization sets satisfying a predetermined reproduction optimization number are obtained; According to the first cutting guide population and the plurality of cutting reproduction optimization sets, cutting guarantee degree maximization optimization is performed, and the cutting adjustment optimization result is generated.

2. The positioning control method for metal cutting according to claim 1, wherein, Multi-dimensional cutting accident deduction is performed according to the cutting monitoring data set, and a cutting accident deduction graph is obtained, comprising: Abnormality detection is performed according to the cutting monitoring data set, and a cutting abnormality detection vector is constructed; Cutting quality accident deduction is performed on the cutting abnormality detection vector according to a cutting quality accident event set, and a first accident deduction result is obtained; cutting machine health accident event set, to obtain a second accident deduction result; cutting safety accident event set, to obtain a third accident deduction result; According to the first accident deduction result, the second accident deduction result and the third accident deduction result, the cutting accident deduction graph is constructed.

3. The positioning control method for metal cutting according to claim 2, wherein According to the cutting monitoring data set, an abnormality detection is performed to construct a cutting abnormality detection vector, including: According to the cutting control scheme, a cutting head positioning fitting is performed to construct a reference positioning dynamic grid; According to the cutting head positioning data, a grid processing is performed to obtain a monitoring positioning dynamic grid; According to the reference positioning dynamic grid, an abnormality detection is performed on the monitoring positioning dynamic grid to obtain a positioning abnormality detection result; According to the cutting control scheme, a fitting grid abnormality detection is performed on the cutting machine state data to obtain a cutting machine abnormality detection result; According to the cutting control scheme, a fitting grid abnormality detection is performed on the workpiece state data to obtain a workpiece abnormality detection result, and the cutting abnormality detection vector is generated by combining the positioning abnormality detection result and the cutting machine abnormality detection result.

4. The positioning control method for metal cutting according to claim 2, wherein According to the cutting quality accident event set, a cutting quality accident deduction is performed on the cutting abnormality detection vector to obtain a first accident deduction result, including: According to the cutting quality accident event set, an accident result feature recognition is performed to determine each accident top-level feature; According to the cutting quality accident event set, a cause tracing is performed on the each accident top-level feature to determine each accident trigger factor; According to the each accident top-level feature and the each accident trigger factor, a logical correlation is performed to construct a cutting quality accident path set; According to the cutting quality accident path set, a disturbance injection is performed by an adversarial sample generator to obtain a cutting quality accident disturbance domain; Based on the Bayesian network, an adversarial training is performed on the cutting quality accident disturbance domain to obtain a cutting quality accident deduction network; The cutting abnormality detection vector is input into the cutting quality accident deduction network to generate the first accident deduction result.

5. The positioning control method for metal cutting according to claim 1, wherein, Based on the cutting evaluation multi-dimensional model, the second cutting adjustment space is guided to reproduce and optimize according to the first cutting guide population to obtain a first cutting reproduction and optimization set, including: According to the first cutting guide population, a difference detection is performed on the second cutting adjustment space to obtain a cutting adjustment difference feature set; According to the cutting adjustment difference feature set, a random variation is performed to obtain a variation vector set; According to the variation vector set, a cross variation is performed on the second cutting adjustment space to obtain a first cutting reproduction set; According to the cutting evaluation multi-dimensional model, a cutting evaluation optimization is performed on the first cutting reproduction set to obtain the first cutting reproduction and optimization set.

6. The positioning control method for metal cutting of claim 1, wherein, The cutting evaluation multi-dimensional constraint includes cutting quality evaluation constraint, cutting machine health evaluation constraint and cutting safety evaluation constraint.

7. The positioning control method for metal cutting according to claim 1, wherein According to the cutting evaluation multi-dimensional index of the cutting evaluation multi-dimensional model, weight distribution is performed, and a cutting guarantee degree calculation model is obtained, the cutting evaluation multi-dimensional index including a cutting quality evaluation index, a cutting machine health evaluation index, and a cutting safety evaluation index.

8. Positioning control system for metal cutting, characterized by, The system comprises: A data acquisition module: when a laser cutting machine cuts a metal workpiece according to a cutting control scheme, a cutting monitoring data set is synchronously acquired, the cutting monitoring data set including cutting head positioning data, cutting machine state data, and workpiece state data; An inference module: according to the cutting monitoring data set, a multi-dimensional cutting accident inference is performed, and a cutting accident inference graph is obtained; A scheme adjustment module: based on the workpiece cutting requirements of the metal workpiece, according to the cutting accident inference graph, a multi-degree-of-freedom adjustment is performed on the cutting control scheme, and a first cutting adjustment space is obtained; A first optimization module: a cutting evaluation multi-dimensional model is introduced to perform cutting evaluation optimization on the first cutting adjustment space, and a second cutting adjustment space is obtained; A second optimization module: a cutting guarantee degree calculation model is introduced to perform cutting guarantee degree analytical optimization on the second cutting adjustment space, and a first cutting guide population meeting a predetermined cutting guarantee degree is obtained; A positioning control module: according to the first cutting guide population, a multi-round reproduction optimization is performed on the second cutting adjustment space, a cutting adjustment optimization result is determined, and the metal workpiece is subjected to optimized positioning control according to the cutting adjustment optimization result.

Images