Laser cutting method, device, medium and program product

Abstract

The embodiment of the invention provides a laser cutting method and device, a medium and a program product, and relates to the technical field of laser machining. The method comprises the steps that plate parameters of a target plate and finished product parameters of a cutting piece are input into a prediction model obtained based on historical cutting data training, and cutting parameters are generated; the cutting parameters are input into a multi-physical-field coupled simulation model, the laser cutting process is simulated, and the cutting effect is verified; and when the cutting effect obtained by simulating laser cutting based on the simulation model meets a preset condition, actual cutting is conducted on the target plate based on the cutting parameters. Based on the method, the step of trial cutting with real materials can be omitted, and the problem of raw material waste during laser cutting can be solved.

Description

Laser cutting methods, equipment, media and process products Technical Field

[0001] This application relates to the field of laser processing technology, and in particular to a laser cutting method, equipment, medium and program product. Background Technology

[0002] A laser cutting machine is a sheet metal cutting device widely used in industrial manufacturing. It can be used to cut metal sheets and other materials. For example, in the production of furniture such as office desks and chairs, a laser cutting machine is used to cut raw metal sheets into pre-defined shapes for assembly or decoration.

[0003] A laser cutting machine typically includes components such as a laser emitter, optical fiber, laser gun (cutting head), and a worktable. For example, a metal sheet is first placed on the worktable, and the worktable moves to feed the metal sheet into the laser cutting machine. After the laser emitter generates laser light, it is transmitted to the laser gun head via optical fiber. The laser gun head focuses the laser light and emits a high-energy laser beam. By moving the laser gun head above the metal sheet, the high-energy laser beam can cut a pre-defined shape into the metal sheet.

[0004] To ensure the precision of the finished product meets requirements, existing laser cutting methods involve several trial cuts before the actual cutting process. These trial cuts are used to adjust the cutting parameters of the laser cutting machine. However, since trial cuts consume some raw materials, the existing method suffers from material waste. Summary of the Invention

[0005] This application provides a laser cutting method, equipment, medium, and program product, which reduces the need for trial cutting with real materials when laser cutting materials, thereby solving the problem of material waste caused by real trial cutting.

[0006] In a first aspect, embodiments of this application provide a laser cutting method, the method comprising:

[0007] Input the plate parameters of the target plate and the finished product parameters of the cut part into the prediction model trained based on historical cutting data to generate cutting parameters;

[0008] The cutting parameters are input into a multiphysics coupled simulation model to simulate the laser cutting process and verify the cutting effect.

[0009] When the cutting effect obtained by simulating laser cutting based on the simulation model meets the preset conditions, the target plate is actually cut based on the cutting parameters.

[0010] In one possible implementation, the method further includes:

[0011] When the cutting effect obtained by simulating laser cutting based on the simulation model does not meet the preset conditions, the cutting parameters are optimized by reinforcement learning algorithm, and the optimized cutting parameters are input into the multi-physics coupled simulation model to simulate the laser cutting process and verify the cutting effect.

[0012] In one possible implementation, simulating the laser cutting process and verifying the cutting effect includes:

[0013] In the simulated laser cutting process, the width of the heat-affected zone on both sides of the cut is determined based on the coupling analysis of the microscopic and macroscopic thermal fields introduced in the simulation model.

[0014] The cutting effect was verified based on the width of the heat-affected zone.

[0015] In one possible implementation, the target sheet material is actually cut based on cutting parameters, including:

[0016] During the actual cutting process, multimodal data is acquired, including at least two of the following: temperature data, optical data, acoustic data, and image data.

[0017] Based on multimodal data, the cutting parameters are dynamically adjusted so that the actual cutting can be performed according to the dynamically adjusted cutting parameters.

[0018] In one possible implementation, the multimodal data includes plasma analysis data for determining the plasma density distribution; based on the multimodal data, the cutting parameters are dynamically adjusted, including:

[0019] Acquire plasma analysis data obtained within a preset range of the real-time monitoring laser gun head;

[0020] The plasma density distribution within a preset range of the laser gun head is determined in real time through plasma analysis data.

[0021] The laser power and / or auxiliary gas flow rate in the cutting parameters are dynamically adjusted according to the plasma density distribution to cope with the plasma shielding effect within the preset range of the laser gun head.

[0022] In one possible implementation, the method further includes:

[0023] Based on multimodal data collected during multiple actual cutting processes, and real-time cutting parameters during multiple cutting processes, a fine-tuning training dataset is constructed.

[0024] Based on the fine-tuning training dataset, the prediction model is fine-tuned using at least one model training method to update the model parameters of the prediction model.

[0025] In one possible implementation, the simulation model of multiphysics coupling includes the multiphysics coupling effects between lasers, materials, gases and devices, and covers the dynamic behavior of thermal fields, plasma fields and molten pool flows.

[0026] Secondly, embodiments of this application provide a laser cutting apparatus, the apparatus comprising:

[0027] The generation module is used to input the board parameters of the target board and the finished product parameters of the cut part into the prediction model trained based on historical cutting data to generate cutting parameters;

[0028] The simulation module is used to input cutting parameters into a multiphysics coupled simulation model to simulate the laser cutting process and verify the cutting effect;

[0029] The cutting module is used to actually cut the target material based on the cutting parameters when the cutting effect obtained by simulating laser cutting based on the simulation model meets the preset conditions.

[0030] In one possible implementation, the apparatus further includes an optimization module, which is used to:

[0031] When the cutting effect obtained by simulating laser cutting based on the simulation model does not meet the preset conditions, the cutting parameters are optimized by reinforcement learning algorithm, and the optimized cutting parameters are input into the multi-physics coupled simulation model to simulate the laser cutting process and verify the cutting effect.

[0032] In one possible implementation, the simulation module is specifically used for:

[0033] In the simulated laser cutting process, the width of the heat-affected zone on both sides of the cut is determined based on the coupling analysis of the microscopic and macroscopic thermal fields introduced in the simulation model.

[0034] The cutting effect was verified based on the width of the heat-affected zone.

[0035] In one possible implementation, the cutting module is specifically used for:

[0036] During the actual cutting process, multimodal data is acquired, including at least two of the following: temperature data, optical data, acoustic data, and image data.

[0037] Based on multimodal data, the cutting parameters are dynamically adjusted so that the actual cutting can be performed according to the dynamically adjusted cutting parameters.

[0038] In one possible implementation, the multimodal data includes plasma analysis data for determining the plasma density distribution; the cutting module is specifically used for:

[0039] Acquire plasma analysis data obtained within a preset range of the real-time monitoring laser gun head;

[0040] The plasma density distribution within a preset range of the laser gun head is determined in real time through plasma analysis data.

[0041] The laser power and / or auxiliary gas flow rate in the cutting parameters are dynamically adjusted according to the plasma density distribution to cope with the plasma shielding effect within the preset range of the laser gun head.

[0042] In one possible implementation, the device further includes a fine-tuning module, which is used for:

[0043] Based on multimodal data collected during multiple actual cutting processes, and real-time cutting parameters during multiple cutting processes, a fine-tuning training dataset is constructed.

[0044] Based on the fine-tuning training dataset, the prediction model is fine-tuned using at least one model training method to update the model parameters of the prediction model.

[0045] In one possible implementation, the simulation model of multiphysics coupling includes the multiphysics coupling effects between lasers, materials, gases and devices, and covers the dynamic behavior of thermal fields, plasma fields and molten pool flows.

[0046] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0047] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0048] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0049] The laser cutting method, equipment, medium, and program products provided in this application pre-trained a prediction model to generate cutting parameters before actual cutting, based on the material parameters of the target sheet and the finished product parameters of the part to be cut. These cutting parameters are not directly applied to the raw material for actual cutting; instead, the laser cutting process is simulated in a multi-physics coupled simulation model based on these parameters. This simulated laser cutting process not only verifies the usability of the model-generated cutting parameters but also avoids consuming real raw materials. Therefore, when the cutting effect obtained from the simulated laser cutting meets preset conditions, the target sheet is then actually cut based on these cutting parameters. This eliminates the need for trial cutting with real materials, solving the problem of raw material waste. Attached Figure Description

[0050] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0051] Figure 1 is a schematic flowchart of the laser cutting method provided in an embodiment of this application;

[0052] Figure 2 is a schematic diagram of the prediction model obtained through training according to an embodiment of this application;

[0053] Figure 3 is a schematic diagram of multiphysics coupling provided in an embodiment of this application;

[0054] Figure 4 is a schematic diagram of the structure of the laser cutting device provided in the embodiment of this application;

[0055] Figure 5 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0056] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0057] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0058] In this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0059] In the embodiments of this application, the use of terms such as "first" and "second" is to distinguish between identical or similar items that have essentially the same function and effect. For example, "first electronic device" and "second electronic device" are merely used to distinguish different electronic devices and do not limit their order of execution. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" do not necessarily imply that they are different.

[0060] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following associated objects have an "or" relationship.

[0061] For example, laser cutting methods can be widely used in scenarios involving the processing of metal raw materials. In industrial manufacturing, for instance, laser cutting can efficiently and precisely cut metal sheets to obtain finished products or parts requiring further processing. Similarly, in the manufacturing of home or office furniture, automotive parts processing, and the production of precision instruments and equipment, laser cutting is used to cut large metal sheets (such as stainless steel and aluminum alloys) into parts of specific shapes and preset sizes.

[0062] For example, in the furniture manufacturing scenario of producing office desks and chairs, a 3000mm × 1500mm rectangular stainless steel sheet needs to be cut into multiple circular pieces with a radius of 200mm for use as the metal frame or decorative parts of the furniture. As another example, in the automotive parts processing scenario, metal sheets requiring high-precision cuts (such as door reinforcements and radiator frames) need to be processed using laser cutting to produce metal sheets that meet the required precision.

[0063] Laser cutting machines achieve non-contact cutting using high-energy laser beams in these scenarios, offering advantages such as high cutting speed, high kerf quality, and the ability to process complex geometries. However, existing laser cutting methods face challenges in real-world raw material testing and cutting parameter optimization, limiting further improvements in raw material consumption control, production efficiency, and cost-effectiveness.

[0064] A laser cutting machine may include a machine housing, an exchange worktable (such as worktable 1 and worktable 2) that alternately transports the sheet metal to be cut into the machine housing, a laser gun head for cutting the sheet metal, and transverse and longitudinal tracks that drive the laser gun head. The laser cutting machine also includes a laser emitter that provides the laser source to the laser gun head, as well as built-in focusing lenses, laser protection plates, slide rails, and lead screws. Furthermore, the laser cutting machine includes a water cooling system for cooling the laser gun head, and a control box, circuit boards, servo mechanisms, and an operation panel for operating and controlling the laser cutting machine.

[0065] The internal housing of a laser cutting machine can accommodate components such as worktables for alternately transporting raw material sheets and laser gun heads. For example, worktables 1 and 2 can alternately transport the sheet material to be cut, improving continuous operation efficiency. The laser system includes a laser emitter (such as a fiber laser), focusing lens, protective plate, slide rail, lead screw, and laser gun head, used to precisely control the position and focus of the laser beam. The cooling system includes a water-cooling device to cool the laser gun head and prevent overheating damage. The control system, which controls the operation of the laser gun head, includes an equipment control box, circuit boards, servo mechanisms, and an operation panel, used to set cutting parameters (such as laser power, cutting speed, focus position, and auxiliary gas parameters).

[0066] Because the sheet metal parameters used in different processing scenarios may vary, as may the required size and shape of the parts to be cut, different processing scenarios require different cutting parameters. Existing laser cutting methods, before actually cutting the parts, will first perform at least one round of trial cutting on the sheet metal to determine the cutting parameters.

[0067] For example, a sheet of raw metal is placed on a worktable and fed into the housing of a laser cutting machine. Based on historical cutting data and manual operating experience, a set of cutting parameters is first set for trial cutting. After the trial cut, the cut parts are observed and measured to see if they meet the design requirements. If not, some or all of the parameters in the first set of cutting parameters need to be adjusted to obtain a second set of cutting parameters, and trial cutting is performed again based on the second set of cutting parameters. This process is repeated one or more times until the cut parts meet the design requirements before mass production of formal cutting begins.

[0068] For example, when cutting stainless steel sheets, multiple trial cuts must be performed using the raw metal sheet to adjust cutting parameters such as laser power and gas flow rate until a satisfactory cut is obtained. This process not only increases production costs but also extends the cutting preparation time, and the problem of wasting raw materials is particularly prominent in small-batch, multi-variety production scenarios.

[0069] It is evident that existing laser cutting methods suffer from raw material waste because they cannot overcome the need for trial cutting of real materials.

[0070] In view of this, embodiments of this application provide a laser cutting method. Before actual cutting, this method first predicts parameters using a pre-trained prediction model based on the material parameters of the target sheet and the finished product parameters of the part to be cut, thereby generating cutting parameters. These cutting parameters are not directly applied to the raw material for actual cutting; instead, the laser cutting process is simulated in a multiphysics coupling simulation model based on these parameters. This simulated laser cutting process not only verifies the usability of the model-generated cutting parameters but also avoids consuming real raw materials. Therefore, when the cutting effect obtained from the simulated laser cutting meets the preset conditions, the target sheet is then actually cut based on these cutting parameters. This eliminates the need for trial cutting with real materials, solving the problem of raw material waste.

[0071] The technical solutions of this application will be described in detail below with reference to specific embodiments. The specific embodiments described below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0072] Figure 1 is a flowchart illustrating the laser cutting method provided in an embodiment of this application. The execution entity of this method can be an electronic device with corresponding data storage and computing capabilities, such as a processor, computer, server, or laser cutting controller. As shown in Figure 1, the method includes:

[0073] S101: Input the plate parameters of the target plate and the finished product parameters of the cut part into the prediction model trained based on historical cutting data to generate cutting parameters.

[0074] For example, the target sheet material can be understood as the sheet material to be cut, such as any metal material within the maximum cutting size range of the laser cutting machine. Sheet material parameters can be understood as the inherent properties of the target sheet material, such as the material (stainless steel, carbon steel, or aluminum alloy), thickness (3mm, 5mm, or 10mm), thermal conductivity, melting point, and surface condition (whether it has a coating). Sheet material parameters influence the cutting difficulty and energy requirements to a certain extent, and are helpful in determining cutting parameters such as laser power and energy output.

[0075] A cut part refers to the finished product obtained after cutting. The finished product parameters of a cut part can be understood as parameters that characterize the desired effect after cutting, such as the parameters corresponding to the quality standards of the cut part. Examples of finished product parameters include the dimensions of the cut part, dimensional tolerances (e.g., ±0.1mm), surface roughness (e.g., Ra≤1.5μm), whether burrs or oxide discoloration are allowed, and the length of the cutting path.

[0076] Figure 2 is a schematic diagram of the prediction model obtained through training according to an embodiment of this application. Historical cutting data can be understood as empirical data obtained from cutting raw materials in the past. Historical cutting data includes, for example, plate parameters, finished product parameters, corresponding optimal cutting parameters, and cutting quality results from historical cutting. As shown in Figure 2, a training dataset can be constructed based on multiple sets of historical cutting data. The training dataset includes N training samples, where N can be any positive integer.

[0077] For example, any training sample may include sample data such as sheet material parameters, finished product parameters, corresponding superior or inferior cutting parameters, and cutting quality results. For example, a training sample may be "Sheet material parameters: rectangular stainless steel + thickness 3mm + tolerance ±0.1mm + no oxide color → Finished product parameters: round + diameter 200mm → Cutting parameters: laser power 3000W + cutting speed 1.2m / min + nitrogen pressure 8bar → Cutting quality: shape qualified + cross-section qualified".

[0078] An initial predictive model can be trained using a training dataset to obtain a generative predictive model. A predictive model can be understood as an algorithmic model trained based on machine learning or deep learning. Types of predictive models include, for example, regression models, random forests, or neural networks.

[0079] Predictive models can learn the mapping relationship between input parameters (such as sheet material parameters and finished product parameters) and output parameters (such as cutting parameters) in historical cutting data through model training. After at least one round of model training, the model parameters are iterated until the model converges, thus the model has learned certain predictive experience and obtained a predictive model. This predictive model can be used for inference and prediction in real-world scenarios to generate the corresponding cutting parameters for the target sheet material and the finished product parameters of the cut part.

[0080] Cutting parameters can be understood as the final configuration parameters that can be used to actually cut the target material. After configuring the cutting parameters in a laser cutting machine, operating instructions for the machine are generated. These instructions control the laser head to perform laser cutting. Different laser cutting machines may include different types of cutting parameters. Examples of cutting parameters include: laser power, cutting speed, type of auxiliary gas (oxygen or nitrogen), gas pressure, focal point position, and pulse frequency.

[0081] For example, supervised learning can be used to train an initial prediction model that can predict and generate cutting parameters based on sheet material parameters and finished product parameters. When training a supervised learning model, multiple sets of historical cutting data can be collected first to construct a training dataset. The training samples in the training dataset are then labeled as positive and negative samples. Positive samples represent training samples corresponding to acceptable cutting quality, while negative samples represent training samples corresponding to unacceptable cutting quality.

[0082] The model parameters (network parameters) of the initial prediction model can be optimized using a loss function (such as cross-entropy) and backpropagation. Training ends when the prediction accuracy of the initial prediction model reaches a preset threshold, resulting in the final prediction model. The trained prediction model can be integrated into the control system of a laser cutting machine for model deployment. During the cutting operation, the prediction model can be invoked to generate cutting parameters.

[0083] Using data-driven methods to predict cutting parameters can avoid the consumption of raw materials through physical trial cutting, reduce the number of trial cuts, and save on raw material costs. Because the prediction model learns from a large amount of data and obtains mapping patterns from historical cutting data, it accumulates experience in predicting cutting parameters. Therefore, it can cover various sheet material parameters and finished product parameters, and the accuracy of the predicted cutting parameters is also high.

[0084] S102, input the cutting parameters into the multi-physics coupled simulation model to simulate the laser cutting process and verify the cutting effect.

[0085] For example, multiphysics coupling can be understood as the mutual coupling between various physical phenomena involved in laser cutting technology. For instance, laser heating (thermal field) can melt the raw material, the auxiliary gas can blow away the molten slag (flow field), and the heat generated by the laser can cause the material to deform (solid mechanical field).

[0086] It is evident that the thermal field, flow field, and solid mechanics field simultaneously influence the laser cutting process, and these effects interact with each other. Therefore, this simulation model, by considering the coupling effects of multiple physics fields, more closely approximates reality and can realistically simulate the laser cutting scenario and process. It should be understood that the coupled multiphysics fields, in addition to the thermal field, flow field, and solid mechanics field, can also include other physical fields, such as electromagnetic fields and plasma fields.

[0087] Virtual laser cutting scenarios can be built using professional simulation tools (such as simulation software). These scenarios can incorporate the physical laws governing the interaction between lasers and materials (such as heat conduction and fluid flow formulas), thus highly replicating the physical process of laser cutting of sheet materials in the real world.

[0088] Constructing a high-precision multiphysics coupled simulation model can accurately simulate the physical process of laser cutting and verify the feasibility of the predicted cutting parameters. When building the simulation model, a basic model can be built first. For example, laser system modeling can be performed, including simulating optical characteristics such as laser power, wavelength, and spot distribution. Plate material modeling can be performed, including modeling thermophysical parameters such as thermal conductivity, absorptivity, and phase transition characteristics. Nozzle and gas modeling can be performed, including simulating the effects of nozzle diameter, gas type, and flow rate on slag removal.

[0089] Following this, core physics coupling models can be built. For example, thermal field modeling can be performed, including simulating laser energy transfer, material melting and vaporization processes, and thermal deformation. Plasma field modeling can be performed, including simulating plasma generation and energy shielding effects induced by high-power lasers. Molten pool flow modeling can be performed, including simulating the dynamic behavior of the molten pool based on the Navier-Stokes equations.

[0090] In addition, equipment and environment models can be built. For example, equipment status modeling can be performed, including simulating changes in equipment status such as laser power fluctuations and optical component contamination. Environmental interference modeling can also be performed, including simulating the effects of workshop temperature, humidity, and airflow on the cutting process.

[0091] For example, if the constructed simulation model lacks the multi-physics coupling effect of at least one of the laser, materials, gas and equipment, it will be difficult to simulate the actual cutting process with high fidelity, which will reduce the credibility of the simulation verification of cutting parameters.

[0092] In one possible implementation, the simulation model of multiphysics coupling includes the multiphysics coupling effects between lasers, materials, gases and devices, and covers the dynamic behavior of thermal fields, plasma fields and molten pool flows.

[0093] For example, the laser can be understood as the heat source for cutting; the material can be understood as the object being cut, such as a metal sheet; and the gas can be understood as the auxiliary medium for cutting, such as oxygen or nitrogen, which can be used for slag removal, cooling, or combustion. The equipment can be understood as the various devices used in the laser cutting process, such as the laser gun head and the worktable. The multiphysics coupling effect between the laser, material, gas, and equipment can be understood as the interaction between the laser, material, gas, and equipment.

[0094] Figure 3 is a schematic diagram of multiphysics coupling provided in an embodiment of this application. As shown in Figure 3, the thermal field can be understood as the spatial distribution of temperature between the material and its surrounding environment and its change over time. The thermal field is the core energy field driving melting, vaporization, and phase change. For example, the high-temperature region at the center of the laser spot and the temperature gradient conducted into the interior of the plate both form thermal fields.

[0095] The plasma field can be understood as the ionized gas region formed after a material is heated to vaporization by a laser, which can influence the thermal field distribution. The dynamic behavior of the molten pool flow can be understood as the flow process of the liquid metal region (i.e., the molten pool) formed after the material melts under the influence of laser force, gas pressure, gravity, etc. For example, being blown towards the bottom of the cut by gas, or the flow trajectory moving with the laser. The dynamic behavior of the molten pool flow can determine the cut quality, such as whether there are burrs and the smoothness of the cut.

[0096] For example, the process of "laser heating of materials, material vaporization to generate plasma, plasma absorbing laser energy, gas blowing away plasma and slag, and equipment movement changing the laser's position" forms a dynamic closed loop of thermal field, plasma field, and molten pool flow, with multiple physical fields forming a coupled interaction.

[0097] By establishing mathematical models and coupling relationships for each physical field, simulation models can be constructed. For example, based on the heat conduction equation, combined with laser energy input, latent heat of material phase transition, plasma radiation heat transfer, and gas convection heat dissipation, a thermal field model can be built. Based on plasma dynamics equations, a plasma field model describing generation, transport, and interaction with the laser can be constructed. Based on Stokes' equations, combined with the driving force of gas pressure, gravity, and thermocapillary forces on the flow of liquid metal, a molten pool flow model can be constructed.

[0098] By defining the coupling relationships between various physical fields, simulation models can be constructed that encompass the multiphysics coupling effects between lasers, materials, gases, and devices, and cover the dynamic behavior of thermal fields, plasma fields, and molten pool flows. For example, the coupling relationship between the thermal field and the plasma field can be defined as: temperature rise triggers material vaporization and ionization. The coupling relationship between the thermal field and the dynamic behavior of molten pool flows can be defined as: temperature gradient generates thermocapillary forces. Based on the individual physical fields and the coupling relationships between them, data modeling and geometric modeling can be performed in simulation tools to construct simulation models.

[0099] In this embodiment, the constructed multiphysics coupling simulation model includes the multiphysics coupling effects between laser, material, gas, and device, and covers the dynamic behavior of thermal field, plasma field, and molten pool flow. Based on this, the feasibility of the generated cutting parameters can be effectively simulated and verified through the multiphysics coupling process between laser, material, gas, and device, improving simulation accuracy and reliability. Thus, high-fidelity simulation can identify problems with cutting parameters in advance, such as insufficient laser power and improper gas flow rate, reducing the failure rate of actual cutting. Simulating the thermal field distribution can also suppress thermal deformation, reducing the error in cut perpendicularity and improving cutting quality.

[0100] For example, when building simulation models, whether it's a basic model, a core model, or an environment model, pre-built modular models can be used for rapid construction. For instance, a modular model library can be pre-built in the simulation software; this library includes a collection of multiple mature models. For example, the modular model library might pre-contain models such as laser cutting machines, various sheet metal models, and laser cutting environment models. When building multiphysics-coupled simulation models, a basic model can be built by calling models already stored in the modular model library. Modifications and additions can then be made to this basic model to quickly construct a multiphysics-coupled simulation model.

[0101] When simulating laser cutting, the predicted cutting parameters can be input into the simulation model to configure its parameters. For example, the core operational commands input for simulating laser cutting include cutting parameters such as laser power, cutting speed, gas pressure, and focal point position. These parameters form the basis for the simulation. After starting the laser cutting simulation, the virtual laser in the simulation model can act on the virtual target material based on the input cutting parameters, mimicking the complete process of melting, slag removal, and kerf formation, and outputting data such as kerf shape and temperature distribution.

[0102] After simulating laser cutting, simulation output results can be obtained. By comparing the simulation output results with actual requirements (finished part parameters, kerf width, kerf perpendicularity, heat-affected zone size, presence of burrs, and whether the cut is complete), it can be verified whether the cutting effect meets the actual cutting requirements.

[0103] S103, when the cutting effect obtained by simulating laser cutting based on the simulation model meets the preset conditions, the target plate is actually cut based on the cutting parameters.

[0104] For example, preset conditions can be understood as judgment conditions preset for the quality requirements of the cut parts and the requirements of the cutting process. For example, preset conditions may include parameter judgment ranges preset based on the finished product parameters and their allowable errors, and may also include process judgment ranges preset for power consumption, material utilization rate, etc. in the cutting process.

[0105] By comparing the simulation output results to see if they fall within the preset ranges, it can be determined whether the cutting effect obtained by the simulated laser cutting meets the preset conditions. For example, if the values ​​of all items in the simulation output are within the preset ranges, the cutting effect can be determined to meet the preset conditions. Conversely, if the value of at least one item in the simulation output is not within the preset ranges, the cutting effect can be determined to not meet the preset conditions.

[0106] When the cutting effect obtained from the simulated laser cutting meets the preset conditions, it indicates that the generated cutting parameters are well-suited to the current sheet material and finished product parameters. When actual cutting is performed based on these parameters, a cutting effect similar to or close to that obtained from the simulated laser cutting can be achieved. Therefore, the target sheet material can be actually cut based on these cutting parameters.

[0107] During actual cutting, the target material can be placed inside the laser cutting machine, and the cutting parameters corresponding to the preset cutting effect can be configured on the laser cutting machine. The laser cutting machine is then started according to the procedure to actually cut the target material, obtaining the cut part.

[0108] The laser cutting method provided in this application first predicts cutting parameters based on the material parameters of the target sheet and the finished product parameters of the part to be cut using a pre-trained prediction model before actual cutting. These cutting parameters are not directly applied to the raw material for actual cutting; instead, the laser cutting process is simulated in a multiphysics coupling simulation model based on these parameters. This simulated laser cutting process not only verifies the usability of the model-generated cutting parameters but also avoids consuming real raw materials. Therefore, when the cutting effect obtained from the simulated laser cutting meets preset conditions, the target sheet is then actually cut based on these cutting parameters. This eliminates the need for trial cutting with real materials, solving the problem of raw material waste.

[0109] For example, when simulating the laser cutting process based on the cutting parameters generated by the prediction model, the cutting effect may meet the preset conditions or it may not. If the cutting effect does not meet the preset conditions, the cutting parameters can be optimized and adjusted, and the simulation and verification can be performed again until the cutting effect that meets the preset conditions is obtained.

[0110] In one possible implementation, the method further includes: when the cutting effect obtained by simulating laser cutting based on the simulation model does not meet the preset conditions, optimizing the cutting parameters through a reinforcement learning algorithm, and inputting the optimized cutting parameters into a multi-physics coupled simulation model to simulate the laser cutting process and verify the cutting effect.

[0111] For example, a reinforcement learning algorithm can be an algorithm that autonomously optimizes itself through an "error-feedback-correction" strategy. If the cutting effect obtained by simulating laser cutting based on a simulation model does not meet the preset conditions, the reinforcement learning algorithm can be used as the main intelligent improvement tool to optimize and adjust the cutting parameters corresponding to the cutting effect that does not meet the preset conditions, so that the cutting effect that meets the preset conditions can be obtained after optimization and adjustment.

[0112] Reinforcement learning algorithms can be understood as intelligent agents that optimize cutting parameters, making decisions on how to adjust these parameters. Examples of reinforcement learning algorithms include Proximal Policy Optimization (PPO) and Deep Q-Network (DQN).

[0113] A multiphysics-coupled simulation model can serve as the optimization environment, receiving cutting parameters output by the agent and returning the corresponding cutting results. These results can then act as feedback in the parameter optimization process. During optimization, the state can be understood as key information in the current optimization process, such as the material parameters of the target sheet, the finished product parameters, and / or preset conditions. Actions can be understood as the agent's adjustments to the cutting parameters, such as increasing or decreasing the cutting power by 100W, or increasing or decreasing the cutting speed by 0.2 m / min.

[0114] Reinforcement learning algorithms also incorporate reward functions. These functions quantify the quality of the cutting effect, guiding the agent to adjust its actions in a direction that satisfies preset conditions. The reward function can include various reward items. For example, basic rewards: 100 points for complete penetration, 200 points deducted for non-penetration; accuracy rewards: 10 points added for every 0.01mm decrease in tolerance (e.g., 100 points for 0.1mm, 150 points for 0.05mm); heat-affected zone rewards: 5 points added for every 0.05mm decrease (e.g., 30 points for 0.3mm, 50 points for 0.2mm). Furthermore, the reward function can include penalty items, such as: deducting 20 points for excessive parameter adjustments (e.g., increasing single laser power beyond 300W) to prevent excessive changes in cutting effect; and deducting 50 points if the target is still not met after optimization (to force rapid algorithm improvement).

[0115] When optimizing cutting parameters using reinforcement learning algorithms, optimization can be performed in a single round or over multiple rounds. For example, based on an initial state (cutting parameters corresponding to cutting effects that do not meet preset conditions), the agent outputs its first adjustment action (such as increasing power to 3200W, reducing speed to 1.3m / min, and adjusting auxiliary gas pressure to 7.5bar), generating the first round of optimized cutting parameters S1. Cutting parameters S1 are then input into a multiphysics-coupled simulation model to re-simulate the laser cutting process, obtaining the simulation output results for cutting parameters S1.

[0116] The simulation output of the cutting parameter S1 is evaluated under preset conditions. If the preset conditions are met, it indicates that the cutting parameter S1 can be used for actual cutting, and the reinforcement learning algorithm optimization is paused. If the preset conditions are not met, it indicates that the cutting parameter S1 is not suitable for actual cutting. The reinforcement learning algorithm optimization can continue to be performed following similar steps to obtain cutting parameters S2, S3, etc., until a cutting parameter that meets the preset conditions is obtained, at which point the reinforcement learning algorithm optimization is paused.

[0117] Because laser cutting is a multi-physics coupling process, it involves multiple types of cutting parameters across multiple dimensions, and these parameters are strongly coupled with each other. Optimizing cutting parameters that do not meet preset conditions using orthogonal experiments or manual experience is difficult because it's hard to determine the direction and extent of adjustments for multiple types of cutting parameters in a short time, resulting in high trial-and-error costs.

[0118] In this embodiment, reinforcement learning algorithms can autonomously explore high-dimensional spaces with various combinations of cutting parameters through continuous interaction between the agent and the simulation model. Furthermore, reinforcement learning algorithms can autonomously optimize based on an "error-feedback-correction" strategy and quickly obtain optimal solutions through reward and penalty mechanisms, i.e., quickly obtain cutting parameters that meet preset conditions. Therefore, the efficiency of obtaining cutting parameters that meet preset conditions can be improved.

[0119] For example, in laser cutting of metal sheets, especially when cutting thin metal sheets (e.g., less than 2mm thick), the metal edges on both sides of the kerf are prone to deformation, affecting the quality of the cut part. Analysis shows that if the range of the heat-affected zone can be effectively controlled during actual cutting, the degree of metal edge deformation can be reduced, thus improving the quality of the cut part.

[0120] In one possible implementation, the laser cutting process is simulated and the cutting effect is verified, including: during the simulated laser cutting process, the width of the heat-affected zone on both sides of the kerf is determined based on the micro- and macro-scale thermal field coupling analysis introduced in the simulation model; and the cutting effect is verified based on the width of the heat-affected zone.

[0121] For example, when constructing a simulation model, a thermal field coupling analysis at both micro and macro scales can be introduced to analyze the width of the heat-affected zone formed during laser cutting. This micro- and macro-scale thermal field coupling analysis can be understood as the analysis of thermal field coupling effects at both the micro-scale (e.g., grain level) and macro-scale (e.g., plate level) levels, and can be implemented using multi-scale thermal analysis algorithms.

[0122] Microscale thermal fields primarily focus on the local thermal behavior of the microstructure of materials (such as grains, grain boundaries, dislocations, and phase interfaces), with scales ranging from micrometers to nanometers. For example, during laser heating cutting, the differences in heat conduction between the grain interior and grain boundaries, the local latent heat generated by phase transformations (such as the transformation of austenite into martensite), and the obstruction of heat diffusion by microscopic defects (such as pores) all contribute to this.

[0123] Macro-scale thermal fields primarily focus on the overall temperature distribution of the board material, with scales ranging from millimeters to meters. Examples include the temperature field along the cutting path and the macro-width of the heat-affected zone. Examples include heat conduction from the high-temperature region (thousands of degrees Celsius) at the center of the laser spot to the interior of the board, macro-scale heat carried away by auxiliary gas blowing, and natural heat dissipation from the edges of the board.

[0124] Since the microscopic and macroscopic thermal fields are not independent but rather exhibit bidirectional coupling feedback, multi-scale thermal field coupling analysis is necessary for both microscopic and macroscopic thermal fields. For example, the macroscopic thermal field determines the thermal environment of the microstructure (e.g., a rise in macroscopic temperature can trigger a microscopic phase transition), while microscopic thermal behaviors (e.g., latent heat of phase transition, differences in grain thermal conductivity, etc.) affect the macroscopic thermal diffusion rate. Coupling analysis simultaneously calculates these interrelated effects, avoiding significant biases that can occur in single-scale analyses.

[0125] In laser cutting, the width of the region where the material does not melt but its properties (such as hardness and toughness) change due to heat can be understood as the heat-affected zone (HAZ). The HAZ width characterizes the extent of the HAZ. The smaller the HAZ width, the smaller the area where material properties change, and the smaller the deformation of the cut edge of the sheet metal. Therefore, when verifying the cutting effect, the HAZ width can be used as a dimension to consider the effect. If the cutting parameters are optimized based on the cutting effect, the optimized cutting parameters can be adjusted in the direction of reducing the HAZ width. For example, when optimizing cutting parameters using reinforcement learning algorithms, the HAZ width can be used as one of the optimization reference factors.

[0126] When constructing a simulation model, one can first define the core variables and boundary conditions at the macroscopic scale, and then define the core objects and key parameters at the microscopic scale. For example, the core variables at the macroscopic scale can include the overall temperature distribution of the plate, the laser heat source model (such as the power density with a Gaussian distribution), macroscopic thermophysical parameters (such as the overall thermal conductivity, macroscopic specific heat capacity, etc.), and the macroscopic width threshold of the heat-affected zone (such as the macroscopic width of the heat-affected zone of carbon steel being the area with a temperature greater than 700℃ from the cut). Boundary conditions can include the energy input of the laser-affected area, the convective heat dissipation coefficient of the auxiliary gas, and the heat conduction between the plate and the worktable.

[0127] For example, core objects at the microscale can include the material's microstructure (such as ferrite grains, austenite grain boundaries, etc.), microscopic thermal behavior (such as intra-grain thermal conductivity, grain boundary thermal conductivity, etc.), latent heat of phase transformation (such as endothermic austenitization, exothermic martensitic transformation, etc.), and microscopic defects (such as changes in local thermal resistance caused by increased dislocation density). Key parameters can include grain size, grain boundary volume fraction, and phase transformation volume fraction.

[0128] After defining the core variables and boundary conditions at the macroscopic scale, and the core objects and key parameters at the microscopic scale, a multi-scale thermal analysis algorithm for coupled analysis of the thermal field at both the microscopic and macroscopic scales can be constructed based on these variables, conditions, and parameters. The construction of this multi-scale thermal analysis algorithm can be achieved by defining macroscopic heat conduction calculation formulas and microscopic structure evolution formulas. For example, the macroscopic heat conduction calculation formula can be an equation concerning the overall thermal conductivity, macroscopic specific heat capacity, and a laser heat source model. The microscopic structure evolution formula can be a set of equations including a grain growth model and a phase transition kinetics model.

[0129] In simulated laser cutting, multi-scale thermal analysis algorithms can be used to perform coupled thermal field analysis and calculations at both micro and macro scales to obtain the width of the heat-affected zone on both sides of the cut. For example, by running a simulation model using simulation software that supports multi-scale modeling, inputting the cutting parameters output by the prediction model, and starting the simulated laser cutting, the simulation software can perform joint calculations based on macroscopic heat conduction calculation formulas and microscopic structure evolution formulas. This can output macroscopic simulation results, such as the spatial distribution of the heat-affected zone at the sheet level and temperature change curves over time; it can also output microscopic simulation results, such as the grain size distribution within the heat-affected zone.

[0130] For example, the width of the heat-affected zone (HAZ) can be determined by the spatial distribution of the HAZ at the sheet metal level and the grain size distribution within the HAZ. The cutting effect can then be verified based on the HAZ width.

[0131] For example, a threshold for the width of the heat-affected zone (HAZ) (e.g., 0.5mm) is set in the preset conditions based on the actual application scenario, such as the material and thickness of the sheet metal. If the width of the HAZ in the simulation output is greater than the threshold, it can be determined that the preset conditions are not met; if the width of the HAZ in the simulation output is less than or equal to the threshold, it can be determined that the preset conditions are met. When the preset conditions are not met because the width of the HAZ is greater than the threshold, the cutting parameters can be adjusted and optimized according to the factors affecting the width of the HAZ.

[0132] In this embodiment of the application, since the width of the heat-affected zone on both sides of the cut can be determined by the micro- and macro-scale thermal field coupling analysis introduced in the simulation model during the simulated laser cutting process, the cutting effect can be verified based on the width of the heat-affected zone.

[0133] Based on this, the evaluation of the cutting effect can be fully combined with the width of the heat-affected zone (HAZ), thus controlling the HAZ width within a preset threshold range. When performing actual cutting based on the cutting parameters corresponding to the cutting effect that meets the preset conditions, the HAZ width can be effectively controlled, reducing the probability of deformation at the cut edge due to thermal effects, lowering the non-compliance rate of the cut parts, and improving the actual cutting quality. This method is particularly effective in improving the yield rate of cut parts from thin sheet metal.

[0134] For example, uncontrollable changes occur during actual cutting, such as variations in sheet material quality, laser system status, and environmental factors. Using fixed cutting parameters indiscriminately can lead to decreased cutting quality or excessive energy consumption. Therefore, this application proposes a scheme to dynamically adjust cutting parameters during actual cutting to overcome these variations and improve cutting quality and energy utilization.

[0135] In one possible implementation, the target board material is actually cut based on cutting parameters, including: during the actual cutting process, acquiring multimodal data, which includes at least two of temperature data, optical data, acoustic data, and image data; and dynamically adjusting the cutting parameters based on the multimodal data to perform the actual cutting according to the dynamically adjusted cutting parameters.

[0136] For example, multimodal data can be understood as a combination of data collected by different types of sensors, reflecting different physical characteristics of the cutting process. Since analysis based on single data may contain biases or omissions, this embodiment employs multimodal data for analysis. Multimodal data can comprehensively depict the state of the cutting process from multiple dimensions. Based on multimodal data, the limitations of single data can be overcome, and cross-validation between multiple modalities can improve the accuracy of the analysis.

[0137] Temperature data can be understood as the temperature distribution and changes in the cutting area, such as the molten pool temperature and the temperature of the cut edge. Temperature data can be collected using an infrared thermal imager or thermocouples. Analysis of the temperature data can determine whether the temperature generated by the laser currently directed at the material is sufficient to cut through it. It can also analyze whether the current temperature is too high, causing cutting deformation. By using temperature data, the laser power can be dynamically adjusted to just cut through the material without causing overheating and deformation, or to avoid energy waste due to excessive laser power.

[0138] Optical data can include data collected based on light signals during the cutting process. Examples include plasma radiation spectra, laser reflectivity, and molten pool luminescence intensity. Optical data can be acquired by spectrometers or photoelectric sensors and can reflect plasma shielding strength, material melting state, etc. By analyzing optical data, cutting parameters can be dynamically adjusted; for example, laser power and gas flow rate can be adjusted to reduce plasma shielding strength and avoid incomplete cutting due to excessively high plasma shielding strength.

[0139] In some high-power laser cutting of thick plates or high-energy-density cutting scenarios, the ionized gas cloud formed after the material is heated to vaporization and further ionized constitutes plasma. A large amount of plasma surrounds the laser action area to form a plasma field, such as the plasma field gathered above and around the molten pool.

[0140] Plasma fields exhibit a plasma shielding effect. This effect can be understood as the absorption and scattering of laser energy by the plasma field, leading to a reduction in the actual laser energy reaching the material surface. This plasma shielding effect can cause instability in the cutting process, resulting in defects such as incomplete penetration, burrs, and ripples, all of which negatively impact cutting quality. To overcome the problems of reduced cutting quality and energy loss caused by the plasma shielding effect, this application proposes a scheme to dynamically adjust the laser power and auxiliary gas flow rate in the cutting parameters based on real-time monitoring of the plasma density distribution.

[0141] In one possible implementation, the multimodal data includes plasma analysis data for determining the plasma density distribution; the cutting parameters are dynamically adjusted based on the multimodal data, including: acquiring plasma analysis data obtained within a preset range of the laser gun head in real time; determining the plasma density distribution within the preset range of the laser gun head through real-time analysis of the plasma analysis data; and dynamically adjusting the laser power and / or auxiliary gas flow rate in the cutting parameters according to the plasma density distribution to cope with the plasma shielding effect within the preset range of the laser gun head.

[0142] For example, plasma analysis data can be one type of data or multiple types of data. For instance, plasma analysis data can be optical data and / or image data.

[0143] For example, by capturing the characteristic radiation spectrum of plasma using a high-resolution spectrometer (such as specific wavelength spectral lines of metal ions, like the 510 nm spectral line of iron ions), the plasma density distribution can be obtained through spectral intensity inversion. Higher spectral intensity indicates higher density, and vice versa.

[0144] For example, by capturing images of the plasma morphology above the molten pool using a high-speed camera (e.g., at a frame rate of 1000fps or higher), and combining this with image segmentation algorithms (e.g., thresholding), the spatial distribution of the plasma can be extracted, thus yielding the plasma density distribution. The plasma density distribution can then be calculated based on the plasma concentration gradient and / or the area of ​​the region within space.

[0145] Plasma analysis data within a preset range of the laser gun head can be acquired using optical or image sensors. This preset range can be a three-dimensional space centered on the laser gun head, encompassing a preset length (e.g., 200mm), width (e.g., 100mm), and height (e.g., 50mm). The preset range can move with the laser gun head, allowing for real-time dynamic monitoring of the plasma density distribution during the cutting process.

[0146] To address the plasma shielding effect within the preset range of the laser gun head, the laser power and / or auxiliary gas flow rate in the cutting parameters can be increased or decreased based on the shielding effect caused by the plasma density distribution. When the plasma shielding effect is high, dynamically increasing the laser power can enhance the laser energy to counteract the absorption and scattering of the laser by the plasma, ensuring that the laser energy reaching the material surface is sufficient to penetrate the material. Increasing the auxiliary gas flow rate can enhance the dispersing effect on the plasma and reduce the shielding effect. When the plasma shielding effect is low, the absorption and scattering of the laser by the plasma are weaker, so the laser power and / or auxiliary gas flow rate can be appropriately reduced.

[0147] For example, if it is determined that the plasma density distribution within a preset range of the laser gun head has a shielding effect on laser energy, such as absorption or scattering, the laser power and auxiliary gas flow rate in the cutting parameters can be increased. If it is determined that the plasma density distribution within a preset range of the laser gun head does not have a shielding effect on laser energy, such as absorption or scattering, the laser power and auxiliary gas flow rate in the cutting parameters can be decreased.

[0148] In this embodiment, by real-time monitoring of plasma density distribution and dynamic adjustment of laser power and / or auxiliary gas flow rate, the plasma shielding effect within a preset range of the laser gun head can be specifically counteracted. This improves cutting stability, reduces cutting quality issues such as "broken cut" or "incomplete cut," avoids rework or scrap rates, and increases cutting efficiency. Especially when laser cutting highly reflective metals (such as copper and aluminum) or high-melting-point metals (such as titanium alloys), these materials are prone to strong plasma shielding effects due to energy reflection. If traditional fixed cutting parameters are used for actual cutting, it is difficult to dynamically address the plasma shielding effect, making stable and efficient cutting difficult. However, the method in this embodiment, targeting metals prone to plasma shielding effects, can effectively solve this problem and achieve better cutting quality by acquiring and analyzing plasma analysis data and combining it with a strategy of dynamically adjusting laser power and / or auxiliary gas flow rate.

[0149] For example, acoustic data can be understood as the data of sound signals generated during cutting, such as the data corresponding to sounds like molten pool boiling, gas jetting, and material fracture. Acoustic data can be collected by a high-definition microphone that follows the movement of the laser gun head. By analyzing sound type, sound wave frequency, loudness changes, etc., it can be determined whether the material can be cut smoothly and whether slag removal is smooth. For example, if abnormal noises are present, burrs may be generated on the cut surface. Based on acoustic data and other multimodal data, a comprehensive judgment can be made on the current actual cutting situation, and cutting parameters such as laser power, laser gun head movement speed, and auxiliary gas flow rate can be dynamically adjusted.

[0150] For example, image data can be understood as images or videos obtained by capturing real-time images of the cutting area. For instance, a high-speed camera can capture images of the molten pool morphology, cut cross-section, slag removal, and spatter during the cutting process. Based on this image data, computer vision processing algorithms can be used for dynamic analysis, providing a basis for dynamically adjusting cutting parameters.

[0151] In this embodiment, since multimodal data including at least two of temperature data, optical data, acoustic data and image data are acquired, cross-verification analysis can be performed through the multimodal data to comprehensively analyze the actual cutting process state, so as to dynamically adjust the cutting parameters, realize dynamic control of the actual cutting, and improve cutting quality and cutting efficiency.

[0152] For example, the multimodal data collected during the actual cutting process is data with high reference value. Based on this multimodal data and the corresponding real-time cutting data, the prediction model can be fed back, enabling data-driven fine-tuning training of the prediction model and achieving incremental training and optimization upgrades of the prediction model.

[0153] In one possible implementation, the method further includes: constructing a fine-tuning training dataset based on multimodal data collected during multiple actual cutting processes and real-time cutting parameters during multiple cutting processes; and fine-tuning the prediction model using at least one model training method based on the fine-tuning training dataset to update the model parameters of the prediction model.

[0154] For example, real-time cutting parameters can be understood as cutting data used during the actual cutting process. This real-time cutting data can be obtained from a database that records data during the actual cutting process. Multimodal data collected during the actual cutting process can also be obtained from the data storage space corresponding to each sensor, or from a database that records the actual cutting data.

[0155] After acquiring multimodal data collected during multiple actual cutting processes, as well as real-time cutting parameters during these processes, data cleaning and normalization can be performed on various types of data. For example, data cleaning can remove erroneous values ​​exceeding the error threshold, and completion operations can supplement accurate data. Furthermore, normalization operations can unify the dimensions of the data, improving its effectiveness in model training.

[0156] When constructing a fine-tuning training dataset, multimodal data can be added to the training samples consisting of sheet material parameters, finished product parameters, and cutting parameters to form fine-tuning training samples. Ground truth labels (positive or negative) can be assigned to each fine-tuning training sample using manual or automatic annotation. By employing supervised or semi-supervised learning models to perform one or more rounds of fine-tuning training on the prediction model, the model parameters can be iteratively optimized. This allows the prediction model to enrich its experience in predicting cutting parameters during training, thereby improving prediction accuracy.

[0157] After fine-tuning the training to obtain a prediction model with updated parameters, this prediction model can be deployed in the actual cutting system. During actual cutting, the multimodal data acquired during the actual cutting process is input into the prediction model, which can then output cutting parameters in real time. These cutting parameters are generated based on the data collected in real time during the cutting process, thus being highly suitable for the current actual cutting and improving cutting quality and efficiency.

[0158] In this embodiment, by using multimodal data collected during multiple actual cutting processes and real-time cutting parameters during multiple cutting processes, a fine-tuning training dataset can be constructed for fine-tuning the prediction model. This helps the prediction model obtain effective features of the predicted cutting parameters from the multimodal data, thereby increasing the generation capability of the prediction model. Consequently, it can generate cutting parameters with better cutting effects during the simulation model cutting stage or during the actual cutting stage, thereby improving laser cutting efficiency and quality and enhancing the product quality of the cut parts.

[0159] Figure 4 is a schematic diagram of the structure of the laser cutting device provided in an embodiment of this application. As shown in Figure 4, this embodiment of the application provides a laser cutting device, which includes:

[0160] The generation module 401 is used to input the board parameters of the target board and the finished product parameters of the cut part into the prediction model trained based on historical cutting data to generate cutting parameters;

[0161] The simulation module 402 is used to input cutting parameters into a multi-physics coupled simulation model to simulate the laser cutting process and verify the cutting effect;

[0162] The cutting module 403 is used to actually cut the target material based on the cutting parameters when the cutting effect obtained by simulating laser cutting based on the simulation model meets the preset conditions.

[0163] In one possible implementation, the apparatus further includes an optimization module, which is used to:

[0164] When the cutting effect obtained by simulating laser cutting based on the simulation model does not meet the preset conditions, the cutting parameters are optimized by reinforcement learning algorithm, and the optimized cutting parameters are input into the multi-physics coupled simulation model to simulate the laser cutting process and verify the cutting effect.

[0165] In one possible implementation, the simulation module 402 is specifically used for:

[0166] In the simulated laser cutting process, the width of the heat-affected zone on both sides of the cut is determined based on the coupling analysis of the microscopic and macroscopic thermal fields introduced in the simulation model.

[0167] The cutting effect was verified based on the width of the heat-affected zone.

[0168] In one possible implementation, the cutting module 403 is specifically used for:

[0169] During the actual cutting process, multimodal data is acquired, including at least two of the following: temperature data, optical data, acoustic data, and image data.

[0170] Based on multimodal data, the cutting parameters are dynamically adjusted so that the actual cutting can be performed according to the dynamically adjusted cutting parameters.

[0171] In one possible implementation, the multimodal data includes plasma analysis data for determining the plasma density distribution; the cutting module 403 is specifically used for:

[0172] Acquire plasma analysis data obtained within a preset range of the real-time monitoring laser gun head;

[0173] The plasma density distribution within a preset range of the laser gun head is determined in real time through plasma analysis data.

[0174] The laser power and / or auxiliary gas flow rate in the cutting parameters are dynamically adjusted according to the plasma density distribution to cope with the plasma shielding effect within the preset range of the laser gun head.

[0175] In one possible implementation, the device further includes a fine-tuning module, which is used for:

[0176] Based on multimodal data collected during multiple actual cutting processes, and real-time cutting parameters during multiple cutting processes, a fine-tuning training dataset is constructed.

[0177] Based on the fine-tuning training dataset, the prediction model is fine-tuned using at least one model training method to update the model parameters of the prediction model.

[0178] In one possible implementation, the simulation model of multiphysics coupling includes the multiphysics coupling effects between lasers, materials, gases and devices, and covers the dynamic behavior of thermal fields, plasma fields and molten pool flows.

[0179] The laser cutting apparatus provided in this application can be used to execute the technical solution of the laser cutting method in any of the above embodiments of this application. Its implementation principle and technical effect are similar, and will not be described again here.

[0180] Figure 5 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. As shown in Figure 5, the electronic device in this embodiment may include: at least one processor 501; and a memory 502 communicatively connected to at least one processor; wherein, the memory 502 stores instructions that can be executed by at least one processor 501, and the instructions are executed by at least one processor 501 to cause the electronic device to perform the method as described in any of the above embodiments.

[0181] Optionally, the memory 502 can be either standalone or integrated with the processor 501.

[0182] The implementation principle and technical effects of the electronic device provided in this embodiment can be found in the foregoing embodiments, and will not be repeated here.

[0183] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the method of any of the foregoing embodiments.

[0184] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the method of any of the foregoing embodiments.

[0185] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.

[0186] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0187] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU) or other general-purpose processors. The processor can also be a Digital Signal Processor (DSP) or an Application Specific Integrated Circuit (ASIC), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0188] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device, and may also be various media that can store program code, such as USB flash drives, portable hard drives, read-only memory (ROM), disks or optical discs.

[0189] The aforementioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof. Examples of storage media include Static Random-Access Memory (SRAM) or Electrically Erasable Programmable Read Only Memory (EEPROM).

[0190] Storage media can be, for example, erasable programmable read-only memory (EPROM) or programmable read-only memory (PROM). Storage media can also be read-only memory (ROM), magnetic storage, flash memory, magnetic disks, or optical disks. Storage media can be any available medium accessible to general-purpose or special-purpose computers.

[0191] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside within an application-specific integrated circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components within an electronic device or host device.

[0192] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0193] The sequence numbers of the embodiments in this application are merely for description and do not represent the superiority or inferiority of the embodiments. Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0194] Based on this understanding, the technical solution of this application, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0195] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

[0196] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0197] It should be further noted that although the steps in the flowchart are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise explicitly stated in this document, there is no strict order requirement for the execution of these steps, and they can be executed in other orders.

[0198] Furthermore, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but may be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but may be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0199] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0200] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0201] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A laser cutting method, characterized in that, The method includes: inputting the plate parameters of the target plate and the finished parameters of the cut part into a prediction model trained based on historical cutting data to generate cutting parameters; inputting the cutting parameters into a multi-physics coupled simulation model to simulate the laser cutting process and verify the cutting effect; and when the cutting effect obtained by simulating laser cutting based on the simulation model meets preset conditions, performing actual cutting on the target plate based on the cutting parameters.

2. The method according to claim 1, characterized in that, The method further includes: if the cutting effect obtained by simulating laser cutting based on the simulation model does not meet the preset conditions, optimizing the cutting parameters through a reinforcement learning algorithm, and inputting the optimized cutting parameters into the multi-physics coupled simulation model to simulate the laser cutting process and verify the cutting effect.

3. The method according to claim 1, characterized in that, The simulated laser cutting process and verification of the cutting effect include: during the simulated laser cutting process, determining the width of the heat-affected zone on both sides of the cut based on the micro- and macro-scale thermal field coupling analysis introduced in the simulation model; and verifying the cutting effect based on the width of the heat-affected zone.

4. The method according to any one of claims 1-3, characterized in that, The actual cutting of the target board material based on the cutting parameters includes: acquiring multimodal data during the actual cutting process, wherein the multimodal data includes at least two of temperature data, optical data, acoustic data, and image data; and dynamically adjusting the cutting parameters according to the multimodal data, so as to perform actual cutting according to the dynamically adjusted cutting parameters.

5. The method according to claim 4, characterized in that, The multimodal data includes plasma analysis data for determining plasma density distribution; the step of dynamically adjusting the cutting parameters based on the multimodal data includes: acquiring plasma analysis data obtained within a preset range of the laser gun head in real time; determining the plasma density distribution within the preset range of the laser gun head through real-time analysis of the plasma analysis data; and dynamically adjusting the laser power and / or auxiliary gas flow rate in the cutting parameters according to the plasma density distribution to address the plasma shielding effect within the preset range of the laser gun head.

6. The method according to claim 4, characterized in that, The method further includes: constructing a fine-tuning training dataset based on multimodal data collected during multiple actual cutting processes and real-time cutting parameters during the multiple cutting processes; and fine-tuning the prediction model using at least one model training method based on the fine-tuning training dataset to update the model parameters of the prediction model.

7. The method according to any one of claims 1-3, characterized in that, The simulation model of multiphysics coupling includes the multiphysics coupling effects between lasers, materials, gases and devices, and covers the dynamic behavior of thermal fields, plasma fields and molten pool flows.

8. An electronic device, characterized in that, include: Memory and processor; The memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-7.

Images