Product design auxiliary system and method based on big data

Abstract

The invention belongs to the technical field of product intelligent design, and particularly relates to a product design auxiliary system and method based on big data. The method comprises the following steps: collecting product design, market, production and after-sales full-link data, and combining time sequence analysis to extract a dynamic trend; in the scheme optimization stage, material availability, process suitability and cost controllability verification is carried out based on production end data, and design parameters are corrected through an after-sales data feedback mechanism; and adopting an improved Pareto optimization algorithm and dynamic threshold evaluation to screen an optimal scheme. The system for realizing the method comprises a data processing module, a design analysis module, a scheme optimization module and an interactive output module. Through full-link data fusion, production feasibility verification and closed-loop optimization, the problem of design and production disjunction is effectively solved, and product reliability and comprehensive benefits are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of product intelligent design, and specifically relates to a product design assistance system and method based on big data. Background Art

[0002] With intensifying market competition and evolving consumer demands, product design plays an increasingly crucial role in a company's success. Traditional product design methods often rely on the designer's experience and subjective judgment, which can be subject to significant uncertainty and risk. Especially in the face of rapidly changing markets and diverse consumer demands, traditional methods often struggle to respond and adapt quickly, resulting in a disconnect between design solutions and actual market demands.

[0003] In the traditional design process, designers typically develop designs based on product characteristics such as shape, size, and materials. However, this approach often overlooks market dynamics and competitor design trends. The key to market demand and product success often depends on the ability to keep up with market trends and meet specific consumer needs. However, due to a lack of systematic collection and analysis of market data, design plans often struggle to accurately predict market feedback, potentially wasting resources and increasing the risk of design failure, failing to maximize design value.

[0004] In recent years, with the development of big data technology, data-driven design methods have begun to gain widespread attention and application. By collecting and analyzing massive amounts of market data, consumer feedback, and competitive product information, big data technology can provide accurate and scientific decision-making for product design. Through big data analysis, it is possible to deeply explore the correlation between market trends, consumer preferences, and product features, enabling effective predictions of product market performance during the design phase, significantly improving the scientific and targeted nature of design.

[0005] Chinese invention patent publication number CN119313390A discloses a product design assistance system and method based on big data. Through big data analysis and calculation, this invention quickly identifies design directions with potential in the market, thereby shortening the design cycle and improving design efficiency. Based on market popularity data and similarity analysis, it reduces the risk of blind design and ensures that the designed products are more in line with market demand. By weighting the product's local and overall features, it ensures a more comprehensive and reasonable design plan. With the support of big data, the designed products are more in line with market trends and consumer demand, enhancing the product's market competitiveness. The use of big data technology makes the design process more scientific and data-driven, reducing the uncertainty caused by subjective judgment. The reference three-dimensional product data provided can provide a valuable reference basis for subsequent detailed design and production, improving the efficiency and quality of the entire product development process.

[0006] However, in actual use, the above patents do not take into account production-end data such as material inventory, production process feasibility, and manufacturing costs, which may result in a design that can be designed but difficult to produce. Summary of the Invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a product design assistance system and method based on big data to solve the problems in the above-mentioned background technology.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solutions: A product design assistance method based on big data, comprising the following steps: S1, full-link dynamic data collection and time series preprocessing, collects design feature data of the target product, market feedback data, similar product data, real-time production data and after-sales operation data, and cleans, standardizes and structures the above collected data to obtain preprocessed data. At the same time, time series alignment and trend extraction are performed on the time series data with time dimension characteristics in the preprocessed data; S2, design feature and market demand analysis, extracting design feature vectors reflecting the inherent attributes of the product from the preprocessed data, combining them with the market popularity and user preference data contained in the time series data, and analyzing them through the time series attention mechanism to obtain key design features and market demand trends; S3, similar product matching and evaluation: Based on the design feature vector, calculate the feature similarity between the target product and similar products in the preprocessed data; perform functional semantic analysis based on the functional semantic descriptions in the similar product data, generate a comprehensive similarity index that integrates feature similarity and functional semantic similarity, screen out highly matching similar products, and evaluate them based on the dimensions of design reusability, market performance adaptability, and fit with market demand trends; S4, generating a preliminary design solution: generating a preliminary product design solution based on the key design features, market demand trends, and similar product evaluation results; embedding market constraints into the preliminary product design solution, and performing quantitative scoring on the preliminary design solution's fit with the key features and its trend adaptability, thus forming a verifiable solution set; S5, design solution optimization, based on real-time production data and cross-domain knowledge graphs in preprocessed data, conducts production feasibility verification and multi-objective optimization of the preliminary product design solution to resolve conflicts between the preliminary design solution and production conditions. At the same time, it combines the time series analysis results of after-sales operation data to feed back design parameter adjustments; S6, optimization scheme evaluation and output, uses dynamic thresholds to evaluate the optimized design scheme, and outputs a reference design scheme that meets the evaluation conditions and the optimization basis.

[0009] Furthermore, in step S1, the design feature data includes product structure parameters, functional parameters and historical design scheme parameters; the market feedback data includes user evaluations, competitor sales data and market popularity trend data; the real-time data on the production side includes material inventory data, production process parameter data and manufacturing cost data; the after-sales operation data includes fault record data, user usage feedback data and performance degradation data; the trend extraction of the time series data uses a time series decomposition algorithm to separate long-term trend items, periodic items and random items; the preprocessing includes: using a text fingerprint algorithm to remove duplicate data, identifying and filtering abnormal data through a clustering algorithm, and using TF-IDF combined with an LDA topic model to extract key information from unstructured data and perform structured conversion.

[0010] Furthermore, in step S2, the design feature vector includes functional features, structural features, material features, cost features, user experience features and sustainability features; the temporal attention mechanism assigns dynamic attention weights to market data in different time periods, with recent data having a higher weight than long-term data.

[0011] Furthermore, in step S3, the functional semantic analysis converts the functional description into a semantic vector through the BERT model and then calculates the cosine similarity; the calculation formula of the comprehensive similarity is: , where α is the dynamic adjustment coefficient, with a value range of 0.4-0.6, and α is dynamically adjusted according to the product type, where the α value of technology-intensive products is 0.4-0.5, and the α value of function-intensive products is 0.5-0.6.

[0012] Furthermore, in step S4, the market constraints include market demand trend constraints, user preference weight constraints, and competitor parameter comparison constraints; the dimensions of the quantitative scoring include key feature coverage, trend matching, and cost controllability.

[0013] Furthermore, in step S5, the production feasibility verification specifically includes: material availability verification, comparing the material requirements of the design scheme with the real-time inventory data, and triggering the retrieval of material alternatives when the inventory is insufficient; process adaptability verification, using the process capability index to evaluate the matching degree between the design dimensional accuracy and the production line processing capacity based on the production process parameter data; cost controllability verification, building a cost estimation model based on the manufacturing cost data, and giving priority to adjusting high-cost features when the cost of the design scheme exceeds the threshold.

[0014] Furthermore, in step S5, the specific method of feeding back the design parameter adjustment through the timing analysis results of the after-sales operation data is to fit the performance attenuation curve through the LSTM timing prediction model and correct the design durability parameters; the association rule algorithm is used to explore the correlation between high-frequency failures and design features, and optimize the design of fault-prone components; the multi-objective optimization adopts an improved Pareto optimization algorithm, and the objective functions include maximizing the market adaptation rate, maximizing the production feasibility and maximizing the cost-effectiveness, and includes a dynamic weight adjustment mechanism, which dynamically adjusts the weight of the objective function according to the real-time data from the production end.

[0015] Furthermore, the triggering conditions of the dynamic weight adjustment mechanism include: when the production end data shows that the material inventory is insufficient, the weight of the material substitution feasibility is increased; the market adaptation rate ,in is the market characteristic weight and .

[0016] Furthermore, in step S6, , where market volatility is calculated based on the ratio of the standard deviation of competing product sales to the mean in the past 30 days, i.e. The benchmark threshold is the average evaluation score of the historical optimal solution for similar products. λ is the adjustment coefficient. The λ value for fast-moving consumer goods is 0.3, and the λ value for durable goods is 0.1. The optimization basis includes the weight ratio of key features, production feasibility analysis, and after-sales data feedback.

[0017] The present invention further provides a product design assistance system based on big data, which is used in the above-mentioned product design assistance method based on big data, comprising: Data processing module, including full-link dynamic data acquisition unit and timing preprocessing unit; Design analysis module, including feature extraction unit, demand trend analysis unit and similar product matching evaluation unit; The solution optimization module includes a production feasibility verification unit, a post-sales data feedback unit, and a multi-objective optimization unit, which interacts with the cross-domain knowledge graph; The interactive output module includes a preliminary solution generating unit, a solution evaluating unit and a human-computer interaction unit, wherein the human-computer interaction unit provides a parameter adjustment interface.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. Through the collaborative work of the full-link dynamic data acquisition module and the time series preprocessing module, comprehensive and dynamic utilization of design data is achieved. Full-link data acquisition covers product design characteristics, market feedback, real-time production data, and after-sales operation data. Combined with the deduplication, anomaly filtering, and time series trend extraction in time series preprocessing, design decisions are based on complete dynamic data patterns.

[0019] 2. The solution optimization module integrates a production feasibility verification unit and a cross-domain knowledge graph, achieving a seamless connection between design and production. The production feasibility verification unit verifies the production suitability of the solution during the design phase through material demand and inventory comparison, process capability index assessment, and cost estimation models. The cross-domain knowledge graph provides related knowledge support such as material substitution and process adaptation, enabling rapid retrieval of alternative material solutions when inventory is insufficient, effectively resolving conflicts between design goals and production conditions. 3. Through a closed-loop design combining after-sales data feedback and a multi-objective optimization unit, we achieve continuous improvement in product reliability and maximize overall benefits. The after-sales data feedback unit converts after-sales data into a basis for adjusting design parameters. The multi-objective optimization unit, based on an improved Pareto algorithm and a dynamic weight adjustment mechanism, balances market adaptability, production feasibility, and cost-effectiveness. It dynamically adjusts optimization weights when material inventory fluctuates to ensure that the solution meets market demand. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flowchart of a product design assistance system and method based on big data of the present invention; Figure 2 This is a system framework diagram of a product design assistance system and method based on big data in the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the present invention, the technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0022] Among them, the accompanying drawings are only for illustrative purposes and represent only schematic diagrams rather than actual pictures, and should not be understood as limiting the present application; in order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the accompanying drawings.

[0023] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present application. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0024] In the description of the present invention, unless otherwise expressly specified or limited, when the term "connection" or the like appears to indicate a connection relationship between components, such term should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be internal communication between two components or an interaction between two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood in specific circumstances.

[0025] like Figure 1-2 Specifically, the present invention discloses a product design assistance method based on big data, which aims to solve the pain points of "disconnection between design and production" and "difficulty in tracing after-sales problems" in traditional product design through full-link data fusion, time series analysis and multi-dimensional optimization, and improve the scientificity and feasibility of design solutions. Its core lies in driving the entire design process with data, combining time series law mining with cross-domain knowledge support, and forming a closed-loop optimization mechanism from data collection to solution output. The specific implementation process is as follows: S1, full-link dynamic data collection and time series preprocessing, collects design feature data of the target product, market feedback data, similar product data, real-time production data and after-sales operation data, and cleans, standardizes and structures the above collected data to obtain preprocessed data. At the same time, time series alignment and trend extraction are performed on the time series data with time dimension characteristics in the preprocessed data; S2, design feature and market demand analysis, extracting design feature vectors reflecting the inherent attributes of the product from the preprocessed data, combining them with the market popularity and user preference data contained in the time series data, and analyzing them through the time series attention mechanism to obtain key design features and market demand trends; S3, similar product matching and evaluation: Based on the design feature vector, calculate the feature similarity between the target product and similar products in the preprocessed data; perform functional semantic analysis based on the functional semantic descriptions in the similar product data, generate a comprehensive similarity index that integrates feature similarity and functional semantic similarity, screen out highly matching similar products, and evaluate them based on the dimensions of design reusability, market performance adaptability, and fit with market demand trends; S4, generating a preliminary design solution: generating a preliminary product design solution based on the key design features, market demand trends, and similar product evaluation results; embedding market constraints into the preliminary product design solution, and performing quantitative scoring on the preliminary design solution's fit with the key features and its trend adaptability, thus forming a verifiable solution set; S5, design solution optimization, based on real-time production data and cross-domain knowledge graphs in preprocessed data, conducts production feasibility verification and multi-objective optimization of the preliminary product design solution to resolve conflicts between the preliminary design solution and production conditions. At the same time, it combines the time series analysis results of after-sales operation data to feed back design parameter adjustments; S6, optimization scheme evaluation and output, uses dynamic thresholds to evaluate the optimized design scheme, and outputs a reference design scheme that meets the evaluation conditions and the optimization basis.

[0026] Among them, in the full-link dynamic data collection and time series preprocessing of step S1, it is necessary to collect product design feature data, market feedback data, similar product data, real-time data from the production end, and after-sales operation data, and clean, standardize, and structure the collected data; at the same time, time series alignment and trend extraction are performed on the time series data. This step is the foundation of the method. By deploying distributed data collection nodes, it realizes the real-time aggregation of multi-source heterogeneous data, covering the entire product life cycle: design feature data includes the structural parameters, functional parameters, and technical specifications of historical design schemes; market feedback data covers user evaluations, competitive product sales data, and market popularity trend data; real-time data from the production end is synchronized from the production execution system (MES) and enterprise resource planning system (ERP), including material inventory data, production process parameter data, and manufacturing cost data; after-sales operation data is obtained through built-in sensors in the product or user feedback channels, including fault record data, user usage feedback data, and performance degradation data. After data collection is complete, multi-dimensional preprocessing is performed. Text fingerprinting algorithms are used to remove duplicate data, such as repeated user reviews on e-commerce platforms. Clustering algorithms (such as K-means) are used to identify and filter out anomalous data, such as abnormal parameter values ​​in production processes that exceed the 3σ range. Unstructured data (such as user feedback text and fault description documents) is converted into structured data using the TF-IDF combined with the LDA topic model to extract key information. For time-series data (such as after-sales performance degradation data and production process parameter fluctuation data), time series decomposition algorithms (such as STL decomposition) are used to align and extract trends, separating long-term trend terms, cyclical terms, and random terms, laying the foundation for subsequent dynamic pattern analysis.

[0027] Next, step S2 involves analyzing design features and market demand. Based on the preprocessed data obtained in step S1, design feature vectors reflecting the inherent properties of the product are extracted from the preprocessed data. Combined with market popularity data and user preference data contained in the time-series data, a temporal attention mechanism is used to analyze key design features and market demand trends. The design feature vectors cover six dimensions: functional characteristics (core functional indicators, completeness of additional functions), structural characteristics (geometric dimensions, component connection methods), material characteristics (raw material type, physical properties), cost characteristics (R&D costs, material costs), user experience characteristics (ease of use, appearance satisfaction), and sustainability characteristics (energy consumption, recycling rate). The temporal attention mechanism assigns dynamic attention weights to market data from different time periods, with recent data being weighted higher than longer-term data to capture the latest demand changes. Time-series modeling identifies patterns in demand fluctuations (such as seasonal preferences and technology iteration directions), ultimately outputting key design features and demand trends.

[0028] Furthermore, in the similar product matching and evaluation of step S3, the feature similarity between the product to be designed and similar products is calculated based on the design feature vector, and the comprehensive similarity is obtained by combining the functional semantic analysis. Similar products with high reference value are screened and evaluated. Feature similarity is calculated using the Euclidean distance or cosine similarity algorithm, focusing on the consistency of core parameters. Functional semantic analysis converts the functional description text of similar products into semantic vectors through the BERT model, and calculates the cosine similarity between the vectors to obtain the functional semantic similarity. Comprehensive similarity is calculated according to the following formula: α is a dynamic adjustment coefficient. Its specific value is verified by the historical matching accuracy of similar products and ranges from 0.4 to 0.6. For technology-intensive products, α ranges from 0.4 to 0.5, focusing on feature parameter matching; for function-intensive products, α ranges from 0.5 to 0.6, focusing on functional consistency. After screening similar products with high overall similarity, an evaluation is conducted based on design reusability, market performance adaptability, and alignment with market demand trends, outputting advantageous features and areas for improvement.

[0029] The value of the dynamic adjustment coefficient α (0.4-0.5 for technology-intensive products and 0.5-0.6 for function-intensive products) is determined based on the dual logic of product core value differences and practical verification: Technology-intensive products (such as industrial inspection drones) whose core value relies on underlying technical parameters and performance indicators. For example, industrial inspection drones (which rely on technical parameters such as flight accuracy and anti-interference level) and precision medical equipment (which rely on inspection accuracy and data processing algorithms) have a technical characteristics weighted at least 60% of the product's competitiveness (such as hardware parameters and core algorithm versions). Functional descriptions serve only as a supplementary expression of technical capabilities. For such products, if α=0.4, the weight of feature similarity (technical parameter matching) is 40%, which can accurately capture core technical differences such as flight accuracy (±0.5m) and anti-interference level (Class B). Verification of 120 projects shows that the matching results are 87% consistent with the correlation of technology iteration, ensuring the technical reference value of special material process adaptation at the production end (such as carbon fiber wing processing accuracy CPK=1.5) and after-sales fault tracing (signal interruption correlation 0.92); if α is increased to 0.6, the weight of functional semantic similarity is too high (60%), and it is easy to match functional descriptions such as "intelligent detection" to low-tech parameter competitors (such as anti-interference level only Class A), resulting in failure of key technology verification.

[0030] Function-intensive products (such as smart home sweeping robots) focus on the user-friendly, directly perceptible functional experience. For example, user decisions are primarily based on functional compatibility for smart home sweeping robots (which rely on features like corner cleaning and battery life) and smart rice cookers (which rely on heating uniformity and recipe compatibility). For these products, functional features (such as "optimized corner cleaning" and "one-touch operation") have a weighting of ≥60% in influencing user choices, while technical parameters (such as motor speed) serve only as support for functional implementation. For such products, when α=0.55 (in the range of 0.5-0.6), the weight of functional semantic similarity (such as matching user needs of "edge and corner cleaning" and "intelligent obstacle avoidance") accounts for 45%. Statistics of 210 cases show that the degree of fit between functional requirements and design features is 88%, effectively screening out functional adaptation features such as "retractable side brush" and "lithium battery capacity optimization"; if α drops to 0.4, the weight of technical features (such as motor speed and battery capacity) is too high (60%), which will weaken the matching priority of high-frequency functional requirements such as "incomplete edge and corner cleaning", causing similar product references to deviate from the core user experience.

[0031] Furthermore, the preliminary design solution generated in step S4 is based on the key design features and market demand trends obtained in step S2, as well as the similar product evaluation results in step S3. Multiple preliminary product design solutions covering dimensions such as function, structure, and materials are generated for further optimization. Market constraints are embedded during solution generation, including market demand trend constraints, user preference weight constraints, and competitive product parameter comparison constraints. At the same time, the preliminary design solution is quantitatively scored for its fit with key features and trend adaptability, with scoring dimensions including key feature coverage, trend matching, and cost controllability, ultimately forming a verifiable solution set.

[0032] Furthermore, design optimization in step S5 is a key step in improving the feasibility and adaptability of the design. Based on real-time production data and a cross-domain knowledge graph from the preprocessed data, the preliminary product design is verified for production feasibility and undergoes multi-objective optimization to resolve conflicts between design goals and production conditions. Simultaneously, the timing analysis results of post-sales operation data are combined to inform design parameter adjustments.

[0033] Among them, the cross-domain knowledge graph covers material friction coefficient, supplier inventory and process adaptation rules, etc., and realizes automatic collection by connecting to supplier APIs, industry standard databases, etc.; new data is verified for compliance through the domain expert rule library to eliminate outliers; an incremental update algorithm is used to expand the graph only for newly added entities and relationships.

[0034] Among them, production feasibility verification includes three aspects: Material availability verification compares the solution material requirements with inventory data, and searches for alternative solutions through the cross-domain knowledge graph when inventory is insufficient; Process adaptability verification uses the process capability index CPK to evaluate the matching degree between design accuracy and production line capability, and CPK ≥ 1.33 is qualified; Cost controllability verification builds a cost estimation model based on manufacturing cost data, and prioritizes adjusting high-cost features when costs exceed thresholds.

[0035] Secondly, after-sales data is fed back into the LSTM time series prediction model to fit the performance attenuation curve and modify the design durability parameters. An association rule algorithm is used to explore the correlation between high-frequency failures and design features to optimize the design of fault-prone components. Multi-objective optimization uses an improved Pareto optimization algorithm. The objective functions include maximizing market adaptability, maximizing production feasibility, and maximizing cost-effectiveness. The market adaptability is calculated using the following formula: ,in is the market characteristic weight and The optimization process includes a dynamic weight adjustment mechanism - dynamically adjusting the weights of each objective function based on real-time data from the production end. When material inventory is insufficient, the weight of material substitution feasibility in the production feasibility objective function is increased.

[0036] Furthermore, the optimization scheme evaluation and output in step S6 uses a dynamic threshold to evaluate the optimized design scheme and output a reference design scheme that meets the evaluation conditions and the optimization basis. The dynamic threshold is calculated according to the following formula: ; Among them, the market volatility is calculated based on the ratio of the standard deviation of the sales volume of competing products in the past 30 days to the mean, that is, The benchmark threshold is the average evaluation score of the best historical solutions for similar products. λ is the adjustment coefficient, with λ=0.3 for fast-moving consumer goods and λ=0.1 for durable goods, to accommodate the market characteristics of different products. The final output is a reference design solution that meets the evaluation criteria and an optimization basis. This optimization basis includes the weighted proportions of key features, production feasibility analysis, and after-sales data feedback, providing comprehensive support for design decisions.

[0037] The rationality of the dynamic threshold calculation rules was verified by the demand differences and market characteristics of different product types. For example, a comparison was made between sweeping robots (a durable good, λ=0.1) and certain fast-moving consumer goods (FMCG) products (such as smart water bottles, λ=0.3). Because sweeping robots have a long purchasing decision cycle and relatively stable market demand, the ratio of the standard deviation to the mean of competing product sales over the past 30 days (market volatility) is typically low. Therefore, using λ=0.1 to calculate the dynamic threshold (dynamic threshold = baseline threshold・(1+λ・market volatility)) prevents short-term market fluctuations from excessively impacting solution evaluation and ensures continued focus on long-term value design features such as "optimized corner cleaning and improved battery life." In contrast, due to the rapid consumer decision-making and volatile market trends of FMCG products (for example, the popularity of the "social attribute function" of smart water bottles exceeds 30% per month), λ=0.3 allows for timely response to market fluctuations and rapid screening of trendy design solutions. In Example 1, the sweeping robot is a durable good. When λ is set to 0.1, the dynamic threshold is 85×(1+0.1×0.17)=86.45 points. This not only refers to the historical optimal solution (benchmark threshold) to ensure the design bottom line, but also fine-tunes the market volatility to adapt to market trend changes such as "increased attention to intelligent obstacle avoidance." This allows the solution evaluation to anchor the long-term value of the product while sensitively capturing demand iterations. Comparative verification shows that this value selection rule improves the accuracy of durable good solution evaluation, with a matching degree of 92% for trendy functions in the fast-moving consumer goods scenario. It accurately adapts to the market feedback logic of different product types, providing a scientific and quantitative basis for solution optimization.

[0038] The present invention also discloses a product design assistance system based on big data, which is used to implement the above-mentioned product design assistance method based on big data. The system is mainly composed of a data processing module, a design analysis module, a solution optimization module and an interactive output module. The modules work together to form a design assistance closed loop.

[0039] Among them, the data processing module is used to execute step S1, including a full-link dynamic data acquisition unit and a timing preprocessing unit: the full-link dynamic data acquisition unit collects full-link data in real time through a multi-source interface; the timing preprocessing unit performs cleaning, standardization and structured conversion on the collected data, and extracts trend features of the timing data through time series analysis technology to provide high-quality data for subsequent modules.

[0040] Secondly, the design analysis module is used to execute steps S2-S3, including a feature extraction unit, a demand trend analysis unit and a similar product matching evaluation unit: the feature extraction unit extracts design feature vectors covering six dimensions; the demand trend analysis unit analyzes market demand through a temporal attention mechanism model and outputs key design features; the similar product matching evaluation unit calculates feature similarity and functional semantic similarity, and screens and evaluates highly matching products according to the comprehensive similarity formula.

[0041] At the same time, the solution optimization module is the core optimization unit of the system, integrating the production feasibility verification unit, the after-sales data feedback unit and the multi-objective optimization unit, and is used to execute step S5: the production feasibility verification unit verifies the production feasibility through material inventory comparison, process capability index evaluation and cost model; the after-sales data feedback unit outputs design parameter adjustment suggestions through the time series prediction model and association rule algorithm; the multi-objective optimization unit adopts the improved Pareto optimization algorithm, with the goal of maximizing market adaptation rate, production feasibility and cost-effectiveness, and adapts to production data changes through a dynamic weight adjustment mechanism.

[0042] Finally, the interactive output module is used to execute steps S4-S6, including a preliminary solution generation unit, a solution evaluation unit and a human-computer interaction unit: the preliminary solution generation unit generates a preliminary solution embedded with market constraints and quantifies the score; the solution evaluation unit uses a dynamic threshold to screen the qualified solutions; the human-computer interaction unit outputs the reference solution and optimization basis, and provides a parameter adjustment interface to support manual optimization.

[0043] Example 1: Taking the design optimization of a household intelligent sweeping robot as an example, the implementation process of the present invention is further explained: During the full-link data collection and time series preprocessing phase, the data processing module collects data from multiple sources: design feature data includes suction power, battery life, and navigation accuracy of existing models; market feedback data covers user reviews on e-commerce platforms over the past year, competitive product sales rankings, and feature search popularity trends; production-end data synchronized from the factory system includes lithium battery inventory, motor assembly process parameters, and unit manufacturing cost targets; after-sales data is obtained through the product app, including fault records, performance degradation data, and user usage scenario distribution. During preprocessing, duplicate reviews are removed, abnormal process parameters are filtered, and user reviews are converted into structured topics; after-sales performance degradation data is decomposed into time series to extract long-term trends, cyclical terms, and random terms to provide data support for subsequent analysis.

[0044] During the design feature and requirements analysis phase, the design analysis module extracted design feature vectors covering six key dimensions: function, structure, materials, cost, user experience, and sustainability. Using a temporal attention mechanism to dynamically weight market data, the analysis identified corner cleaning, battery life, obstacle avoidance accuracy, and humid environment protection as key design features. In the past three months, the user attention weights for corner cleaning and intelligent obstacle avoidance have increased to 0.35 and 0.28, respectively.

[0045] During the similar product matching and evaluation phase, 15 competing products in the same price range were selected and their feature similarity and functional semantic similarity were calculated. Because robot vacuums are function-intensive products, a dynamic adjustment coefficient of 0.55 was used. Two highly compatible reference products were selected based on the comprehensive similarity formula. Their retractable side brushes, optimized lithium battery capacity, and waterproof rating were identified as reusable advantages.

[0046] During the initial design phase, three sets of solutions were generated based on key features and evaluation results of similar products: a performance-first solution with enhanced suction and battery life, an experience-first solution with optimized side brushes and obstacle avoidance algorithms, and a cost-first solution using a single battery and a high-efficiency motor. After embedding market constraints, quantitative scoring was performed, and the experience-first and cost-first solutions formed a verifiable set.

[0047] During the design optimization phase, production feasibility verification revealed insufficient inventory of side brush materials for the experience-first solution. Equivalent alternative materials were retrieved through a cross-domain knowledge graph. Process compatibility verification confirmed acceptable dimensional accuracy. Cost verification ensured cost remained within the required threshold by replacing sensors. During after-sales data feedback, a time series model was used to fit the attenuation curve, suggesting an upgrade to the battery protection level. Association rule analysis revealed that drive wheel sticking was related to the material, leading to a recommendation to replace it with a wear-resistant material. A modified Pareto algorithm was used for multi-objective optimization. Due to limited lithium battery inventory, dynamic weighting adjustments were performed to increase the production feasibility weight, ultimately resulting in the experience-first solution receiving the highest overall score.

[0048] During the optimization solution evaluation and output phase, the calculated dynamic threshold was 86.45 points, and the experience-first solution evaluation score of 89.3 met the target. A reference design solution and optimization rationale were generated, including key feature weights, production feasibility analysis, and after-sales feedback instructions. Designers adjusted the side brush extension length through an interactive interface, and the system updated the parameters in real time to complete the final design.

[0049] It can be seen from this embodiment that the present invention achieves precise adaptation of design solutions to market demands and production conditions through full-link data-driven and multi-dimensional optimization, verifying the practicality of the method and system.

[0050] The above are only embodiments of the present invention, and the circuits, electronic components and modules involved are all prior art, which can be fully implemented by those skilled in the art. It is needless to say that the content protected by this application does not involve improvements to software and methods. Common knowledge such as the specific structures and characteristics known in the scheme are not described in detail here. Ordinary technicians in the field are aware of all common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all prior art in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the field can improve and implement this scheme in combination with their own abilities under the inspiration given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A product design assistance method based on big data, characterized by: The following steps are involved: S1, full-link dynamic data collection and time series preprocessing, collects design feature data of the target product, market feedback data, similar product data, real-time production data and after-sales operation data, and cleans, standardizes and structures the above collected data to obtain preprocessed data. At the same time, time series alignment and trend extraction are performed on the time series data with time dimension characteristics in the preprocessed data; S2, design feature and market demand analysis, extracting design feature vectors reflecting the inherent attributes of the product from the preprocessed data, combining them with the market popularity and user preference data contained in the time series data, and analyzing them through the time series attention mechanism to obtain key design features and market demand trends; S3, similar product matching and evaluation: Based on the design feature vector, calculate the feature similarity between the target product and similar products in the preprocessed data; perform functional semantic analysis based on the functional semantic descriptions in the similar product data, generate a comprehensive similarity index that integrates feature similarity and functional semantic similarity, screen out highly matching similar products, and evaluate them based on the dimensions of design reusability, market performance adaptability, and fit with market demand trends; S4, generating a preliminary design solution: generating a preliminary product design solution based on the key design features, market demand trends, and similar product evaluation results; embedding market constraints into the preliminary product design solution, and performing quantitative scoring on the preliminary design solution's fit with the key features and its trend adaptability, thus forming a verifiable solution set; S5, design optimization, based on real-time production data and cross-domain knowledge graphs in pre-processed data, conducts production feasibility verification and multi-objective optimization on the preliminary product design to resolve conflicts between the preliminary design and production conditions; At the same time, the timing analysis results of after-sales operation data are combined to feed back the design parameter adjustments; S6, optimization scheme evaluation and output, uses dynamic thresholds to evaluate the optimized design scheme, and outputs a reference design scheme that meets the evaluation conditions and the optimization basis.

2. The product design assistance method based on big data according to claim 1, characterized in that: In step S1, the design feature data includes product structure parameters, functional parameters, and historical design scheme parameters; the market feedback data includes user reviews, competitive product sales data, and market popularity trend data; the production end real-time data includes material inventory data, production process parameter data, and manufacturing cost data; The after-sales operation data includes fault record data, user feedback data and performance degradation data; The trend extraction of the time series data adopts a time series decomposition algorithm to separate long-term trend items, periodic items and random items; the preprocessing includes: using a text fingerprint algorithm to remove duplicate data, identifying and filtering abnormal data through a clustering algorithm, and using TF-IDF combined with the LDA topic model to extract key information from unstructured data and perform structured conversion.

3. The product design assistance method based on big data according to claim 1, characterized in that: In step S2, the design feature vector includes functional features, structural features, material features, cost features, user experience features and sustainability features; the temporal attention mechanism assigns dynamic attention weights to market data in different time periods, with recent data having a higher weight than long-term data.

4. The product design assistance method based on big data according to claim 1, characterized in that: In step S3, the functional semantic analysis converts the functional description into a semantic vector through the BERT model and then calculates the cosine similarity; the calculation formula of the comprehensive similarity is: , where α is the dynamic adjustment coefficient, with a value range of 0.4-0.6, and α is dynamically adjusted according to the product type, where the α value of technology-intensive products is 0.4-0.5, and the α value of function-intensive products is 0.5-0.

6.

5. The product design assistance method based on big data according to claim 1, characterized in that: In step S4, the market constraints include market demand trend constraints, user preference weight constraints, and competitor parameter comparison constraints; the dimensions of the quantitative scoring include key feature coverage, trend matching, and cost controllability.

6. The product design assistance method based on big data according to claim 1, characterized in that: In step S5, the production feasibility verification specifically includes: material availability verification, comparing the material requirements of the design scheme with the real-time inventory data, and triggering the retrieval of material alternatives when the inventory is insufficient; process adaptability verification, using the process capability index to evaluate the matching degree between the design dimensional accuracy and the production line processing capacity based on the production process parameter data; cost controllability verification, building a cost estimation model based on the manufacturing cost data, and giving priority to adjusting high-cost features when the cost of the design scheme exceeds the threshold.

7. The product design assistance method based on big data according to claim 1, characterized in that: In step S5, the specific method of feeding back the design parameter adjustment through the timing analysis results of the after-sales operation data is to fit the performance attenuation curve through the LSTM timing prediction model and correct the design durability parameters; the association rule algorithm is used to explore the correlation between high-frequency failures and design features, and optimize the design of fault-prone components; the multi-objective optimization adopts an improved Pareto optimization algorithm, and the objective functions include maximizing the market adaptation rate, maximizing the production feasibility and maximizing the cost-effectiveness, and includes a dynamic weight adjustment mechanism, which dynamically adjusts the weight of the objective function according to the real-time data from the production end.

8. The product design assistance method based on big data according to claim 7, characterized in that: The triggering conditions of the dynamic weight adjustment mechanism include: when the production end data shows that the material inventory is insufficient, the material substitution feasibility weight is increased; the market adaptation rate ,in is the market characteristic weight and .

9. The product design assistance method based on big data according to claim 1, characterized in that: In step S6, , where market volatility is calculated based on the ratio of the standard deviation of competing product sales to the mean in the past 30 days, i.e. The benchmark threshold is the average evaluation score of the historical optimal solution for similar products. λ is the adjustment coefficient. The λ value for fast-moving consumer goods is 0.3, and the λ value for durable goods is 0.

1. The optimization basis includes the weight ratio of key features, production feasibility analysis, and after-sales data feedback.

10. A product design assistance system based on big data, used in the product design assistance method based on big data according to any one of claims 1 to 9, characterized in that: include: Data processing module, including full-link dynamic data acquisition unit and timing preprocessing unit; Design analysis module, including feature extraction unit, demand trend analysis unit and similar product matching evaluation unit; The solution optimization module includes a production feasibility verification unit, a post-sales data feedback unit, and a multi-objective optimization unit, which interacts with the cross-domain knowledge graph; The interactive output module includes a preliminary solution generating unit, a solution evaluating unit and a human-computer interaction unit, wherein the human-computer interaction unit provides a parameter adjustment interface.

Images