Parameter adjustment method and system of laser processing technology, electronic equipment and storage medium

Abstract

The application provides a parameter adjustment method and system of a laser processing technology, an electronic device and a storage medium. The parameter adjustment method of the laser processing technology comprises the following steps: acquiring image data of a cutting section and process parameters; performing feature extraction on the image of the cutting section to obtain multi-dimensional morphological features, wherein the multi-dimensional morphological features comprise slag hanging features, burning features, roughness features and texture trend features; establishing a correlation model of each dimension morphological feature and the corresponding process parameter according to the multi-dimensional morphological features and the process parameters; storing the correlation model of each dimension morphological feature and the corresponding process parameter in a specified database to generate an image feature model library; and inputting real-time image features into the image feature model library to obtain parameter adjustment information.

Description

Technical Field

[0001] This application relates to the technical field of parameter adjustment methods for laser processing technology, and more particularly to a parameter adjustment method, system, electronic device and storage medium for laser processing technology. Background Technology

[0002] With the development of modern precision manufacturing technology, laser processing has been widely used in aerospace, automotive manufacturing, and electronic information fields due to its non-contact, high precision, and high efficiency. In processes such as laser cutting, drilling, or welding, the quality of the cross-section formed by the laser beam acting on the material (such as kerf roughness, slag adhesion, and the extent of the heat-affected zone) directly determines the final performance and pass rate of the workpiece.

[0003] In traditional laser processing, when adjusting the laser processing technology to improve the workpiece qualification rate, it mainly relies on manual judgment and experience analysis to adjust the process, which cannot meet the needs of large-volume continuous laser processing. Summary of the Invention

[0004] This application provides a method, system, electronic device, and storage medium for adjusting parameters in laser processing, to solve the problems existing in related technologies. The technical solution is as follows: In a first aspect, embodiments of this application provide a method for adjusting parameters of a laser processing technology, including: Acquire image data and process parameters of the cut cross-section; Feature extraction is performed on the image of the cut section to obtain multi-dimensional morphological features, including: slag features, overheating features, roughness features, and texture trend features. Based on multi-dimensional morphological characteristics and process parameters, establish a correlation model between morphological characteristics and corresponding process parameters in each dimension; The association model of each dimension of morphological features and corresponding process parameters is stored in a specified database to generate an image feature model library; Real-time image features are input into the image feature model library to obtain parameter adjustment information.

[0005] Secondly, embodiments of this application provide a parameter adjustment method system for laser processing technology, comprising: The first acquisition module is used to acquire image data and process parameters of the cut cross-section; The first module is used to extract features from the image of the cut section to obtain multi-dimensional morphological features, which include: slag features, overheating features, roughness features, and texture trend features. The first module is used to establish a correlation model between morphological features and corresponding process parameters based on multi-dimensional morphological features and process parameters. The first generation module is used to store the association model of morphological features and corresponding process parameters in a specified database to generate an image feature model library. The first module is used to input real-time image features into the image feature model library to obtain parameter adjustment information.

[0006] Thirdly, embodiments of this application provide an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the parameter adjustment method of the laser processing process described above.

[0007] Fourthly, embodiments of this application provide a computer-readable storage medium that stores computer instructions, wherein when the computer instructions are executed on a computer, the methods in any of the above-described embodiments are performed.

[0008] The advantages or beneficial effects of the above technical solutions include at least the following: In this embodiment, the parameter adjustment method for this laser processing technology extracts multi-dimensional morphological features such as slag adhesion, overheating, roughness, and texture trend from the image data of the cut surface. Based on these multi-dimensional morphological features and process parameters, a correlation model is established between each morphological feature and its corresponding process parameter. This correlation model is stored in a designated database to generate an image feature model library. Real-time image features are input into this library to obtain parameter adjustment information. This enables rapid correlation determination of processing quality and process parameters during laser processing, facilitating subsequent process recommendations, peripheral equipment fault diagnosis, and parameter optimization. Based on the morphological features of processing quality, optimized process parameter recommendations are obtained, replacing traditional manual judgment and experience-based analysis. This optimizes the detection and process parameter optimization stages of automated laser processing, making it more objective and automated. It reduces quality variations caused by human factors during laser processing, maintains relatively stable parameter optimization for laser processing, and ensures real-time adjustments based on workpiece quality during processing, guaranteeing the consistency and stability of the output workpiece quality.

[0009] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of this application will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0010] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in this application and should not be construed as limiting the scope of this application.

[0011] Figure 1 This is a flowchart of a parameter adjustment method for a laser processing technology according to an embodiment of this application.

[0012] Figure 2 The image shows a physical workpiece with the slag thickness and distribution density in an embodiment of this application.

[0013] Figure 3 A photograph of a workpiece in an embodiment of this application shows an overheated and identifiable overheating zone.

[0014] Figure 4 A physical image of a workpiece representing the roughness in an embodiment of this application is shown.

[0015] Figure 5 This illustration shows a schematic diagram of the evaluation of the texture as a tangent trend line according to the fitted curve in an embodiment of this application.

[0016] Figure 6 This is a structural block diagram of an electronic device according to an embodiment of the present application. Detailed Implementation

[0017] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0018] Image detection of the cross-sectional quality of materials affected by laser processing beams is currently the technological foundation for realizing automatic process adjustment and process diagnosis. The rapid calibration of process adjustment amounts and schemes through data classification of image processing is an important support for the embodiments of this application. Currently, the construction of an image feature model library is an important technical support for realizing machine learning and processing diagnosis. This application provides a model association between the detection and image feature database for laser cutting quality diagnosis, which is the foundation for realizing quality control based on detection classification and closed-loop recommendation process.

[0019] By using an image feature model library for workpiece quality classification through image analysis, the quality judgment of laser-cut cross-sections and quantitative analysis of quality after peripheral equipment diagnosis were realized. This enabled precise quantitative evaluation of an image-based process optimization association recommendation mechanism, which has constrained the formation of a closed loop for parameter adjustment methods in laser processing technology.

[0020] Figure 1 A flowchart illustrating a parameter adjustment method for a laser processing technology according to an embodiment of this application is shown. Figure 1-5 As shown, a parameter adjustment method for a laser processing technology may include: S110: Acquire image data and process parameters of the cut section; S120: Extract features from the image of the cut section to obtain multi-dimensional morphological features, including: slag features, overheating features, roughness features, and texture trend features; S130: Based on multi-dimensional morphological characteristics and process parameters, establish a correlation model between morphological characteristics and corresponding process parameters in each dimension; S140: Store the association model of each dimension of morphological features and corresponding process parameters into the specified database to generate an image feature model library; S150: Input real-time image features into the image feature model library to obtain parameter adjustment information.

[0021] The parameter adjustment method for laser processing in this embodiment is applicable to the laser processing of common thickness metal materials such as stainless steel, carbon steel, and aluminum alloy. It can also be applied to laser cutting experiments and laser processing production scenarios using different types of lasers and different cutting heads on common thickness materials such as stainless steel, carbon steel, and aluminum alloy.

[0022] In this embodiment, the parameter adjustment method for this laser processing technology extracts multi-dimensional morphological features such as slag adhesion, overheating, roughness, and texture trend from the image data of the cut surface. Based on these multi-dimensional morphological features and process parameters, a correlation model is established between each morphological feature and its corresponding process parameter. This correlation model is stored in a designated database to generate an image feature model library. Real-time image features are input into this library to obtain parameter adjustment information. This enables rapid correlation determination of processing quality and process parameters during laser processing, facilitating subsequent process recommendations, peripheral equipment fault diagnosis, and parameter optimization. Based on the morphological features of processing quality, optimized process parameter recommendations are obtained, replacing traditional manual judgment and experience-based analysis. This optimizes the detection and process parameter optimization stages of automated laser processing, making it more objective and automated. It reduces quality variations caused by human factors during laser processing, maintains relatively stable parameter optimization for laser processing, and ensures real-time adjustments based on workpiece quality during processing, guaranteeing the consistency and stability of the output workpiece quality.

[0023] In step S110, image data and process parameters of the cut section are acquired.

[0024] In this embodiment, laser processing equipment can be driven to perform laser processing on common engineering thickness metal materials such as stainless steel, carbon steel, and aluminum alloys through experiments or pre-processing.

[0025] Due to the presence of bright sparks and stray light at the laser cutting site, this embodiment employs a light source filtering calibration scheme to obtain high-quality original images of the cut surface. Specifically, an industrial camera light source of a specific wavelength illuminates the workpiece cutting surface, and a narrow-band filter is placed in front of the lens to filter out background stray light, retaining only the characteristic wavelength signals reflected from the cutting surface. An industrial-grade high-resolution camera + light (to avoid reflection interference) is used to capture a 90-degree image of the cutting surface, covering key areas such as "slag distribution, overheated areas, and roughness texture." Grayscale conversion and noise reduction are performed using Gaussian filtering (to eliminate random noise). Simultaneously, the core process parameters of the workpiece during processing are read from the CNC controller of the laser cutting machine, mainly including: laser power, processing speed, gas flow rate, sheet thickness, beam type, and line scan laser detection data.

[0026] In step S120, feature extraction is performed on the image of the cut surface to obtain multi-dimensional morphological features, which include: slag features, overheating features, roughness features, and texture trend features.

[0027] In this embodiment, the multi-dimensional morphological features include: slag features, overheating features, roughness features, and texture trend features. That is, by extracting these features from the image of the cut section, we can obtain slag features, overheating features, roughness features, and texture trend features, as detailed below: like Figure 2 As shown, the slag features are obtained by extracting the images of the cut sections separately.

[0028] The acquired original image is preprocessed (e.g., denoising, binarization) and extracted: the slag-covered area at the bottom of the image is segmented, and the slag thickness value (pixel height) and distribution density value (percentage of slag pixels per unit length) are calculated using pixel statistics. The slag thickness value and distribution density value can characterize the slag-covered features.

[0029] For example: read the thickness information of the workpiece plate, and in the image of the cut surface, take the lower edge of the cut surface as the baseline, and extend it upward and downward by a preset pixel height (e.g., 50 pixels each) to form the bottom slag detection zone (Bottom-ROI).

[0030] Image segmentation is performed on the bottom-ROI of the slag detection zone. Since slag is usually formed by the cooling of molten metal, its surface reflectivity differs significantly from the surrounding background (air or grid lines). An adaptive threshold segmentation algorithm (such as the Otsu algorithm) is used to binarize the ROI region. Regions with gray values ​​higher than the threshold are marked as "slag foreground pixels," and the rest are marked as background.

[0031] Morphological closing operations are performed on the binarized image to fill in the tiny voids inside the slag-covered area and connect the broken slag fragments, thereby obtaining a complete slag-covered connected region.

[0032] After obtaining the segmented connected components of the slag coating, specific quantitative indicators are extracted based on pixel statistics. The slag coating feature value M is defined, and the quantitative indicators include the slag coating thickness value (H). avg） and distribution density value (D ensity） Traverse the horizontal direction (X-axis) of the bottom-ROI of the dross detection zone, calculate the total number of pixels belonging to the "dross foreground pixels" in each column, and convert it to physical size (based on the camera's pixel equivalent, e.g., mm / pixel). Calculate the average height of all columns and record it as the dross thickness value. The dross thickness value characterizes the degree of vertical dross of the dross.

[0033] The ratio of the total area of ​​the "slag foreground pixels" in the Bottom-ROI of the slag detection zone to the horizontal width of the ROI is calculated to obtain the distribution density value. The distribution density value characterizes the continuity and density of the slag on the cutting path.

[0034] The dimensionless slag characteristic value S is generated by weighted summation:

[0035] Wherein, α and β are normalized weighting coefficients used to balance the influence of thickness and density on quality evaluation.

[0036] The slag characteristics of the cut section image can be directly determined by the slag thickness and distribution density values.

[0037] like Figure 3 As shown, the overburning features are obtained by extracting the image of the cut section.

[0038] In the laser cutting process, over-burning usually occurs at the moment when the heat input is out of control. It manifests as severe oxidation and melting instability of the material, resulting in huge holes or accumulations on the cut surface.

[0039] In the embodiments of this application, when the core energy density change of the laser exceeds the material's tolerance limit, or when the airflow cannot remove the accumulated heat in time, the reaction heat (especially the exothermic oxidation when oxygen cuts carbon steel) will burst instantly, causing the local temperature of the material to rise sharply and resulting in "runaway melting".

[0040] The most visually significant characteristic of overheating is the drastic change in surface reflectivity, for example, image processing mainly revolves around the grayscale of the overheated image range.

[0041] Gray-scale statistics are performed on the ROI (Region of Interest) of the cut surface. A normal cut surface usually presents a uniform gray texture. Since overheating will produce two extreme phenomena at the same time, two gray-scale threshold ranges are set: the high-intensity zone and the low-intensity zone.

[0042] The extremely bright range corresponds to severe oxide buildup caused by overheating. Due to the rough surface of iron oxide and the potential for glassy reflection, it will appear abnormally bright under certain light sources. The system counts the area of ​​pixels with grayscale values ​​higher than the threshold ThighThigh.

[0043] Extremely dark areas correspond to "holes" or "pits" caused by overheating. These areas appear as black patches in the image because light is trapped (and cannot be reflected back to the camera). The system statistically analyzes grayscale values ​​below a threshold T. low The pixel area.

[0044] Let S be the sum of the pixel areas of these two intervals. gray_abnormal , as a two-dimensional visual feature of overheating.

[0045] Grayscale alone can sometimes be affected by oil or rust, so depth information must be introduced to confirm whether physical deformation has occurred on the surface.

[0046] Using the line structured light (Line Laser) projection in this embodiment, when the red laser line is irradiated on a flat cut surface, it is a straight line; when it is irradiated on an overheated area, the light strip will twist and break.

[0047] Depression: If the laser line is recessed inward (manifested as a shift in the position of the light stripe on the image sensor), it indicates that material is missing at that location, forming a pit. This is the most typical "burn-through" characteristic of overheating.

[0048] Protrusion: If the laser line protrudes outward, it indicates that there is molten material buildup at that location. This is a characteristic of "molten buildup" caused by overheating.

[0049] Calculate the collapse depth D using the principle of triangulation.epth and the height of the protrusion H eight The area of ​​the region exceeding the preset deformation threshold is denoted as S. depth_deform .

[0050] The above method allows for the extraction of overburnt features (collapse depth D) from images of the cut surfaces. epth , Height of protrusion H eight The area of ​​the region exceeding the preset deformation threshold is denoted as S. depth_deform ).

[0051] like Figure 4 As shown, roughness features are obtained by extracting the images of the cut surfaces separately.

[0052] Roughness refers to the minute peaks and valleys on the cut surface, which directly affects the subsequent assembly accuracy and appearance quality of the workpiece. The formation of striations on a laser-machined surface is influenced by various mechanical factors. For precise quantification, the cut surface is divided into three zones along the height direction: Top Zone: Located near the top surface of the plate. The roughness here is mainly affected by the laser beam quality (spot shape) and the initial state of the auxiliary gas flow field.

[0053] Middle Zone: The middle part of the cut surface. The striations here are usually smoother, mainly affected by the cutting speed and the stability of the molten metal flow.

[0054] Bottom Zone: Located near the bottom surface of the plate. This area is most prone to severe drag streaks, mainly affected by slag removal capacity (the interplay between airflow and molten liquid viscosity).

[0055] Since roughness is a microscopic texture, the focus is on analyzing the texture details of the image. High-pass filtering is applied to the acquired cut surface image to enhance the subtle vertical stripe features. Then, based on the plate thickness and pixel height, the image is automatically segmented into three ROIs: upper, middle, and lower.

[0056] For each region, a gray-level co-occurrence matrix is ​​calculated. Contrast and entropy are extracted as surrogate values ​​for roughness: high contrast indicates deep texture grooves and high roughness, while high entropy indicates chaotic texture and poor surface quality. Line-scanned laser data is introduced to detect the frequency and amplitude of minute jitter of the laser line as it passes through the cutting surface stripes. This is used as a ground truth calibration of the image texture features, transforming abstract texture parameters into specific Ra arithmetic mean (micrometers), i.e., roughness features.

[0057] like Figure 5 As shown, texture trend features are obtained by extracting the image of the cut surface.

[0058] In laser cutting, texture trends manifest as "drag lines" or "striations" on the cut surface. These are the trails left by the molten metal as it flows downwards under the influence of the airflow after the laser beam melts the metal. Ideally, if the airflow removes the molten metal very quickly and the processing speed is moderate, the drag lines should be perpendicular to the surface of the sheet (i.e., 90° perpendicular). In actual processing, because the laser cutting head is moving and the molten metal takes time to flow downwards, the position of the molten metal at the bottom is always "lagging" the center of the laser beam at the top. This "phase lag" in position creates the backward-curving drag lines.

[0059] The image of the cut surface is sharpened to enhance the edge contrast of the vertical texture. The Canny edge detection operator or the Sobel vertical operator is used to extract all the edges of the vertical stripes on the cut surface. The resulting lines may be discontinuous and noisy. To eliminate local noise interference and obtain the overall texture direction, the Least Squares Method can be used for fitting. The lower-middle region of the image with the clearest texture is selected. Assuming the dragged line is approximately a straight line (or parabola), its equation is y = kx + b, where y is the vertical height, x is the horizontal position, and k is the slope, i.e., the texture trend feature. The value of k that minimizes the sum of the squared distances from all edge points to this line is found. This k value represents the average tilt angle of the dragged line across the entire image.

[0060] The above embodiments can be used to obtain slag characteristics, overheating characteristics, roughness characteristics, and texture trend characteristics.

[0061] Besides the methods mentioned above, the image of the cut surface can also be input into a trained slag-coating neural network to extract slag-coating features. Similarly, the image can be input into a trained overheating neural network to extract overheating features. Furthermore, the image can be input into a trained roughness neural network to extract roughness features. Finally, the image can be input into a trained texture trend neural network to extract texture trend features. These corresponding trained neural networks can directly yield slag-coating features, overheating features, roughness features, and texture trend features.

[0062] In step S130, a correlation model between the morphological features and corresponding process parameters of each dimension is established based on the multi-dimensional morphological features and process parameters.

[0063] Step S120 yields multi-dimensional morphological features, including slag adhesion features, overheating features, roughness features, and texture trend features.

[0064] For morphological features such as slag adhesion, overburning, roughness, and texture trend, correlation models are established for each dimension of morphological features and corresponding process parameters.

[0065] By correlating the slag adhesion characteristics with the corresponding process parameters, a correlation model between the slag adhesion characteristics and the corresponding process parameters is obtained, as follows: The physical field characteristics of laser cutting change drastically with the power level. Therefore, the current laser power parameter (P) is first read, and combined with the material type (such as carbon steel), plate thickness, and beam type, the corresponding "basic process sub-library" is retrieved from the database.

[0066] For example, if the current value is detected as P=30kW and the material is 20mm carbon steel, then the "30kW-CS-20mm sub-library" is invoked. This sub-library limits the effective range of subsequent fitting coefficients (K).

[0067] To reveal the fluid dynamics and thermodynamics of slag formation, the basic process parameters are transformed into two core competing physical variables: the core energy density change and the gas flow change. The core energy density change (E... density） The calculation formula is as follows: E density =P / V f Where: laser power P and processing speed V f The ratio of core energy density to molten metal density. This value represents the thermal energy absorbed per unit length of cut. The higher the change in core energy density, the more molten metal is produced, and the more abundant the potential "raw material" for slag formation.

[0068] airflow change (F) low） The calculation formula for F: low =Q i (Gas flow rate) Among them, Q i The gas flow rate, the mechanical kinetic energy of the auxiliary gas, and the role of the airflow in providing shear force to blow molten metal out from the bottom of the cut are all considered. The stronger the airflow, the stronger the slag removal capacity. Slag formation is essentially an imbalance between the core energy density change and the airflow change. Based on this, in the selected basic process sub-library, the following correlation model between slag characteristics and corresponding process parameters is established: +

[0069] Where k1, k2, and k3 are fitting coefficients, P is power, and V is... f Q represents the processing speed. i This represents the gas flow rate.

[0070] The coefficient k1 is positive. This indicates that as the change in core energy density increases, the amount of molten liquid increases, and with the gas flow remaining constant, the slag characteristic value S shows a positive correlation trend (more slag).

[0071] This coefficient A negative value indicates that the gas flow rate Q increases with the gas flow rate. i With the increase of , the airflow shearing and slag removal capacity is enhanced, and the slag-hanging characteristic value S shows a negative correlation trend (less slag).

[0072] k3: Intercept term / correction factor. Represents the basic slagging tendency under current material properties (such as viscosity, surface tension).

[0073] For example: The machine is cutting 20mm carbon steel, and the system calculates that the current slag characteristic value S is too high (slag is present). If the model calculation shows... If the value of the item is too large, it is judged as "excessive core energy density change leading to excessive melting", and the output adjustment suggestion is: increase the processing speed V. f .

[0074] If the model calculation shows If the value of the item is too small (small absolute value), it is determined that "insufficient gas flow leads to poor slag discharge". The output adjustment suggestion is to increase the gas flow rate Q. i .

[0075] By correlating the overburning characteristics with the corresponding process parameters, a correlation model between the overburning characteristics and the corresponding process parameters is obtained, as follows: Based on the current basic classification (e.g., 30kW, 20mm carbon steel), a threshold for the change in the core energy density due to overheating is set (E). limit） This threshold is the critical point at which the material experiences thermal runaway.

[0076] The correlation model between overburning characteristics and corresponding process parameters is established as follows:

[0077] When the actual core energy density change Less than E limit At that time, Feature overburn It should be close to 0 (no overheating); When the actual core energy density change exceeds Elimit At that time, the total area of ​​the pixels extracted from the image, Sg ray_abnormal The area S of the region with the preset deformation threshold depth_deform It will rise sharply. If the image features show large areas of "collapsed black spots" (confirmed by both depth and grayscale), and the calculated values ​​at this time... If the temperature is indeed high, it can be diagnosed as overheating caused by excessive changes in core energy density, and instructions to reduce power or significantly increase cutting speed will be automatically generated.

[0078] By correlating the roughness features with the corresponding process parameters, a correlation model between the roughness features and the corresponding process parameters is obtained, as follows: Read the current process parameters and lock the basic model library (e.g., 20mm carbon steel - oxygen cutting - annular spot mode). The same roughness value (e.g., Ra=12.5μm) may indicate insufficient gas pressure on carbon steel, while it may indicate excessive speed on stainless steel. Use the line-scan laser sensor in the device to acquire the true microscopic contour data of the cut surface (arithmetic mean deviation Ra value), and map the dimensionless texture contrast / entropy value extracted by the vision system into physically meaningful micron-level values.

[0079] The zone-based mechanism attribution reveals significant differences in the roughness of laser-processed cross-sections along the height direction. The roughness correlation model in the upper region is dominated by beam and thermal effects, and the texture quality at the top of the cut surface primarily depends on the uniformity of energy distribution at the moment the laser beam enters the material.

[0080] Associated parameters include focus position (F) pos） Laser power (P) and beam mode (B) mode） .

[0081] Correlation model between roughness characteristics of the upper region and corresponding process parameters:

[0082] Among them, F opt The optimal focal position for the current material. This formula expresses: Focal position F pos The further away from the optimal point, or the greater the power fluctuation P fluc The larger the value, the roughness Ra of the upper region. top The worse it is (the higher the value).

[0083] The central region is located at the waist of the cut surface. In this region, the laser beam has completely penetrated the material, and the airflow has not yet experienced severe turbulent divergence.

[0084] In the central region, there is neither interference from the top spot focal effect nor the bottom slag hysteresis effect, making it the region that most objectively reflects the change in core energy density. With gas flow (Q) i The region of the matching relationship.

[0085] Current process parameters include material (e.g., carbon steel), plate thickness (e.g., 10 mm), gas type (e.g., oxygen), and beam type. These parameters determine the baseline for the coefficients (K-values) in the subsequent correlation model.

[0086] Change in core energy density: ; Airflow change: ,in This refers to the gas flow rate; Correlation model between roughness characteristics and corresponding process parameters in the central region:

[0087] Here, coefficient A represents the change in core energy density. In the central region, the change in core energy density primarily affects the sufficiency of melting.

[0088] If A is positively correlated, it means that excess energy leads to overheating and rough texture; if A is negatively correlated, it means that insufficient energy leads to incomplete cutting and the formation of stripes.

[0089] Coefficient B represents the airflow variation. In the central region, the airflow primarily serves to cool and evaporate air, maintaining stability within the cut.

[0090] The constant C is the inherent texture basis of the material.

[0091] The roughness of the lower region of the cut surface (often manifested as wavy lines) is caused by the instability resulting from the interaction between the shear force of the airflow and the surface tension during the downward flow of molten metal.

[0092] Related parameters include processing speed (V) f） Gas flow rate (Q) i） air pressure (P) r） .

[0093] The correlation model between the roughness characteristics of the central region and the corresponding process parameters can be represented by the slag removal ratio (λ), and the specific calculation formula is as follows:

[0094] Among them, when the processing speed V f Increase (the amount of melting increases) while the gas flow rate Q iWhen the ratio fails to increase synchronously, it becomes larger, causing the molten liquid to be unable to drain smoothly, resulting in turbulence and stagnation at the bottom, which in turn increases the surface roughness Ra. bottom It exhibits exponential γ-level deterioration. If severe roughness is detected in the lower region, the correlation model will directly base it on... The relationship is used to calculate how much speed reduction or air pressure increase is needed to make Ra... bottom Returning to the normal range.

[0095] Through the correlation model of the upper, middle and lower regions mentioned above, the correlation model between roughness characteristics and corresponding process parameters is determined.

[0096] By associating texture trend features with corresponding process parameters, a correlation model between texture trend features and corresponding process parameters is obtained, as follows: Based on the feature extraction in step S120, the tangent slope k can be obtained. The drag angle (i.e., the physical mapping of slope k) reflects the ratio of feed rate to slag discharge rate. That is, the correlation model between texture trend features and corresponding process parameters can be:

[0097] Since the material viscosity can be considered constant after primary classification (e.g., selecting carbon steel), the correlation model between texture trend characteristics and corresponding process parameters can be simplified as follows:

[0098] Where γ is the material sensitivity coefficient, a proportionality constant determined by material properties (such as carbon steel and stainless steel in the primary classification) and plate thickness; δ is the inherent deviation constant, the intercept term in linear fitting. V f The function of (processing speed): V f The larger the value, the faster the cutting head moves, the more molten slag falls behind, and the more severe the tilt of the drag line (the slope k increases). Q i The role of (gas flow rate): Q i The larger the value, the greater the downward momentum of the airflow, which can push the molten slag to the bottom more quickly, reduce lag, and straighten the drag line (reduce the slope k).

[0099] Through the above embodiments, correlation models for various morphological features and their corresponding process parameters were determined, including correlation models for slag adhesion features and their corresponding process parameters, overburning features and their corresponding process parameters, roughness features and their corresponding process parameters, and texture trend features and their corresponding process parameters.

[0100] In step S140, the correlation model of each dimension of morphological features and corresponding process parameters is stored in a specified database to generate an image feature model library.

[0101] In this embodiment, the correlation models between various morphological features and corresponding process parameters, such as the correlation model between slag adhesion features and corresponding process parameters, the correlation model between overburning features and corresponding process parameters, the correlation model between roughness features and corresponding process parameters, and the correlation model between texture trend features and corresponding process parameters, which have been determined in step S130, are used to establish a multi-dimensional process expert system database.

[0102] For example: Create multi-level index keys in the database to distinguish different working environments. Establish the skeleton of the image feature model library: Level 1 Index: Material type (e.g., carbon steel, stainless steel, aluminum alloy, etc.); Level 2 Index: Plate thickness (e.g., 6mm, 12mm, 20mm, etc.); Level 3 Index: Gas type (e.g., oxygen O2, nitrogen N2, air); Level 4 Index: Beam type (e.g., beam characteristics determined according to the selected spot pattern or nozzle type).

[0103] This ensures the specialization of the image feature model library. Under the same roughness value, the corresponding process adjustment logic is completely different under the indexes of "20mm carbon steel + oxygen" and "6mm stainless steel + nitrogen".

[0104] In each specific storage unit determined by the above index, the association model of each dimension of morphological features and corresponding process parameters determined in the previous step is stored.

[0105] Slag-coating model data package: stores slag-coating characteristics (such as slag-coating area) and gas flow rate (Q). i） And the threshold relationship of the focal position. Under this operating condition, the maximum allowable slag area threshold, and the response coefficient of airflow adjustment.

[0106] Overburning model data package: stores overburning characteristics (such as melted fillet) and core energy density changes. The upper limit constraint relationship. Under this operating condition, the critical value of the core energy density change when the material overheats.

[0107] Roughness model data package: stores the changes in central roughness and core energy density ( The fitting coefficients for the gas flow rate (Qi) and the roughness of the material are determined. By reading these parameters, it can be determined whether the roughness of the material is primarily determined by energy or by the gas flow.

[0108] Texture trend model data package: stores the slope of texture drag lines and the velocity / airflow ratio ( The linear regression coefficients (i.e., material sensitivity coefficient γ and intercept δ) of the material are used to calculate how much speed adjustment is needed to correct the drag line.

[0109] Once the aforementioned correlation model data for different materials and thicknesses are entered and structured for storage, an image feature model library is generated.

[0110] In step S150, the real-time image features are input into the image feature model library to obtain parameter adjustment information.

[0111] In step S140 above, the image feature model library has already been obtained. In this embodiment, the current machining state of the machine tool is acquired, i.e., the primary classification basis, for example: currently machining 20mm carbon steel using oxygen cutting with a single-layer nozzle. Based on this classification, a set of associated model parameters applicable only to this working condition is retrieved from the image feature model library. What is retrieved is not just data, but also the sensitivity weights for the material. For example, for 20mm carbon steel, the model library records that speed has a significant impact on the drag line (high weight), while for thin stainless steel, it may record that air pressure has a significant impact on burrs. This step establishes the benchmark for subsequent adjustments.

[0112] Reverse diagnostics based on multidimensional features compares real-time extracted image features with correlation models in a retrieved image feature model library to detect the drag line hysteresis (slope) in real time. Based on the correlation model between texture trend features and corresponding process parameters: if the slope is too large (severe hysteresis), the parameters in the image feature model library indicate that the material is primarily dominated by cutting speed. Using the slope sensitivity coefficient in the correlation model between texture trend features and corresponding process parameters, the amount of excess speed that needs to be offset to reduce the slope to 0 (vertical) is calculated. Adjusted speed reduction values ​​are generated (e.g., V needs to be reduced). f 300 mm / min).

[0113] The real-time texture grayscale variance (roughness) of the central region is analyzed based on the correlation model between roughness features and corresponding process parameters: weight coefficients are read from the image feature model library. If the energy term coefficient is much larger than the airflow term coefficient under this condition, it indicates that the current roughness is mainly caused by unstable heat input. Conversely, if the airflow term coefficient is large, it indicates that the roughness is caused by turbulent slag discharge airflow.

[0114] Generate adjustment parameter values: If it is an energy-type roughening: calculate how much laser power or duty cycle needs to be adjusted to stabilize the melting front; if it is a gas flow-type roughening: calculate how much auxiliary gas pressure needs to be adjusted to smooth the gas flow field.

[0115] The area of ​​slag buildup at the bottom or the rounded corners of overheated areas at the top are determined by correlation models between overheating characteristics and corresponding process parameters, as well as correlation models between slag buildup characteristics and corresponding process parameters. Slag buildup directly corresponds to insufficient airflow drag force. The correlation models between slag buildup characteristics and corresponding process parameters calculate how much additional air pressure is needed to generate sufficient momentum to blow away the molten slag in that area.

[0116] Overheating directly corresponds to an overload in the core energy density change. The correlation model between overheating characteristics and corresponding process parameters calculates how much the duty cycle needs to be reduced to bring the heat input back within the material's tolerance threshold.

[0117] After completing the independent calculations for each of the above dimensions, all the adjustment parameter values ​​are compiled into the final parameter adjustment information.

[0118] The parameter adjustment method for laser processing described in this embodiment establishes a correlation model between image features (such as drag line slope and roughness variance) and core process parameters (such as core energy density change and gas flow rate) by constructing an image feature model library. This enables closed-loop intelligent control of the laser cutting process, ensuring the continuous stability of the cut surface. In real-time laser processing, by recognizing the image of the cut surface, even slight quality degradation trends (such as minor changes in the drag line) can be corrected by fine-tuning parameters before defects become irreversible scrap. This significantly reduces reliance on highly skilled operators, minimizes waste, and improves production efficiency.

[0119] In one embodiment of this application, the slag features include slag thickness and distribution density; feature extraction is performed on the image of the cut section to obtain multi-dimensional morphological features, including: Image segmentation is performed on the image of the cut section to obtain the segmented cut section image; The distribution of slag pixels at the bottom of the cut section in the segmented image is statistically analyzed to obtain the slag thickness and distribution density values.

[0120] In the embodiments of this application, the processing range of the cut surface image is locked to the bottom region of the cut surface to obtain a segmented cut surface image. Within the bottom region of the segmented cut surface image, dross appears as irregular protrusions attached to the lower edge of the board. Segmentation is performed using the physical differences in grayscale value (brightness) and texture continuity between the dross and the normal cut surface. Grayscale difference: Due to surface oxidation or different reflection angles, the brightness value of the slag is usually significantly higher or lower than that of the surrounding background (cut-out gaps). A grayscale threshold range is set to binarize the pixels in the bottom area. Pixels that match the brightness characteristics of the slag are marked as "foreground (1)" and the remaining background pixels are marked as "background (0)". A binarized mask image (Segmented Image) is obtained that only retains the outline of the slag shape.

[0121] The slag thickness value refers to the cumulative amount of molten slag in the vertical direction. In the segmented cross-sectional image, each pixel column is scanned along the direction perpendicular to the plate surface (i.e., the Y-axis direction). The number of consecutive pixels marked as "slag" in each column is counted. The number of pixels is multiplied by the physical size represented by a single pixel (e.g., micrometers / pixel) to obtain the actual length of the slag at that location. The maximum or average value among all columns is selected as the slag thickness value.

[0122] The slag thickness represents the length of the molten slag that has fallen downwards. The larger the thickness, the less momentum the downward airflow has to overcome the surface tension of the molten droplets, which is a key factor in determining whether to increase the air pressure.

[0123] The distribution density value refers to the coverage area of ​​slag along the horizontal cutting path, reflecting the stability of processing quality. In the segmented cut cross-section image, the pixel distribution in the bottom region is statistically analyzed along the direction parallel to the cutting path (i.e., the X-axis direction). The proportion of the width of the region containing slag pixels in the horizontal direction to the total sampling width is calculated. That is, the distribution density value = the total pixel area of ​​detected slag; the total pixel area of ​​the bottom region = the total pixel area of ​​the bottom region; the total pixel area of ​​detected slag (or the horizontal coverage length ratio).

[0124] A low distribution density value indicates sporadic, dotted slag (possibly caused by localized impurities); a high distribution density value indicates continuous, elongated slag ("whisker-like" slag). The distribution density value is used to distinguish between sporadic problems and systemic parameter mismatches. A high density value indicates a global deviation in current energy or airflow parameters, necessitating parameter correction from the model library.

[0125] In one embodiment of this application, the process parameters include material type, plate thickness, gas type, beam type, laser power, processing speed, and gas flow rate. The correlation model between each dimension of morphological features and the corresponding process parameters includes the correlation model between slag adhesion features and the corresponding process parameters. Based on the multi-dimensional morphological features and process parameters, the correlation model between each dimension of morphological features and the corresponding process parameters is established as follows: The basic process model library to which the current workpiece belongs is determined based on laser power, material type, plate thickness, gas type, and beam type. Based on the laser power, processing speed, and gas flow rate, the core energy density change and gas flow rate change are determined in the basic process model library. Based on the characteristics of slag adhesion, the change in core energy density, and the change in airflow, a correlation model between the characteristics of slag adhesion and the corresponding process parameters is determined.

[0126] In the embodiments of this application, since the thermophysical properties of different materials (such as carbon steel and aluminum alloy) and different thicknesses are completely different during laser cutting, a general model cannot be applied. First, the current static operating parameters (i.e., the primary classification basis) are read, including material type, plate thickness, gas type, and beam type. Based on these parameters, a specific "basic process model library" is indexed and locked from the database.

[0127] For example, when "20mm carbon steel + oxygen" is detected, this dedicated library is locked. This library contains pre-set benchmark data on the material's specific heat capacity, melting point characteristics, and sensitivity to the heat of reaction with oxygen, providing physical boundaries for subsequent calculations.

[0128] The input parameters (power P, velocity V, flow rate Q) are control variables, but the physical field variables directly determine the cutting quality. Core energy density change (ΔE): Calculated using formulas such as E∝P / V f Calculate the change in heat absorbed per unit volume of material. This reflects fluctuations in melting capacity.

[0129] airflow change (ΔF) air Based on the gas flow rate and nozzle structure, the changes in momentum or shear force acting on the molten slag are calculated. This reflects the fluctuations in slag removal capacity. This step translates the parameter definitions of the laser processing equipment (e.g., how many watts, how much speed) into physical language (e.g., how much heat energy, how much blowing force).

[0130] Using the slag features obtained from image recognition (i.e., the slag thickness / distribution density values ​​statistically analyzed in the above embodiments) as the dependent variable (S), and the process parameters calculated above as the independent variables (X), the following correlation model between slag features and corresponding process parameters is established: +

[0131] Where k1, k2, and k3 are fitting coefficients, P is power, and V is... f Q represents the processing speed. i This represents the gas flow rate.

[0132] The formation of slag is mainly due to the airflow drag force being less than the slag's gravity plus viscosity.

[0133] By fitting the data, a correlation model between the slag adhesion characteristics and the corresponding process parameters is obtained.

[0134] For example, a linear or nonlinear equation is defined that expresses: "When the gas flow rate decreases by 10%, the slag thickness will increase by 0.5 mm." The correlation model between slag characteristics and corresponding process parameters provides recommended directions for eliminating the current slag: should energy (changing slag viscosity) or gas flow (increasing physical thrust) be prioritized?

[0135] By introducing two intermediate physical variables—core energy density and airflow variation—the process parameters are forcibly decoupled from the physical phenomena. This gives the established correlation model between slag characteristics and corresponding process parameters a clear physical meaning. During processing, the laser processing equipment not only knows about slag buildup but also whether it's due to excessive heat input causing the melt to thin or insufficient airflow thrust preventing proper blowing. This allows for the generation of more accurate parameter adjustment strategies, achieving optimal low-slag or slag-free cutting in various processing scenarios.

[0136] In one embodiment of this application, the process parameters include the spot radius, and the correlation model between the morphological features of each dimension and the corresponding process parameters includes the correlation model between the overburning features and the corresponding process parameters. Establishing the correlation model between the morphological features of each dimension and the corresponding process parameters based on the multi-dimensional morphological features and process parameters includes: The change in laser core energy density is determined based on laser power and spot radius; Obtain the boundary conditions for the core energy density change that causes the current material to overheat; The range of overburned core energy density change is determined based on the boundary conditions of the core energy density change. Information is extracted from the transition region of the image of the cut section to obtain grayscale information and image depth information; By identifying grayscale and image depth information, the surface collapse and protrusion areas caused by extreme thermal effects can be determined. The total area of ​​the surface collapse and protrusion regions caused by extreme thermal effects is calculated. Based on the total area, the range of changes in the overburned core energy density, and the change in the laser core energy density, a correlation model between the overburning characteristics and the corresponding process parameters is determined.

[0137] In the embodiments of this application, the thermal core of laser processing depends on the power distribution per unit area. The current laser power (P) and spot radius (r) are read. This is then calculated using formulas such as power density I = P / πr. 2 The laser beam is used to calculate the core energy density acting on the material surface. Core energy density represents how much heat energy the laser beam can inject into the material instantaneously. The smaller the spot size or the higher the power, the greater the energy density and the stronger the heating capacity.

[0138] Different materials (such as carbon steel and stainless steel) have different thermal diffusivity and melting points, and therefore different limits for withstanding heat input. Consult a pre-set material database to obtain the overheating threshold for the current material. For example, for carbon steel, when the heat of reaction plus the heat from the laser exceeds the heat dissipation capacity, self-sustaining combustion (overheating) will occur. The energy density value at this critical point is the boundary condition.

[0139] Based on the above boundary conditions, a numerical range is defined. In-depth analysis of overburning characteristics is only initiated when the calculated energy density falls within this range (typically in the high-energy region close to or exceeding the threshold). This eliminates interference from low-energy cuts (incomplete cuts) and focuses on resolving problems caused by excess energy.

[0140] Identify the upper edge or corner of the cut surface (i.e., the transition zone), as this is where overheating first occurs. Simultaneously extract information from two dimensions: Grayscale information: reflects the degree of oxidation and material changes on the surface (overheated areas usually appear as deep black or special bright spots).

[0141] Depth information: Using 3D cameras or structured light, the microscopic morphology data (Z-axis height) of the cross-section is obtained, reflecting the defects or accumulation of the material.

[0142] Excessive energy density leads to over-vaporization or melting of the material, resulting in pits. Depth information is used to identify the pits, and combined with grayscale information, it is confirmed to be thermal erosion pits, i.e., surface collapse.

[0143] As the molten pool expands uncontrollably, molten metal overflows and solidifies at the edges, forming nodules, or raised areas.

[0144] Extreme thermal effects indicate that the physical cause of these morphologies is heat overload. Surface collapse and protrusion areas can be directly identified through image analysis.

[0145] The total physical area of ​​all identified collapsed and protruding pixels is calculated. The total area represents the severity of overheating; the larger the area, the wider the thermal runaway and the more severe the energy exceedance.

[0146] Using the total overburned area as the dependent variable (result) and the change in core energy density (mainly affected by power and spot size) as the independent variable (cause), data fitting is performed within a preset range. A correlation function is determined. This function expresses the percentage reduction in core energy density (i.e., power or duty cycle) required to reduce the overburned area to zero, thus obtaining a correlation model between overburning characteristics and corresponding process parameters.

[0147] In one embodiment of this application, the process parameters include linear laser detection data, and the correlation model between morphological features and corresponding process parameters includes a correlation model between roughness features and corresponding process parameters. Establishing the correlation model between morphological features and corresponding process parameters based on multi-dimensional morphological features includes: The cut section image is divided along the height direction to obtain the upper region, middle region, and lower region; The grayscale texture in the upper region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the upper region. The grayscale texture in the central region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the central region. The grayscale texture in the lower region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the lower region. Using the linear laser detection data as the verification true value, the roughness level labels of each region are obtained by associating the arithmetic mean deviation of the upper region, the peak-valley height deviation of the upper region, the arithmetic mean deviation of the middle region, the peak-valley height deviation of the middle region, the arithmetic mean deviation of the lower region, and the peak-valley height deviation of the lower region. Based on the roughness grade labels of each region, a correlation model between roughness characteristics and corresponding process parameters is determined.

[0148] In this embodiment, the cross-section processed by laser exhibits different hydrodynamic characteristics in the height direction: Upper region: The area directly affected by the laser beam, where the stripes are usually finer and denser, mainly affected by the spot pattern.

[0149] Central region: a stable flow zone of airflow and molten metal; the stripes represent the average level of processing.

[0150] The lower region is the airflow drag zone, which is prone to producing drag lines and usually has the highest roughness.

[0151] The cut section image can be physically segmented along the height direction (Z-axis) using preset proportions, such as 20% for the upper region, 60% for the middle region, and 20% for the lower region, so that the cut quality at different depths can be evaluated independently.

[0152] Because rough surfaces produce alternating bright and dark streaks under illumination, the grayscale fluctuations in the image directly reflect the physical undulations of the surface. The arithmetic mean deviation and peak-to-valley height deviation were calculated in the upper, middle, and lower regions, respectively. The arithmetic mean deviation is the average of the absolute values ​​of the deviations between the gray values ​​of all pixels within the calculation area and the average gray value. It characterizes the overall contrast of the image texture and corresponds to the physical arithmetic mean deviation of the contour.

[0153] Peak-valley height deviation is the difference between the highest gray level point (highlight ridge) and the lowest gray level point (dark valley) within the calculation area. It characterizes the depth of the stripes and corresponds to the physical micro-irregularity of the ten-point height.

[0154] Line scan laser inspection data can be obtained through a high-precision profilometer or 3D scanning, which are physically measured micron-level roughness values. A mapping relationship is established by aligning the arithmetic mean deviation and peak-valley height deviation of the upper, middle, and lower regions calculated in the previous step with the line scan laser data (physical true values).

[0155] The calculation results are divided into several levels based on the true value (e.g., Level 1 = glossy, Level 5 = severe striping). When the gray-scale arithmetic deviation in the upper region reaches the X value, it actually corresponds to the R value measured by the line scan data. a =6.3μm, which is roughness grade 3.

[0156] Based on the roughness grade labels of each region, the corresponding process parameters are traced back, such as: processing speed V, pulse frequency f, and auxiliary gas pressure Q.

[0157] For example: If the upper region is rough: the model is associated with the focal position (FposFpos) or the beam pattern parameter.

[0158] When the mid-area grade label shows excessively deep periodic texture (large arithmetic deviation), it is associated with the frequency and speed matching ratio.

[0159] The lower region is rough (severe hysteresis stripes): the model is associated with excessive processing speed (V) or insufficient air pressure (Q).

[0160] For example, when the lower region is detected to have a roughness level of 4, reduce the processing speed by 10% or increase the air pressure by 1 bar.

[0161] In this embodiment, by establishing three independent correlation models for the upper, middle, and lower regions, the dominant physical factors corresponding to different depth regions (upper region - focal point, middle region - frequency, lower region - air pressure / velocity) are clearly identified. This allows for precise adjustment of the specific parameter causing defects in a particular region without interfering with other normal process parameters.

[0162] In one embodiment of this application, the process parameters include processing speed and gas flow rate. The correlation model between morphological features and corresponding process parameters includes a correlation model between texture trend features and corresponding process parameters. Establishing the correlation model between morphological features and corresponding process parameters based on multi-dimensional morphological features includes: The least squares method is used to fit the texture trend features to obtain the tangent equation and tangent slope of the fitted curve. Based on the tangent slope, processing speed, and gas flow rate, a correlation model between texture trend characteristics and corresponding process parameters is determined.

[0163] In this embodiment, texture lines on the cut surface (especially the lower region) were identified. These lines physically represent the instantaneous position of the cutting front. The coordinates of the stripe skeleton pixels (i.e., texture trend features) were extracted from the image. The least squares method was used to find the best function match for the data by minimizing the sum of squared errors. Due to the presence of microscopic noise on the actual cut surface, individual stripes may be discontinuous. The least squares method can filter out local noise and fit a smooth curve (usually a parabola or a higher-order polynomial curve) representing the overall molten metal flow path. Ideally, the cutting stripes should be perpendicular to the surface of the sheet (i.e., the slope is 0, or infinitely large, depending on the coordinate system definition). When the cutting speed is too fast or the airflow thrust is insufficient, the molten metal at the bottom will be "dragged" behind, causing the stripes to bend.

[0164] The fitted curve is differentiated to calculate the tangent equation at a specific location (usually the bottom edge of the plate), yielding the tangent slope k. The tangent slope k physically quantifies the drag hysteresis. k≈0 (or nearly vertical): indicates a vertical cutting front and perfect energy-airflow matching. A large k deviation indicates severe bottom hysteresis, a phenomenon known as tailing. The calculated tangent slope k is used as the result, and the processing speed (V) and gas flow rate (Q) are used as the causes to establish a correlation. A faster processing speed (V) results in a faster laser beam advance, preventing the molten slag at the bottom from being expelled in time, thus increasing the tangent slope (increased hysteresis).

[0165] The greater the gas flow rate (Q) (and its corresponding momentum), the greater the downward shear propulsion force on the molten slag, which helps to reduce the tangent slope (correcting hysteresis).

[0166] Construct an inverse model of the function k=f(V,Q). When the tangent slope k is detected to be excessive (stripes are too skewed), the model will output specific adjustment instructions: keep the airflow constant and reduce the processing speed, or keep the speed constant and increase the gas flow rate (or gas pressure). Select the optimal adjustment path based on the current parameter margin.

[0167] In this embodiment, the tangent slope is an extremely sensitive indicator. It can detect imbalances in process parameters (i.e., insufficient airflow) through the precursor of stripe distortion before slag actually forms. This allows for timely optimization of the laser processing equipment parameters to avoid incomplete cutting or rough surfaces. Moreover, the calculated tangent slope of the fitted curve remains stable and accurate, thus ensuring the reliability of process parameter adjustment commands.

[0168] In one embodiment of this application, real-time image features are input into an image feature model library to obtain parameter adjustment information, including: Obtain the cross-sectional image of the cut surface of the workpiece currently being processed; The image of the cut section of the workpiece being processed is extracted to obtain real-time slag features, real-time overheating features, real-time roughness features, and real-time texture trend features. Real-time slag features, real-time overheating features, real-time roughness features, and real-time texture trend features are input into the image feature model library for matching and retrieval to obtain matching results; Based on the matching results and process parameters, determine the parameter adjustment information.

[0169] In this embodiment, at the laser cutting head movement or a specific monitoring station, a vision sensor installed on the side or coaxial with the cutting kerf is used to capture an image of the cut cross-section of the workpiece that has just been cut, i.e., the workpiece currently being processed.

[0170] The image of the cut section of the workpiece being processed is extracted to obtain real-time slag features, real-time overheating features, real-time roughness features, and real-time texture trend features, specifically: The real-time slag-coating feature identifies the pixel area at the bottom of the cross-section and calculates the slag-coating height / thickness (reflecting the mechanical balance in the height direction).

[0171] The real-time overburning feature is to identify the area of ​​collapse and bulge at the top of the cross section, and to calculate the energy density state (reflecting the heat input boundary) in combination with the current spot power.

[0172] The real-time roughness feature calculates grayscale texture statistics for the upper, middle, and lower regions respectively, and maps them to roughness level labels (reflecting the surface smoothness).

[0173] The real-time texture trend feature is obtained by fitting the stripe trajectory using the least squares method and calculating the tangent slope (reflecting the airflow drag lag).

[0174] The image feature model library stores the correlation models between various morphological features and corresponding process parameters, such as the correlation model between slag features and corresponding process parameters, the correlation model between overburning features and corresponding process parameters, the correlation model between roughness features and corresponding process parameters, and the correlation model between texture trend features and corresponding process parameters, which were established through a large number of experiments or self-learning in the above embodiments. It contains the correspondence between various defect morphologies (Features) and process causes (Causes).

[0175] Real-time slag features, real-time overheating features, real-time roughness features, and real-time texture trend features are input into an image feature model library for matching and retrieval to obtain matching results, such as: The current real-time slag feature is large and the real-time texture trend slope is large, indicating insufficient airflow propulsion.

[0176] The current real-time overheating feature is large and the upper roughness is poor, so the matching result is that the focal position is too high and the power is too large.

[0177] The matching result is a physical qualitative assessment of the current cutting quality problem (e.g., whether it is thermal runaway or fluid hysteresis).

[0178] Calculate the specific adjustment amount by combining the current process parameters (the actual P, V, F, and Q values ​​set for the current machine tool) and the matching results (diagnostic conclusions).

[0179] For example, if the matching result is insufficient airflow propulsion and the current air pressure is close to the upper limit, the adjustment information is to reduce the processing speed (V).

[0180] If the matching result is thermal overload, the adjustment information is to reduce the duty cycle or reduce the laser power (P).

[0181] Finally, a set of specific parameter increment instructions is generated (such as...) This is sent to the controller of the laser processing equipment for execution.

[0182] Real-time slag adhesion features, real-time overheating features, real-time roughness features, and real-time texture trend features are input into the image feature model library for matching and retrieval, yielding matching results. The image feature model library can handle complex combined operating conditions. For example, when roughness requires a reduction in speed, but overheating requires an increase in speed (to reduce heat accumulation), the weighting mechanism in the image feature model library can determine the priority adjustment of third-dimensional parameters such as auxiliary gas or pulse frequency based on the severity of the features, thus cleverly resolving the contradictions between parameters.

[0183] Secondly, embodiments of this application provide a parameter adjustment method system for laser processing technology, comprising: The first acquisition module is used to acquire image data and process parameters of the cut cross-section; The first module is used to extract features from the image of the cut section to obtain multi-dimensional morphological features, which include: slag features, overheating features, roughness features, and texture trend features. The first module is used to establish a correlation model between morphological features and corresponding process parameters based on multi-dimensional morphological features and process parameters. The first generation module is used to store the association model of morphological features and corresponding process parameters in a specified database to generate an image feature model library. The first module is used to input real-time image features into the image feature model library to obtain parameter adjustment information.

[0184] In this embodiment, the parameter adjustment method system for this laser processing technology extracts multi-dimensional morphological features such as slag adhesion, overheating, roughness, and texture trend features from the image data of the cut section. Based on these multi-dimensional morphological features and process parameters, a correlation model is established between each morphological feature and its corresponding process parameter. This correlation model is stored in a designated database to generate an image feature model library. Real-time image features are input into the image feature model library to obtain parameter adjustment information. This enables rapid correlation determination of processing quality and processing parameters during laser processing, facilitating subsequent process recommendations, peripheral equipment fault diagnosis, and parameter optimization. Based on the morphological features of processing quality, optimized process parameter recommendations are obtained, replacing traditional manual judgment and experience-based analysis. This optimizes the detection and process parameter optimization stages of automated laser processing, making it more objective and automated. It reduces quality variations caused by human factors during laser processing, maintains relatively stable parameter optimization for laser processing, and ensures real-time adjustments based on workpiece quality during processing, guaranteeing the consistency and stability of the output workpiece quality.

[0185] In one embodiment of this application, the slag features include slag thickness and distribution density; feature extraction is performed on the image of the cut section to obtain multi-dimensional morphological features, including: Image segmentation is performed on the image of the cut section to obtain the segmented cut section image; The distribution of slag pixels at the bottom of the cut section in the segmented image is statistically analyzed to obtain the slag thickness and distribution density values.

[0186] In one embodiment of this application, the process parameters include material type, plate thickness, gas type, beam type, laser power, processing speed, and gas flow rate. The correlation model between each dimension of morphological features and the corresponding process parameters includes the correlation model between slag adhesion features and the corresponding process parameters. Based on the multi-dimensional morphological features and process parameters, the correlation model between each dimension of morphological features and the corresponding process parameters is established as follows: The basic process model library to which the current workpiece belongs is determined based on laser power, material type, plate thickness, gas type, and beam type. Based on the laser power, processing speed, and gas flow rate, the core energy density change and gas flow rate change are determined in the basic process model library. Based on the characteristics of slag adhesion, the change in core energy density, and the change in airflow, a correlation model between the characteristics of slag adhesion and the corresponding process parameters is determined.

[0187] In one embodiment of this application, the process parameters include the spot radius, and the correlation model between the morphological features of each dimension and the corresponding process parameters includes the correlation model between the overburning features and the corresponding process parameters. Establishing the correlation model between the morphological features of each dimension and the corresponding process parameters based on the multi-dimensional morphological features and process parameters includes: The change in laser core energy density is determined based on laser power and spot radius; Obtain the boundary conditions for the core energy density change that causes the current material to overheat; The range of overburned core energy density change is determined based on the boundary conditions of the core energy density change. Information is extracted from the transition region of the image of the cut section to obtain grayscale information and image depth information; By identifying grayscale and image depth information, the surface collapse and protrusion areas caused by extreme thermal effects can be determined. The total area of ​​the surface collapse and protrusion regions caused by extreme thermal effects is calculated. Based on the total area, the range of changes in the overburned core energy density, and the change in the laser core energy density, a correlation model between the overburning characteristics and the corresponding process parameters is determined.

[0188] In one embodiment of this application, the process parameters include linear laser detection data, and the correlation model between morphological features and corresponding process parameters includes a correlation model between roughness features and corresponding process parameters. Establishing the correlation model between morphological features and corresponding process parameters based on multi-dimensional morphological features includes: The cut section image is divided along the height direction to obtain the upper region, middle region, and lower region; The grayscale texture in the upper region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the upper region. The grayscale texture in the central region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the central region. The grayscale texture in the lower region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the lower region. Using the linear laser detection data as the verification true value, the roughness level labels of each region are obtained by associating the arithmetic mean deviation of the upper region, the peak-valley height deviation of the upper region, the arithmetic mean deviation of the middle region, the peak-valley height deviation of the middle region, the arithmetic mean deviation of the lower region, and the peak-valley height deviation of the lower region. Based on the roughness grade labels of each region, a correlation model between roughness characteristics and corresponding process parameters is determined.

[0189] In one embodiment of this application, the process parameters include processing speed and gas flow rate. The correlation model between morphological features and corresponding process parameters includes a correlation model between texture trend features and corresponding process parameters. Establishing the correlation model between morphological features and corresponding process parameters based on multi-dimensional morphological features includes: The least squares method is used to fit the texture trend features to obtain the tangent equation and tangent slope of the fitted curve. Based on the tangent slope, processing speed, and gas flow rate, a correlation model between texture trend characteristics and corresponding process parameters is determined.

[0190] In one embodiment of this application, real-time image features are input into an image feature model library to obtain parameter adjustment information, including: Obtain the cross-sectional image of the cut surface of the workpiece currently being processed; The image of the cut section of the workpiece being processed is extracted to obtain real-time slag features, real-time overheating features, real-time roughness features, and real-time texture trend features. Real-time slag features, real-time overheating features, real-time roughness features, and real-time texture trend features are input into the image feature model library for matching and retrieval to obtain matching results; Based on the matching results and process parameters, determine the parameter adjustment information.

[0191] The functions of each module in each device in the embodiments of this application can be found in the corresponding descriptions in the above methods, and will not be repeated here.

[0192] Figure 6 A structural block diagram of an electronic device according to an embodiment of this application is shown. Figure 6As shown, the electronic device includes a memory 410 and a processor 420. The memory 410 stores instructions that can be executed on the processor 420. When the processor 420 executes the instructions, it implements the parameter adjustment method for the laser processing technology in the above embodiments. The number of memories 410 and processors 420 can be one or more. This electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0193] The electronic device may also include a communication interface 430 for communicating with external devices and exchanging data. The devices are interconnected using different buses and can be mounted on a common motherboard or otherwise as needed. The processor 420 can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). The bus can be divided into address buses, data buses, control buses, etc. For ease of illustration, Figure 6 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0194] Optionally, in a specific implementation, if the memory 410, processor 420 and communication interface 430 are integrated on a single chip, the memory 410, processor 420 and communication interface 430 can communicate with each other through an internal interface.

[0195] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting Advanced Reduced Instruction Set Machines (ARM) architecture.

[0196] This application provides a computer-readable storage medium (such as the memory 410 described above) that stores computer instructions, which, when executed by a processor, implement the method provided in this application.

[0197] Optionally, memory 410 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device, etc. Furthermore, memory 410 may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 410 may optionally include memory remotely located relative to processor 420, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0198] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0199] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0200] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for adjusting parameters in a laser processing technology, characterized in that, include: Acquire image data and process parameters of the cut cross-section; Feature extraction is performed on the image of the cut surface to obtain multi-dimensional morphological features, which include: slag features, overheating features, roughness features, and texture trend features. Based on the multi-dimensional morphological features and the process parameters, establish a correlation model between each morphological feature and the corresponding process parameters; The association model of the morphological features and corresponding process parameters of each dimension is stored in a specified database to generate an image feature model library; Real-time image features are input into the image feature model library to obtain parameter adjustment information; The process parameters include line-scan laser detection data, and the correlation model between the morphological features and corresponding process parameters includes a correlation model between roughness features and corresponding process parameters. Establishing the correlation model between the morphological features and corresponding process parameters based on the multi-dimensional morphological features and the process parameters includes: The cross-sectional image is divided along the height direction to obtain the upper region, middle region, and lower region; The grayscale texture in the upper region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the upper region. The grayscale texture in the central region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the central region. The grayscale texture in the lower region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the lower region. Using the line-scan laser detection data as the verification true value, the roughness level label of each region is obtained by associating the arithmetic mean deviation of the upper region, the peak-valley height deviation of the upper region, the arithmetic mean deviation of the middle region, the peak-valley height deviation of the middle region, the arithmetic mean deviation of the lower region, and the peak-valley height deviation of the lower region. Based on the roughness level labels of each region, a correlation model between roughness characteristics and corresponding process parameters is determined.

2. The method according to claim 1, characterized in that, The slag-coating features include slag thickness and distribution density; the feature extraction of the image of the cut section to obtain multi-dimensional morphological features includes: The image of the cut section is segmented to obtain the segmented cut section image; The distribution of slag pixels at the bottom of the cut section in the segmented cross-section image is statistically analyzed to obtain the slag thickness value and distribution density value.

3. The method according to claim 1, characterized in that, The process parameters also include material type, plate thickness, gas type, beam type, laser power, processing speed, and gas flow rate. The correlation model between each dimension of morphological features and the corresponding process parameters also includes a correlation model between slag adhesion features and the corresponding process parameters. Furthermore, establishing the correlation model between each dimension of morphological features and the corresponding process parameters based on the multi-dimensional morphological features and the process parameters also includes: The basic process model library to which the current workpiece belongs is determined based on the laser power, the material type, the plate thickness, the gas type, and the beam type. In the basic process model library, the core energy density change and airflow change are determined based on the laser power, the processing speed, and the gas flow rate. Based on the slag adhesion characteristics, the core energy density change, and the airflow change, a correlation model between the slag adhesion characteristics and the corresponding process parameters is determined.

4. The method according to claim 1, characterized in that, The process parameters also include laser power and spot radius. The correlation model between the morphological features and corresponding process parameters in each dimension also includes the correlation model between overburning features and corresponding process parameters. The step of establishing the correlation model between the morphological features and corresponding process parameters in each dimension based on the multi-dimensional morphological features and the process parameters also includes: Based on the laser power and the spot radius, the change in laser core energy density is determined; Obtain the boundary conditions for the core energy density change that causes the current material to overheat; Based on the boundary conditions for the change in core energy density, determine the range of change in the overburned core energy density; Information is extracted from the transition region of the image of the cut section to obtain grayscale information and image depth information; The grayscale information and the image depth information are identified to determine the surface collapse area and the protrusion area caused by extreme thermal effects. The total area of ​​the surface collapse and protrusion regions caused by extreme thermal effects is calculated. Based on the total area, the range of changes in the overburned core energy density, and the change in the laser core energy density, a correlation model between the overburning characteristics and the corresponding process parameters is determined.

5. The method according to claim 1, characterized in that, The process parameters also include processing speed and gas flow rate. The correlation model between the morphological features and corresponding process parameters also includes a correlation model between texture trend features and corresponding process parameters. The step of establishing the correlation model between the morphological features and corresponding process parameters based on the multi-dimensional morphological features and the process parameters further includes: The texture trend features are fitted using the least squares method to obtain the tangent equation and tangent slope of the fitted curve. Based on the tangent slope, the processing speed, and the gas flow rate, a correlation model between texture trend features and corresponding process parameters is determined.

6. The method according to claim 1, characterized in that, The step of inputting real-time image features into the image feature model library to obtain parameter adjustment information includes: Obtain the cross-sectional image of the cut surface of the workpiece currently being processed; The image of the cut section of the workpiece being processed is extracted to obtain real-time slag features, real-time overheating features, real-time roughness features, and real-time texture trend features. The real-time slag feature, the real-time overheating feature, the real-time roughness feature, and the real-time texture trend feature are input into the image feature model library for matching and retrieval to obtain the matching results; Based on the matching results and the process parameters, parameter adjustment information is determined.

7. A parameter adjustment method system for laser processing technology, characterized in that, include: The first acquisition module is used to acquire image data and process parameters of the cut cross-section; The first obtaining module is used to extract features from the image of the cut section to obtain multi-dimensional morphological features, which include: slag features, overheating features, roughness features, and texture trend features. The first module is used to establish a correlation model between the multi-dimensional morphological features and the corresponding process parameters based on the multi-dimensional morphological features and the process parameters. The first generation module is used to store the association model of the morphological features of each dimension and the corresponding process parameters into a specified database to generate an image feature model library. The first obtaining module is also used to input real-time image features into an image feature model library to obtain parameter adjustment information; The process parameters include line-scan laser detection data, and the correlation model between the morphological features and corresponding process parameters includes a correlation model between roughness features and corresponding process parameters. Establishing the correlation model between the morphological features and corresponding process parameters based on the multi-dimensional morphological features and the process parameters includes: The cross-sectional image is divided along the height direction to obtain the upper region, middle region, and lower region; The grayscale texture in the upper region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the upper region. The grayscale texture in the central region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the central region. The grayscale texture in the lower region is calculated to obtain the arithmetic mean deviation and the peak-valley height deviation of the lower region. Using the line-scan laser detection data as the verification true value, the roughness level label of each region is obtained by associating the arithmetic mean deviation of the upper region, the peak-valley height deviation of the upper region, the arithmetic mean deviation of the middle region, the peak-valley height deviation of the middle region, the arithmetic mean deviation of the lower region, and the peak-valley height deviation of the lower region. Based on the roughness level labels of each region, a correlation model between roughness characteristics and corresponding process parameters is determined.

8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A computer-readable storage medium storing computer instructions that, when executed by a processor, implement the method as described in any one of claims 1-6.

Images