Artificial intelligence driven foam special-shaped die cutting optimization system

Abstract

The invention relates to the field of intelligent production, and provides an artificial intelligence-driven foam special-shaped die cutting optimization system, which comprises the following steps of: firstly, extracting a set of to-be-processed part drawings from obtained foam special-shaped die cutting order information; performing part die-cutting planning dynamic analysis on the set of the to-be-processed part drawings and the size information of the foam material based on an artificial intelligence data analysis algorithm to generate a foam special-shaped die-cutting typesetting scheme; and finally cutting the foam material based on the generated foam special-shaped die-cutting typesetting scheme. In this way, the reasonability and efficiency of foam special-shaped die cutting typesetting scheme construction can be effectively improved, and maximization of the material utilization rate can be achieved while the influence on delivery time efficiency is reduced.

Description

Technical Field

[0001] The present application relates to the field of intelligent manufacturing, and more particularly, to an artificial intelligence-driven optimization system for die-cutting of special-shaped foam materials. Background Art

[0002] In precision industrial fields such as electronic products, automotive manufacturing, and medical devices, foam materials, as important buffering, sealing, and insulating components, play a crucial role in the precision die-cutting of special-shaped parts. With the accelerating product iteration speed and the growing personalized demand, the geometric complexity of special-shaped foam parts has increased significantly, involving features such as multi-curvature profiles, nested holes, and asymmetric topological structures, which pose higher requirements for die-cutting layout technology.

[0003] In the traditional production process, layout engineers need to manually plan the arrangement of parts within the limited space of the sheet based on two-dimensional drawings. This process highly relies on manual experience for spatial fitting and waste material estimation. Especially when facing large-volume orders containing dozens of special-shaped parts, operators need to repeatedly perform graphic translation, rotation, and gap adjustment, often taking several hours or even days to complete the basic layout plan. This manual-dominated operation mode has inherent limitations: First, there are visual errors in the human eye's judgment of the spatial fit of complex geometric boundaries, resulting in unexpected waste of material gaps during actual cutting; Second, the experience-driven trial-and-error layout is difficult to exhaust all possible combination methods, especially when dealing with special-shaped parts with complementary profiles, it is easy to miss the optimal nesting plan; Third, manual adjustment is difficult to respond in real time to the dynamic changes of order parameters. When customers temporarily modify part sizes or increase or decrease quantities, the original layout plan often needs to be completely redone, seriously affecting the delivery timeliness.

[0004] Therefore, an artificial intelligence-driven optimization solution for die-cutting of special-shaped foam materials is expected. Summary of the Invention

[0005] To solve the above technical problems, the present application is proposed. Embodiments of the present application provide an artificial intelligence-driven optimization system for die-cutting of special-shaped foam materials.

[0006] According to one aspect of the present application, there is provided an artificial intelligence-driven optimization system for die-cutting of special-shaped foam materials, which includes:

[0007] A part drawing extraction module, configured to obtain die-cutting order information of special-shaped foam materials and extract a set of part drawings to be processed from the die-cutting order information of special-shaped foam materials:

[0008] A dimension information acquisition module, configured to obtain the dimension information of the foam material;

[0009] The part die-cutting dynamic analysis module is used to perform dynamic analysis on the collection of the drawings of the parts to be processed and the dimensional information of the foam material to obtain a foam special-shaped die-cutting layout plan. Among them, the part die-cutting dynamic analysis module includes: a dynamic response encoding unit, which is used to perform image-text-based decision state perception dynamic response encoding on the collection of the drawings of the parts to be processed and the dimensional information of the foam material to obtain part die-cutting planning dynamic response encoding features; a layout plan generation unit, which is used to obtain the foam special-shaped die-cutting layout plan based on the part die-cutting planning dynamic response encoding features and the dimensional information.

[0010] A cutting module, which is used to cut the foam material based on the foam special-shaped die-cutting layout plan.

[0011] Compared with the prior art, the artificial intelligence-driven foam special-shaped die-cutting optimization system provided by the present application first extracts the collection of the drawings of the parts to be processed from the obtained foam special-shaped die-cutting order information, and then performs dynamic analysis on the collection of the drawings of the parts to be processed and the dimensional information of the foam material based on the data analysis algorithm of artificial intelligence to generate a foam special-shaped die-cutting layout plan, and finally cuts the foam material based on the generated foam special-shaped die-cutting layout plan. In this way, the rationality and efficiency of the construction of the foam special-shaped die-cutting layout plan can be effectively improved, and the maximization of material utilization rate can be achieved while reducing the impact on the delivery time limit. Description of the Drawings

[0012] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0013] Figure 1 It is a system block diagram of an artificial intelligence-driven foam special-shaped die-cutting optimization system according to an embodiment of the present application.

[0014] Figure 2 It is a block diagram of a part die-cutting dynamic analysis module in an artificial intelligence-driven foam special-shaped die-cutting optimization system according to an embodiment of the present application.

[0015] Figure 3 It is a block diagram of a dynamic response encoding unit in an artificial intelligence-driven foam special-shaped die-cutting optimization system according to an embodiment of the present application.

[0016] Figure 4It is a block diagram of a part - material size dynamic response sub - unit in an artificial - intelligence - driven optimized system for die - cutting special - shaped foam according to an embodiment of the present application. Detailed implementation manners

[0017] Hereinafter, exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0018] In the field of precision industry, such as electronic products, automobile manufacturing, and medical equipment, foam materials are widely used due to their buffering, sealing, and insulating properties. The precision die - cutting of special - shaped parts is crucial. With the accelerating product iteration and increasing personalized requirements, the geometric complexity of special - shaped parts has been significantly improved, often including features such as multi - curvature profiles, nested holes, and asymmetric structures, which pose higher requirements for die - cutting layout technology.

[0019] Traditional layout methods rely on engineers to manually plan the part layout based on two - dimensional drawings and complete the nesting by operations such as translation, rotation, and gap adjustment. This method highly depends on manual experience. Especially when dealing with large - batch orders containing multiple special - shaped parts, it is time - consuming and inefficient. Its limitations are mainly reflected in three aspects: First, there are errors in the human eye's judgment of complex geometric boundaries, which easily lead to material waste; second, the trial - and - error nesting method is difficult to exhaust all combinations, especially when dealing with complementary profiles, often missing the optimal solution; third, manual adjustment is difficult to quickly respond to order changes. When the part size or quantity changes, the layout plan often needs to be redesigned, seriously affecting production efficiency and delivery timeliness.

[0020] To address the above technical problems, the technical concept of this application is to construct an end-to-end optimized system for special-shaped die-cutting of foam through artificial intelligence technology. Its core processing flow is divided into three stages: geometric feature analysis, dynamic programming decision-making, and intelligent layout generation. First, the system extracts the drawings of special-shaped parts from the order and performs high-precision analysis on the geometric features of complex contours, capturing irregular shape characteristics such as multi-scale curvature and hole nesting relationships that are difficult to characterize by traditional algorithms, and generating quantifiable shape features of the parts to be processed. Subsequently, combined with the semantic coding information of the foam board size, by dynamically analyzing and simulating the decision-making logic of human layout, a geometric complementarity relationship between parts is established in the feature space, and the correlation mapping between the part arrangement priority and the board size is dynamically evaluated based on the non-linear spectral aggregation mechanism. Finally, the abstract features are mapped to a specific layout plan in the physical space to maximize the board utilization rate while meeting the cutting process gap constraints. This solution replaces manual empirical trial and error with automated feature learning and intelligent decision-making, can generate a material arrangement plan close to the theoretical limit within a few minutes, significantly improve the nesting density of special-shaped foam parts, and reduce the waste rate; at the same time, the system has dynamic response capabilities, can quickly adapt to order parameter changes or new part types, and reduce production delays caused by repeated manual adjustments.

[0021] Figure 1 FIG. is a system block diagram of an artificial intelligence-driven optimized system for special-shaped die-cutting of foam according to an embodiment of the present application. As Figure 1 shown, in the artificial intelligence-driven optimized system 100 for special-shaped die-cutting of foam, it includes: a part drawing extraction module 110 for obtaining the information of the special-shaped die-cutting order of foam and extracting a set of drawings of parts to be processed from the information of the special-shaped die-cutting order of foam; a size information acquisition module 120 for obtaining the size information of the foam material; a part die-cutting dynamic analysis module 130 for performing dynamic analysis of part die-cutting planning on the set of drawings of parts to be processed and the size information of the foam material to obtain a special-shaped die-cutting layout plan of the foam; and a cutting module 140 for cutting the foam material based on the special-shaped die-cutting layout plan.

[0022] In the embodiment of the present application, the part drawing extraction module 110 is configured to obtain the information of the foam special-shaped die-cutting order and extract the set of part drawings to be processed from the information of the foam special-shaped die-cutting order. It should be understood that the order information is the starting basis for the entire die-cutting processing flow. The foam special-shaped die-cutting order contains various requirements of the customer for the parts, such as important information like the quantity, size specifications, precision requirements, delivery time, etc. of the parts. Only by obtaining the order information can the goals and scopes of the subsequent work be clarified. The part drawings to be processed are the key materials for specifically guiding the die-cutting processing, and they detail the geometric features such as the shape and size of the parts. Extracting the set of part drawings to be processed from the order information is because these drawings are the basic data sources for subsequent geometric feature analysis, layout, etc. Without accurate part drawings, it is impossible to effectively analyze and process the parts, and it is even more impossible to achieve a reasonable die-cutting layout. In this way, after obtaining the set of part drawings to be processed, it is possible to perform high-precision analysis on the geometric features of the complex contours therein, capture irregular shape characteristics such as multi-scale curvatures and hole nesting relationships, and then generate quantifiable part shape characteristics to be processed, preparing for intelligent layout.

[0023] The following is a detailed elaboration of a specific implementation process for "obtaining the information of the foam special-shaped die-cutting order and extracting the set of part drawings to be processed from the information of the foam special-shaped die-cutting order":

[0024] Obtaining the information of the foam special-shaped die-cutting order is the starting point of the entire implementation process. In the actual production scenario, the sources of order information are diverse. Enterprises usually establish special order receiving channels, such as docking with the customer's information system and realizing the automatic transmission of order information through electronic data interchange (EDI) technology. At the same time, some customers also place orders through traditional methods such as mail and fax, and this part of the order information also needs to be entered into the internal order management system in a timely manner. The order information is rich in content and contains many key elements, such as the quantity, size specifications, precision requirements, delivery time, etc. of the parts. These information are crucial for the formulation and execution of the production plan. Among them, the part drawings are the core materials, which detail the geometric features such as the shape and size of the parts and are the key basis for subsequent processing.

[0025] Extracting a set of drawings of parts to be processed from the orders of foam profile die-cutting is a complex process. First, professional information parsing tools are used to preliminarily process the order information. These tools can identify various data formats in the order, whether it is a structured data table or the text information contained in the document, and can accurately extract it. For example, using optical character recognition (OCR) technology, the text information in scanned paper orders or electronic documents can be converted into editable text for further analysis. For the part drawings contained in the order, corresponding reading methods will be adopted according to their formats. If it is a common two-dimensional drawing format, such as DWG, DXF, etc., the powerful functions of computer-aided design (CAD) software will be used for reading. CAD software can accurately analyze the graphic elements in the drawing, including lines, curves, polygons, etc., and at the same time can obtain key information such as dimension markings and tolerance requirements. For three-dimensional drawings, such as STEP, IGES and other formats, special three-dimensional modeling software or conversion tools are needed to convert them into a format convenient for subsequent processing, and important data such as the geometric shape, size and assembly relationship of the parts will be retained during the conversion process.

[0026] During the process of extracting the drawings, it is also necessary to screen and verify the drawings. Since the order information may contain various different types of files, and not all files are the drawings of parts to be processed, it is necessary to accurately screen out the drawings related to this order from numerous files according to the key information in the order, such as part numbers, names, etc. The screened drawings cannot all be directly used for production and still need to be strictly verified. The verification content includes checking the integrity of the drawing to ensure that the drawing contains a complete description of the part shape and there are no missing parts; at the same time, the accuracy of the dimension markings should be checked to see if the dimension values are clear, accurate, and there are no contradictions or omissions. In addition, the technical requirements in the drawing, such as surface roughness, material requirements, etc., also need to be carefully reviewed to ensure compliance with the production process and customer requirements. If problems are found in the drawing during the verification process, it is necessary to communicate with the customer in time to obtain accurate drawing information to avoid production errors caused by drawing errors, resulting in waste of resources and delivery delays.

[0027] After the drawings are screened and verified, format conversion and storage of the drawings are also required. Since different production processes and software systems may have different requirements for drawing formats, for the convenience of subsequent processing, the drawings are usually converted into a unified format suitable for system processing. For example, converting 2D drawings into a vector graphics format (SVG), which has the advantages of small file size, scalability, and no distortion, and is convenient for viewing and editing on different software and devices. For 3D drawings, they can be converted into lightweight formats suitable for subsequent feature extraction, such as JT, 3MF, etc. These formats reduce the data volume while retaining the key geometric information of the parts, improving the processing efficiency. The converted drawings will be stored in the enterprise's production database. By establishing an association with the order information, it is convenient for subsequent production processes to quickly call. During the storage process, the drawings will also be classified and managed, classified according to order numbers, part types, etc., for easy searching and use.

[0028] In the embodiment of the present application, the dimension information acquisition module 120 is used to acquire the dimension information of the foam material. Correspondingly, considering that the dimension information of the foam material is an important constraint condition for die-cutting layout. To reasonably arrange the parts to be processed on a limited foam board, it is necessary to know the specific dimensions of the foam material, such as length, width, etc. Without obtaining this information, the layout work will lack clear boundaries and scope, and it will be impossible to accurately plan the placement positions of the parts, resulting in the inability to implement the layout plan. Moreover, the dimensions of the foam material are closely related to the dimensions and quantities of the parts to be processed. Different-sized foam materials have a direct impact on the number of parts that can be accommodated and the arrangement methods. Only by obtaining the dimension information of the foam material can we comprehensively consider how to achieve the optimal part arrangement on the foam board according to the set of drawings of the parts to be processed, and avoid situations such as material waste or inability to meet production requirements caused by mismatched material dimensions.

[0029] In the embodiment of the present application, the part die-cutting dynamic analysis module 130 is used to perform dynamic analysis on the die-cutting planning of parts for the set of drawings of the parts to be processed and the dimension information of the foam material to obtain a foam special-shaped die-cutting layout plan. Specifically, Figure 2 It is a block diagram of the part die-cutting dynamic analysis module in the artificial intelligence-driven foam special-shaped die-cutting optimization system according to the embodiment of the present application. As Figure 2 shown, the part die-cutting dynamic analysis module 130 includes: a dynamic response encoding unit 131, which is used to perform image-text-based decision state perception dynamic response encoding on the set of drawings of the parts to be processed and the dimension information of the foam material to obtain part die-cutting planning dynamic response encoding features; a layout plan generation unit 132, which is used to obtain the foam special-shaped die-cutting layout plan based on the part die-cutting planning dynamic response encoding features and the dimension information.

[0030] In the embodiment of the present application, the dynamic response encoding unit 131 is configured to perform image-text based decision state perception dynamic response encoding on the set of drawings of the parts to be processed and the size information of the foam material to obtain the dynamic response encoding features of the part die-cutting plan. Specifically, Figure 3 It is a block diagram of the dynamic response encoding unit in the artificial intelligence-driven foam special-shaped die-cutting optimization system according to the embodiment of the present application. As Figure 3 shown, the dynamic response encoding unit 131 includes: a processed part shape feature extraction sub-unit 1311, configured to extract shape feature information from each drawing of the parts to be processed in the set of drawings of the parts to be processed to obtain a set of shape feature vectors of the parts to be processed; a material size semantic encoding sub-unit 1312, configured to perform semantic encoding on the size information of the foam material to obtain a foam material size semantic encoding vector; a part to be processed - material size dynamic response sub-unit 1313, configured to perform decision state perception-based part die-cutting plan dynamic response encoding on the set of shape feature vectors of the parts to be processed and the foam material size semantic encoding vector to obtain a part die-cutting plan dynamic response encoding vector as the dynamic response encoding features of the part die-cutting plan.

[0031] In an embodiment of the present application, the processing part shape feature extraction subunit 1311 is used to extract shape feature information from each drawing of the parts to be processed in the set of drawings of the parts to be processed to obtain a set of shape feature vectors of the parts to be processed. Specifically, in an embodiment of the present application, the processing part shape feature extraction subunit is used to: use a capsule network-based shape feature extractor to extract shape feature information from each drawing of the parts to be processed in the set of drawings of the parts to be processed to obtain a set of shape feature vectors of the parts to be processed. Accordingly, considering that the geometric features of special-shaped parts often have complex morphological characteristics such as multi-scale curvature changes, asymmetric topological structures, and nested holes, traditional feature extraction methods based on artificial experience or rule algorithms are difficult to accurately quantify their spatial relationships. For example, for foam parts with wavy edges or internal hollow structures, the determination of their effective contours requires not only the identification of the curvature changes of the outer boundaries, but also the analysis of the relative position relationship between the holes and the main structure. When dealing with such tasks, traditional convolutional neural networks (CNNs) are prone to losing local geometric details due to pooling operations, and cannot effectively characterize the equivalent features after the parts are rotated and translated, which makes it difficult for subsequent typesetting algorithms to explore the geometric complementarity between parts, seriously affecting the optimization potential of nesting density. Based on this, in the technical solution of the present application, a shape feature extractor based on a capsule network is used to extract shape feature information from each drawing of the parts to be processed in the set of drawings of the parts to be processed to obtain a set of shape feature vectors of the parts to be processed. It is worth noting that the shape feature extractor based on a capsule network is essentially to simulate the hierarchical cognitive process of humans on geometric shapes through the principle of bionics. The capsule network establishes spatial associations between neuron capsules through a dynamic routing mechanism, and uses the posture matrix to encode the local coordinate system and global position relationship of the parts, which can decouple key geometric elements such as curvature gradients, hole boundaries, and symmetry axes from the original drawings. For example, for special-shaped foam parts with a spiral profile, the capsule network can capture multi-level features from microscopic curvature fragments to macroscopic spiral directions layer by layer, and retain the directionality and relative proportional relationship of each feature component through vectorized output. This feature expression method breaks through the limitations of traditional contour point cloud or rasterized image processing, allowing the geometric essence of irregular shapes to be mapped to high-dimensional vector space with high fidelity.

[0032] In the embodiment of the present application, the material size semantic encoding subunit 1312 is used to perform semantic encoding on the size information of the foam material to obtain a foam material size semantic encoding vector. Specifically, in the embodiment of the present application, the material size semantic encoding subunit includes: using a semantic encoder based on text convolution to perform semantic encoding on the size information of the foam material to obtain the foam material size semantic encoding vector. Correspondingly, considering that the size information of the foam board is not a simple set of length and width numerical values, but a composite data carrier that implies material usage boundaries, cutting process constraints, and layout space characteristics. In traditional layout systems, size parameters are usually directly input into the algorithm in discrete numerical form, which cannot effectively represent engineering semantic information such as the edge gap tolerance of the board, the stress distribution sensitive area, and the cutting tool path limit. For example, when there are local thickness non-uniformities or edge reinforcement requirements in the board, the simple geometric size data cannot convey the actual impact of these physical characteristics on the layout, resulting in the arrangement scheme generated by the subsequent algorithm may violate the actual production constraints and require manual secondary correction. This disconnection between low-dimensional numerical expressions and engineering semantics severely restricts the adaptability of the intelligent layout system to material characteristics. Based on this, the present application performs semantic encoding on the size information of the foam material to obtain a foam material size semantic encoding vector. In particular, in a specific example of the present application, a semantic encoder based on text convolution can be used to perform semantic encoding on the size information of the foam material to obtain the foam material size semantic encoding vector. The foam material size semantic encoding vector obtained by the semantic encoder based on text convolution contains richer semantic information about the foam material size. These semantic information are not only the numerical size of the size, but also include potential semantics such as the relationship between sizes and the impact of sizes on processing technology. Specifically, the encoder adopts parallel operations of multi-scale convolution kernels to mine local correlations for the structured features of size parameters (such as length-width ratio, thickness grading, tolerance range, etc.), and captures the non-linear interaction relationships between parameters in different dimensions through the feature extraction ability of the convolution layer. For example, for a composite size description labeled as "1200mm × 800mm ± 5mm / thickness 2.5mm (elastic zone occupancy 30%)", the convolution kernel automatically learns the co-constraint relationship between the length-width fluctuation range and the thickness parameter within the sliding window, and simultaneously identifies the restriction rules of the elastic zone occupancy on the edge arrangement density, and finally generates a high-dimensional semantic vector that integrates material physical characteristics and process constraints. This processing method elevates the board size from a simple geometric description to a feature expression that includes engineering semantics such as the weight distribution of the cuttable area and the material deformation tolerance.

[0033] In the embodiment of the present application, the to-be-processed part - material size dynamic response subunit 1313 is configured to perform decision state perception-based part die-cutting planning dynamic response encoding on the set of shape feature vectors of the to-be-processed part and the semantic encoding vector of the foam material size to obtain a part die-cutting planning dynamic response encoding vector as the part die-cutting planning dynamic response encoding feature. Specifically, Figure 4 is a block diagram of the to-be-processed part - material size dynamic response subunit in the artificial intelligence-driven foam special-shaped die-cutting optimization system according to the embodiment of the present application. As Figure 4 shown, the to-be-processed part - material size dynamic response subunit 1313 includes: a part die-cutting planning decision-related feature response secondary subunit 1313-1, configured to perform non-linear decision response on each shape feature vector of the to-be-processed part in the set of the semantic encoding vector of the foam material size and the set of shape feature vectors of the to-be-processed part to obtain a set of implicit encoding vectors of the dynamic attributes of the part die-cutting planning decision point nodes; a part die-cutting planning core feature extraction secondary subunit 1313-2, configured to perform core component feature extraction on the set of implicit encoding vectors of the dynamic attributes of the part die-cutting planning decision point nodes to obtain a set of core component encoding vectors of the part die-cutting planning decision point; a part die-cutting planning core feature fusion secondary subunit 1313-3, configured to perform adaptive fusion on the set of core component encoding vectors of the part die-cutting planning decision point to obtain the part die-cutting planning dynamic response encoding vector.

[0034] It should be understood that there is a complex non-linear correlation between the geometric features of the shaped parts and the dimensional constraints of the sheet material. Traditional nesting algorithms based on rule engines usually adopt a priori nesting priority strategies (such as arranging parts in descending order of area). However, when faced with complex shaped parts such as multi-curved profiles and nested holes, static rules are difficult to dynamically balance multi-objective conflicts such as geometric complementarity between parts, sheet edge avoidance requirements, and cutting process gaps. For example, when two parts with concave profiles have the potential for mirror-symmetrical nesting, manual experience may overlook their combined optimization opportunities due to visual limitations, while traditional algorithms are limited by fixed decision-making logic and cannot dynamically explore such implicit associations in the feature space. Therefore, in the technical solution of this application, a dynamic response encoding of the part die-cutting plan for decision state perception is performed on the set of shape feature vectors of the parts to be processed and the semantic encoding vector of the foam material dimensions to obtain a dynamic response encoding vector of the part die-cutting plan. Specifically, this method realizes the deep interaction of multi-modal features and the transmission of engineering semantics by constructing an implicit state space. First, a non-linear decision response is performed on the semantic encoding vector of the foam material dimensions (representing the spatial semantics of the available area of the sheet) and each part shape feature vector (encoding geometric topological characteristics), and an implicit encoding vector of the dynamic attributes of the decision point nodes is generated through feature interaction. These encoding vectors reconstruct the relationship between the parts and the sheet in the hidden space. For example, a potential matching relationship is established between the concave curvature feature of a certain part and the convex shape of the remaining space at the edge of the sheet. Subsequently, by calculating the class neighborhood matrix and degree matrix of the dynamic attributes of the decision point nodes, the local structural relationship between the decision points is explicitly modeled: the neighborhood matrix captures the geometric complementarity between parts (such as the affinity of concave-convex profiles), and the degree matrix quantifies the influence weight of each decision point on the global layout target (such as core parts need to be laid out first). The spectral decomposition based on the Laplacian matrix further reveals the low-dimensional structure of the data manifold, mapping the discrete decision points in the high-dimensional feature space to the continuous core component encoding vectors in the spectral domain. For example, a cluster of part rotation angle features adapted to the long axis direction of the sheet is separated. Finally, these spectral domain features are dynamically integrated through an adaptive fusion mechanism to generate a dynamic query response encoding vector that can reflect both the global optimization target and be compatible with local engineering constraints. For example, the dense arrangement tendency of small-sized parts is strengthened in the edge area of the sheet, while the layout flexibility of large-sized shaped parts is retained in the central area.

[0035] Specifically, in the embodiment of this application, the part die-cutting plan decision-related feature response secondary subunit 1313-1 is used to perform a non-linear decision response on each part shape feature vector in the set of the semantic encoding vector of the foam material dimensions and the part shape feature vectors of the parts to be processed to obtain a set of implicit encoding vectors of the dynamic attributes of the part die-cutting plan decision point nodes, which can be expressed by the following non-linear response formula:

[0036] V2 = {v 21 ,v22 ,..., v 2i ,..., v 2n}

[0037]

[0038] V 1,2 = {v 1,21 , v 1,22 ,..., v 1,2i ,..., v 1,2n}

[0039] Among them, V2 is the set of shape feature vectors of the parts to be processed, and v 21 , v 22 , v 2i and v 2n are the 1st, 2nd, ith, and nth shape feature vectors of the parts to be processed in the set of shape feature vectors of the parts to be processed respectively. V1 is the semantic coding vector of the size of the foam material, is matrix multiplication, W i and b i are the decision response weight matrix and the decision response bias vector corresponding to v 2i respectively. sigmoid is the activation function, and v 1,21 , v 1,22 , v 1,2i and v 1,2n are the 1st, 2nd, ith, and nth implicit coding vectors of the dynamic attributes of the part die-cutting planning decision point nodes in the set of implicit coding vectors of the dynamic attributes of the part die-cutting planning decision point nodes respectively. V 1,2 is the set of implicit coding vectors of the dynamic attributes of the part die-cutting planning decision point nodes.

[0040] It should be understood that there is a complex non-linear association between the semantic coding vector of the foam material size and the shape feature vector of the part to be processed, and it is difficult for traditional methods to dynamically capture these relationships. Through non-linear decision response processing, the spatial semantics of the foam material can be deeply interacted with the geometric topological characteristics of the part to generate a set of implicit coding vectors of the dynamic attributes of the part die-cutting planning decision point nodes. These coding vectors can reconstruct the association relationship between the part and the plate in the hidden space. For example, a potential matching association can be established between the concave curvature feature of the part and the convex shape of the remaining space at the edge of the plate, thus providing a richer feature representation for subsequent analysis. Generally speaking, through non-linear mapping and feature interaction, the generated implicit coding vectors can effectively separate the originally indistinguishable features, simplify the complex decision boundary, and lay a reliable analysis data foundation for dynamic layout optimization.

[0041] Specifically, in the embodiment of the present application, the secondary subunit 1313-2 for extracting the core features of part die-cutting planning includes: a tertiary subunit for calculating the dynamic attribute class neighborhood matrix, which is used to calculate the dynamic attribute class neighborhood matrix of the part die-cutting planning decision point nodes based on the set of implicit coding vectors of the dynamic attributes of the part die-cutting planning decision point nodes, and can be expressed by the following formula:

[0042]

[0043]

[0044] where v 1,2j is the j-th implicit coding vector of the dynamic attributes of the part die-cutting planning decision point nodes in the set of implicit coding vectors of the dynamic attributes of the part die-cutting planning decision point nodes, ‖·‖ 2 is the Euclidean norm of the calculation vector, arccosh is the inverse hyperbolic cosine function, A 11 , A n1 , A 1n , A ij and A nn are the eigenvalues at each position in the dynamic attribute class neighborhood matrix of the part die-cutting planning decision point nodes respectively, and A is the dynamic attribute class neighborhood matrix of the part die-cutting planning decision point nodes;

[0045] A tertiary subunit for calculating the dynamic attribute degree matrix, which is used to calculate the dynamic attribute degree matrix of the part die-cutting planning decision point nodes based on the set of implicit coding vectors of the dynamic attributes of the part die-cutting planning decision point nodes, and can be expressed by the following formula:

[0046]

[0047] where is the square of the calculation vector one-norm, n is the number of vectors in V 1,2 minus one, D1, D i and D n are the eigenvalues at each position on the diagonal in the dynamic attribute degree matrix of the part die-cutting planning decision point nodes respectively, and D is the dynamic attribute degree matrix of the part die-cutting planning decision point nodes;

[0048] A tertiary subunit for calculating the dynamic attribute Laplacian matrix, which is used to calculate the dynamic attribute Laplacian matrix of the part die-cutting planning decision point nodes based on the dynamic attribute class neighborhood matrix of the part die-cutting planning decision point nodes and the dynamic attribute degree matrix of the part die-cutting planning decision point nodes, and can be expressed by the following formula:

[0049] L = D - A

[0050] where L is the dynamic attribute Laplacian matrix of the part die-cutting planning decision point nodes;

[0051] The matrix spectral decomposition three - level subunit is used to perform spectral decomposition on the Laplacian matrix of the dynamic attributes of the part die - cutting planning decision - point nodes to obtain a set of core component encoding vectors of the part die - cutting planning decision - point nodes, which can be expressed by the following formula:

[0052]

[0053] Among them, Spectral Decomposition(L) is the operation of performing spectral decomposition on L, U is the part die - cutting planning eigenvector matrix, that is, the set of core component encoding vectors of the part die - cutting planning decision - point nodes, x1, x2 and x k are respectively the 1st, 2nd, and k - th core component encoding vectors of the part die - cutting planning decision - point nodes in the set of core component encoding vectors of the part die - cutting planning decision - point nodes, diag(λ1, λ2, …, λ k ) is the part die - cutting planning eigenvalue diagonal matrix with elements λ1, λ2 and λ k on the diagonal of the matrix, λ1, λ2 and λ k are respectively the eigenvalues corresponding to x1, x2 and x k , and Λ is the part die - cutting planning eigenvalue diagonal matrix.

[0054] Correspondingly, in order to capture the geometric complementarity between parts (such as the affinity of concave - convex contours) and local structure relationships, it is necessary to calculate the part die - cutting planning decision - point node dynamic - attribute class neighborhood matrix based on the set of implicit encoding vectors of the part die - cutting planning decision - point nodes. This neighborhood matrix explicitly encodes the local connection relationships between data points into a graph structure, which can provide the internal structure of the data for subsequent spectral analysis. And the generated part die - cutting planning decision - point node dynamic - attribute class neighborhood matrix can quantify the state semantic correlation degree between parts, such as the affinity of two parts with the potential of mirror - symmetric nesting in the feature space, which can provide local structure information support for global layout optimization.

[0055] It should be understood that in die-cutting planning, the layout priority and global influence weight of parts need to be clearly quantified. By calculating the node dynamic attribute degree matrix of the die-cutting planning decision points of parts, the connection number of each die-cutting planning decision point node of the part can be recorded, reflecting the "importance" and "centrality" of the node in the graph structure. This quantification method enables the Laplacian matrix to be adaptively adjusted according to the connection strength of the die-cutting planning decision point nodes of parts, so as to more accurately reflect the internal structural characteristics of the data. Specifically, the node dynamic attribute degree matrix of the die-cutting planning decision points of parts records the connection strength of each die-cutting planning decision point node in the form of a diagonal matrix, enabling the implicit correlation degree of each node to be stacked and quantified. This expression of importance ensures that in the subsequent spectral decomposition process, the characteristics of the key die-cutting planning decision point nodes of parts can be preferentially retained, explicitly reflecting the global influence weight and local constraint balance of the part layout.

[0056] Correspondingly, the construction of the Laplacian matrix is the core link of spectral graph theory, and its role is to transform the graph structure information into an algebraic representation form. By calculating the node dynamic attribute Laplacian matrix of the die-cutting planning decision points of parts based on the node dynamic attribute neighborhood matrix and the node dynamic attribute degree matrix of the die-cutting planning decision points of parts, the local structure and global relationship between data points can be comprehensively expressed, providing rich spectral map information for the subsequent spectral decomposition. That is, the generated node dynamic attribute Laplacian matrix of the die-cutting planning decision points of parts captures the balance relationship between the geometric complementarity between parts and the global influence weight, providing an accurate algebraic basis for the subsequent spectral decomposition and ensuring that complex geometric relationships can be condensed and expressed in a low-dimensional space.

[0057] It should be understood that spectral decomposition is the key step to reveal the information encoded in the Laplacian matrix. By performing spectral decomposition on the node dynamic attribute Laplacian matrix of the die-cutting planning decision points of parts, the discrete decision points (such as the complex association between the geometric topology characteristics of parts and the semantic of the sheet size) in the high-dimensional feature space can be mapped to the continuous node dynamic attribute core component coding vectors of the die-cutting planning decision points of parts in the spectral domain. The key to this step is to transform the complex geometric features (such as the affinity of concave-convex contours) into low-dimensional representations through non-linear dimensionality reduction, so as to provide a condensed feature representation for dynamic query response. For example, the part rotation angle feature clusters adapted to the long axis direction of the sheet are separated, taking into account the cutting process gap and edge avoidance requirements while optimizing the nesting layout.

[0058] Preferably, in another example of the present application, the matrix spectral decomposition three-level subunit is configured to: perform matrix topology optimization based on a double subspace on the furniture material decision point state Laplacian matrix to obtain an optimized furniture material decision point state Laplacian matrix; perform spectral decomposition on the optimized furniture material decision point state Laplacian matrix to obtain a set of core component coding vectors of the furniture material decision point.

[0059] Specifically, the furniture material decision point state Laplacian matrix is defined as the union representation of all limit points in the topological space, which can capture the global correlation characteristics of the dynamic attributes of the part die-cutting planning decision point nodes. In order to improve its robustness and the expression ability of topological features, it is necessary to apply a topological completion method to optimize this matrix. Specifically, first, the connected region features of the part die-cutting planning decision point node dynamic attribute class neighborhood matrix and the part die-cutting planning decision point node dynamic attribute class degree matrix are extracted through pseudo-inverse operations, that is, let:

[0060]

[0061] Then the matrices M1 and M2 can respectively represent the partition subspace and the loop subspace of the furniture material decision point state Laplacian matrix, which can explicitly express the local connection relationship and node importance in the graph structure. Based on this, the furniture material decision point state Laplacian matrix is optimized through partition subspace completion and loop subspace completion to obtain an optimized furniture material decision point state Laplacian matrix, which can be expressed as:

[0062]

[0063] In this way, the topological conserved quantity in the structure information of the graph is extracted through the double completion operation of the partition subspace and the loop subspace, thereby significantly enhancing the topological invariance strength of the original furniture material decision point state Laplacian matrix in spectral analysis. Then, spectral decomposition is performed on the optimized furniture material decision point state Laplacian matrix to obtain a set of core component coding vectors of the furniture material decision point. Since the optimized furniture material decision point state Laplacian matrix can more explicitly capture the local geometric complementarity (such as the edge avoidance requirement of the board) and global constraints (such as maximizing space utilization) of the furniture material decision point state, it can make the core feature expression of the furniture material decision point state obtained by spectral decomposition more stable, reduce noise interference, and thus effectively improve the effect of spectral decomposition.

[0064] Specifically, in the embodiment of the present application, the part die-cutting planning core feature fusion two-level subunit 1313-3 is configured to adaptively fuse the set of core component coding vectors of the part die-cutting planning decision point to obtain the part die-cutting planning dynamic response coding vector, which can be expressed by the following adaptive fusion formula:

[0065]

[0066] Among them, x i is the i-th core component coding vector of the part die-cutting planning decision point in the set of core component coding vectors of the part die-cutting planning, AF(U) is the adaptive fusion operation on U, W 2i and b 2i are the corresponding fusion weight matrix and fusion bias vector respectively, softmax is the softmax function, is the corresponding part die-cutting planning scoring weight vector, a i is the corresponding part die-cutting planning feature weight value, mask is the masking operation, τ is the preset threshold, w ri is the corresponding part die-cutting planning feature mask weight value of x i and v f is the dynamic response coding vector of the part die-cutting planning.

[0067] It should be understood that the die-cutting planning of special-shaped parts needs to dynamically integrate various features to adapt to different layout requirements. By adaptively fusing the set of core component coding vectors of the part die-cutting planning decision point, the weights of each feature component can be dynamically adjusted, focusing on the most relevant features and suppressing noise. Specifically, the dynamic response coding vector of the part die-cutting planning generated by adaptive fusion can not only reflect the global optimization goal (such as maximizing the utilization rate of the sheet space), but also be compatible with local engineering constraints (such as the edge avoidance requirement). By dynamically allocating weights, it is ensured that the complementary information of the feature components is fully utilized, and thus a robust and discriminative dynamic response coding vector of the part die-cutting planning can be provided for complex die-cutting planning.

[0068] In an embodiment of the present application, the layout scheme generation unit 132 is used to obtain the foam shaped die-cutting layout scheme based on the dynamic response coding feature of the part die-cutting planning and the size information. Specifically, in an embodiment of the present application, the layout scheme generation unit is used to: input the dynamic response coding vector of the part die-cutting planning and the size information into the foam shaped die-cutting planner based on AIGC to obtain the foam shaped die-cutting layout scheme. In particular, the foam shaped die-cutting layout scheme includes the specific position and angle of each part on the foam sheet. It should be understood that the dynamic response coding vector of the part die-cutting planning has integrated the shape characteristics of the parts to be processed, the semantic information of the foam material size, and the relevant decision information obtained through dynamic analysis, and the size information further clarifies the specific specifications of the foam material. The foam shaped die-cutting planner based on AIGC (artificial intelligence generated content) needs these comprehensive and processed information as input to comprehensively consider various factors and generate a foam shaped die-cutting layout scheme that meets actual production needs. A single piece of information cannot cover all the key elements in the die-cutting process. Only by combining the two can the planner make more reasonable decisions. It is worth mentioning that AI GC technology has strong generation and learning capabilities, and can learn and reason based on large amounts of input data and patterns. The dynamic response coding vector of part die-cutting planning contains complex features and decision-making information. AI GC's foam special-shaped die-cutting planner can use its algorithms and models to conduct in-depth analysis and processing of this information, and explore the potential rules and optimal combinations. At the same time, combined with size information, the planner can generate a variety of possible layout schemes within the actual size range of the foam material, and select the best one from them, which is difficult to achieve with traditional methods.

[0069] The following is a detailed description of a specific implementation process of "inputting the dynamic response coding vector of the part die-cutting planning and the size information into the foam special-shaped die-cutting planner based on AI GC to obtain the foam special-shaped die-cutting layout plan":

[0070] First, for the dynamic response coding vector of the part die-cutting plan, due to its high-dimensional and complex characteristics, it needs to be standardized. This is to unify the scale of the data and avoid affecting the accurate analysis of the data by the subsequent planner due to the large difference in the magnitude of data in different dimensions. For example, some features may have large numerical values, while others have small numerical values. Without standardization, the planner may over-focus on the features with large magnitudes during processing, resulting in biased analysis results. For the dimension information, since its sources and formats may vary, it needs to be converted into a structured data format that the planner can recognize and process. For example, for dimension information in the form of text descriptions, it needs to be converted into a combination form containing numerical values and semantic labels through a specific parsing algorithm, so that the planner can understand the dimension values and the meaning of related process constraints therein.

[0071] Next is to build and train an AI GC-based foam special-shaped die-cutting planner. Currently, the Transformer architecture has become an ideal choice for building the planner due to its powerful ability to process sequence data and capture long-sequence dependencies. In the planner, the multi-head attention mechanism plays a key role. It enables the planner to simultaneously focus on different parts of the dynamic response coding vector of the part die-cutting plan and the dimension information, and accurately capture the complex correlation relationships between them. To enable the planner to learn effective layout patterns and rules, a large amount of historical foam die-cutting order data is required. These data cover various part drawings, foam material dimensions, and corresponding successful layout schemes. Before using these data, strict data cleaning and annotation work must be carried out. Data cleaning mainly removes noise in the data, such as handling missing values and outliers, to ensure the accuracy and integrity of the data. Annotation adds key information labels to the data so that the planner can clearly understand the meaning represented by different data during the learning process. After completing data cleaning and annotation, the data is divided into a training set, a validation set, and a test set. The training set is used to train the planner, and during the training process, hyperparameters such as the learning rate and the number of layers are continuously adjusted to optimize the model performance. At the same time, the validation set is used to evaluate the model during the training process, and the model is adjusted according to the evaluation results to prevent the model from overfitting or underfitting. Finally, a final performance test is carried out on the test set to ensure that the planner has good generalization ability and accuracy.

[0072] After the planner is trained, the preprocessed part die-cutting planning dynamic response encoding vector and dimension information can be input into it. The generation module inside the planner will start generating multiple possible die-cutting layout plans for the foam profile based on this input information, combined with the knowledge and patterns learned during the training process. Each plan will contain detailed key information such as the specific position and angle of the parts on the foam sheet. However, there may be some unreasonable situations in these generated plans, so preliminary screening is required. The screening process mainly checks whether the plan meets the basic production requirements, such as whether the parts exceed the boundaries of the foam sheet, whether there are obvious unreasonable layout situations such as overlap between parts, and these unreasonable plans are eliminated.

[0073] The layout plans after preliminary screening still need to be comprehensively evaluated. The evaluation starts from multiple important aspects. Among them, the material utilization rate is a key indicator, which is directly related to the production cost and is measured by calculating the ratio of the effective utilization area of the foam material to the total area in each plan. At the same time, the feasibility of the cutting process cannot be ignored. It is necessary to evaluate whether the layout of the parts in the plan is convenient for actual cutting operations, which involves factors such as the length of the cutting path and the number of tool changes. If the cutting path is too long or the tool changes are frequent, it will increase the production cost and production time. In addition, the production efficiency is also an important content of the evaluation, considering how to layout the parts to make the production process more efficient and reduce the production cycle. If the evaluation results show that the expected standards are not met, the evaluation information needs to be fed back to the planner. The planner will adjust the generation strategy according to these feedback information, regenerate and screen the layout plans, and so on in a cycle until the optimal layout plan that meets the requirements of high material utilization rate, feasible cutting process, high production efficiency, etc. is obtained.

[0074] In the embodiment of the present application, the cutting module 140 is used to cut the foam material based on the die-cutting layout plan for the foam profile. Specifically, in the embodiment of the present application, the cutting module is used to: convey the die-cutting layout plan for the foam profile to the die-cutting equipment control system to control the die-cutting equipment to perform cutting based on the die-cutting layout plan for the foam profile. Particularly, the die-cutting layout plan for the foam profile is the optimal plan obtained through comprehensive analysis and intelligent planning of various information such as the shape characteristics of the parts and the dimensions of the foam material. The die-cutting equipment control system needs to execute the cutting task according to the specific layout plan. Conveying the layout plan to the control system can achieve the automatic connection from plan planning to actual production operation, thereby producing foam profile parts that meet the design requirements, avoiding errors and low efficiency problems that may occur in manual information transmission, ensuring the smooth progress of the entire production process, and ensuring that parameters such as the shape and dimensions of the parts meet the order requirements, so as to provide high-quality components for subsequent product assembly and use.

[0075] In summary, the artificial intelligence-driven optimized system 100 for special-shaped die-cutting of foam is elucidated according to the embodiments of the present application. First, a set of drawings of parts to be processed is extracted from the obtained order information of special-shaped die-cutting of foam. Then, based on the data analysis algorithm of artificial intelligence, a dynamic analysis of the die-cutting plan for the parts is carried out on the set of drawings of parts to be processed and the size information of the foam material to generate a special-shaped die-cutting layout plan for the foam. Finally, the foam material is cut based on the generated special-shaped die-cutting layout plan for the foam. In this way, the rationality and efficiency of constructing the special-shaped die-cutting layout plan for the foam can be effectively improved, and the maximum utilization rate of the material can be achieved while reducing the impact on the delivery time limit.

Claims

1. An artificial intelligence-driven optimization system for special-shaped die-cutting of foam, characterized in that, Including: A part drawing extraction module, configured to obtain the information of the foam special-shaped die-cutting order and extract a set of part drawings to be processed from the information of the foam special-shaped die-cutting order: A size information acquisition module, configured to obtain the size information of the foam material; A part die-cutting dynamic analysis module, configured to perform dynamic analysis on the set of part drawings to be processed and the size information of the foam material for part die-cutting planning to obtain a foam special-shaped die-cutting layout plan. Among them, the part die-cutting dynamic analysis module includes: a dynamic response encoding unit, configured to perform image-text based decision state perception dynamic response encoding on the set of part drawings to be processed and the size information of the foam material to obtain part die-cutting planning dynamic response encoding features; a layout plan generation unit, configured to obtain the foam special-shaped die-cutting layout plan based on the part die-cutting planning dynamic response encoding features and the size information; A cutting module, configured to cut the foam material based on the foam special-shaped die-cutting layout plan.

2. The artificial intelligence-driven optimized system for special-shaped die-cutting of foam, as claimed in claim 1, wherein The dynamic response encoding unit includes: A processed part shape feature extraction sub-unit, configured to extract shape feature information from each processed part drawing in the set of processed part drawings to obtain a set of processed part shape feature vectors; A material size semantic encoding sub-unit, configured to perform semantic encoding on the size information of the foam material to obtain a foam material size semantic encoding vector; A processed part - material size dynamic response sub-unit, configured to perform decision state perception part die-cutting planning dynamic response encoding on the set of processed part shape feature vectors and the foam material size semantic encoding vector to obtain a part die-cutting planning dynamic response encoding vector as the part die-cutting planning dynamic response encoding feature.

3. The artificial intelligence-driven optimized system for special-shaped die-cutting of foam according to claim 2, wherein, The processed part shape feature extraction sub-unit is configured to: use a shape feature extractor based on a capsule network to extract shape feature information from each processed part drawing in the set of processed part drawings to obtain the set of processed part shape feature vectors.

4. The artificial intelligence-driven optimized system for special-shaped die-cutting of foam according to claim 3, wherein, The material size semantic encoding sub-unit includes: using a semantic encoder based on text convolution to perform semantic encoding on the size information of the foam material to obtain the foam material size semantic encoding vector.

5. The artificial intelligence-driven optimized system for special-shaped die-cutting of foam, according to claim 4, wherein The processed part - material size dynamic response sub-unit includes: A part die-cutting planning decision-related feature response secondary sub-unit, configured to perform non-linear decision response on the foam material size semantic encoding vector and each processed part shape feature vector in the set of processed part shape feature vectors to obtain a set of part die-cutting planning decision point node dynamic attribute implicit encoding vectors; A part die-cutting planning core feature extraction secondary sub-unit, configured to perform core component feature extraction on the set of part die-cutting planning decision point node dynamic attribute implicit encoding vectors to obtain a set of part die-cutting planning decision point core component encoding vectors; A part die-cutting planning core feature fusion secondary sub-unit, configured to perform adaptive fusion on the set of part die-cutting planning decision point core component encoding vectors to obtain the part die-cutting planning dynamic response encoding vector.

6. The artificial intelligence-driven optimized system for special-shaped die-cutting of foam according to claim 5, wherein The secondary subunit for extracting the core features of the part die-cutting plan includes: The tertiary subunit for calculating the dynamic attribute class neighborhood matrix, which is used to calculate the dynamic attribute class neighborhood matrix of the part die-cutting plan decision point nodes based on the set of implicit coding vectors of the dynamic attributes of the part die-cutting plan decision point nodes; The tertiary subunit for calculating the dynamic attribute class degree matrix, which is used to calculate the dynamic attribute class degree matrix of the part die-cutting plan decision point nodes based on the set of implicit coding vectors of the dynamic attributes of the part die-cutting plan decision point nodes; The tertiary subunit for calculating the dynamic attribute Laplacian matrix, which is used to calculate the dynamic attribute Laplacian matrix of the part die-cutting plan decision point nodes based on the dynamic attribute class neighborhood matrix of the part die-cutting plan decision point nodes and the dynamic attribute class degree matrix of the part die-cutting plan decision point nodes; The tertiary subunit for matrix spectral decomposition, which is used to perform spectral decomposition on the dynamic attribute Laplacian matrix of the part die-cutting plan decision point nodes to obtain the set of core component coding vectors of the part die-cutting plan decision point.

7. The artificial intelligence-driven optimized system for special-shaped die-cutting of foam, according to claim 6, wherein The tertiary subunit for matrix spectral decomposition is used to: Perform topology optimization based on the matrix double subspace on the dynamic attribute Laplacian matrix of the part die-cutting plan decision point nodes to obtain the optimized dynamic attribute Laplacian matrix of the part die-cutting plan decision point nodes; Perform spectral decomposition on the optimized dynamic attribute Laplacian matrix of the part die-cutting plan decision point nodes to obtain the set of core component coding vectors of the part die-cutting plan decision point.

8. The artificial intelligence-driven optimized system for special-shaped die-cutting of foam, according to claim 7, wherein The typesetting scheme generation unit is used to: input the dynamic response coding vector of the part die-cutting plan and the dimension information into the AIGC-based foam special-shaped die-cutting planner to obtain the foam special-shaped die-cutting typesetting scheme.

9. The artificial intelligence-driven optimized system for special-shaped die-cutting of foam according to claim 8, wherein The cutting module is used to: convey the foam special-shaped die-cutting typesetting scheme to the die-cutting equipment control system to control the die-cutting equipment to perform cutting based on the foam special-shaped die-cutting typesetting scheme.

Images