Laser cutting machine capable of realizing digital high-precision control

Abstract

The invention discloses a laser cutting machine for realizing digital high-precision control, which relates to the technical field of laser cutting machines and comprises a path analysis module, a path segmentation module, an initial parameter configuration module, a feature extraction and risk assessment module, an intelligent score prediction module and a parameter dynamic regulation and control module. The path analysis module is used for reading a complete cutting path generated by a CAD drawing; and the path segmentation module is used for uniformly dividing the whole cutting path into a plurality of cutting sub-segments according to a preset fixed length. According to the method, the heat accumulation risk degree of each cutting sub-section is automatically predicted by means of the trained machine learning model, and the actual laser power and the cutting speed are intelligently regulated and controlled according to the prediction scoring result, so that the heat of a high accumulation risk area can be timely and effectively controlled and reduced; therefore, the probability of occurrence of serious deformation such as warping and bulging of materials is remarkably reduced, and comprehensive improvement of the size precision, the surface quality and the yield of machined parts is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of laser cutting machines, and particularly to a laser cutting machine that realizes digital high-precision control. Background Art

[0002] A laser cutting machine that realizes digital high-precision control refers to the precise management and real-time regulation of the laser cutting process through advanced digital technologies (such as computer numerical control systems, sensor technologies, and intelligent algorithms), so as to achieve high-precision control of parameters such as cutting paths, speeds, and powers. This control method can automatically generate cutting paths according to CAD drawings, and dynamically adjust the laser focus and motion trajectories through a feedback system to ensure that the dimensional accuracy, edge smoothness, and material utilization rate of the cutting reach the best level. Digital high-precision control not only improves processing efficiency and product quality, but also supports flexible processing of complex graphics and various materials, and is widely used in fields such as metal processing, automobile manufacturing, aerospace, and the electronics industry.

[0003] The existing technologies have the following deficiencies: In the existing technologies, when a laser cutting machine performs real-time cutting according to the cutting path of a CAD drawing, in order to improve processing efficiency, the laser power and cutting speed are usually set relatively high. However, when a high-accumulation cutting area appears during the cutting process, the heat dissipation problem is usually ignored. If the high cutting speed continues to be maintained, heat accumulates concentratedly in this area and is difficult to dissipate in time, easily leading to serious deformation problems such as material warping or bulging, and ultimately resulting in unqualified cutting quality.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a laser cutting machine that realizes digital high-precision control. By dividing the cutting path generated by CAD into multiple complete cutting sub-segments, and real-time collecting the path data of each segment to construct a feature vector, and automatically predicting the heat accumulation risk degree of each cutting sub-segment with a trained machine learning model, and intelligently regulating the actual laser power and cutting speed according to the prediction scoring result, so that the heat in the high-accumulation risk area can be effectively controlled and reduced in time, thereby significantly reducing the occurrence probability of serious deformations such as material warping and bulging, ensuring the overall improvement of the dimensional accuracy, surface quality, and yield rate of the processed parts, and at the same time avoiding the efficiency loss and uncertainty brought by human intervention, improving production stability and the level of automated processing, so as to solve the problems in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions: A laser cutting machine for realizing digital high-precision control, including a path parsing module, a path segmentation module, an initial parameter configuration module, a feature extraction and risk assessment module, an intelligent scoring prediction module, and a parameter dynamic regulation module; The path parsing module reads the complete cutting path generated by the CAD drawing; The path segmentation module evenly divides the entire cutting path into multiple cutting sub-segments according to a preset fixed length (such as several millimeters or fine-tuned according to material characteristics). During the segmentation process, ensure that the geometric features in the corresponding CAD original drawing of each cutting sub-segment are complete; After the initial parameter configuration module completes the division of the path cutting sub-segments, it enters the initial configuration stage, and sets a high laser power and a high cutting speed as the default startup parameters for the entire cutting process for laser cutting; The feature extraction and risk assessment module collects the cutting path data of each cutting sub-segment in real time in the host computer software, constructs a feature vector through the cutting path data, and evaluates the high heat accumulation risk of the current cutting sub-segment; The intelligent scoring prediction module analyzes the constructed feature vector through a pre-trained machine learning model, outputs a cutting accumulation degree score, and predicts the cutting path accumulation degree of each cutting sub-segment based on the scoring result; The parameter dynamic regulation module, according to the scoring result output by the machine learning model, sends power and speed adjustment instructions to the machine tool numerical control system in real time, and intelligently regulates the actual laser power and the actual cutting speed to ensure that the real-time temperature of the cutting area does not exceed the deformation temperature threshold of the material.

[0007] Preferably, the steps of reading the CAD drawing to generate a complete cutting path are as follows: Import the CAD drawing file, and identify the geometric elements in the drawing through the CAD parsing unit; Based on the topological structure analysis, identify the relative position relationship, connection order, and layer information of each graphic, and screen out the path elements that need to perform laser cutting according to the process requirements; Perform preprocessing operations on the extracted geometric path to ensure that the path is coherent, without breaks and without redundant repeated line segments; Convert the geometric path information into a format recognizable by the numerical control system to form an executable complete cutting path.

[0008] Preferably, the specific steps of dividing the entire cutting path into multiple cutting sub-segments according to a preset fixed length are as follows: After completing the extraction of the geometric path of the CAD drawing and the optimization of the path sorting, calculate the total length of the entire continuous cutting path; According to the preset segmentation length parameter, divide along the geometric distance of the path starting from the path origin in sequence, and mark the end of each cutting sub-segment when a set length is reached. If a certain path segment is a curve or contains complex corners, use an interpolation algorithm for equidistant fitting to ensure that the segmentation points fall on the actual cutting trajectory.

[0009] Preferably, construct a feature vector through the cutting path data, and the specific steps are as follows: Extract the indicators reflecting the high concentration of the cutting path from the cutting path data of each collected cutting sub-segment. Among them, the extracted indicators include the laser scanning time per unit area and the adjacent distance between cutting paths. After in-depth analysis of the extracted indicators, generate a path scanning time density reference value and a path adjacent distance reference value respectively, and use the analyzed path scanning time density reference value and path adjacent distance reference value as the feature vector to evaluate the high heat accumulation risk of the current cutting sub-segment.

[0010] Preferably, after analyzing the path scanning time density reference value and the path adjacent distance reference value through a pre-trained machine learning model, output a cutting accumulation degree score, record the score result as the accumulation index, and predict the cutting path accumulation degree of each cutting sub-segment based on the accumulation index.

[0011] Preferably, according to the score result output by the machine learning model, intelligently adjust the actual laser power and the actual cutting speed, and the specific steps are as follows: Calculate the accumulation index of the current cutting sub-segment and its difference from the accumulation index threshold, and based on the abnormal calculation, determine the specific amplitude of reducing the laser power and the cutting speed, specifically: Use an exponential function with a penalty factor and combine the concept of "zero threshold truncation" for calculation. The calculation formula is: , Where: The accumulation index of the current cutting sub-segment output by the machine learning model at time t, Is the accumulation index threshold, Is the deformation temperature threshold, Is the difference between the accumulation index of the current cutting sub-segment and the accumulation index threshold, indicating the degree of accumulation of the current cutting sub-segment exceeding the threshold. If , then , no power reduction or speed reduction operation is performed, And Are the initial set high laser power and high cutting speed during laser cutting respectively, And Are the adjustment sensitivity coefficients used to control the attenuation slopes of power and speed, And It represents the actual laser power and actual cutting speed that need to be executed in the current cutting sub-segment; After calculating the actual laser power and actual cutting speed, the actual laser power and the actual cutting speed are used as a new set of instruction sets and are sent to the numerical control system of the machine tool in real time, causing the actual laser power and cutting speed to change accordingly; With the help of an internal temperature sensor, continuously track the temperature change trend of the current cutting sub-segment, and observe whether it approaches or exceeds the deformation temperature threshold of the material. If the temperature is still within the safe range, continue to execute the current adjustment plan; if there are signs of temperature increase, then increase or the value of, and send a secondary correction instruction.

[0012] Preferably, the specific steps for deeply analyzing the laser scanning time per unit area of each cutting sub-segment to generate a path scanning time density reference value are as follows: In each cutting sub-segment, construct a time density function to describe the coverage distribution of the laser scanning behavior in the cutting sub-segment range on the spatial heat input. The time density function is defined as follows: , where: represents the scanning time intensity per unit area at the coordinate point , n is the number of all trajectory segments in this cutting sub-segment, is the dwell time of the laser trajectory on the k-th trajectory segment, is an indicator function. When the point falls within the heat-affected area covered by the trajectory segment k, , otherwise it is 0; Generate a path scanning time density reference value based on the time density function. The generated expression is: , where: A is the physical coverage area of the current cutting sub-segment, is the area weight adjustment factor, is the non-linear amplification factor of heat input, is the path scanning time density reference value, indicating the overall path scanning heat density level of this cutting sub-segment.

[0013] Preferably, the specific steps for deeply analyzing the adjacent distance between cutting paths to generate a path adjacent distance reference value are as follows: For any two paths i and j in the same cutting sub-segment, calculate the minimum Euclidean distance between them, and introduce a local proximity function to amplify the influence of short distances on heat accumulation and suppress the interference of long distances on the evaluation result. The expression of the local proximity function is as follows: , where: is the local proximity between paths i and j within the same cutting sub-segment, is the sensitivity coefficient, which is used to control the decay rate of proximity with the increase of distance, and ; is the non-linear exponent, which is used to strengthen or weaken the influence of distance on proximity, ; After obtaining the set of local proximities of all path pairs within the cutting sub-segment, it is synthesized into a path proximity reference value reflecting the overall agglomeration level through an aggregation function, and the following non-linear aggregation method is adopted: , where: is the path proximity reference value, and the symbol ∏ represents the product of all path pairs one by one.

[0014] In the above technical solution, the technical effects and advantages provided by the present invention are as follows: The present invention divides the cutting path generated by CAD into multiple complete cutting sub-segments, and real-time collects the path data of each segment to construct a feature vector. With the help of a trained machine learning model, it automatically predicts the heat accumulation risk degree of each cutting sub-segment, and intelligently adjusts the actual laser power and cutting speed according to the prediction score result, so that the heat in the high accumulation risk area can be controlled and reduced in time and effectively, thereby significantly reducing the occurrence probability of serious deformations such as material warping and bulging, ensuring the overall improvement of the dimensional accuracy, surface quality and yield rate of the processed parts, and at the same time avoiding the efficiency loss and uncertainty brought by human intervention, improving the production stability and the level of automated processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0016] Figure 1 is a schematic diagram of the modules of a laser cutting machine for realizing digital high-precision control according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Now, the exemplary embodiments will be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that this disclosure will be more complete and comprehensive, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0018] The present invention provides as Figure 1A laser cutting machine for realizing digital high-precision control as shown includes a path parsing module, a path segmentation module, an initial parameter configuration module, a feature extraction and risk assessment module, an intelligent scoring prediction module, and a parameter dynamic regulation module; The path parsing module reads the complete cutting path generated from the CAD drawing; The process of reading the CAD drawing to generate the complete cutting path mainly includes several key steps such as drawing parsing, geometric information extraction, path planning, and format conversion. First, the system imports the CAD drawing file (common formats such as DXF, DWG, SVG, etc.), and the CAD parsing unit identifies the geometric elements in the drawing, including basic graphics such as lines, arcs, polylines, and closed contours. Subsequently, based on the topological structure analysis, the relative position relationship, connection order, and layer information of each graphic are identified, and the path elements that need to perform laser cutting are screened out according to the process requirements. Then, preprocessing operations such as continuity check, direction unification, and path sorting are performed on the extracted geometric path to ensure that the path is coherent, without breaks, and without redundant repeated line segments. Finally, this geometric path information is converted into a format recognizable by the numerical control system (such as G code or other intermediate representation languages) to form an executable complete cutting path for subsequent processing flows such as path segmentation and intelligent control.

[0019] The path segmentation module evenly divides the entire cutting path into multiple cutting sub-segments according to a preset fixed length (such as several millimeters or fine-tuned according to material characteristics). During the segmentation process, the geometric features in the corresponding CAD original drawing of each cutting sub-segment are ensured to be complete; The specific steps of dividing the entire cutting path into multiple cutting sub-segments according to the preset fixed length are as follows: First, after completing the extraction and sorting optimization of the geometric path of the CAD drawing, calculate the total length of the entire continuous cutting path; then, according to the preset segmentation length parameter (for example, 10 mm per segment, or dynamically set according to material thermal conductivity, plate thickness, etc.), divide it sequentially along the path starting point according to the geometric distance of the path. Each time a set length is reached, it is marked as the end point of a cutting sub-segment; if a certain path segment is a curve or contains complex corners, an interpolation algorithm is also required for equidistant fitting to ensure that the segmentation point falls on the actual cutting trajectory; during the division process, the system also needs to record information such as the start and end point coordinates, path type, and angle change of each cutting sub-segment to ensure that each cutting sub-segment is geometrically continuous and convenient for subsequent heat accumulation analysis and dynamic control call.

[0020] After the path cutting sub-segment division is completed, the initial parameter configuration module enters the initial configuration stage and sets high laser power and high cutting speed as the default start parameters for the entire cutting process for laser cutting; This setting needs to be based on the current demand for efficiency in industrial processing and is suitable for rapid cutting needs in most low-heat risk areas. Laser power, cutting speed, and auxiliary gas parameters are all loaded into the CNC program at this stage, and the cutting task enters the execution state with this initial configuration. High power + high speed is the current mainstream strategy for increasing productivity in laser cutting. This step ensures that the cutting start has a high-efficiency starting point by standardizing the initial parameter configuration. The core goal of this step is to ensure that the system is "fast as usual" and achieve maximum processing throughput in areas that do not affect quality, while reserving interfaces for subsequent intelligent adjustments.

[0021] The feature extraction and risk assessment module collects the cutting path data of each cutting sub-segment in real time in the host computer software, constructs a feature vector through the cutting path data, and assesses the high heat accumulation risk of the current cutting sub-segment; The intelligent scoring prediction module analyzes the constructed feature vector through a pre-trained machine learning model, outputs a cutting accumulation score, and predicts the cutting path accumulation of each cutting sub-segment based on the score result;

[0022] Construct a feature vector through cutting path data. The specific steps are as follows: Indicators reflecting high concentration of cutting paths are extracted from the collected cutting path data of each cutting sub-segment, wherein the extracted indicators include the laser scanning time within a unit area and the proximity distance between cutting paths. After in-depth analysis of the extracted indicators, path scanning time density reference values ​​and path proximity reference values ​​are generated respectively. The analyzed path scanning time density reference values ​​and path proximity reference values ​​are used as feature vectors to evaluate the high heat accumulation risk of the current cutting sub-segment.

[0023] A longer laser scanning time per unit area of ​​a cutting sub-segment usually indicates that the path concentration of the cutting sub-segment is high. The reason is that the longer the scanning time per unit area, the denser the movement trajectory of the laser head in the area, and there may be path overlap, return, frequent curvature changes, or small structure aggregation, which causes the laser to stay in the area for a longer time, resulting in greater heat input accumulation. Especially in scenarios with complex patterns, fine contours, and hole array structures, high scanning time often means that multiple cutting trajectories are concentrated in the same area, with short heat dissipation paths, difficult heat diffusion, and easy formation of local high-temperature areas. Therefore, the scanning time per unit area not only reflects the complexity and density of the path, but can also be used as an effective indicator to determine whether the cutting sub-segment has the risk of heat accumulation and the "high concentration" of the cutting path.

[0024] The specific steps for performing in-depth analysis on the laser scanning time per unit area of ​​each cutting sub-segment to generate a path scanning time density reference value are as follows: In each cutting sub-segment, a time density function is constructed to describe the coverage distribution of the spatial heat input by the scanning behavior of the laser within the cutting sub-segment range. The time density function is defined as follows: , where: represents the scanning time intensity per unit area at the coordinate point , n is the number of all trajectory segments (or laser movement segments) within this cutting sub-segment, is the dwell time of the laser trajectory on the k-th trajectory segment, is an indicator function. When the point falls within the heat-affected area covered by the trajectory segment k, , otherwise it is 0; The core of this step lies in establishing a "heat field coverage function" for the time response of a spatial point to the laser heat input, revealing the characteristics of the heat coverage density distribution of the path within the region, and accurately depicting the uneven distribution of heat input in space.

[0025] Based on the time density function, a reference value of the path scanning time density is generated. The generated expression is: , where: A is the physical coverage area of the current cutting sub-segment, is the area weight adjustment factor ( when, the smaller the area, the larger the reference value of the path scanning time density, reflecting the high heat accumulation risk in a small area), is the non-linear amplification factor of heat input ( when, emphasizing the influence of the heat input peak area), is the reference value of the path scanning time density, indicating the overall path scanning heat density level of this cutting sub-segment; The physical coverage area of a cutting sub-segment refers to the projected area or effective influence range occupied by the actual geometric region corresponding to a cutting sub-segment on the workpiece material during the cutting task. It is not simply the path length or path boundary, but the area obtained by appropriately defining and quantifying the material region affected by this laser cutting movement, such as the local shape enclosed by this path segment or the smallest closed region including the heat-affected range. The size of this area can be used to evaluate the concentration of laser energy input and heat distribution within this region, thereby more accurately judging the heat accumulation risk.

[0026] By performing power integration of the time density function over the entire cutting sub-segment region and normalizing it by the area, the local heat density peak can be comprehensively transformed into a global heat risk assessment indicator. The higher the reference value of the path scanning time density, the more concentrated the scanning time per unit area, the more obvious the heat accumulation trend, and the greater the deformation risk. Using , Two factors regulate the non-linear sensitivity and area sensitivity, and the model response can be dynamically adjusted according to different materials or thicknesses.

[0027] From the path scan time density reference value, it can be seen that the larger the performance value of the path scan time density reference value generated by the in-depth analysis of the laser scan time per unit area of each cutting sub-segment, the higher the aggregation degree of the cutting sub-segment and the more obvious the heat accumulation trend. Because the path scan time density reference value is an index obtained by weighted integration of the laser scan time per unit area, it reflects the energy input density and spatial coverage intensity of the laser in a specific area. When the path scan time density reference value is large, it means that the laser stays concentrated in the area of the cutting sub-segment, the path is dense or overlaps repeatedly, resulting in a large local heat input and difficult to spread, and there is a significant risk of heat accumulation; on the contrary, if the path scan time density reference value is small, it means that the laser scans are sparsely distributed in space, the path distribution is uniform, the heat input is dispersed, and the heat accumulation is not obvious. Therefore, the path scan time density reference value can be used as an effective criterion for high heat aggregation risk.

[0028] For each cutting sub-segment in the laser cutting path, if the distance between different cutting paths within the cutting sub-segment is very close (that is, the paths are very close to each other), it usually means that the cutting trajectory distribution in this area is very dense, belonging to a high path aggregation degree area. In such an area, the laser irradiates multiple times in a short time at adjacent or repeated positions, resulting in continuous superposition of local heat and difficult to dissipate in time, and it is easy to form a heat accumulation phenomenon. This high-density path layout is common in structures such as small hole arrays, complex engraving patterns or multi-closed contours. Due to limited heat diffusion, materials are more likely to have deformation problems such as local overheating, warping, and bulging. Therefore, "the adjacent distance of the cutting path is relatively low" is a key geometric feature parameter for evaluating the aggregation degree of the cutting sub-segment.

[0029] The specific steps for generating the path adjacent distance reference value by in-depth analysis of the adjacent distance between cutting paths are as follows: For any two paths i and j in the same cutting sub-segment, calculate the minimum Euclidean distance between them , to measure the degree of proximity between these two paths within this cutting sub-segment, a local proximity function is introduced to amplify the influence of short distances on heat accumulation and suppress the interference of long distances on the evaluation results. The expression of the local proximity function is as follows: , where: is the local proximity between path i and j within the same cutting sub-segment, is the sensitivity coefficient, used to control the decay rate of the proximity with the increase of distance, and ; is the non-linear exponent, used to strengthen or weaken the influence of distance on proximity, ; When the minimum Euclidean distance between two paths i and j is very small (the paths are extremely close), is close to 1, indicating a very high risk of heat accumulation for this pair of paths; conversely, if the minimum Euclidean distance between two paths i and j is large, then this exponential decays rapidly and the proximity approaches 0, indicating that the two paths have relatively limited thermal influence; Through this function, each pair of paths can be accurately quantified to obtain a set of proximities for all pairwise combinations of paths within the cut sub-segment ; This step converts the basic geometric information of "distance" into an exponential weight that conforms to the law of heat diffusion, highlighting the contribution of extremely close distances to heat accumulation while weakening the interference of long distances, providing a fine-grained basis for subsequent integration to form a global "path proximity reference value".

[0030] After obtaining the set of local proximities for all pairs of paths within the cut sub-segment, they are synthesized into a path proximity reference value reflecting the overall aggregation level through an aggregation function, using the following non-linear aggregation method: , where: is the path proximity reference value, and the symbol ∏ represents the product of all pairs of paths one by one, The closer it is to 1, the higher the heat aggregation of the corresponding pair of paths. Most of the product factors will decrease accordingly. When there is at least one pair of paths with an extremely short distance within the cut sub-segment, most of the product factors will approach 0, causing the entire path proximity reference value to quickly approach 1. And when the distances of all pairs of paths are far, are all small, and the product result will remain relatively large, thus keeping the path proximity reference value at a low level; This step effectively captures the contribution of "local high proximity" to the overall heat accumulation risk of the cut sub-segment through non-linear aggregation, ensuring that even if there are only a few pairs of paths that are extremely close, the final path proximity reference value can be pushed up, achieving an accurate estimation of the aggregation degree of the cut sub-segment.

[0031] From the path adjacent distance reference value, it can be seen that the larger the performance value of the path adjacent distance reference value generated by the in-depth analysis of the adjacent distance between cutting paths, the denser the spatial distribution between the cutting paths in the cutting sub-segment, the higher the aggregation degree, and thus the more obvious the heat accumulation trend. This is because the path adjacent distance reference value is obtained by non-linearly amplifying the minimum distance between each pair of paths and then integrating through a global aggregation function. It can sensitively capture the areas with extremely small path spacings in the cutting sub-segment. When the path adjacent distance reference value is close to 1, it indicates that there is at least one group or even multiple groups of paths with extremely close distances, and the laser passes through these areas frequently, resulting in the continuous superposition of local heat in a short period of time, forming a significant heat accumulation risk; conversely, when the path adjacent distance reference value is low, it indicates that the path distribution is relatively sparse, and the thermal energy has sufficient time and space to diffuse, and the material is more likely to maintain thermal equilibrium during the cutting process. Therefore, the level of the path adjacent distance reference value can be used as an important basis for measuring the heat accumulation risk level of the cutting sub-segment.

[0032] After analyzing the path scanning time density reference value and the path adjacent distance reference value through a pre-trained machine learning model, a cutting aggregation degree score is output, and this scoring result is recorded as the aggregation index. Based on the aggregation index, the cutting path aggregation degree of each cutting sub-segment is predicted.

[0033] A pre-trained machine learning model refers to a model that, before being formally applied to actual cutting path analysis, has been fully trained and optimized using specific algorithms on the basis of a large amount of historical data, simulation data, or labeled data. This enables the model to effectively evaluate, analyze, and predict input features such as "path scan time density reference value" and "path adjacent distance reference value". During this training stage, technicians first collect various data samples related to cutting paths, such as the actual cutting feedback of different materials (such as stainless steel, aluminum alloy, carbon steel, etc.), CAD drawings with various shape complexities, and the cutting effects presented under different laser power, cutting speed, and auxiliary gas pressure conditions. Then, these data are cleaned, sorted, and labeled to ensure that the information received by the model is complete and accurate. Before model training, the type of machine learning algorithm to be used is determined, such as traditional algorithms like RandomForest, Gradient Boosting Decision Trees (XGBoost, LightGBM, etc.), or deep learning architectures (such as fully connected neural networks, convolutional neural networks, etc.), and corresponding hyperparameters are set for the model. During the training process, technicians use the "path scan time density reference value" and "path adjacent distance reference value" of a large number of samples as inputs, and the actually measured or expert-labeled "cutting defect rate", "thermal deformation degree", or "heat concentration distribution" as labels. By continuously iteratively calculating the loss function and performing gradient descent or decision tree splitting, the model gradually adjusts its internal parameters to maximize the learning of the rule of "under what path overlap and adjacent distance conditions, high heat accumulation, material warping, or other defects will occur in the cutting area". Through such a large amount of data training, the model can automatically extract useful information from the complex parameters and features and form an internal feature weight or rule system. After training is completed, technicians will verify and test the model on a batch of independent data sets to evaluate its accuracy, recall rate, or other metrics in predicting high accumulation risks and determining the level of accumulation. If the accuracy meets the requirements, or after repeated optimization, it reaches the established engineering standards, then this model will be solidified as the "pre-trained machine learning model". During this process, the model not only simply learns how to view the "path scan time density reference value" and "path adjacent distance reference value", but also implicitly understands other potential associated information, such as the overall topological structure of the cutting path, local curvature changes, and the fine-tuning mode of the cutting speed. In this way, when the model is truly deployed to the production environment, as long as a series of numerical values reflecting the characteristics of the current cutting segment are input to it, it can quickly output a "cutting accumulation degree score" or "accumulation risk assessment result", and after new data continuously pours in, it can further improve its adaptability and prediction accuracy to the real scenario through online learning or regular updates.

[0034] In practical applications, this "pre-trained machine learning model" is usually deployed in the control system of laser cutting or the host computer management software, and is tightly integrated with the CAD parsing module, the cutting path planning module, and the real-time monitoring module. When the machine starts cutting, the system first divides the CAD path according to the established logic to obtain several cutting sub-segments, and calculates and extracts the core feature indicators of "path scanning time density reference value" and "path adjacent distance reference value" for each cutting sub-segment. Subsequently, these feature indicators, together with other auxiliary information (such as material thickness, current laser power, cutting speed, ambient temperature, etc.), are packed into an input vector and fed into the model. Based on the internal rules learned from a large amount of training data, the model will output a "cutting accumulation degree score", that is, the so-called accumulation index, in a very short time. This score can accurately measure the risk of heat accumulation in the local area under the conditions of the current cutting sub-segment. In short, the "pre-trained machine learning model" plays a role in this system not only as a static algorithm, but also as a key hub throughout the entire process of data acquisition, feature extraction, decision-making and control, and self-improvement. It is responsible for combining the complex and changeable cutting working conditions with a large amount of historical experience, automatically identifying those high-risk cutting sub-segments that are prone to excessive heat concentration and causing quality defects, and giving targeted control suggestions to help the overall system achieve the best balance in the two dimensions of "efficiency and quality".

[0035] The machine learning model is not specifically limited here. It can realize the comprehensive analysis of the path scanning time density reference value and the path adjacent distance reference value to generate the accumulation index Any deep learning model can be used. To implement the technical solution of the present invention, the present invention provides a specific implementation method; the expression for generating the accumulation index is: , where , are the preset proportionality coefficients of the path scanning time density reference value and the path adjacent distance reference value respectively, and , are both greater than 0.

[0036] The preset proportionality coefficient refers to: when constructing the calculation formula of the accumulation index, in order to reasonably measure the influence weights of two different feature indicators on the final score, two non-zero positive coefficients set artificially or according to experience, data analysis, etc., which are and .

[0037] Specifically, these two coefficients correspond to: : Path scanning time density reference value 's weight; : Path adjacent distance reference value 's weight.

[0038] Because the path scanning time density reference value and the path adjacent distance reference value represent path features of different types and dimensions (one reflects the cutting time per unit area, and the other reflects the path spatial proximity). Their influence degrees on the cutting heat accumulation risk are not exactly the same. Therefore, it is necessary to reconcile and balance through a "preset proportional coefficient" so that the two indicators can play their roles according to their actual influence when calculating the total cutting accumulation degree score. This coefficient can be set based on experience, determined by the expert scoring method, or obtained through machine learning model training, and ensure that 、 , to ensure that both features have a positive contribution to the final score and the scoring system is reasonable and effective.

[0039] From the accumulation index, the larger the performance value of the path scanning time density reference value generated by deeply analyzing the laser scanning time per unit area of each cutting sub-segment, and the larger the performance value of the path adjacent distance reference value generated by deeply analyzing the adjacent distance between cutting paths, that is, the larger the performance value of the accumulation index generated when predicting the cutting path accumulation degree of each cutting sub-segment through a pre-trained machine learning model, it indicates that the degree of concentration of cutting paths in this cutting sub-segment is higher and the heat accumulation trend is more obvious. On the contrary, it indicates that the degree of concentration of cutting paths in this cutting sub-segment is lower and the heat accumulation trend is less obvious.

[0040] The parameter dynamic regulation module, according to the scoring results output by the machine learning model, issues power and speed adjustment instructions to the machine tool numerical control system in real time to intelligently regulate the actual laser power and the actual cutting speed to ensure that the real-time temperature of the cutting area does not exceed the deformation temperature threshold of the material; After the system is started, the upper computer software will first calculate the cutting accumulation degree score of each cutting sub-segment in real time according to the pre-trained machine learning model, and this score is the accumulation index , reflecting the possible heat concentration risk in the current cutting sub-segment. To make the subsequent regulation more targeted, it is necessary to set an accumulation index threshold before startup by process experts or through a large amount of historical experimental data to determine whether the degree of heat accumulation has reached the high-risk level. In addition, to establish a closer connection between the model score and the actual process, it is also necessary to configure the deformation temperature threshold of the material. Once the predicted temperature trend in the cutting area may exceed this upper limit, power reduction or speed reduction intervention must be carried out. The parameter meanings are as follows: : The accumulation index of the current cutting sub-segment output by the machine learning model at time t; : The threshold of the accumulation index set under experience or data-driven, used to distinguish the "safe zone" and the "high heat accumulation risk zone"; : The deformation temperature threshold. Once the local temperature exceeds this value, irreversible damages such as warping and bulging are likely to occur.

[0041] The function of this step is: on the one hand, it provides a unified scale for measuring the heat accumulation risk for subsequent regulation; on the other hand, by collecting and caching the accumulation index corresponding to each cutting sub-segment , it ensures that the system can compare with the threshold at any time and make timely responses.

[0042] Calculate the accumulation index of the current cutting sub-segment and its difference from the accumulation index threshold, and based on the anomaly calculation, determine the specific amplitudes of reducing the laser power and cutting speed, specifically: Use an exponential function with a penalty factor and combine the "zero threshold truncation" concept for calculation. The calculation expression is: , Where: is the accumulation index of the current cutting sub-segment and its difference from the accumulation index threshold, representing the degree of accumulation of the current cutting sub-segment exceeding the threshold. If , then , no power and speed reduction operations are performed, and are respectively the high laser power and high cutting speed set initially during laser cutting, and are the adjustment sensitivity coefficients, used to control the attenuation slopes of the power and speed. The larger the value, the more sensitive it is to the part exceeding the threshold, and represent the actual laser power and actual cutting speed that need to be executed in the current cutting sub-segment; The function of this step is that through an exponential decay calculation mode, it can dynamically, smoothly and controllably reduce the actual power and speed after crossing the threshold. This can not only quickly suppress the excessive local heat accumulation, but also avoid the "sudden high and low" violent fluctuations when approaching the threshold, taking into account both the processing efficiency and the cutting quality.

[0043] After calculating the actual laser power and actual cutting speed, the host computer software will send the actual laser power and the actual cutting speed As a new round of instruction sets, they are sent to the numerical control system of the machine tool in real time, causing corresponding changes in the actual laser power and cutting speed. At the same time, the system uses an internal temperature sensor to continuously track the temperature change trend of the current cutting sub-segment, and focuses on observing whether it approaches or exceeds the deformation temperature threshold of the material. If the temperature is still within the safe range, the current adjustment plan continues to be executed; if there are signs of further temperature increase, the system can further increase or the value, and send a secondary correction instruction to avoid irreversible warping or damage.

[0044] The core significance of this step lies in: Closed-loop execution: The host computer converts the algorithm output into specific numerical control system instructions to complete the closed-loop from "algorithm decision-making" to "physical execution"; Temperature monitoring: By monitoring the proximity of the deformation temperature threshold, ensure that all adjustment decisions are centered around the core goal of "safety without deformation"; Elastic response: In actual production, factors such as workpiece thickness, material properties, and environmental temperature will affect the degree of heat accumulation. Real-time monitoring and multiple adjustments enable the system to have an adaptive ability to timely resolve the thermal shock risk brought by the "high accumulation area".

[0045] In summary, while ensuring processing efficiency, through an innovative closed-loop control mechanism, the risk of deformation caused by local overheating is minimized to the greatest extent, realizing intelligent and refined management of high-power and high-speed laser cutting.

[0046] The above solution realizes a digital high-precision laser cutting control method based on intelligent perception and dynamic feedback regulation. Its prominent beneficial effects are: it can effectively prevent overheating deformation problems in high heat accumulation areas on the basis of ensuring the overall processing efficiency; specifically, by dividing the cutting path generated by CAD into multiple complete cutting sub-segments, and real-time collecting data of each segment path to construct feature vectors, automatically predicting the degree of heat accumulation risk of each cutting sub-segment with a trained machine learning model, and intelligently regulating the actual laser power and cutting speed according to the prediction score results, so that the heat in high accumulation risk areas can be timely and effectively controlled and reduced, thus significantly reducing the occurrence probability of serious deformations such as material warping and bulging, ensuring the comprehensive improvement of the dimensional accuracy, surface quality and yield rate of the processed parts, and at the same time avoiding the efficiency loss and uncertainty brought by human intervention, improving production stability and the level of automated processing.

[0047] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by technicians in this field according to the actual situation.

[0048] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims described above.

[0049] Only some exemplary embodiments of the present invention have been described above by way of illustration. Without doubt, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims

1. A laser cutting machine that realizes digital high-precision control, characterized in that: It includes path parsing module, path segmentation module, initial parameter configuration module, feature extraction and risk assessment module, intelligent scoring prediction module and parameter dynamic control module; Path parsing module, reads the complete cutting path generated by CAD drawings; The path segmentation module evenly divides the entire cutting path into multiple cutting sub-segments according to a preset fixed length, ensuring that the geometric features in the CAD original image corresponding to each cutting sub-segment are complete; Initial parameter configuration module, which performs initial configuration and sets high laser power and high cutting speed for laser cutting as the default startup parameters for the entire cutting process; The feature extraction and risk assessment module collects the cutting path data of each cutting sub-segment in real time, constructs a feature vector based on the cutting path data, and assesses the high heat accumulation risk of the current cutting sub-segment; The intelligent scoring prediction module analyzes the constructed feature vector through a pre-trained machine learning model, outputs a cutting accumulation score, and predicts the cutting path accumulation of each cutting sub-segment based on the score result; The parameter dynamic control module sends power and speed adjustment instructions to the machine tool CNC system in real time according to the scoring results output by the machine learning model, and intelligently controls the actual laser power and actual cutting speed to ensure that the real-time temperature of the cutting area does not exceed the deformation temperature threshold of the material.

2. A laser cutting machine realizing digital high-precision control according to claim 1, characterized in that: Read the CAD drawing to generate a complete cutting path. The specific steps are as follows: Import CAD drawing files and identify the geometric elements in the drawings through the CAD parsing unit; Based on topological structure analysis, the relative position relationship, connection order and layer information of each graphic are identified, and the path elements that need to be laser cut are selected according to the process requirements; Preprocess the extracted geometric path to ensure that the path is continuous, has no breaks, and has no redundant or repeated line segments; Convert the geometric path information into a format that can be recognized by the CNC system to form an executable complete cutting path.

3. The laser cutting machine realizing digital high-precision control according to claim 1, characterized in that: The specific steps of dividing the entire cutting path into multiple cutting sub-segments according to the preset fixed length are as follows: After completing the geometric path extraction and path sorting optimization of the CAD drawing, the total length of the entire continuous cutting path is calculated; According to the preset segment length parameters, the path is divided along the starting point according to the geometric distance of the path. Every time a set length is reached, it is marked as the end point of a cutting sub-segment; If a certain section of the path is a curve or contains complex corners, use the interpolation algorithm to perform equidistant fitting to ensure that the segmentation points fall on the actual cutting trajectory.

4. The laser cutting machine realizing digital high-precision control according to claim 1, characterized in that: Construct a feature vector through cutting path data. The specific steps are as follows: Indicators reflecting high concentration of cutting paths are extracted from the collected cutting path data of each cutting sub-segment, wherein the extracted indicators include the laser scanning time within a unit area and the proximity distance between cutting paths. After in-depth analysis of the extracted indicators, path scanning time density reference values ​​and path proximity reference values ​​are generated respectively. The analyzed path scanning time density reference values ​​and path proximity reference values ​​are used as feature vectors to evaluate the high heat accumulation risk of the current cutting sub-segment.

5. The laser cutting machine realizing digital high-precision control according to claim 4, characterized in that: After analyzing the path scanning time density reference value and the path proximity reference value through a pre-trained machine learning model, a cutting accumulation score is output. The score result is recorded as the accumulation index, and the cutting path accumulation of each cutting sub-segment is predicted based on the accumulation index.

6. The laser cutting machine realizing digital high-precision control according to claim 5, characterized in that: According to the scoring results output by the machine learning model, the actual laser power and actual cutting speed are intelligently controlled. The specific steps are as follows: Calculate the accumulation index of the current cutting sub-segment and its difference with the accumulation index threshold, and calculate the specific amplitude of lowering the laser power and cutting speed based on the abnormality, specifically: Using an exponential function with a penalty factor and combining it with the concept of "zero threshold cutoff", the calculation expression is: , in: The accumulated index of the current cut sub-segment output by the machine learning model at time t, is the accumulation index threshold, is the deformation temperature threshold, is the accumulation index of the current cutting sub-segment and its difference with the accumulation index threshold, indicating the accumulation degree of the current cutting sub-segment exceeding the threshold. ,but , without any power reduction or speed reduction operation. and They are the high laser power and high cutting speed initially set during laser cutting. and To adjust the sensitivity coefficient, used to control the attenuation slope of power and speed, and It indicates the actual laser power and actual cutting speed that need to be executed in the current cutting sub-segment; After the actual laser power and actual cutting speed are calculated, the actual laser power The actual cutting speed As a new round of instruction set, it is sent to the CNC system of the machine tool in real time, causing the actual laser power and cutting speed to change accordingly; With the help of the internal temperature sensor, the temperature change trend of the current cutting sub-segment is continuously tracked to observe whether it approaches or exceeds the deformation temperature threshold of the material. If the temperature is still in the safe range, the current adjustment plan will continue to be implemented; if there is a sign of temperature increase, the temperature will be increased again. or value and send a secondary correction instruction.

7. The laser cutting machine realizing digital high-precision control according to claim 4, characterized in that: The specific steps for performing in-depth analysis on the laser scanning time per unit area of ​​each cutting sub-segment to generate a path scanning time density reference value are as follows: In each cutting sub-segment, a time density function is constructed to describe the coverage distribution of the spatial heat input by the laser scanning behavior within the cutting sub-segment. The time density function is defined as follows: ,in: Indicates coordinate points The scanning time intensity per unit area at , n is the number of all trajectory segments in the cutting sub-segment, is the residence time of the laser trajectory on the kth trajectory segment, is the indicator function, when the point When it falls within the heat affected zone covered by trajectory segment k, , otherwise 0; Generate the path scan time density reference value based on the time density function, and the generated expression is: , where: A is the physical coverage area of ​​the current cut sub-segment, is the area weight adjustment factor, is the nonlinear amplification factor of heat input, It is the path scanning time density reference value, which indicates the path scanning thermal density level of the entire cutting sub-segment.

8. The laser cutting machine realizing digital high-precision control according to claim 4, characterized in that: The specific steps for performing in-depth analysis on the proximity distances between cutting paths to generate path proximity reference values ​​are as follows: For any two paths i and j in the same cutting subsegment, calculate the minimum Euclidean distance between them , a local proximity function is introduced to amplify the effect of short distance on heat accumulation and suppress the interference of long distance on the evaluation results. The expression of the local proximity function is as follows: ,in: is the local proximity between paths i and j in the same cutting subsegment, is a sensitivity coefficient that controls the rate at which proximity decays with distance, and ; It is a nonlinear index used to strengthen or weaken the effect of distance on proximity. ; After obtaining the local proximity set of all path pairs in the cut sub-segment, it is synthesized into a path proximity reference value reflecting the overall aggregation level through an aggregation function, using the following nonlinear aggregation method: ,in: is the reference value of path distance, and the symbol ∏ indicates that The one-by-one product of .

Images