Packaging structure simulation analysis and regulation system and method based on multi-objective optimization

Abstract

The invention belongs to the technical field of packaging design, and provides a packaging structure simulation analysis and regulation system and method based on multi-objective optimization. The method comprises the following steps: respectively extracting fuzzy constraint information and clear constraint information from design demand information input by a user; automatically generating a plurality of first packaging structure design schemes based on the fuzzy constraint information, performing simulation analysis with optimization difficulty analysis as a target on each first packaging structure design scheme, and determining the first packaging structure design scheme with the lowest optimization difficulty as a second packaging structure design scheme; on the basis of the second packaging structure design scheme, a third packaging structure design scheme is obtained based on clear constraint information optimization. According to the method, the adaptive scheme is generated based on the fuzzy constraint, the low-optimization-difficulty scheme is screened, and precise optimization is performed in combination with the clear constraint, so that the problems of demand conversion and constraint deficiency are solved, the multi-dimensional target is balanced, scheme modification and resource waste can be reduced, and the design efficiency and accuracy are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of packaging design, in particular to a packaging structure simulation analysis and regulation system and method based on multi-objective optimization. BACKGROUND

[0002] With the promotion of digitalization and intelligentization, the field of packaging design is undergoing profound changes. Currently, various online design platforms and AI design tools are emerging, providing users with a variety of packaging box design solutions. For example, some online design platforms support users to complete packaging box design independently through template dragging, parameter adjustment, etc.; other platforms use artificial intelligence technology to automatically generate design drawings according to user needs, greatly shortening the design cycle and reducing the design threshold. However, these innovative technologies still face significant defects in practical application.

[0003] However, in the process of existing AI generating packaging box design drawings, there is a common problem of lack of demand transformation and constraint mechanism. Most platforms only take the explicit parameter-type requirements input by users as a single guide, failing to transform abstract functional requirements (such as load-bearing, moisture-proof, portability), style preferences (such as minimalist style, national trend style), and use scenarios (such as e-commerce transportation, gift giving) into explicit and executable constraint conditions. This extensive generation mode leads to a serious disconnection between the AI output design drawings and user expectations, such as designing high-end gift packaging into cartoon style, or generating structures that cannot meet the actual transportation protection requirements, causing repeated modification of design schemes and waste of resources.

[0004] From the perspective of technical implementation, existing AI design tools have significant deficiencies in the construction and execution of constraint mechanisms. On the one hand, some platforms lack fine-grained control over the AI generation process, failing to accurately map user requirements into model-recognizable technical parameters and visual rules, leaving the AI generation behavior in a "borderless" state; on the other hand, even a few platforms that introduce constraints are mostly limited to simple style or size restrictions, making it difficult to balance multiple dimensions such as functionality, cost, and environmental protection. For example, in the trade-off between lightweight and protection performance, AI may generate overly simplified structures due to lack of effective constraints, leading to product damage during transportation; or in the application of environmentally friendly materials, the generated design drawings cannot match the process requirements of recyclable materials due to the lack of explicit material property constraints.

[0005] In summary, there is an urgent need for an accurate and efficient packaging structure design solution in the existing technology. SUMMARY

[0006] To this end, the present application provides a packaging structure simulation analysis and regulation method and system based on multi-objective optimization, an electronic device, a computer storage medium, and a computer program product to solve at least one of the above technical problems.

[0007] In a first aspect, the present application provides a packaging structure simulation analysis and regulation method based on multi-objective optimization, comprising the following method steps: obtaining fuzzy constraint information and clear constraint information from user input design requirement information; the fuzzy constraint information and the clear constraint information each include multiple design target information; automatically generating a plurality of first packaging structure design schemes based on the fuzzy constraint information, performing simulation analysis on each of the first packaging structure design schemes with optimization difficulty analysis as the target, and determining the first packaging structure design scheme with the lowest optimization difficulty as a second packaging structure design scheme; obtaining a third packaging structure design scheme based on the second packaging structure design scheme and the clear constraint information.

[0008] In a second aspect, the present application provides a packaging structure simulation analysis and regulation system based on multi-objective optimization, the system comprising an extraction unit, a scheme generation unit and a scheme optimization unit; The extraction unit obtains fuzzy constraint information and clear constraint information from user input design requirement information; the fuzzy constraint information and the clear constraint information each include multiple design target information; The scheme generation unit automatically generates a plurality of first packaging structure design schemes based on the fuzzy constraint information, performs simulation analysis on each of the first packaging structure design schemes with optimization difficulty analysis as the target, and determines the first packaging structure design scheme with the lowest optimization difficulty as a second packaging structure design scheme; The scheme optimization unit obtains a third packaging structure design scheme based on the second packaging structure design scheme and the clear constraint information.

[0009] In a third aspect, the present application provides an electronic device, comprising at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the computer program is executed by the processor to implement the method according to any one of the preceding aspects.

[0010] In a fourth aspect, the present application provides a computer storage medium storing a computer program executable by a processor to implement the method according to any one of the preceding aspects.

[0011] In a fifth aspect, the present application provides a computer program product comprising a computer program executable by a processor to implement the method according to any one of the preceding aspects.

[0012] The application generates and optimizes the scheme in stages by separating the fuzzy and clear constraints: first, generate the adaptive scheme based on the fuzzy constraints and screen the low optimization difficulty scheme, and then combine the clear constraints for accurate optimization. In this way, the demand conversion and constraint missing problems are solved, and the multi-dimensional target is balanced, which can reduce scheme modification and resource waste, and significantly improve the design efficiency and accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the application, and therefore should not be regarded as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0014] Figure 1 is a flow diagram of a packaging structure simulation analysis and regulation method based on multi-objective optimization disclosed by the embodiments of the application; Figure 2 is a structure diagram of a generative adversarial network disclosed by the embodiments of the application; Figure 3 is a structure diagram of a packaging structure simulation analysis and regulation system based on multi-objective optimization disclosed by the embodiments of the application. DETAILED DESCRIPTION

[0015] The embodiments of the application will be described below by specific embodiments. Those skilled in the art can easily understand other advantages and effects of the application from the content disclosed in the specification. Obviously, the described embodiments are part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of the application.

[0016] In addition, the technical features involved in different embodiments of the application described below can be combined with each other as long as there is no conflict.

[0017] As shown in Figure 1 The embodiments of the application disclose a packaging structure simulation analysis and regulation method based on multi-objective optimization, which comprises the following method steps: S10, the fuzzy constraint information and the clear constraint information are respectively extracted from the design requirement information input by the user; the fuzzy constraint information and the clear constraint information both include a plurality of design target information.

[0018] The user-input design requirement information usually contains multi-dimensional targets, such as requirements for the protection performance (e.g., compression strength, impact resistance) of the package, cost control (e.g., material usage, processing complexity), environmental friendliness (e.g., proportion of recyclable materials, degradation rate), appearance style (e.g., minimalist style, retro style), etc. Among these requirements, some can be directly quantified as specific parameters, and some are expressed in an abstract manner.

[0019] Among them, the clear constraint information refers to design targets that can be defined by specific numerical values or explicit standards, such as "package load capacity ≥ 5 kg", "material cost ≤ 2 yuan per unit", "size is 20 cm × 15 cm × 10 cm", etc. This type of information has clear boundaries and verifiability. The fuzzy constraint information refers to design targets that are difficult to quantify directly and rely on subjective judgment, such as "the package needs to reflect a high-end feel", "the structure should be easy for the elderly to operate", "the overall style is coordinated and unified", etc. This type of information needs to be converted into executable design parameters through subsequent processing.

[0020] Through the structured analysis of user requirements, the above two types of constraint information are extracted, and each type of constraint information contains multiple interrelated design targets (e.g., the clear constraints may include load capacity, size, cost, etc., and the fuzzy constraints may include style, ease of use, brand recognition, etc.).

[0021] S20, based on the fuzzy constraint information, a plurality of first packaging structure design schemes are automatically generated, and simulation analysis is performed on each first packaging structure design scheme with optimization difficulty as the target, and the first packaging structure design scheme with the lowest optimization difficulty is determined as the second packaging structure design scheme.

[0022] First, based on the fuzzy constraint information extracted in S10, a plurality of first packaging structure design schemes are automatically generated using generative AI or parametric design tools. These packaging structure design schemes need to meet the targets in the fuzzy constraints, such as generating different structure schemes using gilding process and matte materials for "high-end feel" requirements, and generating various layout schemes with easy-to-tear openings and carrying straps for "ease of use" requirements. During the generation process, constraint rule library is introduced (such as mapping "high-end feel" to "color saturation ≤ 30%, process includes more than one surface treatment"), to ensure that the scheme conforms to the overall orientation of the fuzzy constraints.

[0023] Subsequently, simulation analysis is performed on each first packaging structure design scheme with optimization difficulty as the target. Optimization difficulty is mainly evaluated through the following dimensions: Adjustability of structural parameters: whether the adjustment of parameters such as material thickness, corner radius, etc. will cause a chain reaction; Conflict degree of multi-objective: such as whether the cost will significantly increase when improving the protection performance, and the difficulty of trade-off between the two; Process feasibility: whether the design scheme is suitable for the existing production line, the complexity of modifying the process.

[0024] The above dimensions are quantitatively scored by a simulation tool, and finally the first packaging structure design scheme with the lowest optimization difficulty (i.e. high structure adjustability, small target conflict, and strong process adaptability) is determined as the second packaging structure design scheme, so as to reduce the resource consumption in the subsequent optimization process and improve the design efficiency.

[0025] S30, based on the second packaging structure design scheme, the clear constraint information is optimized to obtain a third packaging structure design scheme.

[0026] For the second packaging structure design scheme with lower optimization difficulty, the clear constraint information extracted in S10 is used as a hard index for multi-objective optimization of the scheme. For example, if the clear constraint includes "material cost ≤ 2 yuan", "compression strength ≥ 50N", and "recyclable material ratio ≥ 80%", the optimization is performed in the following way: Adjust the structure parameters: such as reducing the material thickness of the non-bearing area and increasing the reinforcing rib of the key part; Replace the material combination: such as using "kraft paper + honeycomb core" composite structure to improve the strength under the premise of maintaining the cost; Optimize the process details: such as changing the full wrapping design to local splicing to reduce the processing cost while meeting the recycling requirements.

[0027] In the optimization process, a multi-objective optimization algorithm (such as genetic algorithm, particle swarm algorithm) is used to solve the clear constraint targets cooperatively, so as to meet the single target while avoiding the performance loss of other targets. Finally, a third packaging structure design scheme that meets all the quantitative standards of the clear constraint information is generated, realizing the balance of packaging structure in multiple dimensions such as function, cost, and process.

[0028] The present application separates fuzzy and clear constraints, generates and optimizes schemes in stages: first, generate an adaptive scheme based on fuzzy constraints and select a low optimization difficulty scheme, and then combine clear constraints for precise optimization. This setting not only solves the problem of demand conversion and constraint loss, but also balances multiple dimensions, reduces scheme modification and resource waste, and significantly improves design efficiency and accuracy.

[0029] Optionally, the fuzzy constraint information and the clear constraint information are respectively extracted from the user input design requirement information, including: Use a natural language processing model to perform semantic analysis on the user input design requirement information and extract the quantitative indicators and non-quantitative descriptions contained therein; The quantitative indicators are determined as clear constraint information, and the quantitative indicators at least include packaging size, load-bearing parameter, material cost threshold, and recyclable material proportion, which are directly verifiable numerical targets; The non-quantitative descriptions are determined as fuzzy constraint information, and the non-quantitative descriptions at least include style preference, operation convenience, and visual coordination, which are abstract targets depending on subjective judgment.

[0030] Firstly, a pre-trained natural language processing model (such as a BERT model) is used to perform semantic analysis on the text requirement input by the user, and two types of information are extracted from the requirement through word segmentation, entity recognition, and relationship extraction: one type is quantitative indicators containing specific numerical values (such as “packaging size 20 cm × 15 cm × 10 cm”, “load-bearing ≥ 5 kg”), and the other type is non-quantitative descriptions without explicit numerical values (such as “simple and atmospheric style”, “convenient for children to open”).

[0031] Secondly, the extracted information is classified: all quantitative indicators are classified into clear constraint information, which has clear verification standards, such as “material cost ≤ 2 yuan per unit”, “recyclable material proportion ≥ 80%”, etc., which can be directly used as hard parameters for subsequent optimization.

[0032] At the same time, non-quantitative descriptions are classified into fuzzy constraint information, which depends on subjective judgment, such as “overall visual coordination”, “reflecting high-end texture”, etc., which need to be mapped to executable design parameters (such as “color contrast ≤ 40%”, “including gold stamping process”) through a subsequent rule library.

[0033] Finally, the two types of constraint information are aggregated respectively to form a constraint set containing multiple design targets, providing structured input for subsequent scheme generation.

[0034] Optionally, a plurality of first packaging structure design schemes are automatically generated based on the fuzzy constraint information, comprising: The fuzzy constraint information is mapped to an executable design parameter combination through a pre-set constraint rule library, and the constraint rule library includes associated visual design rules belonging to style preference constraints, associated structure feature rules belonging to functional experience constraints, and associated identification layout rules belonging to brand identification constraints; A generative adversarial network is used to generate a diversified initial design scheme set that satisfies the fuzzy constraints, with the design parameter combination as the conditional input; The initial design scheme set is subjected to feasibility filtering to eliminate schemes that violate basic physical rules and process standards, and a plurality of first packaging structure design schemes are obtained.

[0035] Firstly, the abstract fuzzy constraint information is converted into specific design parameters through a pre-set constraint rule library. The rule library at least includes the following three types of mapping rules: Correlation visual design rules (style preference class constraints): For example, map "high-end feel" to "color saturation ≤ 30%, use gold stamping / UV process, proportion of sans-serif font ≥ 80%"; Correlation structural feature rules (functional experience class constraints): For example, map "easy for the elderly to operate" to "easy-tear opening width ≥ 5mm, opening force ≤ 20N, corner radius ≥ 3mm"; Correlation identification layout rules (brand identification class constraints): For example, map "highlight brand LOGO" to "LOGO area proportion ≥ 15%, position centered / top, use brand standard color".

[0036] Through the above mapping operation, the problem of fuzzy constraints being difficult to quantify can be solved, providing executable parameter boundaries for AI generation, and ensuring that the design direction meets the user's intention.

[0037] Then, using a GAN architecture, as shown in Figure 2 the mapped design parameter combination is input as a condition (Condition), and through the adversarial training of the generator (Generator) and the discriminator (Discriminator), a three-dimensional packaging structure model that meets the constraint conditions is generated. For example: input "high-end feel + portability" parameter combination, the generator outputs a packaging box with magnetic seal, handle and matte material; by adjusting the parameter weights (such as "high-end feel" weight 0.8, "cost control" weight 0.6), the balance of multiple objectives is achieved.

[0038] Random noise (Latent Space) and parameter perturbation can also be introduced to generate multiple design variants under the same constraint conditions, ensuring the diversity of the initial design scheme set.

[0039] In addition, to avoid generating schemes that are not manufacturable or have performance defects, reduce the cost of subsequent optimization (manual or automatic optimization), and improve design efficiency, physical rule checking and process standard checking are also needed.

[0040] Basic physical rule checking: Check whether each scheme in the initial design scheme set violates the basic physical rules. For example, structural stability, material adaptability. Specifically, through finite element analysis, verify whether the deformation of the packaging box under the load condition exceeds the safety threshold (such as ≤ 5%); and ensure that the selected material (such as kraft paper) can meet the strength requirement (such as compressive strength ≥ 50N).

[0041] Process standard checking: Filter schemes that meet the manufacturing process, for example: folding line spacing ≥ 3mm, avoid sharp angle folding that cannot be achieved by the process; the distance between the printing area and the die cutting edge ≥ 2mm, to prevent the pattern from being cut.

[0042] The embodiment can break through the limitation of traditional AI design which only relies on explicit numerical values by converting subjective requirements into objective design parameters through a rule base. Meanwhile, combining the creativity of GAN and the constraint of the rule base, the diversity of the initial design scheme can be ensured, and the initial design scheme obtained can meet the design requirements of the user. In addition, the feasibility filtering mechanism is used to exclude unfeasible schemes in advance, to break through the whole process from design to production, and to reduce the trial and error cost.

[0043] Optionally, the generative adversarial network is adopted to generate a diversified initial design scheme set meeting the fuzzy constraints with the design parameter combination as a conditional input, comprising: The generator of the generative adversarial network receives the design parameter combination as a conditional input, and introduces dynamic random noise; wherein the variance of the dynamic random noise is dynamically adjusted according to the abstract degree of the fuzzy constraint information; A parameter perturbation module is arranged in the middle layer of the generator to randomly fine-tune the design element parameters; wherein the fine-tuning amplitude is dynamically allocated based on the priority of each fuzzy constraint information; The generated scheme is double-checked by the discriminator, including checking whether it meets the fuzzy constraints and checking the feature difference degree between the schemes, and when the difference degree is lower than a preset threshold, the generator is fed back to increase the random noise variance or the parameter fine-tuning amplitude; The above process is iterated until a preset number of schemes are generated, forming a diversified initial design scheme set.

[0044] When generating the initial design scheme set, more abundant variant schemes need to be generated for the abstract constraints, for example, generating multiple implementation paths such as matte, gilding and embossing for “high-end feeling” to avoid falling into a single local optimum. Specifically: In the generator of the generative adversarial network (GAN), the design parameter combination is taken as a conditional input (Condition), and Gaussian random noise is introduced . Among them, the noise variance is dynamically determined by the abstract degree of the fuzzy constraint, and the noise variance is positively correlated with the abstract degree. Specifically, the higher the abstract degree, the larger the noise variance , and vice versa.

[0045] The quantification of the abstract degree can be realized through three-level semantic analysis, specifically: (1) The NLP model is used to perform word segmentation and part-of-speech tagging on the fuzzy constraint text, and extract core components such as nouns, verbs and adjectives. For example, for “simple style and easy to open with one hand”, the keywords “simple style” and “one-handed opening” are extracted, and the modifier “easy to” is extracted.

[0046] (2) The following two quantitative indicators are calculated: Keyword abstraction: Calculate the semantic distance between keywords and entity words (such as "size", "material") through pre-trained word vector models (such as Word2Vec), the greater the distance, the higher the abstraction (such as "high-end feeling" distance from entity words 0.82, "rounded design" distance 0.35); Adjective density: Calculate the proportion of adjectives and keywords (such as "extremely high-end texture" with an adjectival proportion of 100%, "slightly rounded" with an adjectival proportion of 50%).

[0047] (3) Weighted sum of the two indicators (for example, the weight is 0.5), get the abstract score ([0,1] interval), for example, "high-end feeling" gets 0.78, "rounded design" gets 0.32.

[0048] As shown in the following Table 1 example: At the same time, set the parameter perturbation module in the middle layer of the generator (such as after the convolution layer), randomly fine-tune the design element parameters : Wherein, is a uniformly distributed random vector, ; The perturbation coefficient is inversely proportional to the constraint priority: low-priority constraints (such as "secondary visual elements") get larger values (such as 0.2), high-priority constraints (such as "brand logo position") get smaller values (such as 0.05).

[0049] Through the above application mechanism of the dynamic random noise and parameter perturbation signal of the application, more bold exploration of secondary constraints can be realized under the premise of guaranteeing core constraints, for example, while fixing the LOGO position, allowing the packaging corner radius to change within a large range.

[0050] Next, for the scheme generated by the generator, the discriminator simultaneously performs two tasks: Constraint compliance check: Evaluate whether the generated scheme meets the fuzzy constraints (such as judging whether the "high-end feeling" related features meet the standard through the feature extractor); Difference check: Calculate the structural similarity (such as Hausdorff distance, shape descriptor difference) between the generated schemes, and trigger feedback when the similarity exceeds the threshold (such as 85%).

[0051] The feedback adjustment strategy is: when the difference is not enough, dynamically increase the random noise variance or the parameter perturbation coefficient , for example: , wherein, is the current diversity, is the learning rate (e.g., 0.1).

[0052] The above process is iterated continuously until a preset number N of schemes (e.g., 50) are generated, and the scheme set satisfies: average diversity ≥ preset threshold (e.g., 70%); constraint compliance rate ≥ preset threshold (e.g., 90%).

[0053] The final scheme set is stored according to the diversity clustering, for example: Each scheme cluster in it represents a design direction (e.g., minimalist, retro), ensuring diversity coverage.

[0054] This embodiment dynamically adjusts the search strategy according to the abstract degree and priority of the constraints, solving the problem of insufficient diversity of traditional GAN under multiple constraints. At the same time, through double-checking feedback loop, it simultaneously evaluates the constraint compliance and scheme diversity, ensuring that the generated scheme meets the requirements and avoids duplication, thereby improving design efficiency.

[0055] Optionally, the simulation analysis of each first packaging structure design scheme for the purpose of optimizing difficulty analysis includes: calling a preset multi-dimensional optimization difficulty evaluation index and a corresponding simulation analysis model; wherein the multi-dimensional optimization difficulty evaluation index includes a structure parameter adjustability index, a multi-objective conflict degree index, and a process adaptability index; using the simulation analysis model to process the corresponding first packaging structure design scheme respectively, obtaining a corresponding quantitative score, and calculating a comprehensive optimization difficulty score corresponding to the first packaging structure design scheme based on each quantitative score.

[0056] An evaluation index system including three core dimensions is constructed, and the optimization difficulty is decomposed into quantifiable sub-indices: structure parameter adjustability index: evaluating the degree of freedom of design parameter adjustment, for example, whether adjusting the material thickness will cause a chain reaction (such as adjusting the strength change simultaneously with the reinforcing rib).

[0057] multi-objective conflict degree index: quantifying the degree of contradiction between each design objective, for example, the conflict between improving protection performance (increasing material thickness) and reducing cost (reducing material usage).

[0058] process adaptability index: evaluating the matching degree of the design scheme with the existing production line, for example, whether the folding structure is suitable for the existing die-cutting equipment.

[0059] For each of the above evaluation indices, a special simulation model is used for quantitative analysis: Parameter sensitivity analysis model: Through finite element analysis (FEA) or parametric modeling, calculate the influence coefficient matrix of parameter adjustment on other parameters. For example: . Wherein, is the adjustment parameter, is the affected parameter. The greater the influence coefficient, the more significant the impact of adjusting on , the lower the adjustability, and vice versa.

[0060] Target correlation analysis model: Through historical design data or experimental design (DOE), calculate the Pearson correlation coefficient between each target: . Wherein , are different design targets, such as cost and strength; closer to -1, the higher the conflict, and vice versa.

[0061] Process compatibility analysis model: Based on the production process knowledge base, evaluate the manufacturability of the design scheme. For example: whether the folding line spacing meets the minimum machining precision of the equipment (such as ≥3mm); whether the material combination is suitable for the existing printing process (such as the adhesion of UV ink to different materials).

[0062] The quantitative scores of each index are weighted and summed, and the comprehensive score = ω1×adjustability score + ω2×adjustability score conflict score + ω3×process adaptability score. The weighted weights of each quantitative score can be dynamically adjusted based on user demand priority. For example: When the user emphasizes "quick production", adjust ω3; When the user emphasizes "performance optimization", adjust ω2. It can be understood that the first packaging structure design scheme with the lowest optimization difficulty is the one with the lowest comprehensive score, which is taken as the second packaging structure design scheme.

[0063] This embodiment decomposes the abstract "optimization difficulty" into three dimensions of structure, target, and process, solving the problem of relying on subjective judgment in traditional methods; Moreover, the weighted weights of the quantitative scores can be adjusted according to the user demand priority, so that the evaluation result is more suitable for the actual application scene, to enhance the pertinence of the scheme.

[0064] As shown in Figure 3 , the embodiment of the present application also provides a packaging structure simulation analysis and control system 10 based on multi-objective optimization, which comprises an extraction unit 11, a scheme generation unit 12, and a scheme optimization unit 13; The extraction unit 11 extracts the fuzzy constraint information and clear constraint information from the design requirement information input by the user; The fuzzy constraint information and the clear constraint information both include a plurality of design target information; The scheme generation unit 12 automatically generates a plurality of first packaging structure design schemes based on the fuzzy constraint information, performs simulation analysis on each first packaging structure design scheme for the purpose of optimization difficulty analysis, and determines the first packaging structure design scheme with the lowest optimization difficulty as a second packaging structure design scheme; The scheme optimization unit 13 optimizes a third packaging structure design scheme based on the second packaging structure design scheme and the clear constraint information.

[0065] Optionally, the extraction unit 11 is configured to: perform semantic analysis on the design requirement information input by the user using a natural language processing model, and extract quantitative indicators and non-quantitative descriptions contained therein; determine each quantitative indicator as clear constraint information, and the quantitative indicators at least include packaging size, bearing parameter, material cost threshold, and recyclable material proportion, which are directly verifiable numerical targets; determine each non-quantitative description as fuzzy constraint information, and the non-quantitative descriptions at least include style preference, operation convenience, and visual coordination, which are abstract targets depending on subjective judgment.

[0066] Optionally, the scheme generation unit 12 is configured to: map the fuzzy constraint information to an executable design parameter combination through a preset constraint rule library, and the constraint rule library includes associated visual design rules belonging to style preference constraints, associated structure feature rules belonging to function experience constraints, and associated identification layout rules belonging to brand identification constraints; adopt a generative adversarial network to generate a diversified initial design scheme set satisfying the fuzzy constraint as a conditional input of the design parameter combination; perform feasibility filtering on the initial design scheme set to eliminate schemes violating basic physical rules and process standards, and obtain a plurality of first packaging structure design schemes.

[0067] Optionally, the scheme generation unit 12 is configured to: The generative adversarial network is adopted to generate a diversified initial design scheme set satisfying the fuzzy constraint as a conditional input of the design parameter combination, including: a generator of the generative adversarial network receives the design parameter combination as a conditional input, while introducing dynamic random noise; wherein the variance of the dynamic random noise is dynamically adjusted according to the abstraction degree of the fuzzy constraint information; a parameter perturbation module is arranged in the middle layer of the generator to randomly fine-tune the design element parameters; wherein the fine-tuning amplitude is dynamically allocated based on the priority of each fuzzy constraint information; The generated scheme is double-checked by the discriminator, including checking whether it meets the blur constraint and checking the feature difference between schemes, and when the difference is lower than a preset threshold, feedback is fed back to the generator to increase the random noise variance or the parameter fine-tuning amplitude. Iterate the above process until a preset number of schemes are generated, forming a diversified initial design scheme set.

[0068] Optionally, the scheme generation unit 12 is configured to: Call a preset multi-dimensional optimization difficulty evaluation index and a corresponding simulation analysis model, wherein the multi-dimensional optimization difficulty evaluation index includes a structure parameter adjustability index, a multi-objective conflict degree index and a process adaptability index. The simulation analysis model is used to process the corresponding first packaging structure design scheme to obtain a corresponding quantitative score, and a comprehensive optimization difficulty score corresponding to the first packaging structure design scheme is calculated based on the quantitative scores.

[0069] The embodiment of the application also provides an electronic device, which comprises at least one processor, a memory and a computer program stored in the memory and executable on the at least one processor, and the computer program is executed by the processor to implement the method according to any one of the preceding embodiments.

[0070] The embodiment of the application also provides a computer storage medium, which stores a computer program executable by a processor to implement the method according to any one of the preceding embodiments.

[0071] The embodiment of the application also provides a computer program product, which contains a computer program executable by a processor to implement the method according to any one of the preceding embodiments.

[0072] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in the above description. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the application.

[0073] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements shall be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A packaging structure simulation analysis and regulation method based on multi-objective optimization, characterized in that: The method comprises the following steps: obtain fuzzy constraint information and clear constraint information from user input design requirement information; the fuzzy constraint information and the clear constraint information each comprise a plurality of design target information; automatically generate a plurality of first packaging structure design schemes based on the fuzzy constraint information, and perform simulation analysis on each first packaging structure design scheme to analyze optimization difficulty, and determine the first packaging structure design scheme with the lowest optimization difficulty as a second packaging structure design scheme; optimize a third packaging structure design scheme based on the second packaging structure design scheme and the clear constraint information.

2. The packaging structure simulation analysis and regulation method based on multi-objective optimization according to claim 1, characterized in that: obtaining fuzzy constraint information and clear constraint information from user input design requirement information comprises: using a natural language processing model to perform semantic analysis on the user input design requirement information, and extracting quantitative indicators and non-quantitative descriptions contained therein; determining each quantitative indicator as clear constraint information, and the quantitative indicators at least comprise packaging size, bearing parameter, material cost threshold, and recyclable material proportion, which are directly verifiable numerical targets; determining each non-quantitative description as fuzzy constraint information, and the non-quantitative descriptions at least comprise style preference, operation convenience, and visual coordination, which are abstract targets depending on subjective judgment.

3. The packaging structure simulation analysis and regulation method based on multi-objective optimization according to claim 1, characterized in that: automatically generating a plurality of first packaging structure design schemes based on the fuzzy constraint information comprises: mapping the fuzzy constraint information to an executable design parameter combination through a preset constraint rule library, and the constraint rule library comprises associated visual design rules belonging to style preference constraints, associated structure feature rules belonging to function experience constraints, and associated identification layout rules belonging to brand identification constraints; using a generative adversarial network to generate a diversified initial design scheme set satisfying the fuzzy constraint as a conditional input of the design parameter combination; performing feasibility filtering on the initial design scheme set to eliminate schemes violating basic physical rules and process standards, and obtaining a plurality of first packaging structure design schemes.

4. The packaging structure simulation analysis and regulation method based on multi-objective optimization according to claim 3, characterized in that: The method of using a generative adversarial network to generate a diversified initial design scheme set satisfying the fuzzy constraint as a conditional input of the design parameter combination comprises: a generator of the generative adversarial network receives the design parameter combination as a conditional input, and simultaneously introduces dynamic random noise; wherein the variance of the dynamic random noise is dynamically adjusted according to the abstraction degree of the fuzzy constraint information; a parameter perturbation module is arranged in the middle layer of the generator to randomly fine-tune design element parameters; wherein the fine-tuning amplitude is dynamically allocated based on the priority of each fuzzy constraint information; the generated scheme is double-checked by a discriminator, including checking whether it conforms to the fuzzy constraint and checking the feature difference degree between schemes, and when the difference degree is lower than a preset threshold, the generator is fed back to increase the random noise variance or the parameter fine-tuning amplitude; iterating the above process until a preset number of schemes are generated to form a diversified initial design scheme set.

5. The packaging structure simulation analysis and regulation method based on multi-objective optimization according to claim 4, characterized in that: performing simulation analysis on each first packaging structure design scheme to analyze optimization difficulty comprises: The preset multi-dimensional optimization difficulty evaluation index and the corresponding simulation analysis model are called, wherein the multi-dimensional optimization difficulty evaluation index includes a structure parameter adjustability index, a multi-objective conflict degree index, and a process adaptability index; The simulation analysis model is used to process the corresponding first packaging structure design scheme respectively, to obtain a corresponding quantitative score, and a comprehensive optimization difficulty score corresponding to the first packaging structure design scheme is calculated based on the quantitative scores.

6. A packaging structure simulation analysis and regulation system based on multi-objective optimization, characterized in that, The system includes an extraction unit, a scheme generation unit, and a scheme optimization unit. The extraction unit extracts fuzzy constraint information and clear constraint information from user input design requirement information; The fuzzy constraint information and the clear constraint information each include a plurality of design target information; The scheme generation unit automatically generates a plurality of first packaging structure design schemes based on the fuzzy constraint information, performs simulation analysis on each first packaging structure design scheme for optimization difficulty analysis, and determines the first packaging structure design scheme with the lowest optimization difficulty as a second packaging structure design scheme; The scheme optimization unit optimizes a third packaging structure design scheme based on the second packaging structure design scheme and the clear constraint information.

7. The packaging structure simulation analysis and regulation system based on multi-objective optimization of claim 6, wherein: The extraction unit is configured to: use a natural language processing model to perform semantic analysis on user input design requirement information, and extract quantitative indicators and non-quantitative descriptions contained therein; determine each quantitative indicator as clear constraint information, wherein the quantitative indicators at least include packaging size, load-bearing parameter, material cost threshold, and recyclable material proportion, which are directly verifiable numerical targets; determine each non-quantitative description as fuzzy constraint information, wherein the non-quantitative descriptions at least include style preference, operation convenience, and visual coordination, which are abstract targets dependent on subjective judgment.

8. An electronic device, comprising: The electronic device includes at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the computer program is executed by the processor to implement the method of any one of claims 1-5.

9. A computer storage medium, characterized in that: The computer storage medium stores a computer program executable by a processor to implement the method of any one of claims 1-5.

10. A computer program product, characterised in that: The computer program product includes a computer program executable by a processor to implement the method of any one of claims 1-5.

Images