A digital twin-based intelligent control method

Abstract

This invention discloses an intelligent control method based on digital twins, specifically relating to the field of digital twin technology. The method includes S1: multi-dimensional data acquisition; S2: construction of a digital twin sheet material model; S3: generation of a nesting scheme; S4: iterative shearing; S5: comprehensive quality assessment; and S6: adaptive optimization. By constructing a digital twin sheet material model integrating geometric dimensions, surface quality, and material properties, this invention employs a multi-objective optimization algorithm when making nesting decisions. Through a dynamic iterative closed loop of nesting, shearing, scanning, and re-nesting, and scientifically determining the termination timing of sheet material cutting based on the residual value index, it can improve the utilization rate of thin metal sheets, significantly reduce material waste, and thus maximize material utilization while ensuring the quality of thin metal sheet shearing. This simultaneously meets the dual requirements of cost reduction and quality improvement.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and more specifically, to an intelligent control method based on digital twins. Background Technology

[0002] Metal sheet shearing lines are a type of metal processing equipment that integrates automation and high precision. They are mainly used to cut coiled metal sheets, such as steel, aluminum, and copper, into narrow strips of specific widths or fixed-length plates. By controlling the layout of the metal sheet shearing line, the efficiency and precision of metal sheet production can be improved, thereby promoting the high-quality development of the metal processing industry.

[0003] Currently, the layout and scheduling control of metal sheet shearing lines mainly relies on a combination of image acquisition and simple algorithms. That is, an industrial camera is used to acquire images of the sheet surface, and after obtaining the approximate outline of the sheet based on image recognition, a wiring and shearing scheme is generated using simple layout logic such as greedy algorithms and rectangular nesting. Then, the shearing operation is executed manually or by a semi-automatic control system.

[0004] However, it still has some drawbacks in actual use. First, the existing simple nesting algorithm only takes whether the area of ​​the board can accommodate the template as the core judgment criterion, without considering key factors such as the adaptability of the template geometry and the reusability of the remaining area of ​​the board, which easily generates a large amount of irregular waste. At the same time, the remaining boards after cutting may still have utilization value and are directly discarded, which to a certain extent leads to serious material waste, resulting in low material utilization and significantly increasing the cost of raw materials for production.

[0005] Secondly, due to the defects such as scratches, dents, and oxide spots that easily occur on the surface of metal sheets during rolling, storage, and transportation, there are significant differences in the roughness of different areas. However, existing technologies only focus on the geometric dimensions of the sheet material and do not conduct precise collection and quantitative analysis of surface quality. This approach can easily lead to the problem of using high-quality materials for low-end applications and using low-quality materials for high-end applications, ultimately resulting in an increased defect rate of finished products, increased rework and scrap costs, and an inability to achieve the dual goals of optimizing quality and utilization. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide an intelligent control method based on digital twins, which constructs a digital twin model through multi-dimensional data acquisition and iteratively cuts the generated nesting scheme to achieve a dual-objective balance between maximizing material utilization and optimizing cutting quality, effectively solving the problems raised in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] S1: Multi-dimensional data acquisition: Real-time acquisition of three types of raw data—geometric dimensions, surface quality, and material properties—of a single metal sheet in the area to be sheared, according to predefined data specifications;

[0009] S2: Digital Twin Sheet Model Construction: After preprocessing the geometric dimensions, surface quality, and material properties, the mass value of the metal sheet to be sheared is calculated by weighting. This establishes a sheet mass distribution map, which is then used to divide the metal sheet into quality regions and establish a sample quality requirement database.

[0010] S3: Generate a layout plan: Retrieve the set of templates to be cut from the template quality requirement database, including the geometric dimensions and minimum quality threshold requirements of each template, and use an improved genetic algorithm to optimize the layout, thereby outputting the optimal layout plan;

[0011] S4: Iterative shearing: The optimal layout scheme is sent to the shearing equipment, and the shearing parameters are collected in real time. After shearing, the remaining thin plate is rescanned, the digital twin plate model is updated, the remaining value index is calculated, and the layout termination judgment is made accordingly.

[0012] S5: Comprehensive Quality Assessment: After shearing, the metal sheet is inspected for quality, including dimensional accuracy and cut quality, and the comprehensive quality coefficient of the sheared metal sheet is calculated accordingly.

[0013] S6: Adaptive Optimization: Based on the digital twin sheet material model and the comprehensive quality coefficient corresponding to the sheared metal sheet, a comprehensive performance evaluation is performed. Then, the sheet state is used as the state space, the layout scheme is used as the action space, and the comprehensive performance evaluation index is used as the reward to update the layout scheme and perform adaptive optimization.

[0014] The technical effects and advantages of this invention are as follows:

[0015] 1. This invention constructs a digital twin plate model that integrates geometric dimensions, surface quality, and material properties, enabling quantitative and visualized full-domain perception of plate quality. This provides a reliable data foundation for subsequent refined layout decisions. Furthermore, a multi-objective optimization algorithm is employed when executing layout decisions, accurately matching the plate quality distribution and sample quality requirements to achieve refined production management. This maximizes material utilization while ensuring the quality of metal sheet shearing, thus meeting the dual requirements of cost reduction and quality improvement.

[0016] 2. This invention, through a dynamic iterative closed loop of nesting, shearing, scanning, and re-nesting, scientifically determines the timing of terminating the cutting of sheet metal based on the residual value index. This improves the utilization rate of thin metal sheets and greatly reduces material waste, making it particularly suitable for processing precious metal sheets. After shearing, this invention performs a comprehensive quality assessment and feeds the results back to adaptive optimization, forming a complete quality closed loop from quality prediction to performance verification to strategy optimization. This achieves traceability, assessability, and improveability of product quality, significantly enhancing the level of quality control. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0018] Figure 2 This is a schematic diagram illustrating the construction of the digital twin board model of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] As attached Figure 1-2 The present invention illustrates an intelligent control method based on digital twins. A specific implementation of this invention includes the following steps:

[0021] S1: Multi-dimensional data acquisition: Real-time acquisition of three types of raw data—geometric dimensions, surface quality, and material properties—of a single metal sheet in the area to be sheared, according to predefined data specifications.

[0022] In this embodiment, it should be specifically noted that, according to the predefined data specifications, the metal sheet is marked with sampling points according to a grid with a sampling length. The sampling length can be 8mm, and the grid can be 500mm×500mm. The geometric dimensions of the individual metal sheet in the area to be sheared are collected by a size measurement terminal, including the length, width, thickness, and flatness of the metal sheet. The length is the maximum extension dimension of the metal sheet along the X-axis, the width is the maximum extension dimension of the metal sheet along the Y-axis, the thickness is the thickness of the metal sheet along the Z-axis, and the flatness is the maximum deviation value between each sampling point on the surface of the metal sheet and the ideal plane. Specifically, the flatness can be obtained by placing the metal sheet on a horizontal plane and monitoring the angle between the metal sheet and the horizontal plane through the size monitoring terminal.

[0023] Surface quality is assessed by acquiring optical images of the thin plate surface using a machine vision terminal. Image processing algorithms are then used to analyze the surface features of the metal sheet, including surface roughness and defect rate. The surface roughness measurement process involves sampling points on the metal sheet surface using the average surface height baseline as a reference. A stylus sensor slides uniformly along the surface, and the displacement of the stylus due to microscopic surface undulations is converted into a continuous electrical signal. The device samples the electrical signal at a fixed frequency and calculates the absolute value of the distance between each sampling point and the contour centerline. Within the sampling length, the arithmetic mean of these absolute distance values ​​is taken to obtain the surface roughness. Surface defect rate is obtained by automatically identifying defects at sampling points using an AI algorithm based on the optical images of the thin plate surface, and then dividing the area of ​​the surface defect at the sampling point by the surface area of ​​the thin plate.

[0024] Material properties include the hardness of the metal sheet and the material composition deviation. The hardness of the metal sheet is determined by applying pressure to the sampling points of the metal sheet sample in the area to be sheared using a material property sensing terminal. After maintaining the pressure for 15 seconds, the hardness of the metal sheet sample is automatically read. The material composition deviation is the difference between the actual composition of the metal sheet and the standard composition. Specifically, the intensity of the characteristic spectral lines of the metal sheet is measured by the material property sensing terminal and compared with the intensity of the spectral lines of the standard sample to calculate the actual element content. Then, the actual content is compared with the standard composition of the sheet to obtain the material composition deviation.

[0025] It should be added that the element content mentioned above refers to the specific weight percentage of the main chemical elements that make up the metal sheet, which mainly depends on the material of the metal sheet itself. For example, if the metal sheet is stainless steel, the main element content is iron; if the metal sheet is aluminum alloy, the main element content is aluminum.

[0026] In a more specific application of the present invention, the size measurement terminal is used to measure the geometric dimensions of the metal sheet. Specifically, it can be a laser scanner. The laser scanner can perform high-precision measurement on the metal sheet and can achieve non-contact measurement, which can effectively reduce the probability of scratches on the surface of the metal sheet during high-speed detection, ensure the original quality of the material, and improve the continuity and efficiency of the production line.

[0027] Machine vision terminals are used to acquire optical images of thin metal sheets and to obtain surface features of the thin metal sheets based on the optical images. Specifically, they can be a combination of acoustic imaging devices and stylus sensors. The acoustic imaging devices can provide optical images of the thin metal sheets through acoustic imaging technology, helping to identify scratches, dents and other defects on the sheets. The stylus sensors can convert the surface defects of the thin metal sheets into electrical signals that are difficult to measure directly, thus greatly improving the detection efficiency.

[0028] Material property sensing terminals are used to sense the material properties of thin metal sheets. Specifically, they can be hardness testers and spectrometers. Since both hardness testers and spectrometers are portable devices, they can adapt to the needs of industrial sites when shearing thin metal sheets. Hardness testers can accurately identify hardness fluctuations in different areas of the sheet, providing a basis for subsequent shearing. Spectrometers can accurately capture minute differences in the composition of the sheet, which can prevent batches of sheets from being scrapped due to composition issues.

[0029] S2: Digital Twin Sheet Model Construction: After preprocessing the geometric dimensions, surface quality, and material properties, the mass value of the metal sheet to be sheared is calculated by weighting. This establishes a sheet mass distribution map, which is then used to divide the metal sheet into quality regions and establish a sample quality requirement database.

[0030] It needs to be explained that the preprocessing operation specifically involves normalizing the geometric dimensions, surface quality, and material properties, including normalizing both positive and negative indices. The normalization of positive indices is specifically expressed as: (xx min ) / (x max -x min ), where x represents the data corresponding to the positive index of the metal sheet, x min and x max These represent the minimum and maximum values ​​of the data at each sampling point, respectively. The normalization of the negative index is specifically expressed as: (x max -x) / (x max -x min Positive indicators include the flatness of geometric dimensions and the hardness of metal sheets, while negative indicators include the surface roughness and defects of surface quality and the material composition deviation of material properties.

[0031] In this embodiment, the construction of the thin plate mass distribution map is specifically described as follows:

[0032] After preprocessing, the flatness of each sampling point is extracted, and the flatness of each sampling point is compared with the standard flatness to screen out qualified sampling points. The number of qualified sampling points is then extracted, and the number of qualified sampling points is compared with the total number of sampling points to obtain the geometric dimension compliance rate.

[0033] The surface quality of each sampling point is extracted, and then the average value of the surface roughness and defect degree after preprocessing is calculated to obtain the surface quality score. Similarly, the material properties of each sampling point are extracted to obtain the material property score.

[0034] The quality value of each sampling point is calculated by weighted summation based on the geometric dimension compliance rate, surface quality score, and material property score, and is labeled as Qc. The weights are set according to production requirements. For example, the weights corresponding to the geometric dimension compliance rate, surface quality score, and material property score are 0.3, 0.4, and 0.3, respectively.

[0035] Based on the mass values ​​of all sampling points across the entire plate, a mass distribution map of the thin metal plate corresponding one-to-one with the location of the thin metal plate is generated.

[0036] It should be further explained that the quality zone division is based on the generated thin plate quality distribution map. According to the preset quality threshold, the thin plate is divided into different quality zones, including the high-quality zone, the qualified zone, and the degraded zone. The high-quality zone is Qc≥Q1, which means that the surface of this zone is smooth, without defects, and the material properties are uniform. It is used to cut the samples with the highest requirements for appearance and performance. The qualified zone is Q2≤Qc<Q1, which means that there are slight defects or performance deviations in this zone. It is used to cut the samples with general quality requirements. The degraded zone is Qc<Q2, which means that there are obvious defects or the material properties are not up to standard. It is only used to cut the samples with the lowest quality requirements or as scrap material for recycling. For example, Q1 and Q2 can be 0.9 and 0.7, respectively.

[0037] It should be added that the quality zone division results are visualized in the digital twin model using different color layers and stored as matrix data. The template quality requirement database is used to store the quality threshold requirements Qr for templates for different purposes. For example, suppose metal sheet cutting is performed on automotive parts. Automotive parts include templates for automotive exterior panels, interior structural parts, packaging liners, and other parts. The minimum quality threshold requirement Qr for the automotive exterior panel template is 0.9, so this part must be laid out using metal sheets from the high-quality zone. The minimum quality threshold requirement Qr for the interior structural part template is 0.75, so this part must be laid out using metal sheets from the acceptable zone or above. The minimum quality threshold requirement Qr for the packaging liner template is 0.25, so this part must be laid out using metal sheets from the degraded zone or above. The minimum quality threshold requirement Qr for other parts templates is 0, so this part can be laid out in any zone.

[0038] S3: Generate a nesting scheme: Retrieve the set of templates to be cut from the template quality requirement database, including the geometric dimensions and minimum quality threshold requirements of each template, and use an improved genetic algorithm to optimize the nesting, thereby outputting the optimal nesting scheme.

[0039] In this embodiment, the optimal nesting scheme generation operation needs to be explained in detail as follows:

[0040] The template quality requirement database retrieves all templates for the current production task based on the current production order, generating a set of templates to be cut. These templates are then sorted by area from largest to smallest, which improves the initial convergence speed of subsequent nesting algorithms. These templates are denoted as {S1, S2, ..., S...}. n}, where each template S i The information includes its geometry and minimum quality threshold requirements;

[0041] An improved genetic algorithm is used to optimize the layout scheme. After the algorithm iteration terminates, the optimal solution set is finally output and labeled as {Scheme 1, Scheme 2, ..., Scheme n}. The schemes are sorted in descending order according to their comprehensive fitness. The first executable layout scheme is selected as the optimal layout scheme. The optimal layout scheme includes the specific position of each sample on the thin plate, the corresponding quality area, the spacing between samples, and the material utilization rate and quality compliance rate of the scheme.

[0042] It should be explained that the improved genetic algorithm addresses the shortcomings of traditional genetic algorithms in plate cutting problems, such as premature convergence and weak local search capabilities, by making several key improvements. Specifically, the optimization of the nesting scheme includes:

[0043] Extract the arrangement order of the templates in the set of templates to be cut, and determine the rotation angle of each template relative to its placement reference point, where the placement reference point can be the lower left corner of the template, and the rotation angle includes 0°, 90°, 180°, and 270°.

[0044] For the current template S i Based on its minimum mass threshold requirement, its placement area is limited to all corresponding mass areas in the thin plate mass distribution map, and the optimal orientation of the position is found according to the rotation angle, with the goal of minimizing wasted space.

[0045] After the sample is laid out, the material utilization rate and quality matching rate of the thin plate are calculated respectively. The material utilization rate is obtained by comparing the total area of ​​all laid samples with the original total area of ​​the thin plate. The quality matching rate is obtained by comparing the number of quality-matched samples with the total number of samples. The quality-matched sample refers to the sample laid out in the area that meets its minimum quality requirements.

[0046] Weighting coefficients, denoted as α and β, are assigned to material utilization rate and quality matching rate, respectively. The weighting coefficients are dynamically adjusted according to the production task priority. For example, if the production task priority is quality, then α=0.3 and β=0.7; if the production task priority is utilization rate, then α=0.7 and β=0.3; if the production task priority is both quality and utilization rate, then α=0.5 and β=0.5. Then, the weights of material utilization rate and quality matching rate are added together for the thin plate to calculate the overall adaptability of the layout scheme.

[0047] A crossover and mutation strategy is adopted. When the number of iterations reaches a preset maximum value, the algorithm stops iterating and outputs the optimal solution set.

[0048] It should be added that the crossover mutation strategy includes adaptive crossover probability and local mutation. The adaptive crossover probability reduces the crossover probability for high-fitness nesting schemes to retain high-quality genes, and increases the crossover probability for low-fitness nesting schemes to promote gene updates. During crossover, only the position codes of the templates are exchanged, without changing the rotation angle, to avoid the size exceeding the tolerance after template rotation. Local mutation randomly selects the position of one template for fine-tuning. The fine-tuning can be a small translation of the template or replacement of the adjacent edge. The mutation probability is set to 0.05 to prevent the algorithm from getting trapped in local optima. After mutation, it is necessary to check whether the template exceeds the material range or overlaps with other templates. If not, the mutation is repeated.

[0049] S4: Iterative shearing: The optimal nesting scheme is sent to the shearing equipment, and the shearing parameters are collected in real time. After shearing, the remaining thin plate is rescanned, the digital twin plate model is updated, the remaining value index is calculated, and the nesting termination judgment is made accordingly.

[0050] In this embodiment, it should be specifically explained that S4 performs cutting using a cutting device according to the optimal nesting scheme.

[0051] It should be added that before the shearing equipment performs the shearing, the shearing process needs to be simulated in the digital twin material model to verify whether the template positions overlap and whether the tool path is reasonable. If there are any problems, return to S3 to re-optimize the layout scheme.

[0052] In the more specific applications mentioned above, the shearing equipment is used to shear thin plates according to the optimal layout scheme, and is equipped with sensors that can collect shearing parameters in real time. Specifically, an encoder can be used to collect the shearing speed, a pressure sensor can be used to collect the tool pressure, and a flow sensor can be used to collect the coolant flow rate.

[0053] It should be further explained that the specific operation for calculating the surplus value index is as follows:

[0054] After cutting according to the optimal layout scheme, the remaining thin plate is immediately scanned through the S1 size measurement terminal, machine vision terminal, and material property sensing terminal to update the digital twin plate model.

[0055] The mass value of the remaining thin plate is resampled and calculated. If the mass value of the remaining thin plate decreases due to shearing, its mass range is adjusted.

[0056] The remaining area is calculated based on the geometric dimensions of the remaining sheet, the average remaining mass value is calculated based on the mass values ​​of each sampling point, and the remaining area is multiplied by the average remaining mass value to obtain the remaining value index.

[0057] Set a value threshold and compare the residual value index with the value threshold. If the residual value index is less than the value threshold, stop the sampling. If the residual value index is greater than the value threshold, perform a second sampling.

[0058] S5: Comprehensive Quality Assessment: After shearing, the metal sheet is inspected for quality, including dimensional accuracy and cut quality, and the comprehensive quality coefficient of the sheared metal sheet is calculated accordingly.

[0059] In this embodiment, the comprehensive quality assessment is specifically described as follows: After shearing, the dimensional accuracy and cut quality of each metal sheet are inspected. The dimensional accuracy is calculated by comparing the absolute difference between the area of ​​the sheared metal sheet and the area of ​​the design template with the accuracy tolerance. The cut quality is obtained by detecting the burr height and perpendicularity of the cut using a laser profilometer, and comparing them with the allowable height and allowable perpendicularity, respectively. Based on the degree of influence of dimensional accuracy and cut quality, the comprehensive quality coefficient corresponding to the sheared metal sheet is evaluated. The accuracy tolerance, allowable height, and allowable perpendicularity are all set before shearing the metal sheet.

[0060] It should be added that the degree of influence of dimensional accuracy and cut quality can be set based on experience, taking into account the impact of quality requirements of different metal sheets in different scenarios. For example, for internal structural component templates, precise dimensions are required to improve vehicle structural safety, so the degree of influence of dimensional accuracy is higher. For other part templates, cut quality is required to ensure service life, so the degree of influence of cut quality is higher.

[0061] S6: Adaptive Optimization: Based on the digital twin sheet material model and the comprehensive quality coefficient corresponding to the sheared metal sheet, a comprehensive performance evaluation is performed. Then, the sheet state is used as the state space, the layout scheme is used as the action space, and the comprehensive performance evaluation index is used as the reward to update the layout scheme and perform adaptive optimization.

[0062] In this embodiment, it should be specifically explained that the comprehensive performance evaluation is calculated by combining the comprehensive quality coefficient with its corresponding comprehensive fitness to obtain the comprehensive performance evaluation index, which is specifically expressed as follows:

[0063] ,

[0064] Where E represents the comprehensive performance evaluation index, Cm represents the comprehensive quality coefficient corresponding to the sheared metal sheet, and Tc represents the comprehensive adaptability under the optimal layout scheme.

[0065] It should be further explained that the adaptive optimization, based on the reinforcement learning algorithm, obtains the current state of the thin sheet, including the digital twin model of the current thin sheet, and obtains the action space according to the current layout scheme. The comprehensive performance evaluation index is used as the reward. This allows the system to observe the state of the next batch of metal thin sheets and update the layout scheme accordingly. Specifically:

[0066] ,

[0067] Where π(s) represents the optimal layout scheme found in the state space, s represents the state space, a represents the action space, and E represents the reward. According to the above formula, under the current digital twin board model of thin plate, the goal is to maximize all the comprehensive performance evaluation indices to be obtained in the future among multiple layout schemes.

[0068] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0069] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent control method based on digital twinning, characterized in that, The method comprises the following steps: S1: multi-dimensional data acquisition: real-time acquisition of three types of original data of geometric size, surface quality and material properties of a single metal sheet in the cutting region according to predefined data specifications; S2: digital twin sheet model construction: after preprocessing and weighting of the geometric size, surface quality and material properties, the quality value of the metal sheet to be cut is calculated, the sheet quality distribution atlas is established, and the metal sheet is divided into quality regions based on the sheet quality distribution atlas, and a sample quality requirement database is established; S3: generating a nesting scheme: the sample set to be cut is retrieved from the sample quality requirement database, including the geometric size and minimum quality threshold requirement of each sample, and an improved genetic algorithm is used for nesting optimization, and thus an optimal nesting scheme is output; The optimal nesting scheme generation operation is as follows: All the sample plates of the current production task are retrieved from the sample plate quality requirement database according to the current production order, thus generating a set of sample plates to be cut, and the set of sample plates to be cut is sorted by area from large to small and marked as {S1, S2, …, S n}, wherein the information of each sample plate S i includes its geometric size and the minimum quality threshold requirement; An improved genetic algorithm is used for nesting scheme optimization, and after the algorithm iteration is terminated, the final optimal solution set is output, marked as {solution1, solution2, …, solutionn}, and arranged in descending order according to the comprehensive fitness of the nesting scheme, and thus the first executable nesting scheme is selected as the optimal nesting scheme, wherein the optimal nesting scheme includes the specific position of each sample on the sheet, the corresponding quality region, the spacing between the samples, and the material utilization rate and quality compliance rate of the scheme; The nesting scheme optimization specifically includes: Extracting the sample arrangement order in the sample set to be cut, and determining the rotation angle of each sample relative to its discharge reference point; S i , according to its minimum quality threshold requirement, its placeable region is limited within all corresponding quality regions in the sheet quality distribution map, and the optimal orientation of the position is found according to the rotation angle; After the sample is discharged, the material utilization rate and the quality matching rate of the sheet are calculated, respectively, wherein the material utilization rate is obtained by comparing the total area of all discharged samples with the original total area of the sheet, and the quality matching rate is obtained by comparing the number of quality-matched samples with the total number of samples; Assigning weight coefficients to the material utilization rate and the quality matching rate, respectively, marked as α and β, and dynamically adjusting the weight coefficients according to the production task priority, and then adding the weights of the material utilization rate and the quality matching rate to calculate the comprehensive fitness of the nesting scheme; Using the crossover mutation strategy, the iteration is stopped when the number of iterations reaches the preset maximum value, and the optimal solution set is output; S4: iterative cutting: based on the optimal nesting scheme, the cutting equipment is issued, and the cutting parameters are collected in real time, the remaining sheet is rescanned after cutting, and the digital twin sheet model is updated, the remaining value index is calculated, and the nesting termination judgment is made; S5: comprehensive quality evaluation: after cutting, the metal sheet is subjected to quality detection, including size accuracy and cut quality, and thus the comprehensive quality coefficient of the cut metal sheet is calculated; S6: adaptive optimization: based on the digital twin sheet model and the comprehensive quality coefficient of the cut metal sheet, the comprehensive performance evaluation is performed, and then the sheet state is taken as the state space, the nesting scheme is taken as the action space, and the comprehensive performance evaluation index is taken as the reward, the nesting scheme is updated, and adaptive optimization is performed.

2. The intelligent control method based on digital twinning according to claim 1, characterized in that: The geometric size of the single metal sheet in the cutting region is collected by a size measurement terminal, including the length, width, thickness and flatness of the metal sheet; The surface quality is obtained by a machine vision terminal to obtain an optical image of the surface of the sheet, and the surface characteristics of the metal sheet, including surface roughness and defect degree, are analyzed by an image processing algorithm; The material characteristics include the hardness and material composition deviation of the metal sheet.

3. The intelligent control method based on digital twinning according to claim 1, characterized in that: The sheet quality distribution map is constructed as follows: The flatness of each sampling point after pretreatment is extracted, and the flatness of each sampling point is compared with the standard flatness to screen the qualified sampling points, thereby extracting the number of qualified sampling points, and comparing the number of qualified sampling points with the total number of sampling points to obtain the geometric size compliance rate; The surface quality of each sampling point is extracted, and the average value of the surface roughness and defect degree after pretreatment is calculated, thereby obtaining the surface quality score. Similarly, the material characteristics of each sampling point are extracted, thereby obtaining the material characteristic score; The quality value of each sampling point is calculated based on the weighted sum of the geometric size compliance rate, the surface quality score and the material characteristic score, and is marked as Qc; Based on the quality values of all sampling points on the whole plate surface, a sheet quality distribution map corresponding to the position of the metal sheet is generated.

4. The intelligent control method based on digital twinning according to claim 1, characterized in that: The residual value index is calculated as follows: Immediately after cutting according to the optimal nesting scheme, each terminal of S1 scans the remaining sheet, thereby updating the digital twin plate model; The quality value of the remaining sheet is recalculated, and if the quality value of the remaining sheet decreases due to cutting, the quality area is adjusted; The residual area is calculated according to the geometric size of the remaining sheet, the residual average quality value is calculated according to the quality value of each sampling point, and the residual area is multiplied by the residual average quality value to obtain the residual value index; A value threshold is set, and the residual value index is compared with the value threshold. If the residual value index is less than the value threshold, the nesting is stopped. If the residual value index is greater than the value threshold, the nesting is performed again.

5. The intelligent control method based on digital twinning according to claim 1, characterized in that: The comprehensive quality evaluation is as follows: after cutting, the dimensional accuracy and cut quality of each metal sheet are detected, wherein the dimensional accuracy is calculated according to the absolute difference between the area of the cut metal sheet and the area of the design template, and compared with the accuracy tolerance to obtain the cut quality, the cut quality is detected by a laser profiler to detect the cut burr height and the cut perpendicularity, and compared with the allowed height and the allowed perpendicularity respectively to obtain the cut quality; according to the influence degree of the dimensional accuracy and the cut quality, the comprehensive quality coefficient of the cut metal sheet is evaluated.

6. The intelligent control method based on digital twinning according to claim 1, characterized in that: The adaptive optimization is based on the reinforcement learning algorithm, which obtains the current sheet state, including the digital twin plate model of the current sheet, and obtains the action space according to the current nesting scheme, takes the comprehensive performance evaluation index as the reward, thereby observes the state of the next batch of metal sheets, and updates the nesting scheme, which is specifically represented as: ， Wherein π(s) represents the optimal nesting scheme searched in the state space, s represents the state space, a represents the action space, and E represents the reward. According to the above formula, the maximum of all comprehensive performance evaluation indexes obtained in the future is searched among multiple nesting schemes under the digital twin plate model of the current sheet. The adaptive optimization is based on the reinforcement learning algorithm, which obtains the current sheet state, including the digital twin plate model of the current sheet, and obtains the action space according to the current nesting scheme, takes the comprehensive performance evaluation index as the reward, thereby observes the state of the next batch of metal sheets, and updates the nesting scheme, which is specifically represented as: Wherein π(s) represents the optimal nesting scheme searched in the state space, s represents the state space, a represents the action space, and E represents the reward. According to the above formula, the maximum of all comprehensive performance evaluation indexes obtained in the future is searched among multiple nesting schemes under the digital twin plate model of the current sheet.

Images