Waterproof material laser cutting path optimization method based on deep learning

Abstract

The invention relates to the technical field of laser cutting, in particular to a waterproof material laser cutting path optimization method based on deep learning, which comprises the following steps: acquiring basic size data, thickness change rate distribution data and preloading stress distribution data of a target test piece, and identifying key / non-key path sections through discretized grid units; cutting is conducted on the non-critical path section through basic cutting parameters, edge size data are monitored in real time, and the parameters are dynamically adjusted to form verification cutting parameters; screening a first confidence path segment based on stable cutting quality, and extracting verification parameters to construct a confidence collection data set; key cutting parameters are determined through parameter fluctuation analysis and correlation calculation, and a confidence data set is formed; cutting the key path section by adopting the parameter; and finally, establishing a deep learning model based on the historical confidence data set, and outputting basic cutting parameters of the new test piece. According to the method, self-adaptive optimization of cutting parameters and paths is achieved, and the cutting precision of the waterproof material test piece is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of laser cutting technology, and in particular to a method for optimizing the laser cutting path of a waterproof material based on deep learning. Background Art

[0002] Waterproof material test specimen preparation equipment is mainly used for the preparation of waterproof material test specimens, which can ensure the standardization of test specimens during the preparation process. Currently, mature waterproof material test specimen preparation methods mostly use templates or cutters to process test specimens, which have many defects: the cutting equipment lacks effective control over material deformation, tool wear and other conditions during cutting, resulting in large deviations in the size of the cut test specimens, which cannot meet the millimeter-level processing requirements of the test standards; burrs on the formed test specimens are common, affecting the physical and chemical properties of the waterproof material and greatly affecting the reliability of the test results; the waterproof material testing parameters are complicated, and the requirements for the shape and size of the test specimens are different. The existing cutting equipment is limited to simple straight line cutting, and it is difficult to complete the processing of complex graphic test specimens; the test specimen preparation cycle is long, resulting in an extension of the test cycle and a reduction in the utilization rate of the test equipment, increasing equipment investment and labor costs, and reducing production efficiency. At the same time, the existing automated cutting equipment can only achieve basic cutting control, lacks intelligent optimization and self-learning capabilities, and cannot optimize cutting paths and parameters based on material properties, cutting history data, etc. Operators still need to manually adjust a large number of parameters, which is cumbersome and error-prone. At the same time, the software is difficult to be compatible with new testing standards and material types, and is difficult to upgrade, limiting the long-term use value of the equipment.

[0003] Laser cutting offers the following advantages: it completely eliminates the effects of template wear and deformation, reducing template replacement costs. Laser cutting utilizes computer graphics and allows for one-step specimen preparation, significantly improving work efficiency. Furthermore, the process is safe and stable, with accurate and reliable precision, effectively ensuring the accuracy of specimen preparation and test data. Therefore, a laser cutting path optimization method for waterproof materials that can self-learn data is of great value in preparing waterproof material test specimens. Summary of the Invention

[0004] To this end, the present invention provides a deep learning-based laser cutting path optimization method for waterproof materials, which is used to overcome the problem of poor cutting accuracy caused by the inability to adjust the cutting laser parameters according to the cutting accuracy in real time in the existing technology.

[0005] To achieve the above objectives, the present invention provides a method for optimizing the laser cutting path of waterproof materials based on deep learning, comprising:

[0006] Step S1, obtaining basic size data, thickness change rate distribution data, and preload stress distribution data of a target test piece of waterproof material, and determining a critical path segment and a non-critical path segment of a cutting path of the target test piece;

[0007] Step S2, determining basic cutting parameters based on the basic size data of the target specimen, cutting the non-critical path segment of the target specimen with the basic cutting parameters, and monitoring the cutting edge size data of the target specimen in real time;

[0008] Step S3, based on the obtained cutting edge dimension data of the target specimen and the sample preparation preset data, the basic cutting parameters are adjusted to obtain verification cutting parameters, and the laser cutting of the non-critical path segment is completed with the verification cutting parameters;

[0009] Step S4, determining a first trusted path segment based on the cutting edge size data of the non-critical path segment, and extracting the corresponding verification cutting parameters to form a trusted data set;

[0010] Step S5, determining the correlation degree of the verification cutting parameters based on the fluctuation degree of the verification cutting parameters in the trusted data set, obtaining key cutting parameters and forming a trusted data set;

[0011] Step S6, performing laser cutting on the critical path segment using the key cutting parameters in the confidence data set to obtain a target test piece;

[0012] Step S7: establishing a deep learning optimization model based on the plurality of confidence data sets, and outputting the basic cutting parameters of subsequent test pieces.

[0013] Furthermore, the step S1 includes:

[0014] Step S11, obtaining the basic dimension data of the target specimen, including the cutting path and thickness reference value;

[0015] Step S12, obtaining the thickness change rate distribution data of the sample where the target specimen is located;

[0016] Step S13, performing finite element analysis based on the detection working conditions of the target specimen and the basic dimension data to generate the preload stress distribution data;

[0017] Step S14 , determining the critical path segments and non-critical path segments of the cutting path according to the thickness change rate distribution data and the preload stress distribution data.

[0018] Furthermore, the step S14 includes:

[0019] Step S141, discretizing the cutting path into a plurality of grid units, wherein the grid resolution of each grid unit is set according to the thickness reference value;

[0020] Step S142 , performing binary classification on each grid cell, wherein if the judgment condition of the thickness change rate threshold or the stress threshold is met, it is marked as a critical grid cell, otherwise it is marked as a non-critical grid cell;

[0021] Step S143: Based on the test conditions of the target specimen, the grid cells corresponding to the test stress area are forcibly marked as key grid cells;

[0022] Step S144, removing isolated single-point key grid cells;

[0023] Step S145 : merging adjacent grid units of the same type to form continuous critical path segments and non-critical path segments.

[0024] Furthermore, the step S145 includes:

[0025] Step S1451, performing a morphological closing operation on the key grid cells to fill the gaps;

[0026] Step S1452: Perform connected domain analysis on key grid cells and merge discrete segments whose spacing is less than a preset merging threshold.

[0027] Step S1453: output a topologically continuous closed path segment.

[0028] Furthermore, in step S144, the isolated single-point key grid unit is a single key grid unit in which a predetermined number of grid units preceding and following the cutting path are all non-critical.

[0029] Furthermore, the step S1 further includes:

[0030] Step S15: when the length of the longest non-critical path segment among the non-critical path segments is less than the minimum representation length, generating a compensating cutting segment in the longest non-critical path segment of the target specimen;

[0031] Step S16: merging the compensating cutting segment into the non-critical path segment to form an extended non-critical path segment, so that the length of the extended non-critical path segment meets the minimum characterization length requirement.

[0032] Furthermore, step S3 includes:

[0033] Step S31, comparing the real-time monitored cutting edge dimension data with the sample preparation preset data one by one, and calculating the corresponding several dimension deviation values;

[0034] Step S32: When a predetermined number of consecutive dimensional deviation values ​​are greater than an allowable tolerance range, adjusting the basic cutting parameters, wherein the basic cutting parameters are adjusted based on the direction and magnitude of the dimensional deviation value, and the basic cutting parameters include laser power and cutting speed;

[0035] Step S33 , completing the cutting of the current non-critical path segment with the adjusted verification cutting parameters, and continuously acquiring the corresponding relationship between the cutting edge dimension data of the target specimen and the corresponding verification cutting parameters during the cutting process.

[0036] Furthermore, the step S4 includes:

[0037] Step S41, screening the continuous sections in the non-critical path segments where the cutting edge dimension data is continuously within the preset tolerance band, and marking them as first confidence path segments;

[0038] Step S42, extracting the set of verified cutting parameters in the first trusted path segment, recording them as trusted cutting parameters, and extracting the cutting edge dimension data corresponding to the trusted cutting parameters, recording them as trusted dimension data;

[0039] Step S43: storing the trusted cutting parameters and the trusted size data in spatial order to form a trusted data set, and recording a mapping relationship between the trusted cutting parameters and the trusted size data.

[0040] Furthermore, the step S5 includes:

[0041] Step S51, dividing each of the first confidence path segments according to the change of the confidence cutting parameter to form a plurality of second confidence path segments;

[0042] Step S52, calculating the correlation coefficient between the trusted cutting parameter and the corresponding trusted size data in each of the second trusted path segments to determine the degree of association between each trusted cutting parameter and the corresponding trusted size data;

[0043] Step S53 , eliminating the trusted cutting parameters whose correlation coefficients are less than a preset correlation coefficient threshold, obtaining the key cutting parameters and forming the trusted data set.

[0044] Furthermore, the step S7 includes:

[0045] Step S71, collecting the confidence data sets and corresponding cutting quality rating data sets of multiple groups of target test pieces;

[0046] Step S72, constructing a supervised learning model with the cutting parameters as input and the cutting quality rating as output;

[0047] Step S73: When the basic size data of the new specimen is input, the optimized initial values ​​of the cutting parameters are output.

[0048] Compared with the existing technology, the beneficial effect of the present invention lies in that, based on the functional characteristics differences of different areas of the waterproof specimen, the present invention uses a hierarchical processing mechanism of critical / non-critical path segments to ensure ultra-high precision cutting of critical areas (such as stress areas and sealing surfaces), while using non-critical segments to quickly complete parameter self-learning, greatly reducing the overall debugging time, and establishing a closed-loop system of dynamic parameter adjustment, confidence data extraction, and deep learning optimization, so that the cutting process has the ability to continuously evolve, and significantly improves the adaptability and stability under complex working conditions.

[0049] Furthermore, the present invention accurately identifies the key path segments that affect the waterproof performance through a dual-threshold judgment mechanism of thickness change rate and stress distribution, preloads stress data based on finite element analysis, and forcibly marks highly sensitive areas in combination with detection conditions to ensure that special cutting strategies are adopted in areas of material thickness mutation and stress concentration, avoiding defects such as melt nodules and cracks from the source. At the same time, a continuous cutting path is generated through morphological closing operations and connected domain analysis to eliminate parameter oscillations caused by data noise, ensure the energy stability of cutting in key areas, and enable the test specimens to demonstrate true and reliable performance in subsequent tests.

[0050] Furthermore, the present invention achieves intelligent improvements in cutting efficiency by implementing dynamic parameter adjustment and confidence data extraction in non-critical path segments. Through a real-time closed-loop "cutting-monitoring-feedback" system, power and speed parameters are automatically optimized based on the direction and magnitude of dimensional deviations, and a confidence data set is constructed based on continuous stable segments. By compensating for the expansion of cutting segments and screening for parameter correlation, the data in non-critical segments is ensured to be statistically significant, avoiding ineffective debugging while providing highly reliable parameters for critical segments. This transforms traditional trial-and-error parameter optimization into a highly efficient, data-driven process, significantly shortening the new product development cycle.

[0051] Furthermore, the present invention endows the cutting system with the ability to continuously evolve by constructing a deep learning model of historical confidence data sets. Using optimized cutting parameters and quality ratings as training samples, an intelligent mapping relationship from basic dimensions to initial parameters is established to solve the cold start problem under new materials and new working conditions. The model automatically outputs initial parameters that adapt to specific thicknesses and material properties, greatly reducing the need for manual intervention. At the same time, through periodic data iterative updates, the system has the robustness to cope with material batch fluctuations and environmental changes, and maintains high-precision cutting levels for a long time. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of the method for optimizing the laser cutting path of waterproof materials based on deep learning of the present invention;

[0053] Figure 2 This is a flowchart of step S1 of the method for optimizing the laser cutting path of waterproof materials based on deep learning of the present invention;

[0054] Figure 3 This is a flowchart of step S14 of the method for optimizing the laser cutting path of waterproof materials based on deep learning of the present invention;

[0055] Figure 4 Schematic diagram of step S1 of the waterproof material laser cutting path optimization method based on deep learning of the present invention.

[0056] In the figure: 11-isolated single-point critical grid unit; 21-critical grid unit; 22-critical path segment; 31-non-critical grid unit; 32-non-critical path segment; 41-compensating cutting segment; 42-deletion segment; a1-first front side grid unit; a2-second front side grid unit; a3-third front side grid unit; b1-first rear side grid unit; b2-second rear side grid unit; b3-third rear side grid unit. DETAILED DESCRIPTION

[0057] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0058] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0059] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0060] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0061] See also Figure 1As shown, it is a flow chart of a method for optimizing a laser cutting path of a waterproof material based on deep learning according to the present invention; an embodiment of the present invention provides a method for optimizing a laser cutting path of a waterproof material based on deep learning, comprising:

[0062] Step S1, obtaining basic size data, thickness change rate distribution data, and preload stress distribution data of a target test piece of waterproof material, and determining a critical path segment and a non-critical path segment of a cutting path of the target test piece;

[0063] See also Figures 2 to 4 As shown, they are respectively a flow chart of step S1 of the method for optimizing the laser cutting path of waterproof materials based on deep learning of the present invention, a flow chart of step S14 of the method for optimizing the laser cutting path of waterproof materials based on deep learning of the present invention, and a schematic diagram of step S1 of the method for optimizing the laser cutting path of waterproof materials based on deep learning of the present invention. Specifically, step S1 includes:

[0064] Step S11, obtaining the basic dimension data of the target specimen, including the cutting path and thickness reference value;

[0065] In the implementation, the reference thickness value of the specimen is obtained from the CAD drawing of the target specimen, in millimeters (mm), denoted as T; and the preset cutting path of the specimen;

[0066] Step S12, obtaining the thickness change rate distribution data of the sample where the target specimen is located;

[0067] In the implementation, the surface of the sample where the target specimen is located is scanned as a whole by a laser thickness gauge, and the thickness change rate distribution data of the sample where the target specimen is located is calculated and recorded as R ij , preferably, the detection step length of the thickness change rate distribution data is 0.1mm to 1mm;

[0068] Step S13, performing finite element analysis based on the detection working conditions of the target specimen and the basic dimension data to generate the preload stress distribution data;

[0069] In practice, the stress distribution of the specimen in the corresponding test condition is simulated by finite element analysis, and the equivalent stress cloud diagram is output to obtain the preload stress distribution data in megapascals (Mpa), which is recorded as σ ij .

[0070] In this embodiment, the finite element analysis is specifically as follows:

[0071] A three-dimensional solid model is established based on the basic dimensional data of the target specimen, and the model boundary strictly follows the specimen contour; the constitutive model parameters of the waterproof material are input, including the elastic modulus, Poisson's ratio, and yield limit; based on the type of test condition, the load conditions are set according to the GB / T328 standard, and the load application area is preset in the model. For example: for tensile test conditions, clamping surface areas are generated at both ends of the specimen to simulate fixture constraints; for water pressure test conditions, a normal uniformly distributed pressure load is applied to the specimen surface; static structural analysis is performed, and an equivalent stress cloud diagram is output to obtain preload stress distribution data.

[0072] Those skilled in the art will understand that the testing conditions include the lap seam peeling test, tensile performance test, tear performance test, impermeability test and other tests that require load loading as specified in GB / T328 "Test Methods for Building Waterproofing Membranes", which will not be repeated here.

[0073] Those skilled in the art will appreciate that the finite element analysis technology is an existing mature technology, and its specific operating details are not described here in detail.

[0074] It is understandable that the thickness mutation area of ​​the waterproof material is prone to generate melt nodules due to heat accumulation during the laser cutting process, and the laser cutting defects in the high stress area may become the source of cracks, which will interfere with the accuracy of subsequent detection.

[0075] Step S14 : determining the critical path segment 22 and the non-critical path segment 32 of the cutting path according to the thickness change rate distribution data and the preload stress distribution data.

[0076] Specifically, step S14 includes:

[0077] Step S141, discretizing the cutting path into a plurality of grid units, wherein the grid resolution of each grid unit is set according to the thickness reference value;

[0078] In practice, the calculation formula for the grid resolution is specifically:

[0079] δ=2-0.1×T, and δ≥1mm,

[0080] Where δ is the grid resolution, that is, the side length of a single grid unit, in millimeters (mm).

[0081] It is understandable that the grid resolution needs to match the laser processing accuracy (slightly larger than the line width of the cutting laser, which usually does not exceed 1mm) to avoid path deviation due to discrete errors; at the same time, the thinner the waterproof material, the more sensitive it is to the laser cutting processing parameters, and the grid resolution should be increased.

[0082] Step S142 , performing binary classification on each grid cell, wherein if the judgment condition of the thickness change rate threshold or the stress threshold is met, it is marked as a key grid cell 21 , otherwise it is marked as a non-key grid cell 31 ;

[0083] In practice, the thickness change rate distribution data R ij and preload stress distribution data σ ij According to the corresponding spatial coordinates, it is mapped to the grid unit, and each grid unit is judged. When the thickness change rate distribution data R is satisfied, ij Greater than or equal to the thickness change rate threshold R0, or preload stress distribution data σ ij When the value is greater than or equal to any of the stress thresholds σ0, the grid unit is marked as a critical grid unit 21, otherwise it is marked as a non-critical grid unit 31; wherein, the thickness change rate threshold R0 is preferably 0.05; the stress threshold σ0 is determined according to the yield limit of the waterproof material, preferably, σ0 is 0.3 times the yield limit of the waterproof material.

[0084] It's understandable that the thickness change rate threshold, R0, is based on the critical value for melt buildup, calibrated with historical experimental data. When the thickness change rate is ≥5%, the risk of melt accumulation due to uneven heat conduction increases significantly. The stress threshold, σ0, corresponds to the crack initiation stress threshold. Exceeding this value significantly increases the probability of cutting defect propagation. This dual-conditional logic ensures that every risk factor is captured.

[0085] Step S143: Based on the test conditions of the target specimen, the grid cells corresponding to the test stress area are forcibly marked as key grid cells 21;

[0086] It is understandable that the stress-bearing area to be tested is the core sensitive area for specimen failure. For example, in a water pressure test, the sealing surface is subjected to a water pressure of 0.5 MPa, and micron-level cutting defects can cause leakage, affecting the experimental results. Therefore, the priority of mandatory marking of the stress-bearing area to be recorded as a key grid unit 21 is higher than the binary classification, ensuring that these areas are always cut with high precision to avoid the particularity of the working conditions being missed by the general threshold.

[0087] Step S144, removing isolated single-point key grid cells 11;

[0088] Specifically, in step S144 , the isolated single-point key grid unit 11 is a single key grid unit 21 in which a predetermined number of grid units are continuous before and after the single key grid unit 21 in the cutting path direction and are all non-critical.

[0089] In implementation, the cutting path is discretized into point sequences based on the grid cells, specifically:

[0090] P k ,k=1,2,...,N;

[0091] Traverse the point sequence P k All points in the k If the corresponding grid cell is a key grid cell, but the first n grid cells and the last n grid cells in the cutting path direction are non-key grid cells, then the key grid cell is determined to be the isolated single-point key grid cell, and the grid cell is re-marked as a non-key grid cell, where n≥10, and preferably, n is 12.

[0092] In this embodiment, for the isolated single-point critical grid unit 11, in the cutting path direction, there are the first front side grid unit a1, the second front side grid unit a2, the third front side grid unit a3, the first rear side grid unit b1, the second rear side grid unit b2 and the third rear side grid unit b3, all of which are non-critical grid units 31. Therefore, the grid unit is determined to be an isolated single-point critical grid unit 11, and the grid unit is remarked as a non-critical grid unit 31.

[0093] It is understandable that isolated single-point critical grids are usually data noise rather than real risk areas, and will interfere with the merging of critical path segments 22. If the isolated single-point critical grids are not removed, it will cause frequent power switching and thus cause cutting heat accumulation, which is not conducive to the path optimization of laser cutting.

[0094] Step S145 : merging adjacent grid cells of the same type to form continuous critical path segments 22 and non-critical path segments 32 .

[0095] Specifically, step S145 includes:

[0096] Step S1451, performing a morphological closing operation on the key grid cell 21 to fill the gap;

[0097] In implementation, the morphological closing operation is specifically as follows: first, a dilation operation is performed on the key grid cell 21 in the direction of the cutting path (preferably, the boundary is expanded by 3 to 5 grid cells), and then an erosion operation is performed in the direction of the cutting path (the number of grid cells of the contraction boundary is the same as that of the expansion boundary);

[0098] It can be understood that this operation is similar to "bridge building". During the laser cutting process of waterproof materials, due to the accuracy limitations of the thickness measuring equipment, false gaps may appear between the critical grid units 21. The gaps formed by such non-critical grid units 31 will lead to a large number of discontinuous critical path segments 22 and non-critical path segments 32, making subsequent steps difficult to implement; first, an expansion operation is performed to expand the boundary of the critical grid unit 21 outward by a certain number of grid units so that the false gap is permanently closed, and then an corrosion operation is performed to restore the original range of the critical grid unit 21 to ensure the continuity of the critical path segment 22 and the non-critical path segment 32.

[0099] Step S1452 , performing connected domain analysis on the key grid unit 21 , merging discrete segments whose spacing is less than a preset merging threshold;

[0100] In implementation, all key grid cells 21 are identified and spacing detection is performed: if the spacing between two key grid cells 21 is less than or equal to a preset merging threshold, they are merged into a single continuous segment, wherein the preset merging threshold is preferably the length of 2 grid cells.

[0101] It can be understood that the small gaps that still exist after the morphological closing operation are processed by connected domain analysis (similar to hole filling in image processing), so as to avoid frequent fluctuations in laser energy caused by changes in cutting laser parameters due to interruption of the critical path segment 22.

[0102] Step S1453: output a topologically continuous closed path segment.

[0103] In implementation, the processed mesh is reorganized into continuous path segments by type, and a closure check is performed to ensure that all path segments are connected end to end. The output is a vector path including critical path segments 22 and non-critical path segments 32 that can be executed by the laser cutting system.

[0104] Specifically, the step S1 further includes:

[0105] Step S16: when the length of the longest non-critical path segment 32 among the non-critical path segments 32 is less than the minimum representation length, generating a compensating cutting segment 41 in the longest non-critical path segment of the target specimen;

[0106] In practice, the lengths of all non-critical path segments 32 are calculated, and the longest segment is identified, whose length is recorded as L max ; If L max If it is less than L0, a compensating cutting segment 41 and a deleting segment 42 are generated outside the path segment, wherein L0 is the minimum characterization length in millimeters (mm). Preferably, L0 is 80 mm.

[0107] The calculation formula of the compensation cutting section 41 is specifically:

[0108] L b =L0-L max +L d ,

[0109] Among them, L b The length of the compensation cutting section 41 is expressed in millimeters (mm); L d The length of the deleted segment 42 in millimeters (mm)

[0110] Step S17 : Incorporating the compensating cut segment 41 into the non-critical path segment 32 to form an extended non-critical path segment 32 , so that the length of the extended non-critical path segment 32 meets the minimum characterization length requirement.

[0111] In implementation, the compensating cut segment 41 is connected to the original non-critical path segment 32 , and the deleted segment 42 is deleted at the same time; the entire path after the merger is uniformly marked as the extended non-critical path segment 32 .

[0112] In practice, after completing the laser cutting of the target specimen, the deleted non-critical path segment 32 should be repeatedly executed to cut off the extra portion added by the extended non-critical path segment 32 , which will not be described in detail here.

[0113] It can be understood that the minimum characterization length is the critical value for the stability of the cutting parameters calibrated based on historical experimental data. When the length is lower than this, the laser power / speed cannot reach a steady state, the data fluctuates greatly, and the data samples obtained are small, which makes it difficult to characterize the quality data of laser cutting in the non-critical path segment 32. Therefore, it needs to be extended to ensure that the cutting quality data is statistically significant.

[0114] Step S2, determining basic cutting parameters based on the basic size data of the target specimen, cutting the non-critical path segment 32 of the target specimen with the basic cutting parameters, and monitoring the cutting edge size data of the target specimen in real time;

[0115] In the implementation, the basic cutting parameters are determined according to the specimen thickness T and the material type (see Table 1 below). The basic cutting parameters include the basic cutting speed and the basic laser power. The basic cutting speed is recorded as V0, in meters per minute (m / min); the basic laser power is recorded as P 0， The unit is watt (W); the non-critical path segment 32 is cut at P0 and V0, and the cutting edge size is monitored in real time using a laser displacement sensor, with a measurement point collected every 0.1 mm; the cutting edge size data includes the slit width and edge verticality, the slit width is recorded as W, in millimeters (mm); the edge verticality is recorded as θ, in degrees (°).

[0116] Table 1 Basic cutting parameter query table

[0117]

[0118] It can be understood that the above-mentioned basic cutting parameter query table is a parameter reference table summarized based on the characteristics of polymer waterproof membranes and industrial laser cutting practices. The data is based on industry experience, literature research and historical experimental calibration. Technical personnel in this field can make adaptive adjustments according to actual conditions, which will not be repeated here.

[0119] Step S3, based on the obtained cutting edge dimension data of the target specimen and the sample preparation preset data, the basic cutting parameters are adjusted to obtain verification cutting parameters, and the laser cutting of the non-critical path segment 32 is completed using the verification cutting parameters;

[0120] Specifically, step S3 includes:

[0121] Step S31, comparing the real-time monitored cutting edge dimension data with the sample preparation preset data one by one, and calculating the corresponding several dimension deviation values;

[0122] In implementation, the sample preparation preset data of the slit width W is W0, in millimeters (mm), preferably, W0 is 0.8 mm; the sample preparation preset data of the edge verticality θ is θ0, in degrees (°), preferably, θ0 is 90°.

[0123] The size deviation value includes the slit width deviation value, recorded as △W, in millimeters (mm); and the edge verticality deviation value, recorded as △θ, in degrees (°), where:

[0124] ΔW=W-W0,

[0125] Δθ=θ-θ0.

[0126] Step S32: When a predetermined number of consecutive dimensional deviation values ​​are greater than an allowable tolerance range, adjusting the basic cutting parameters, wherein the basic cutting parameters are adjusted based on the direction and magnitude of the dimensional deviation value, and the basic cutting parameters include laser power and cutting speed;

[0127] In this embodiment, the allowable tolerance range of the slit width deviation value is △W≤±0.1mm; the allowable tolerance range of the edge verticality deviation value is △θ≤±2°.

[0128] In this embodiment, the adjustment rules for the basic cutting parameters are specifically as follows:

[0129] When 5 consecutive data points satisfy △W>0, and △W is greater than the allowable tolerance range, adjust the laser power, P 新 =0.95×P0;

[0130] When 5 consecutive data points satisfy △W<0, and △W is greater than the allowable tolerance range, adjust the laser power, P 新 =1.05×P0;

[0131] When 5 consecutive data points satisfy the condition that △θ is greater than the allowable tolerance range, adjust the cutting speed, V 新 =0.9×V0; where P 新 is the adjusted laser power, V 新is the cutting speed after adjustment, P0 is the laser power before adjustment, and V0 is the cutting speed before adjustment.

[0132] It can be understood that the core principle of dynamic parameter adjustment is based on the causal relationship between cutting quality defects and cutting laser energy input. When real-time monitoring detects deviation in cutting edge size, the system automatically analyzes the defect type and its physical cause, and adjusts the laser parameters accordingly: When the kerf is too wide (positive deviation), this is usually caused by excessive laser power leading to excessive material melting. In principle, excessive energy input will expand the molten pool and the liquid material will flow to the sides, thereby widening the kerf. In this case, the system will proportionally reduce the laser power, shrinking the molten pool size by reducing heat input. When the kerf is too narrow (negative deviation), this is usually due to insufficient power resulting in insufficient penetration. When the energy is insufficient to completely vaporize the material, the melt cannot be fully discharged, forming a narrow kerf. The system will then increase the power accordingly to enhance the material vaporization capacity and ensure that the kerf reaches the preset width. When the edge is tilted (perpendicularity deviation), the main cause is cutting speed mismatch. Excessive speed will result in insufficient residence time of the laser beam in the material, resulting in asymmetric heat conduction and a tilted incision. In this case, the system will reduce the cutting speed and extend the laser action time to make the heat distribution more uniform. At the same time, a continuous multi-point trigger mechanism (the default is 5 consecutive points out of tolerance) is used to distinguish real defects from instantaneous interference. The deviation of a single data point may be caused by local impurities in the material or detection noise, while the deviation of multiple consecutive points indicates the existence of systematic parameter imbalance. The continuous multi-point trigger mechanism ensures the timeliness of adjustment and avoids excessive parameter oscillation.

[0133] Step S33 , completing the cutting of the current non-critical path segment 32 with the adjusted verification cutting parameters, and continuously acquiring the corresponding relationship between the cutting edge dimension data of the target specimen and the corresponding verification cutting parameters during the cutting process.

[0134] With the new parameter P 新 and V 新 Continue laser cutting to obtain verification cutting parameters, and establish a cutting parameter-cutting quality (size deviation value) mapping table, that is, the correspondence between the cutting edge size data and the corresponding verification cutting parameters.

[0135] Step S4, determining a first trusted path segment based on the cutting edge dimension data of the non-critical path segment 32, and extracting the corresponding verification cutting parameters to form a trusted data set;

[0136] Specifically, the step S4 includes:

[0137] Step S41 , screening a continuous section in the non-critical path segment 32 where the cutting edge dimension data is continuously within a preset tolerance band, and marking the section as a first confidence path segment;

[0138] In implementation, the continuous sections in which the cutting edge dimension data in the non-critical path section 32 is continuously within the preset tolerance band are screened. When the dimension data of 50 consecutive measurement points (corresponding to 5 mm length) are all within the tolerance band, the continuous section is marked as the first trust path section.

[0139] In this embodiment, the preset tolerance zone for the slit width deviation value is △W≤±0.08mm; the preset tolerance zone for the edge verticality deviation value is △θ≤±1°.

[0140] It can be understood that only a continuous and stable cutting section that is long enough (50 consecutive measurement points) can indicate that the cutting parameters used in this section are reliable and adapt to the material properties. By setting the minimum length requirement (5mm), accidental qualified points are avoided from being mistakenly accepted, ensuring that the parameters are statistically significant; for the preset tolerance band, the allowable tolerance range that is stricter than the target specimen size tolerance requirement should be selected, so that the cutting accuracy of the edge size data cut by the verification cutting parameters can meet the requirements, and can characterize the cutting parameters as the optimal energy input combination under the corresponding target specimen.

[0141] Step S42, extracting the set of verified cutting parameters in the first trusted path segment, recording them as trusted cutting parameters, and extracting the cutting edge dimension data corresponding to the trusted cutting parameters, recording them as trusted dimension data;

[0142] In implementation, for each first trusted path segment, the verification cutting parameters corresponding to the segment are extracted and recorded as trusted cutting parameters, including laser power and cutting speed values, which are recorded as a laser power-cutting speed combination; the cutting edge dimension data corresponding to the trusted cutting parameters are extracted and recorded as trusted dimension data.

[0143] It can be understood that the cutting accuracy of the edge size data cut by the verification cutting parameters corresponding to the first confidence path segment can meet the requirements, which can characterize that the cutting parameters are suitable for the specific material properties and are the optimal energy input combination under the corresponding target specimen.

[0144] Step S43: storing the trusted cutting parameters and the trusted size data in spatial order to form a trusted data set, and recording a mapping relationship between the trusted cutting parameters and the trusted size data.

[0145] In implementation, the parameter set is stored in spatial position order to form a trusted data set, specifically: the data is arranged in the order of the cutting path, and each data unit contains: starting coordinates, ending coordinates, corresponding trusted cutting parameters and corresponding trusted size data. At the same time, its topological relationship is simplified, and adjacent first trusted path segments are automatically connected, and non-adjacent first trusted path segments retain position marks.

[0146] Step S5, determining the correlation degree of the verification cutting parameters based on the fluctuation degree of the verification cutting parameters in the trusted data set, obtaining key cutting parameters and forming a trusted data set;

[0147] Specifically, step S5 includes:

[0148] Step S51, dividing each of the first confidence path segments according to the change of the confidence cutting parameter to form a plurality of second confidence path segments;

[0149] In implementation, the first confidence path segment is finely segmented, specifically: the laser power value and cutting speed value are scanned at intervals of 0.1 mm along the cutting path, and when the laser power changes by ≥20% or the cutting speed changes by ≥30%, it is marked as a segmentation point; at the segmentation point, the original first confidence path segment is divided into new sub-segments, namely the second confidence path segment.

[0150] It is understandable that although the overall quality of the first confidence path segment is qualified, there may be parameter fine-tuning inside it. By segmenting according to the parameter change points, it is ensured that the cutting parameters in each second confidence path segment are not completely consistent, providing available data samples for subsequent correlation analysis.

[0151] Step S52, calculating the correlation coefficient between the trusted cutting parameter and the corresponding trusted size data in each of the second trusted path segments to determine the degree of association between each trusted cutting parameter and the corresponding trusted size data;

[0152] In implementation, the average value of the cutting speed in each second confidence path segment and the average value of the laser power in each second confidence path segment are calculated; the average value of the slit width deviation ΔW in each second confidence path segment and the average value of the verticality deviation Δθ in each second confidence path segment are calculated; and then the correlation between ΔW and the laser power and the correlation between Δθ and the cutting speed of each second confidence path are calculated.

[0153] In this embodiment, the Pearson correlation coefficient is specifically used to calculate the above correlation degree, and the correlation coefficient r1 between the laser power and ΔW and the correlation coefficient r2 between the cutting speed and Δθ are obtained.

[0154] It can be understood that the correlation shows the specific influence mechanism of cutting parameters on cutting quality, where when r1>0, power change causes the cutting seam to become wider, and when r1<0, power change causes the cutting seam to become narrower; when r2>0, speed increase leads to improved verticality, and when r2<0, speed increase leads to increased inclination.

[0155] Step S53 , eliminating the trusted cutting parameters whose correlation coefficients are less than a preset correlation coefficient threshold, obtaining the key cutting parameters and forming the trusted data set.

[0156] In implementation, the preset correlation coefficient threshold is 0.7. For r1, the correlation between laser power and ΔW, the cutting parameters corresponding to |r1|>0.7 are retained. For r2, the correlation between cutting speed and Δθ, the cutting parameters corresponding to r2>0.7 are retained. Other cutting parameters are eliminated, and the cutting parameters with the largest correlation among the retained cutting parameters are compared and calculated to obtain the key cutting parameters, and then the confidence data set is obtained.

[0157] It is understandable that the setting of the correlation coefficient threshold of 0.7 is based on the principle of statistical significance. Eliminating low-correlation parameters can avoid overfitting and ensure that the recommended parameters are universal.

[0158] Step S6, laser cutting the critical path segment 22 using the key cutting parameters in the confidence data set to obtain a target test piece;

[0159] Step S7: establishing a deep learning optimization model based on the plurality of confidence data sets, and outputting the basic cutting parameters of subsequent test pieces.

[0160] Specifically, step S7 includes:

[0161] Step S71, collecting the confidence data sets and corresponding kerf quality rating data sets of multiple groups of target test pieces;

[0162] In the implementation, a complete confidence data set containing at least 50 sets of historical cutting specimens is collected. Each data set contains: optimized cutting parameters (laser power / cutting speed combination) and corresponding quality rating (pass / fail).

[0163] The specific grading standards for seam quality are:

[0164] Qualified: slit width ±0.02mm and verticality ±0.3°;

[0165] Unqualified: Any indicator is out of tolerance (slit width or verticality);

[0166] The data is stored as a structured database, with each record containing metadata such as material type and thickness.

[0167] Step S72, constructing a supervised learning model with the cutting parameters as input and the cutting quality rating as output;

[0168] In implementation, the model uses a standard multi-layer perceptron (MLP) neural network. The input layer encodes the basic size data and material type; the output layer is the recommended initial laser power and the recommended initial cutting speed; the training process uses 70% historical data for training, 15% for validation set parameter adjustment, and 15% for test set evaluation.

[0169] Step S73: When the basic size data of the new specimen is input, the optimized initial values ​​of the cutting parameters are output.

[0170] In this embodiment, when basic size data of a new specimen is input, the output optimized cutting parameter initial values ​​are used as basic cutting parameters to cut the non-critical path segment 32, so as to continuously optimize the cutting path and the cutting parameters corresponding to the cutting path.

[0171] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A method for optimizing laser cutting paths for waterproof materials based on deep learning, characterized in that: include: Step S1, obtaining basic size data, thickness change rate distribution data, and preload stress distribution data of a target test piece of waterproof material, and determining a critical path segment and a non-critical path segment of a cutting path of the target test piece; Step S2, determining basic cutting parameters based on the basic size data of the target specimen, cutting the non-critical path segment of the target specimen with the basic cutting parameters, and monitoring the cutting edge size data of the target specimen in real time; Step S3, based on the obtained cutting edge dimension data of the target specimen and the sample preparation preset data, the basic cutting parameters are adjusted to obtain verification cutting parameters, and the laser cutting of the non-critical path segment is completed with the verification cutting parameters; Step S4, determining a first trusted path segment based on the cutting edge size data of the non-critical path segment, and extracting the corresponding verification cutting parameters to form a trusted data set; Step S5, determining the correlation degree of the verification cutting parameters based on the fluctuation degree of the verification cutting parameters in the trusted data set, obtaining key cutting parameters and forming a trusted data set; Step S6, performing laser cutting on the critical path segment using the key cutting parameters in the confidence data set to obtain a target test piece; Step S7: establishing a deep learning optimization model based on the plurality of confidence data sets, and outputting the basic cutting parameters of subsequent test pieces.

2. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 1, characterized in that: The step S1 comprises: Step S11, obtaining the basic dimension data of the target specimen, including the cutting path and thickness reference value; Step S12, obtaining the thickness change rate distribution data of the sample where the target specimen is located; Step S13, performing finite element analysis based on the detection working conditions of the target specimen and the basic dimension data to generate the preload stress distribution data; Step S14 , determining the critical path segments and non-critical path segments of the cutting path according to the thickness change rate distribution data and the preload stress distribution data.

3. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 2, characterized in that: The step S14 includes: Step S141, discretizing the cutting path into a plurality of grid units, wherein the grid resolution of each grid unit is set according to the thickness reference value; Step S142 , performing binary classification on each grid cell, wherein if the judgment condition of the thickness change rate threshold or the stress threshold is met, it is marked as a critical grid cell, otherwise it is marked as a non-critical grid cell; Step S143: Based on the test conditions of the target specimen, the grid cells corresponding to the test stress area are forcibly marked as key grid cells; Step S144, removing isolated single-point key grid cells; Step S145 : merging adjacent grid units of the same type to form continuous critical path segments and non-critical path segments.

4. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 3, characterized in that: The step S145 includes: Step S1451, performing a morphological closing operation on the key grid cells to fill the gaps; Step S1452: Perform connected domain analysis on key grid cells and merge discrete segments whose spacing is less than a preset merging threshold. Step S1453: output a topologically continuous closed path segment.

5. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 3, characterized in that: In step S144, the isolated single-point key grid unit is a single key grid unit in which a predetermined number of consecutive grid units are all non-key types in the cutting path direction.

6. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 2 or 4, characterized in that: The step S1 further includes: Step S15: when the length of the longest non-critical path segment among the non-critical path segments is less than the minimum representation length, generating a compensating cutting segment in the longest non-critical path segment of the target specimen; Step S16: merging the compensating cutting segment into the non-critical path segment to form an extended non-critical path segment, so that the length of the extended non-critical path segment meets the minimum characterization length requirement.

7. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 4 or 5, characterized in that: The step S3 comprises: Step S31, comparing the real-time monitored cutting edge dimension data with the sample preparation preset data one by one, and calculating the corresponding several dimension deviation values; Step S32: When a predetermined number of consecutive dimensional deviation values ​​are greater than an allowable tolerance range, adjusting the basic cutting parameters, wherein the basic cutting parameters are adjusted based on the direction and magnitude of the dimensional deviation value, and the basic cutting parameters include laser power and cutting speed; Step S33 , completing the cutting of the current non-critical path segment with the adjusted verification cutting parameters, and continuously acquiring the corresponding relationship between the cutting edge dimension data of the target specimen and the corresponding verification cutting parameters during the cutting process.

8. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 6, characterized in that: The step S4 includes: Step S41, screening the continuous sections in the non-critical path segments where the cutting edge dimension data is continuously within the preset tolerance band, and marking them as first confidence path segments; Step S42, extracting the set of verified cutting parameters in the first trusted path segment, recording them as trusted cutting parameters, and extracting the cutting edge dimension data corresponding to the trusted cutting parameters, recording them as trusted dimension data; Step S43: storing the trusted cutting parameters and the trusted size data in spatial order to form a trusted data set, and recording a mapping relationship between the trusted cutting parameters and the trusted size data.

9. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 7, characterized in that: The step S5 comprises: Step S51, dividing each of the first confidence path segments according to the change of the confidence cutting parameter to form a plurality of second confidence path segments; Step S52, calculating the correlation coefficient between the trusted cutting parameter and the corresponding trusted size data in each of the second trusted path segments to determine the degree of association between each trusted cutting parameter and the corresponding trusted size data; Step S53 , eliminating the trusted cutting parameters whose correlation coefficients are less than a preset correlation coefficient threshold, obtaining the key cutting parameters and forming the trusted data set.

10. The method for optimizing the laser cutting path of waterproof materials based on deep learning according to claim 9, characterized in that: The step S7 comprises: Step S71, collecting the confidence data sets and corresponding cutting quality rating data sets of multiple groups of target test pieces; Step S72, constructing a supervised learning model with the cutting parameters as input and the cutting quality rating as output; Step S73: When the basic size data of the new specimen is input, the optimized initial values ​​of the cutting parameters are output.

Images