Intelligent product design method and system based on big data and axiom design

Abstract

The invention relates to the technical field of intelligent product design, and discloses an intelligent product design method and system based on big data and axiom design. According to the method, user comment data is analyzed, and an improvement-needing degree quantitative index fusing attention and satisfaction is constructed to determine a design priority; taking the index as input, driving an axiom design theory to perform zigzag mapping and decomposition, guiding design decoupling according to data priority in the process, and outputting a design parameter set meeting the independence axiom; and finally, quantitatively evaluating the effectiveness of different design elements through a hierarchical generation scheme and user testing, and feeding back to a front end to form a closed loop. According to the method, automatic and precise conversion from subjective user feedback to an objective and non-coupling design scheme is achieved, the core problems that in a traditional design process, requirements are fuzzy, experience is relied on, and attribution is difficult to verify are solved, and design scientificity and efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of design optimization, specifically to a product intelligent design method and system based on big data and axiomatic design. Background Technology

[0002] With the development of big data and artificial intelligence technologies, data-driven design has become an important trend in product development. Currently, methods for using user reviews, market feedback, and other data to guide design optimization mainly fall into two categories: one is demand identification technology based on sentiment analysis and hotspot mining, which aims to extract user concerns and sentiment tendencies from massive amounts of text; the other is solution generation technology based on parametric modeling and intelligent optimization algorithms, which automatically generates candidate solutions by searching the design space. These technologies have improved the objectivity of design to some extent, but still have significant limitations.

[0003] The problems in the existing technology are mainly reflected in the following three aspects: First, in the requirement transformation stage, traditional sentiment analysis often outputs discrete sentiment tags or macro-level popularity scores, lacking a quantitative indicator system that can comprehensively assess users' "dissatisfaction level" and "attention intensity" to accurately indicate design priorities, resulting in ambiguous design inputs. Second, in the design generation stage, whether relying on algorithmic numerical optimization or designer experience-based decisions, it is difficult to systematically guarantee the coordination and non-conflictality of the internal parameters of the final design solution, i.e., satisfying the independence axiom, often leading to difficulties in subsequent engineering implementation. Finally, in the validation stage, existing methods mostly focus on evaluating the overall solution, failing to quantitatively isolate and attribute the specific contributions of different design elements to the user experience, resulting in a lack of precise directional guidance for design iteration.

[0004] Therefore, how to build an intelligent design system that can connect the entire process of "precise demand quantification - design logic generation - effect attribution verification" and where each link is closely coupled and mutually reinforcing has become a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a product intelligent design method and system based on big data and axiomatic design.

[0006] According to a first aspect of the present invention, a product intelligent design method based on big data and axiomatic design is proposed, comprising: Based on user review data for the target product, determine quantitative metrics to indicate design priorities; Based on axiomatic design theory, quantitative indicators are mapped and decomposed to determine the set of design parameters that satisfy the independence axiom. Product solutions with different optimization levels are generated based on the set of design parameters, and each product solution is verified through user testing to evaluate the effectiveness of the design elements.

[0007] According to some embodiments, in the method of the first aspect of the present invention, mapping and decomposing quantitative indicators based on axiomatic design theory includes: The portion of the quantitative indicators that meets the preset conditions is transformed into one or more top-level functional requirements. Based on the top-level functional requirements, the design parameters that meet the current functional requirements are found layer by layer, and the next layer of functional requirements are derived from the design parameters. The coupling relationship between functional requirements is analyzed in each mapping process.

[0008] According to some embodiments, in the method of the first aspect of the present invention, analyzing the coupling relationship between functional requirements includes: A design matrix is ​​constructed for judgment; when coupling is determined to exist, the design parameters are adjusted according to the priority order indicated by the quantitative indicators to eliminate or reduce coupling.

[0009] According to some embodiments, in the method of the first aspect of the present invention, adjusting the setting of design parameters according to the priority order indicated by the quantitative indicators specifically includes: prioritizing the adjustment of design parameters corresponding to the higher priority parts of the quantitative indicators.

[0010] According to some embodiments, in the method of the first aspect of the present invention, determining a quantitative indicator for indicating design priority based on user review data includes: Multiple product features that users care about are identified from user review data; based on user review data, the degree of improvement required for each product feature is calculated to form a quantitative indicator.

[0011] According to some embodiments, in the method of the first aspect of the present invention, calculating the degree of improvement required for each product feature includes: Calculate the level of attention given to product features in reviews; calculate user satisfaction with product features; and calculate the degree of improvement needed based on the level of attention and satisfaction.

[0012] According to some embodiments, in the method of the first aspect of the present invention, calculating user satisfaction with product features includes: A sentiment analysis model specifically designed for product features is constructed. The model is trained based on benchmark words and word vectors matched to product features. The sentiment analysis model is then used to analyze the sentiment tendencies of reviews regarding product features in order to calculate satisfaction.

[0013] According to some embodiments, the method of the first aspect of the present invention further includes: dynamically adjusting the calculation logic or weight of quantitative indicators based on the effectiveness evaluation results of different design elements obtained from user testing verification.

[0014] According to a second aspect of the present invention, a product intelligent design system based on big data and axiomatic design is proposed for performing the method as described in the first aspect of the present invention, the system comprising: The indicator determination module is used to determine quantitative indicators for indicating design priorities based on user review data; the theoretical mapping module, connected to the indicator determination module, is used to take the quantitative indicators as input, map and decompose them based on axiomatic design theory, and output a set of design parameters that satisfy the independence axiom; the verification and evaluation module, connected to the theoretical mapping module, is used to generate product solutions with different optimization levels based on the set of design parameters, and verify each solution through user testing to evaluate the effectiveness of the design elements.

[0015] According to some embodiments, in the system of the second aspect of the present invention: The theoretical mapping module also includes a coupling analysis and decoupling unit, which is used to construct the coupling relationship between the design matrix analysis functional requirements in each mapping process, and when coupling is determined, guide the setting and adjustment of design parameters to achieve decoupling based on the priority indicated by the quantitative indicators obtained from the indicator determination module. The verification and evaluation module is also connected to the indicator determination module, which feeds back the validity results of the design elements obtained from the quantitative evaluation to the indicator determination module, so as to dynamically adjust the calculation logic or weight of the quantitative indicators.

[0016] Compared with existing technologies, this invention provides a product intelligent design method and system based on big data and axiomatic design, which has the following beneficial effects: 1. To address the problem that user review data is difficult to transform into clear and sortable design inputs, this invention establishes a quantitative indicator system with "degree of improvement" as the core (the calculation formula is (1-satisfaction)×attention). This enables the accurate identification of core design pain points with "high attention and low satisfaction" from massive subjective feedback and provides data-driven priority ranking for design decisions.

[0017] 2. To address the issues of lack of theoretical constraints and potential parameter conflicts in data-driven design processes, this invention uses quantitative indicators as input to drive axiomatic design theory through a zigzag mapping and decomposition. It also guides decoupling based on data priority, thereby achieving an optimal design scheme that outputs a set of independent internal parameters and meets engineering rigor, thus eliminating design coupling risks in principle.

[0018] 3. To address the inaccuracy of general sentiment analysis models in evaluating professional product features, this invention constructs a feature-specific sentiment analysis model based on benchmark words and word vectors for each identified product feature. This enables precise quantification of complex, context-dependent sentiment tendencies in user reviews, ensuring the accuracy of demand insights.

[0019] 4. To address the problem of the disconnect between design verification results and front-end requirements analysis, which hinders continuous optimization, this invention dynamically feeds back the effectiveness results of design elements quantitatively evaluated in user testing to the requirements quantification module to adjust the calculation logic or weights. This achieves an adaptive closed loop of "design-verification-learning-optimization," enabling the system to have the ability to continuously iterate and evolve intelligently. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating an embodiment 1000 of the intelligent product design method based on big data and axiomatic design according to the present invention. Figure 2 for Figure 1 A flowchart illustrating step S2 in embodiment 1000; Figure 3 for Figure 2 The flowchart of the sub-step S22A for analyzing the function key coupling relationship in step S2; Figure 4 for Figure 1 A flowchart illustrating step S1 in embodiment 1000; Figure 5 for Figure 4 A flowchart illustrating sub-step S12 within step S1; Figure 6 for Figure 5 The flowchart of sub-step S122 further includes the following sub-step S122; Figure 7 This is a flowchart illustrating an embodiment 2000 of the intelligent product design method based on big data and axiomatic design according to the present invention. Figure 8 This is a schematic diagram of an embodiment 3000 of the intelligent product design system based on big data and axiomatic design according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Figure 1 This is a flowchart illustrating an embodiment 1000 of the intelligent product design method based on big data and axiomatic design according to the present invention. Figure 1 As shown, Example 1000 includes steps S1-S3.

[0023] In step S1, a product intelligent design system based on big data and axiomatic design (hereinafter referred to as the system) determines quantitative indicators to indicate design priorities based on user review data of the target product. The main function of step S1 is to transform the massive, unstructured user reviews obtained from e-commerce platforms and other channels into a structured, quantifiable product feature evaluation system, and ultimately generate a priority indicator that can scientifically guide the allocation of design resources, laying the objective foundation for the entire data-driven design process.

[0024] In some specific embodiments, step S1 includes: identifying multiple product features that users care about from user review data; and, based on the user review data, calculating the degree of improvement required for each product feature to form a quantitative indicator.

[0025] In step S2, the system maps and decomposes the quantitative indicators based on axiomatic design theory to determine the set of design parameters that satisfy the independence axiom. Step S2 is the core of the design generation phase of this solution. Its main function is to transform the quantitative indicators representing market demand generated in step S1 into a series of specific, feasible, and internally coordinated design parameters through rigorous engineering theory. This ensures that the final solution not only meets user expectations but also possesses engineering optimality and robustness.

[0026] Optionally, the implementation process of step S2 specifically includes: transforming the part of the quantitative indicators that meets the preset conditions into one or more top-level functional requirements; based on the top-level functional requirements, decomposing and mapping by finding the design parameters that meet the current functional requirements layer by layer, and deriving the next layer of functional requirements according to the design parameters, and analyzing the coupling relationship between functional requirements in each mapping process.

[0027] In step S3, product solutions with different optimization levels are generated based on the design parameter set, and each product solution is verified through user testing to evaluate the effectiveness of the design elements. In step S3, the system transforms the design parameter set output in step S2 into specific solutions that can be intuitively perceived and evaluated by users, and through scientific comparative experiments, accurately quantifies the independent contribution of each design element to the final user experience, thereby completing the closed-loop verification from parameter assumptions to effect verification.

[0028] Optionally, in some specific embodiments, in step S3, the system completes the scheme generation and hierarchy construction. The main principle is to use the hierarchical control variable method to systematically generate a set of product schemes with logical progression to isolate the influence of different design elements.

[0029] Specifically, in step S3, the system automatically or assistedly generates three core solutions based on the set of design parameters: Baseline Scheme A: Represents the original or existing market state without applying the optimized design parameters.

[0030] Single-element optimization scheme B: Based on scheme A, only one type of design parameter is applied, such as only appearance aesthetic parameters: color, texture, and style.

[0031] Full-element optimization scheme C: Applying all design parameters, representing the complete output of this design, including all optimizations such as appearance and functional layout.

[0032] In step S3, during the scheme generation process, the design parameter set is used as input to drive 3D modeling software or a parametric design platform to automatically render or generate visual representations of different schemes. The system needs to preset rule templates for scheme generation to ensure that, except for the target variable, other background conditions are consistent.

[0033] Optionally, in some specific embodiments, in step S3, the system performs user testing and data collection, which collects quantitative feedback data on multiple solutions through the subjective evaluation of the target user group.

[0034] Specifically, in step S3, the system integrates or invokes online survey tools to distribute questionnaires to the selected target users. The questionnaire presents a comparison of options A, B, and C, requiring users to rate the options from multiple dimensions, such as overall satisfaction, aesthetics, practicality, and compatibility with home decor. Simultaneously, open-ended questions can be set to collect qualitative feedback. All rating data is automatically collected and stored in a structured format.

[0035] Finally, in step S3, the system performs effectiveness evaluation and contribution analysis, using statistical analysis methods to compare the score differences between options, thereby attributing the value of design elements. The data analysis module automatically calculates the mean score and standard deviation of each option across all dimensions. Key analyses include: Option B vs. Option A: The score difference can be attributed to the independent contribution of the "appearance aesthetic elements".

[0036] Option C vs. Option B: The score difference can be attributed to the independent contribution of "core elements such as functional layout".

[0037] Optionally, in step S3, the system generates a "Design Element Effectiveness Evaluation Report" which is clearly presented in the form of charts, such as: "The functional layout optimization contributed +1.52 points to the overall satisfaction improvement (accounting for 70% of the total improvement value)".

[0038] Figure 2 for Figure 1 A flowchart illustrating step S2 in embodiment 1000. (See attached diagram.) Figure 2 As shown, step S2 includes steps S21-S22.

[0039] In step S21, the system transforms the portion of the quantitative indicators that meets preset conditions into one or more top-level functional requirements. Step S21 translates the data language into the language of design theory, that is, converts numerical values ​​into functional requirements, providing a clear starting point for subsequent systematic decomposition.

[0040] Optionally, step S21 establishes a mapping relationship between "product features and functional requirements" based on a set of preset conversion rules. The core logic is to define the product features that users most urgently want to improve as the top-level design goals that must be achieved to realize those features.

[0041] In some specific embodiments, the operation process and configuration of step S21 include: (1) Data Input and Threshold Judgment: The system reads the product feature quantitative analysis indicators output in step S1. The user or system presets the threshold for the degree of improvement. All features whose degree of improvement exceeds the threshold will be automatically marked as high-priority improvement items.

[0042] (2) Rule matching and requirement generation: The system calls the built-in conversion rule library. Each rule is in the form of: "IF Product Feature = 'X' THEN Top-Level Functional Requirement (FR) = 'Implement Y Function / Performance'". For example, the rule library may contain: "IF Feature = 'Functional Layout' THEN FR = 'Provide modular, pet behavior-compliant space planning'". The system automatically matches rules based on the identified features to generate an initial list of top-level functional requirements.

[0043] (3) Requirements Integration and Confirmation: The generated initial requirements list will be presented to the designer for confirmation and fine-tuning. The designer can merge requirements with similar meanings, or supplement a few key aesthetic and brand requirements that are not directly indicated by the data based on the design vision. Finally, a clear "Top-Level Functional Requirements (FRs) List" will be output.

[0044] In step S22, the system decomposes and maps the functional requirements based on the top-level functional requirements by finding the design parameters that meet the current functional requirements layer by layer, and deriving the next layer of functional requirements based on the design parameters. In each mapping process, the coupling relationship between functional requirements is analyzed.

[0045] Optionally, in step S22, the system expands the top-level functional requirements step by step, decomposing them into progressively progressive design parameters that can ultimately be implemented, and ensuring the independence of each parameter in the process to avoid design conflicts.

[0046] Optionally, in step S22, the system strictly follows the "zigzag mapping" process and "independence axiom" of axiomatic design theory. The idea is that in order to satisfy a functional requirement (FR), a design parameter (DP) must be found; and the introduction of this DP will generate new, more detailed lower-level FRs, and so on, until it is decomposed to the executable underlying DP.

[0047] Specifically, the specific operation process of step S22 includes: First, a hierarchical decomposition is performed: Starting with the "Top-Level Functional Requirements List," for each Functional Requirements (FR), the designer or system auxiliary tool proposes a possible Functional Requirements (DP). For example, to meet FR1 "provide a reasonable layout for pet activity space," DP1 "adopt a multi-layered, three-dimensional integrated structure" is proposed. Subsequently, based on DP1's "multi-layered, three-dimensional structure," the next level of FRs are derived, such as FR11 "must include a separate feeding area," FR12 "must include a toilet area," etc. This process is visualized in the software interface in the form of a hierarchical tree diagram.

[0048] Then, the design matrix is ​​constructed and coupling is determined: In each mapping relationship, the system automatically constructs a design matrix. This matrix is ​​used to formally describe the influence relationship between FRs and DPs. For example, it determines whether FR11 (feeding area) and FR12 (toilet area) are implemented by the same DP (e.g., "underlying platform partitioning"), thus creating coupling. The design matrix is ​​typically generated and stored in the background in the form of triples (influence, no influence, negative influence) or numerical form.

[0049] According to such Figure 2 The embodiments shown in this invention have the following beneficial effects: 1. Traditional design requirements definition relies on brainstorming or client interviews, which is subjective and divergent. This invention, through rule-based transformation based on quantitative data, makes the design starting point objective, precise, and strictly aligned with market pain points, ensuring from the source that the design does not deviate from the core needs of users.

[0050] 2. The design decision-making process is transformed from vague experience-based deduction into a visible, traceable, and verifiable logical deduction chain. Each final design parameter can be traced back to the specific user needs it is intended to satisfy, greatly improving the rigor and transparency of the design process.

[0051] Figure 3 for Figure 2The flowchart of the sub-step S22A for analyzing the function key coupling relationship in step S2 is shown below. Figure 3 As shown, step S22A includes steps SA1-SA2. The main function of step S22A is to provide an objective and automated decision-making basis to guide decoupling when the design matrix shows that coupling exists (i.e., one DP affects multiple FRs, or FRs interfere with each other), so as to satisfy the independence axiom in the optimal way.

[0052] Optionally, in step SA1, the system constructs a design matrix for judgment. Specifically, the system algorithm analyzes the design matrix. If the matrix is ​​a non-diagonal or triangular matrix, it determines that the current design has "coupling" and identifies the mutually coupled FRs groups and related DPs. In step SA2, when coupling is determined to exist, the design parameters are adjusted according to the priority order indicated by the quantitative indicators to eliminate or reduce coupling. Optionally, adjusting the design parameters according to the priority order indicated by the quantitative indicators in step SA2 specifically includes prioritizing the adjustment of design parameters corresponding to the higher priority parts of the quantitative indicators.

[0053] In some specific embodiments, in step SA2, the system does not rely on the designer's experience to try various decoupling solutions, but instead introduces the quantitative indicators from step S1 as the basis for decision-making. The decoupling algorithm follows these instructions: Priority sorting: Read the improvement level values ​​of the original product features corresponding to each FR involved in the coupling.

[0054] Decision Execution: Prioritize ensuring the optimal achievement of design goals related to the highest priority functional objectives (FRs). Decoupling strategies may include: adjusting the order of functional design goals (DPs), re-dividing functional boundaries, or introducing a separate functional design goal (DP) for high-priority FRs. The system may provide several alternative decoupling schemes and their predictive evaluations of FR satisfaction for designers to make decisions.

[0055] Optionally, in specific operations, the implementation of the instruction includes: "In the coupling group {FRa, FRb}, since the feature associated with FRa has a higher priority than FRb, priority is given to ensuring that DPa satisfies the specificity of FRa, and an independent parameter DPb' is found or added for FRb."

[0056] In traditional axiomatic design, decoupling is the biggest challenge, relying entirely on the designer's personal experience and repeated trial and error, resulting in low efficiency and unpredictable outcomes. According to... Figure 3The implementation method shown creatively breaks the deadlock in design logic by prioritizing data, ensuring that decoupled decisions always serve the core user needs. This makes the final design not only theoretically satisfy the axiom of independence but also commercially optimal, meaning resources are tilted towards the greatest pain points. This is equivalent to equipping axiomatic design theory with intelligent decision navigation, elevating it from a theoretical framework to a highly efficient automated design engine.

[0057] Figure 4 for Figure 1 A flowchart illustrating step S1 in embodiment 1000. (See attached diagram.) Figure 4 As shown, step S1 includes steps S11-S12.

[0058] In step S11, the system identifies multiple product features that users are interested in from user review data. The main function of step S11 is to automatically extract product dimensions that users commonly focus on and discuss from thousands of reviews, and establish a product feature system for subsequent quantitative analysis.

[0059] Optionally, step S11 is basically implemented using an unsupervised text topic clustering model. The principle is to treat each comment as a document set composed of words, and to discover recurring word patterns across all documents through a statistical model. Each recurring pattern is considered to represent a potential topic, that is, a product feature that a user is interested in.

[0060] In some specific embodiments, the specific operation process and configuration for identifying product features in step S11 include: (1) Data preparation and input: The preprocessed comment corpus was used as input. Each comment had been segmented and irrelevant words were removed.

[0061] (2) Model Training and Configuration: The LDA topic model is used for training. In implementation, the key configuration parameter is the preset number of topics. This can be determined by combining the perplexity index with business understanding; for example, initially setting it to 10-15 topics for multiple training comparisons. After the model runs, it will output a list of the most probable feature words for each topic.

[0062] (3) Feature Interpretation and System Construction: Analyze the high-frequency words under each theme, and have domain experts or designers summarize and name them to form understandable primary and secondary product features. For example, a theme containing words such as "stable", "shaky", "sturdy" and "falling apart" can be named the primary feature "quality", which includes the secondary feature "stability".

[0063] Traditional methods rely on manually preset keywords or category templates, resulting in narrow coverage and subjectivity. This invention, however, uses unsupervised machine learning to automatically discover features, comprehensively and objectively capturing user-generated concerns, even those unexpected by designers, ensuring the completeness and authenticity of demand insights.

[0064] In step S12, based on user review data, the system calculates the degree of improvement required for each product feature to form a quantitative indicator. Optionally, step S12 aims to assign a numerical value representing the urgency of improvement to each identified product feature.

[0065] In some specific embodiments, in step S12, the system first calculates the attention given to the product features in the reviews; then calculates the user's satisfaction with the product features; and finally calculates the degree of improvement required based on the attention given and the satisfaction level.

[0066] Figure 5 for Figure 4 A flowchart illustrating sub-step S12 within step S1. (See attached flowchart.) Figure 5 As shown, step S12 includes steps S121-S123.

[0067] In step S121, the system calculates the attention given to a product feature in the comments to measure the popularity or prevalence of a product feature in user discussions.

[0068] In some specific implementations, attention is a metric based on word frequency statistics. In the system, the program iterates through all comments, counting the total number of times keywords related to a particular product feature appear. For ease of comparison, this count is typically divided by the total number of comments or normalized to obtain a relative attention score. For example, if the term "functional layout" appears 3,500 times in 10,000 comments, its attention score is significantly higher than that of the feature "packaging," which only appears 500 times.

[0069] In step S122, the system calculates the user's satisfaction with the product features. Optionally, step S122 specifically includes: constructing a dedicated sentiment analysis model for the product features, the model being trained based on benchmark words and word vectors matched for the product features; and using the dedicated sentiment analysis model to analyze the sentiment tendency of comments regarding the product features in order to calculate satisfaction.

[0070] In step S123, the system calculates the degree of improvement required based on attention and satisfaction, in order to synthesize a single priority indicator that can scientifically guide design actions.

[0071] Optionally, in step S123, the system calculates the required improvement level using the formula: (1 - Satisfaction Level) × Attention Level. The system automatically reads the attention level generated in S121 and the satisfaction level generated in S122, and substitutes them into the formula to complete the calculation.

[0072] Low satisfaction alone might indicate a niche problem; high attention alone might suggest only mentioned issues with general satisfaction. Combining both is essential to identifying the true pain point where "most people are dissatisfied." This invention's formula creatively multiplies "dissatisfaction intensity (1 - satisfaction level)" by "problem breadth (attention level)," automatically and objectively identifying the product features most in need of priority improvement. For example, a feature with moderately low satisfaction but extremely high attention will have a much higher improvement requirement than a feature with extremely low satisfaction but no user engagement. This provides design teams with indisputable data-driven decision-making support.

[0073] Figure 6 for Figure 5 The flowchart of sub-step S12, which further includes sub-step S122, is shown below. Figure 6 As shown, step S122 includes steps S122a-S122b.

[0074] In step S122a, the system constructs a sentiment analysis model specific to product features. The model is trained based on benchmark words and word vectors matched to product features. The core innovation of this step lies in constructing an analysis model that deeply understands the product design context, rather than using general-purpose tools.

[0075] Specifically, in step S122a, the system completes dynamic matching and calibration of benchmark words based on the design semantic knowledge base, including: after initialization, dynamic expansion based on the built-in design semantic knowledge base. (The design semantic knowledge base includes associated features, design parameters, and functions. Using the word vectors obtained through training, the system automatically mines design domain terms that are highly semantically similar to the initial benchmark words (e.g., expanding from "refined" to "perfectly fitted"), and adds them to the benchmark word set. At the same time, the system automatically detects and resolves polarity conflicts of polysemous words in the design context, ensuring that the emotional polarity of each word under a specific feature is unique and accurate. For example, it determines that "heavy" is positive under the "material" feature and negative under the "portability" feature.)

[0076] General sentiment dictionaries cannot handle the differences in terminology and contextual ambiguity in the design field. The design of step S122a in this invention, through dynamic learning and calibration, enables the model to possess a precise, rich, and adaptive design evaluation vocabulary system, ensuring the accuracy of sentiment judgment from the source. This is the foundation for achieving high-precision satisfaction calculation.

[0077] In some specific embodiments, in step S122a, the system also completes word vector training enhanced with corpus from the design domain: the corpus used to train the word vector model, in addition to massive user comments, also deeply integrates professional design documents and a historical design case library. Through this domain-adaptive enhancement training, the semantic relationships in the word vector space are upgraded from everyday associations to associations that imply design logic and trade-offs. For example, "thin and light" and "structural strength" will establish a specific association in the vector space.

[0078] General word vectors cannot capture the deep semantics of design trade-offs such as sacrificing stability for aesthetic purposes. The design of step S122a in this invention enables the model to not only understand the surface meaning of words, but also to perceive the underlying design intent and constraints, making word vectors the semantic foundation for carrying design knowledge and providing a potential bridge for subsequent mapping from problems to design parameters.

[0079] In step S122b, the system uses a dedicated sentiment analysis model to analyze the sentiment tendencies in the reviews regarding product features in order to calculate satisfaction.

[0080] Optionally, in step S122b, the system implements a multi-dimensional emotion intensity calculation and aggregation rule for design decision-making, the specific implementation process of which includes: (1) Design Dimension Sentiment Deconstruction: For a comment involving multiple design dimensions, the model does not give an overall sentiment score, but instead initiates a sentiment deconstruction process. The system uses named entity recognition and dependency parsing to break down complex sentences. For example, if the comment content is "The appearance is very stylish, but the installation instructions are too obscure", the sentiment of "stylish" is attributed to the "appearance" feature, and the sentiment of "obscure" is attributed to the "instruction manual" sub-dimension.

[0081] (2) Satisfaction Aggregation Based on Design Weights: When calculating the overall satisfaction of a certain feature, it is not a simple average of all relevant sentiment scores. The system adopts weighted aggregation, and the weights are dynamically determined by two factors: A. Design relevance strength of sentiment words: Sentiment words that are more directly related to core design elements have higher weight. For example, describing "loose hinges" has a higher weight than simply saying "poor quality".

[0082] B. Experiential Representations of Reviewers: Reviews from experienced users may be given higher credibility weight. Optionally, experienced users may infer credibility from factors such as the length of historical reviews and the use of technical terminology.

[0083] The design of step S122b in this invention upgrades the sentiment analysis results from a crude positive / negative judgment to a structured problem diagnosis report that can directly guide specific design modifications. It distinguishes between primary and secondary contradictions and identifies the differences in focus between professional and ordinary users, making the final calculated satisfaction index more accurate and more instructive for action.

[0084] According to such Figure 6 The embodiment shown in this invention, a proprietary sentiment analysis model, successfully transforms a general natural language processing tool into a design perceptron that deeply understands the product design context, semantics, and decision-making needs by introducing a design knowledge base, performing domain-enhanced training, and implementing design-oriented refined computational rules. This represents a concentrated manifestation of the deep innovation and creative labor involved in this solution based on technological integration, and its output of high-fidelity satisfaction data ensures the accurate and effective operation of the entire subsequent axiomatic design process.

[0085] Figure 7 This is a flowchart illustrating an embodiment 2000 of the intelligent product design method based on big data and axiomatic design according to the present invention. Figure 7 As shown, Embodiment 2000 includes steps S201-S204. Specifically, steps S201-S203 and... Figure 1 Steps S1-S3 in Embodiment 1000 are the same and will not be repeated here.

[0086] Optionally, the method in Embodiment 2000 further includes: step S204, dynamically adjusting the calculation logic or weight of the quantitative indicators based on the effectiveness evaluation results of different design elements obtained from user testing and verification. Step S204 feeds back the knowledge obtained from the empirical verification in step S203 to the initial analysis stage of the system, forming an intelligent closed loop that can continuously learn from historical verification results and self-optimize, making the system more accurate and intelligent as the number of iterations increases.

[0087] Optionally, the specific implementation process of step S204 includes: dynamically correcting the requirement quantification model based on the assumption that "the more effective the design element, the more likely its corresponding original user requirement importance may be underestimated," in order to achieve result feedback and weight adjustment. Specifically: First, the system reads the effectiveness evaluation results of different design elements from the S3 report. For example, the contribution of the "functional layout" element is as high as 70%.

[0088] Then, the system traces which product features this element originally originated from. For example, "functional layout" and "spatial rationality".

[0089] Ultimately, the system determines the automatic adjustment instruction: according to preset rules, increase the weight coefficients of these features in the "degree of improvement needed" calculation in step S1. For example, increase the weight of the "functional layout" feature's attention (T) in the formula from 1.0 to 1.3, or directly introduce a gain factor based on historical validity into its satisfaction (Q) calculation. The new weight coefficients are written into the configuration database of the "demand quantification module" for data analysis of the next round of new products or iterations.

[0090] Optionally, in some specific embodiments, step S204 further includes dynamic optimization of the computational logic, the principle of which is that long-term accumulated validation data can be used to train higher-order meta-models and optimize the algorithm of the initial quantization index itself.

[0091] Specifically, the system uses multiple correlations between "quantitative indicators of product features and the effectiveness of design elements" as training data, and leverages machine learning models to discover more accurate predictive relationships. In the future, the system may automatically suggest adjustments to the calculation formula for "degree of improvement needed" or add new sentiment analysis dimensions.

[0092] Traditional design processes are open-loop, ending with the completion of a single project, failing to generate organizational-level design knowledge assets. A / B testing data is rarely used to refine the requirements analysis model itself. According to... Figure 7 The embodiment shown in this invention creatively establishes a reverse data link from terminal verification to initial analysis, realizing the adaptive evolution of the system, thereby enabling: Design insights are becoming increasingly accurate: systems are becoming better at identifying the key features that truly drive satisfaction from user reviews.

[0093] Resource allocation is becoming increasingly optimized: Priorities adjusted based on historical success data can more accurately guide design teams to focus their efforts on areas with "high returns".

[0094] Forming core data assets: What enterprises accumulate is not only a single successful solution, but also a continuously evolving intelligent decision-making model of "user needs - design action - market validation", which constitutes a sustainable competitive barrier.

[0095] Figure 8 This is a schematic diagram of an embodiment 3000 of the intelligent product design system based on big data and axiomatic design according to the present invention. Figure 8 As shown, Example 3000 includes an index determination module 301, a theoretical mapping module 302, and a verification and evaluation module 303.

[0096] The system shown in Embodiment 3000 of this invention is not a simple combination of existing tools, but rather a fully intelligent design entity with complete "perception-decision-verification-evolution" capabilities through a series of innovative module designs that deeply couple data insights, theoretical derivation, and empirical verification. The following will focus on how each module specifically carries out and realizes the core innovations of this solution.

[0097] In some specific embodiments, such as embodiment 3000, a system with a closed-loop architecture consisting of three core modules forming a unidirectional flow and bidirectional feedback is constructed. The overall system architecture and data flow include: (1) Forward design flow: index determination module 301 → theoretical mapping module 302 → verification and evaluation module 303, completing the whole process from data to solution.

[0098] (2) Reverse optimization flow: Verification and evaluation module 303 → index determination module 301, to realize system self-learning based on empirical results.

[0099] Data flows along this path, knowledge is accumulated in this cycle, and the system continues to evolve.

[0100] Optionally, the indicator determination module 301 is used to determine quantitative indicators for indicating design priorities based on user review data. Specifically, the indicator determination module 301 is implemented through the following design: (1) Feature-specific sentiment analysis engine: Instead of using a general sentiment dictionary, the module integrates a dynamic benchmark word matching and calibration mechanism as well as a word vector model trained with domain-specific corpus. This enables it to understand that "heavy" is positive when evaluating "material" and negative when evaluating "portability", thus achieving high-precision, contextualized sentiment intensity calculation, which is the basis for generating reliable quantitative indicators.

[0101] (2) Intelligent index synthesizer: The module automatically synthesizes the final index based on the original formula: Degree of improvement required = (1 - Satisfaction) × Attention level. Its innovation lies in the fact that the calculation process is not fixed. Its internal weight parameters are dynamically adjusted by the feedback signal from the verification and evaluation module 303 (see below), which reflects the system's adaptive capability.

[0102] The theoretical mapping module 302, connected to the index determination module 301, is used to take quantitative indices as input, map and decompose them based on axiomatic design theory, and output a set of design parameters that satisfy the independence axiom. Optionally, the theoretical mapping module 302 includes a coupling analysis and decoupling unit, which is used to construct the coupling relationship between the functional requirements of the design matrix analysis in each mapping process, and when coupling is determined to exist, guide the setting and adjustment of design parameters to achieve decoupling based on the priority indicated by the quantitative indices obtained from the index determination module.

[0103] In some specific embodiments, the theoretical mapping module 302 includes a coupling analysis and decoupling unit: this unit is the core that distinguishes this system from traditional axiomatic design software or CAD tools. Optionally, the coupling analysis and decoupling unit automatically constructs a design matrix in each level of the zigzag mapping, and when functional requirement coupling is detected, it does not leave the user to blindly try and fail. Its innovative decision-making logic is: to immediately query the indicator determination module 301 to obtain the original "degree of improvement" priority data corresponding to the involved functional requirements, and automatically generate or recommend decoupling solutions based on this priority order. This transforms the subjective process of decoupling into an objective, data-driven engineering decision for the first time.

[0104] The verification and evaluation module 303, connected to the theoretical mapping module 302, is used to generate product solutions with different optimization levels based on the design parameter set, and to verify each solution through user testing to evaluate the effectiveness of the design elements. Optionally, the verification and evaluation module 303 is also connected to the indicator determination module 301, which is used to feed back the effectiveness results of the design elements obtained from the quantitative evaluation to the indicator determination module, so as to dynamically adjust the calculation logic or weight of the quantitative indicators.

[0105] Specifically, the key implementation of the verification and evaluation module 303 in embodiment 3000 includes: The tiered scheme generator and contribution analysis engine include a pre-defined tiered control variable experiment template: "Baseline Scheme - Single-Element Optimization Scheme - Full-Element Optimization Scheme." It can automatically or assistedly generate this series of schemes and, through integrated statistical analysis tools, quantify and isolate the independent contribution values ​​of different design elements to user satisfaction. This surpasses traditional overall A / B testing, achieving a microscopic insight into design value.

[0106] Feedback Actuator: The most strategically significant innovation of this module lies in its direct feedback channel with the indicator determination module 301. It transforms the conclusions of the "contribution analysis" into specific adjustment instructions and sends them to the indicator determination module 301. The latter then dynamically adjusts the computational weight of "functional layout" related features in the quantitative model accordingly. This completes the intelligent closed loop, enabling the system to learn from each design practice and become increasingly accurate with use.

[0107] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0108] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A product intelligent design method based on big data and axiomatic design, characterized in that, include: Based on user review data for the target product, determine quantitative metrics to indicate design priorities; Based on axiomatic design theory, the quantitative indicators are mapped and decomposed to determine the set of design parameters that satisfy the independence axiom. Based on the set of design parameters, product solutions with different optimization levels are generated, and each product solution is verified through user testing to evaluate the effectiveness of the design elements.

2. The method according to claim 1, characterized in that, The mapping and decomposition of the quantitative indicators based on axiomatic design theory includes: The portion of the quantitative indicators that meets the preset conditions is transformed into one or more top-level functional requirements; Based on the top-level functional requirements, the design parameters that meet the current functional requirements are found layer by layer, and the next layer of functional requirements are derived from the design parameters. In each layer of mapping, the coupling relationship between functional requirements is analyzed.

3. The method according to claim 2, characterized in that, The coupling relationships between the analytical function requirements include: Construct a design matrix for judgment; When coupling is detected, the design parameters are adjusted according to the priority order indicated by the quantitative indicators to eliminate or reduce the coupling.

4. The method according to claim 3, characterized in that, The step of adjusting the design parameter settings according to the priority order indicated by the quantitative indicators specifically includes: Prioritize adjusting the design parameters corresponding to the higher priority components of the quantitative indicators.

5. The method according to claim 1, characterized in that, The determination of quantitative indicators for indicating design priorities based on user review data includes: Multiple product features that users are interested in were identified from the user review data; Based on the user review data, for each of the product features, a value indicating the degree of improvement is calculated to form the quantitative indicator.

6. The method according to claim 5, characterized in that, For each of the aforementioned product features, the calculation of the required degree of improvement includes: Calculate the level of attention given to the product features in the comments; Calculate user satisfaction with the product features; The required improvement value is calculated based on the level of attention and the level of satisfaction.

7. The method according to claim 6, characterized in that, The calculation of user satisfaction with the product features includes: Construct a sentiment analysis model specific to the product features, the model being trained based on benchmark words and word vectors matched to the product features; The proprietary sentiment analysis model is used to analyze the sentiment tendencies in reviews regarding the product features in order to calculate satisfaction.

8. The method according to claim 1, characterized in that, The method further includes: Based on the effectiveness evaluation results of different design elements obtained through user testing, the calculation logic or weight of the quantitative indicators are dynamically adjusted.

9. A product intelligent design system based on big data and axiomatic design, characterized in that, The system for performing the method as described in any one of claims 1-8 includes: The metrics determination module is used to determine quantitative metrics for indicating design priorities based on user review data. The theoretical mapping module, connected to the index determination module, is used to take the quantitative index as input, perform mapping and decomposition based on axiomatic design theory, and output a set of design parameters that satisfy the independence axiom. The verification and evaluation module, connected to the theoretical mapping module, is used to generate product solutions with different optimization levels based on the set of design parameters, and to verify each solution through user testing to evaluate the effectiveness of the design elements.

10. The system according to claim 9, characterized in that: The theoretical mapping module also includes a coupling analysis and decoupling unit, which is used to construct the coupling relationship between the design matrix analysis functional requirements in each mapping process, and when it is determined that there is coupling, guide the setting and adjustment of design parameters to achieve decoupling based on the priority indicated by the quantitative indicators obtained from the indicator determination module. The verification and evaluation module is also connected to the indicator determination module, and is used to feed back the validity results of the design elements obtained from the quantitative evaluation to the indicator determination module so as to dynamically adjust the calculation logic or weight of the quantitative indicators.

Images