Vehicle body selection design method and system based on perceptual engineering

Abstract

The invention provides a perceptual engineering-based vehicle body selection design method. The perceptual engineering-based vehicle body selection design method comprises the following steps of: acquiring vehicle body data and comment information of a whole vehicle from data permitted to be accessed; analyzing and sorting the obtained comment text data by using a natural language processing technology to form a database; collecting user expectations, and performing semantic reasoning and analysis on the collected user expectations based on the formed database to obtain product performance semantic weights of the user expectations; and performing quantitative matching on the obtained product performance semantic weight expected by the user and the obtained vehicle body data by using the perceptual engineering model, and determining the vehicle body data corresponding to the vehicle model most conforming to the user expectation. And finally, the product with the highest customer expectation score is determined to correspondingly obtain the target product, the designed optimal product scheme is determined, and the vehicle body design efficiency and reliability are improved.

Description

Technical Field

[0001] This application belongs to the field of vehicle body design, and in particular relates to a vehicle body selection design method and system based on kinetic engineering. Background Technology

[0002] With the rapid pace of automotive updates and iterations, not only traditional automakers but also many emerging players have joined the fray. As technology advances rapidly and consumers pursue personalization, only by shortening the automotive development cycle as much as possible can a company stand out in the fierce market competition. Therefore, it is crucial to improve automotive design efficiency while meeting consumers' personalized needs. Summary of the Invention

[0003] This application provides a vehicle body selection design method and system based on kinetic engineering, which can solve the problem in existing methods of how to improve the efficiency of vehicle design while meeting consumers' personalized needs for vehicles.

[0004] Firstly, this application provides a vehicle body selection design method based on kinetic engineering, including: Obtain the vehicle's body data and comment information from the data that is granted access; Natural language processing technology is used to parse and organize the acquired comment text data to form a database; User expectations are collected, and semantic reasoning and analysis are performed on the collected user expectations based on the formed database to obtain the semantic weights of product performance expected by users. By using the Kansei Engineering model, the semantic weights of the product performance expected by users are quantitatively matched with the acquired vehicle body data to determine the vehicle body data corresponding to the model that best matches the user's expectations.

[0005] Optionally, the step of using natural language processing technology to parse and organize the acquired comment text data to form a database includes: Perform text preprocessing on the comment text data; Based on the text preprocessing results, feature extraction and structural analysis are performed on the text; Based on the results of feature extraction and structural analysis, the text is classified and clustered. Based on the classification and clustering results, association analysis and pattern recognition are performed; Post-processing of the association analysis and pattern recognition results yields parsed and organized text data; The parsed and organized text data is stored in a preset manner to form the database.

[0006] Optionally, the process of collecting user expectations and performing semantic reasoning and analysis on the collected user expectations based on the formed database to obtain the semantic weights of product performance based on user expectations includes: The user's expressed expectations are subjected to text mining processing against the database, and the processed user expectation text is output. Based on the processed user expectation text, identify key words for product performance, evaluate and analyze product performance based on these key words, and output the performance evaluation and analysis results. Using the semantic differential method, a semantic evaluation quantification table for the designed product is constructed based on the performance evaluation analysis results, and the semantic evaluation quantification table is output. Based on the semantic evaluation quantification table and user language expression, we can mine the user's expected priority of product performance and assign expected scores to obtain the semantic weight of the product performance expected by the user.

[0007] Optionally, the process of performing text mining processing on the user's expressed expectations and the database to output processed user expectation text includes: The user's expressed expectations are compared with the text data in the database using word segmentation. Based on the word segmentation results, a thesaurus is constructed; Based on the constructed thesaurus, the segmented text data is assigned categories; Integrate the category assignment results and output the processed user-expected text.

[0008] Optionally, based on the processed user expectation text, the process identifies key words representing product performance, evaluates and analyzes product performance based on these key words, and outputs performance evaluation and analysis results, including: From the processed user expectation text, filter and identify key words that characterize product performance; Based on the aforementioned key words, the number of positive and negative reviews of the product performance corresponding to each key word is counted. The total number of evaluations of product performance corresponding to the key term is counted; The evaluation weight of the product's performance is determined based on the number of positive reviews, the number of negative reviews, and the total number of evaluations. The performance evaluation weights of all products are integrated, and the performance evaluation analysis results are output.

[0009] Optionally, based on the semantic evaluation quantification table and user language expression, the process of mining the user's expected priority of product performance and assigning expected scores to obtain the semantic weight of the user's expected product performance includes: Based on the language users use to express their expectations, summarize the product performance aspects that users focus on. Based on the degree of user concern regarding the performance of each product, the performance aspects of the products are prioritized. Assign a desired score to each priority level for product performance aspects, with the desired score ranging from 0 to 1, and assigning a score of 0 to product performance aspects not mentioned by the user. Adjust the expected scores for each product's performance aspects so that the sum of the expected scores for all product performance aspects is 1; Based on the adjusted expected score, determine and output the semantic weights of the product performance expected by the user.

[0010] Optionally, the step of using a sensory engineering model to quantitatively match the semantic weights of the product performance expected by the user with the acquired vehicle body data to determine the vehicle body data corresponding to the model that best matches the user's expectations, as the basis for vehicle body design, includes: Define a quantitative model relating each vehicle model to its various performance subjective evaluations; Based on the emphasis of user input expectations and the purpose requirements of vehicle body design, and combined with the semantic weights of product performance obtained from user expectations, the weights of each product performance are determined. Based on the defined quantitative model and the determined weights of each product performance, the correlation coefficient between existing vehicle models and user expectations is calculated. The preference correlation coefficient is verified by multiplying the quantification results of each vehicle model in the semantic evaluation quantification table with the determined weights of each product performance and then summing the results. Select the car models with the highest preference correlation coefficient, extract the corresponding vehicle body data for each model, and output it.

[0011] Optionally, the quantitative model defining the relationship between each vehicle model and its various performance subjective evaluations includes: The design elements of the vehicle model are used as classification-independent parameters, and dummy variables are defined to represent these classification-independent parameters. Analyze the degree of influence of each of the classification independent parameters on the vehicle product performance; Based on the degree of influence of each of the aforementioned independent parameters on the vehicle product performance, the intrinsic relationship between design elements and product performance is determined; Based on the aforementioned inherent relationships, a quantitative model is constructed and output regarding the relationship between vehicle models and various subjective performance evaluations.

[0012] Optionally, the calculation of the correlation coefficient between existing vehicle models and user preferences based on the defined quantitative model and the determined weights of each product performance includes: From the acquired vehicle body data, extract the design element data of existing vehicle models; The extracted design element data is correlated with the user's emotional expectation score to determine the degree of correlation between the design element data and the user's emotional expectation score. Based on the degree of correlation, calculate the correlation coefficient of user expectation preferences for each vehicle model. Output the correlation coefficients of user expectations and preferences for all vehicle models.

[0013] Secondly, this application provides a sensible ergonomic vehicle body selection design system, comprising: The acquisition module retrieves the vehicle's body data and comment information from the data that is allowed to be accessed. The database module uses natural language processing technology to parse and organize the acquired comment text data to form a database. The reasoning module collects user expectations and performs semantic reasoning and analysis on the collected user expectations based on the formed database to obtain the semantic weights of the product performance expected by the user. The matching module uses a sensory engineering model to quantitatively match the semantic weights of the product performance expected by the user with the acquired vehicle body data, and determines the vehicle body data corresponding to the model that best matches the user's expectations.

[0014] As can be seen from the above technical solution, the vehicle body selection design method and system based on kanji science provided in this application adopts a combination of text analysis and kanji science for comprehensive analysis. Based on the established model, the target product corresponding to the product with the highest customer expectation score is finally determined, and the optimal product solution for design is determined, thereby improving the efficiency and reliability of vehicle body design. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a vehicle body selection design method based on Kansei Engineering in an embodiment of this application.

[0017] Figure 2 This is a schematic diagram of a vehicle body selection design system based on kinetic engineering, as described in an embodiment of this application. Detailed Implementation

[0018] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not limiting, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without such specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0019] This application provides a vehicle body selection design method based on kinetic engineering, such as... Figure 1 As shown, it includes: 101: Obtain the vehicle's body data and comment information from the data that is granted access; 102: Utilize natural language processing technology to parse and organize the acquired comment text data to form a database; 103: Collect user expectations and perform semantic reasoning and analysis on the collected user expectations based on the formed database to obtain the semantic weight of product performance based on user expectations; 104: Using the Kansei Engineering model, the semantic weights of the product performance expected by the user are quantitatively matched with the acquired vehicle body data to determine the vehicle body data corresponding to the model that best matches the user's expectations.

[0020] This application provides a vehicle body selection design method based on kanji science. It uses a combination of text analysis and kanji science for comprehensive analysis. Based on the established model, it finally determines the target product corresponding to the product with the highest customer expectation score, and determines the optimal product solution for the design, thereby improving the efficiency and reliability of vehicle body design.

[0021] In this application embodiment, a vehicle body selection design method based on kinetic engineering is proposed. Its core concept is to transform the user's abstract kinetic expectations into specific basic data for vehicle body design through a closed-loop process of data acquisition, text parsing, semantic reasoning, and quantitative matching.

[0022] The following provides a detailed description of this application. In this embodiment, the data accessed under license is from data providers such as automobile manufacturers, authorized automotive data platforms, and compliant user review platforms, which have explicitly authorized access to such data sources. It should be noted that licensed access must be implemented through an authentication mechanism, such as API interface key authentication and data access agreement signing, to ensure the compliance of data acquisition. This application is not limited to this; the method of data acquisition can be flexibly adjusted according to the data source type. It can utilize automated tools to selectively scrape publicly available data from authorized platforms, or it can receive survey data actively submitted by users. As long as the data acquisition process has clear licensing grounds and the data content is strongly related to vehicle body design, it falls within the scope of this method.

[0023] Comments are textual feedback from users on their experience with a car, which can be found in the comment sections of car forums and car review platforms. These comments can include user evaluations of aspects such as the car's space, appearance, comfort, and safety.

[0024] In the quantitative matching stage using the sensory engineering model, in this embodiment, the semantic weight of the product performance expected by the user can be used as a demand benchmark. The vehicle body data of each model can be substituted into the model to calculate the degree of fit between the model and the user's expectations. For example, if the user has a high semantic weight for spatial performance, the model can prioritize space-related parameters such as the wheelbase and seat spacing of the vehicle to ensure that the matching result meets the user's core needs.

[0025] Optionally, the step of using natural language processing technology to parse and organize the acquired comment text data to form a database includes: Perform text preprocessing on the comment text data; Based on the text preprocessing results, feature extraction and structural analysis are performed on the text; Based on the results of feature extraction and structural analysis, the text is classified and clustered. Based on the classification and clustering results, association analysis and pattern recognition are performed; Post-processing of the association analysis and pattern recognition results yields parsed and organized text data; The parsed and organized text data is stored in a preset manner to form the database.

[0026] In this embodiment of the application, text preprocessing of the comment text data can eliminate data noise. For example, during preprocessing, meaningless special symbols such as @#￥ and repeated interjections such as consecutive "hahaha" can be filtered out, and the text can be uniformly encoded to avoid parsing errors caused by differences in encoding formats. This application does not impose any restrictions on this.

[0027] Based on the text preprocessing results, when performing feature extraction and structural analysis on the text, in this embodiment of the application, feature extraction can focus on core words related to vehicle body performance, such as wheelbase, sound insulation, and fastback design. Information that is of reference value to vehicle body design can be selected through word frequency statistics, keyword weight calculation, and other methods. Structural analysis can clarify the relationship between evaluation objects and evaluation tendencies in the text through syntactic analysis technology, such as rear seat space, front sound insulation effect, and negative effects.

[0028] Optionally, the process of collecting user expectations and performing semantic reasoning and analysis on the collected user expectations based on the formed database to obtain the semantic weights of product performance based on user expectations includes: The user's expressed expectations are subjected to text mining processing against the database, and the processed user expectation text is output. Based on the processed user expectation text, identify key words for product performance, evaluate and analyze product performance based on these key words, and output the performance evaluation and analysis results. Using the semantic differential method, a semantic evaluation quantification table for the designed product is constructed based on the performance evaluation analysis results, and the semantic evaluation quantification table is output. Based on the semantic evaluation quantification table and user language expression, we can mine the user's expected priority of product performance and assign expected scores to obtain the semantic weight of the product performance expected by the user.

[0029] In this embodiment of the application, user expectations are collected, and semantic reasoning and analysis are performed on the collected user expectations based on the formed database to obtain the semantic weight of the product performance of the user expectations. This can transform the user's abstract emotional expectations, such as wanting a car with good space, into quantifiable performance weights.

[0030] In this embodiment of the application, when collecting user expectations, a variety of interaction methods can be provided, such as text input boxes, voice input to text conversion, and demand survey questionnaires, to facilitate users in expressing their needs.

[0031] In this embodiment of the application, when the user's expressed expectations are processed by text mining against the database and the processed user expectation text is output, keywords in the user's expectations, such as spatial sound insulation, can be matched with the core feature words of each performance dimension in the database to clarify the vehicle body performance corresponding to the user's expectations. At the same time, the user's vague expression can be completed based on the association rules in the database, such as completing "spacious" as "ample rear legroom", to ensure that the processed user expectation text has a clear performance orientation.

[0032] Based on the semantic evaluation quantification table and user language expression, this embodiment of the application mines the user's expected priority of product performance and assigns expected scores. Priority can be determined by analyzing the language characteristics of the user's expression of expectations, such as keyword frequency, use of emphatic words, and order of expression. For example, performance with high frequency of mention, use of emphatic tones like "most important" or "must be met," and appearing early in the expression order have higher priority. For instance, if a user repeatedly mentions space and says they value space most, then space performance has a higher priority than other performance aspects. When assigning expected scores, a rule can be followed where the score range is 0-1, and performance not mentioned by the user is assigned a score of 0. Further adjustments can be made to ensure that the sum of the expected scores for all performance aspects is 1. The final set of scores represents the semantic weight of the product performance expected by the user.

[0033] Optionally, the process of performing text mining processing on the user's expressed expectations and the database to output processed user expectation text includes: The user's expressed expectations are compared with the text data in the database using word segmentation. Based on the word segmentation results, a thesaurus is constructed; Based on the constructed thesaurus, the segmented text data is assigned categories; Integrate the category assignment results and output the processed user-expected text.

[0034] In this embodiment of the application, a sub-process is performed to mine the user's expressed expectations and the database, and output the processed user expectation text. This process associates the user's expectations with the database text, eliminates discrepancies and ambiguities in expression, and provides a clear textual basis for the subsequent determination of key words.

[0035] In this embodiment, when performing word segmentation on the user's expressed expectations and the text data in the database, a professional word segmentation dictionary adapted to the automotive body field can be used to ensure that professional terms such as wheelbase, sloping roof angle, and seat spacing are not mis-segmented, while filtering out stop words without actual semantic meaning, retaining core vocabulary. It should be noted that the word segmentation process can also accommodate mixed Chinese and English expressions, such as a user inputting "I want a spacious SUV," ensuring that words like "SUV space" are accurately segmented to avoid semantic loss. For example, the user's expectation text "I want a car with spacious rear seats" can be segmented into "I want a car with spacious rear seats," and then stop words like "I want a car" can be filtered out, retaining the core vocabulary of "spacious rear seats."

[0036] Based on the word segmentation results, when constructing a thesaurus, in this embodiment of the application, words with the same or similar meanings can be grouped into the same category, unifying the expression format of users and database text, and solving the problem of different expressions for the same need. For example, words like "large space," "spacious space," and "sufficient space" can be grouped into one synonym group, while words like "good sound insulation," "excellent sound insulation effect," and "soundproofing stick" can be grouped into another synonym group, ensuring that the same need is not misjudged as different needs during subsequent analysis. It should be noted that the thesaurus can be updated regularly to include newly emerging user expressions, ensuring coverage of diverse user expressions if space is not limited.

[0037] Based on the constructed thesaurus, when assigning categories to the segmented text data, in this embodiment of the application, a mapping relationship between synonym groups and vehicle body performance categories can be established first. For example, the synonym group "large / spacious" corresponds to space performance, and "good / excellent sound insulation" corresponds to comfort performance. Then, the segmented user expectation text and the database text are assigned to the corresponding performance categories according to the mapping relationship.

[0038] Optionally, based on the processed user expectation text, the process identifies key words representing product performance, evaluates and analyzes product performance based on these key words, and outputs performance evaluation and analysis results, including: From the processed user expectation text, filter and identify key words that characterize product performance; Based on the aforementioned key words, the number of positive and negative reviews of the product performance corresponding to each key word is counted. The total number of evaluations of product performance corresponding to the key term is counted; The evaluation weight of the product's performance is determined based on the number of positive reviews, the number of negative reviews, and the total number of evaluations. The performance evaluation weights of all products are integrated, and the performance evaluation analysis results are output.

[0039] In this embodiment of the application, based on the processed user expectation text, key words for product performance are identified, and the product performance is evaluated and analyzed based on these key words, outputting the performance evaluation and analysis results. The core of this process is to clarify the specific performance points that users care about, and to quantify the user evaluation of these performances in conjunction with database comments, providing a basis for the subsequent construction of a semantic evaluation quantification table.

[0040] Based on the aforementioned key words, when counting the number of positive and negative comments on the product performance corresponding to the key words, in this embodiment of the application, all comment texts containing the key words in the database can be retrieved, and positive and negative expressions can be distinguished by sentiment analysis technology. For example, comments containing positive words such as spacious, ample, and good are classified as positive comments, while comments containing negative words such as cramped, oppressive, and poor are classified as negative comments.

[0041] When calculating the total number of evaluations for a product performance corresponding to a specific keyword, in this embodiment, the total number of evaluations refers to the number of all comments containing that keyword, including positive, negative, and ambiguous comments. This reflects the level of user attention to that performance point and the sample size of the evaluations. It should be noted that a minimum threshold for the total number of evaluations can be set. If the total number of evaluations for a certain keyword is lower than the threshold, it can be determined that the sample size for that performance point is insufficient, requiring supplementary database searches or an expanded search scope to ensure the reliability of subsequent evaluation weight calculations.

[0042] When determining the evaluation weight of a product's performance based on the number of positive comments, negative comments, and total number of evaluations, in this embodiment of the application, the evaluation weight can comprehensively consider the proportion of positive comments and the total number of evaluations. For example, the higher the proportion of positive comments and the more total number of evaluations, the higher the evaluation weight, and vice versa. For instance, a keyword with a high number of positive comments and a high total number of evaluations may have a significantly higher evaluation weight than a keyword with a high proportion of positive comments but a low total number of evaluations.

[0043] Optionally, based on the semantic evaluation quantification table and user language expression, the process of mining the user's expected priority of product performance and assigning expected scores to obtain the semantic weight of the user's expected product performance includes: Based on the language users use to express their expectations, summarize the product performance aspects that users focus on. Based on the degree of user concern regarding the performance of each product, the performance aspects of the products are prioritized. Assign a desired score to each priority level for product performance aspects, with the desired score ranging from 0 to 1, and assigning a score of 0 to product performance aspects not mentioned by the user. Adjust the expected scores for each product's performance aspects so that the sum of the expected scores for all product performance aspects is 1; Based on the adjusted expected score, determine and output the semantic weights of the product performance expected by the user.

[0044] In this embodiment, based on the semantic evaluation quantification table and the user's language expression, the user's expected priority of product performance is mined and assigned expected scores to obtain the semantic weight of the product performance expected by the user. This process can convert the user's emphasis on each performance into a calculable score, ensuring that the performance that the user cares about most can be given priority in subsequent quantitative matching.

[0045] In this embodiment, when summarizing the product performance aspects that users prioritize based on their language when expressing expectations, keyword extraction and language feature analysis can be used to identify performance aspects that users repeatedly mention and emphasize. For example, if a user mentions space multiple times in expressing expectations and uses emphatic words such as "most important and must be met," space performance can be summarized as a performance aspect that the user prioritizes. If the user only briefly mentions sound insulation without emphasis, comfort performance can be classified as a secondary priority. It should be noted that the summarization process can be combined with the processed user expectation text to avoid omitting implicit key needs of the user. For example, if a user mentions family travel, space performance may be implicitly considered a priority.

[0046] Specifically, when prioritizing product performance aspects based on users' level of attention to each aspect, this embodiment of the application can employ a multi-dimensional comprehensive ranking method. This method uses keyword frequency as a basis, combined with the use of emphatic words and the order of expression to adjust priority: performance aspects with higher frequency of mention, more emphatic usage, and earlier appearance in the order of expression receive higher priority. For example, if a user mentions space first, then sound insulation, and finally appearance, the priority ranking would be: space performance > comfort performance (sound insulation) > appearance performance.

[0047] Optionally, the step of using a sensory engineering model to quantitatively match the semantic weights of the product performance expected by the user with the acquired vehicle body data to determine the vehicle body data corresponding to the model that best matches the user's expectations, as the basis for vehicle body design, includes: Define a quantitative model relating each vehicle model to its various performance subjective evaluations; Based on the emphasis of user input expectations and the purpose requirements of vehicle body design, and combined with the semantic weights of product performance obtained from user expectations, the weights of each product performance are determined. Based on the defined quantitative model and the determined weights of each product performance, the correlation coefficient between existing vehicle models and user expectations is calculated. The preference correlation coefficient is verified by multiplying the quantification results of each vehicle model in the semantic evaluation quantification table with the determined weights of each product performance and then summing the results. Select the car models with the highest preference correlation coefficient, extract the corresponding vehicle body data for each model, and output it.

[0048] In this embodiment, the semantic weights of the product performance expected by the user are quantitatively matched with the acquired vehicle body data using a sensible engineering model to determine the vehicle body data corresponding to the model that best matches the user's expectations. Specifically, when defining the quantitative model between each vehicle model and its various performance sensible evaluations, the vehicle body design elements can be used as core parameters affecting performance. By analyzing the correlation between these parameters and the performance sensible evaluations, a mathematical model is constructed.

[0049] In this embodiment, all models can be sorted in descending order of preference correlation coefficient, and the model with the highest coefficient can be selected as the model that best matches the user's expectations. Then, the core body data of the model can be extracted, including structural parameters and performance-related parameters, such as the sloping roof angle and seat spacing, and organized into a structured document output.

[0050] Optionally, the quantitative model defining the relationship between each vehicle model and its various performance subjective evaluations includes: The design elements of the vehicle model are used as classification-independent parameters, and dummy variables are defined to represent these classification-independent parameters. Analyze the degree of influence of each of the classification independent parameters on the vehicle product performance; Based on the degree of influence of each of the aforementioned independent parameters on the vehicle product performance, the intrinsic relationship between design elements and product performance is determined; Based on the aforementioned inherent relationships, a quantitative model is constructed and output regarding the relationship between vehicle models and various subjective performance evaluations.

[0051] In this embodiment, when design elements of the vehicle model are used as category-independent parameters and dummy variables are defined to represent these parameters, design elements directly related to vehicle body performance can be selected as category-independent parameters to avoid including non-core elements, such as paint color, which can be adjusted later and has little impact on performance. Dummy variables are numerical identifiers used to represent different states of design elements. For example, dummy variable X1 represents wheelbase, where X1=1 represents a wheelbase within a certain range, and X1=2 represents another range. X2 represents the sloping roof angle, where X2=1 represents a small sloping roof angle, and X2=2 represents a large sloping roof angle, ensuring that design elements can be used in model calculations. It should be noted that the definition of dummy variables can be combined with the engineering feasibility range of the design elements to avoid parameter states exceeding actual mass production capabilities.

[0052] Based on the degree of influence of each of the aforementioned independent parameters on the vehicle's performance, when determining the intrinsic relationship between design elements and product performance, this application embodiment can summarize the pattern of changes in design element parameters leading to changes in perceived performance. For example, it can be determined that an increase in wheelbase leads to improved perceived space performance, a decrease in the sloping roof angle leads to improved perceived headroom performance, and an increase in sound insulation thickness leads to improved perceived comfort performance. Simultaneously, it can also identify the correlated effects between design elements, such as an increase in wheelbase potentially leading to an increase in turning radius, providing constraints for model construction.

[0053] Based on the aforementioned inherent connections, when constructing and outputting a quantitative model of the relationship between vehicle models and various performance subjective evaluations, in this embodiment of the application, the dummy variables and influence weights of design elements can be substituted into mathematical expressions to form a quantitative model. For example, the spatial performance subjective evaluation score = X1 × W1 + X3 × W3, where X1 is the wheelbase dummy variable, W1 is the wheelbase influence weight; X3 is the seat spacing dummy variable, W3 is the seat spacing influence weight; the comfort performance subjective evaluation score = X4 × W4, where X4 is the sound insulation cotton thickness dummy variable, W4 is the sound insulation cotton thickness influence weight. When outputting the model, a parameter description and influence weight table constraint can be included, clarifying the meaning of each variable, the basis for determining the weight values, and the engineering feasibility range of the parameters, facilitating accurate retrieval during subsequent calculations. It should be noted that the quantitative model can be periodically optimized, incorporating new vehicle body design data and user evaluations to update the influence weights, ensuring the model's timeliness and accuracy.

[0054] Optionally, the calculation of the correlation coefficient between existing vehicle models and user preferences based on the defined quantitative model and the determined weights of each product performance includes: From the acquired vehicle body data, extract the design element data of existing vehicle models; The extracted design element data is correlated with the user's emotional expectation score to determine the degree of correlation between the design element data and the user's emotional expectation score. Based on the degree of correlation, calculate the correlation coefficient of user expectation preferences for each vehicle model. Output the correlation coefficients of user expectations and preferences for all vehicle models.

[0055] In this embodiment, the user's perceptual expectation score is based on the performance expectation targets determined by the user's semantic weights. For example, if the user's expectation score for spatial performance is level 4, correlation analysis can determine whether the design element data of a certain car model can meet this expectation target. If the spatial performance score corresponding to the wheelbase data of the car model reaches level 4, the correlation is high; if it only reaches level 2, the correlation is low. For example, if the user's expectation score for spatial performance is level 4, and the wheelbase data of a certain car model is substituted into the quantification model, the spatial performance score is level 4, indicating that the wheelbase of the car model has a high correlation with the user's perceptual expectation; the spatial performance score of another car model is level 3, indicating a moderate correlation.

[0056] Based on the degree of correlation, when calculating the user expectation preference correlation coefficient for each vehicle model, in this embodiment of the application, the degree of correlation can first be converted into scores for each performance, such as 4-5 points for high correlation, 3 points for medium correlation, and 1-2 points for low correlation. Then, each performance score is multiplied by its corresponding performance weight, and the sum is obtained to obtain the preference correlation coefficient.

[0057] Furthermore, current technologies for using public opinion data in automotive body design lack the technology to automatically analyze comment sections such as forums. However, the results obtained by analyzing comment sections are actually more important for body design, especially comment sections with final discussion results. The deepening of debates in these sections can guide other users' opinions and attract more users to agree with the views. Based on this, this application further provides a technical solution for sorting out the conclusions of comment sections.

[0058] Specifically, the interaction in the comment section currently involves multiple dimensions. Generally, we first sort out the hierarchical relationships to form a data hierarchy definition table as shown in Table 1, and then identify each comment based on the data hierarchy: Table 1 - Data Hierarchy Definition Table

[0059] This approach allows us to derive the direct evaluation results of most potential users or existing users based on the comments. However, the current challenge lies in analyzing each valid interactive comment. Due to the imprecise language used in the interactive comments, many references are not clearly mentioned, making it difficult to effectively analyze comment interactions. Consequently, public opinion analysis is limited by the data scope and cannot cover the comment area, resulting in significant errors in the analysis results.

[0060] Based on this, the embodiments of this application can first anchor entity association clues through the hypothetical connection relationship between multiple statements, then input the clues and context into a large model to generate candidate resolution results, and finally filter the optimal solution through vehicle body domain rule verification.

[0061] Specifically, based on the semantic logic of the vehicle body comment interaction chain, it can be assumed that the current statement containing reference has a specific relationship with the upstream statement, and the range of candidate entities can be locked through this relationship.

[0062] First, assume the connection type is defined. 1. Assumption of Continuation of the Same Topic: When the target sentence and the upstream continuating sentence discuss the "same topic about the car body," the referent in the target sentence must be the car body entity already mentioned in the continuating sentence, without introducing a new dimension of the car body topic. From the perspective of sentence characteristics, firstly, the continuating sentence must contain a clear car body entity, such as "rear seat space" or "front seat"; secondly, the target sentence will contain keywords that reflect the "continuation relationship," such as "also," "still," "same," and "agree." These words can be used to determine that the target sentence is a continuation of the topic of the continuating sentence, rather than a new topic.

[0063] Example: The continuation sentence (level L0) is "The rear seat space is small, and three people are cramped." This sentence explicitly mentions the physical "rear seat space" of the vehicle. The target sentence (level L1) is "It also causes knees to hit the front seats." The word "also" indicates a continuation relationship, meaning that the target sentence continues the topic related to "rear seat space." Therefore, the referent of "it" should be locked from "rear seat space" in L0, which conforms to the logic of the continuation assumption of the same topic.

[0064] 2. Logical Causal Assumption: The target sentence and an upstream statement (which could be a "cause" or a "result") have a clear "causal logical relationship." In this case, the referent in the target sentence needs to be located based on the causal relationship. The referent may be an entity in the result sentence. The core is to identify the entity based on the "causal association." Keywords that reflect causal logic will appear in the statement. One type is direct causal words, such as "because," "leading to," "therefore," "causing," etc., which can directly determine the causal relationship. The other type is indirect logical words, such as "problem," "cause," "improvement," etc., which can indirectly link to the causal logic of "a certain entity causing a certain problem" or "a certain problem requires the improvement of a certain entity." Simultaneously, the upstream and target sentences must revolve around the causal relationship of "the same car body problem," without any disconnect in the topic.

[0065] Example: The cause sentence (level L0) states "The short wheelbase leads to cramped rear space." This sentence clearly demonstrates the causal relationship through the word "leads to," where "wheelbase" is the causal entity that causes "cramped rear space." The target sentence (level L1) states "It is the main problem for family outings." Combining the causal logic of L0, "it" must refer to the core cause of the problem. Therefore, we lock onto "wheelbase" in L0, which conforms to the referential logic of the logical causal assumption.

[0066] 3. Entity Reuse Assumption: The pronouns in the target sentence do not refer to new vehicle body entities, but rather are "reuses" of vehicle body entities already explicitly mentioned in the upstream "original sentence." No new vehicle body entity topics will be introduced throughout the process; the discussion will only revolve around the entities in the original sentence, such as proposing improvement suggestions or supplementing user experience. From the perspective of sentence content, the original sentence must contain specific and explicit vehicle body entities, such as "rear seat rails" or "sound insulation thickness," rather than abstract terms like "space" or "comfort." Simultaneously, the target sentence will not contain new vehicle body entity vocabulary; it will only use pronouns such as "it" or "should" to refer to the entities in the original sentence, and the discussion content will be directly related to the entities in the original sentence. For example, if the original sentence mentions "entity attributes," the target sentence will mention "entity improvements."

[0067] Example: The original sentence (level L0) is "The length of the rear seat rail is insufficient," which explicitly mentions the specific vehicle body entity "rear seat rail." The target sentence (level L1) is "It needs to be lengthened by 5cm." There are no new vehicle body entity words in the sentence; it is only referred to by "it." Furthermore, "lengthening by 5cm" is an improvement suggestion for the "length of the rear seat rail," which is a discussion of entity reuse in the original sentence and conforms to the logic of the entity reuse assumption.

[0068] 4. Supplementary Explanation Assumption: The target sentence's role is to provide "detailed supplementation" to the vehicle body entity mentioned in the upstream "core sentence." Through specific descriptions, such as usage scenarios and functional performance, it makes the entity in the core sentence more concrete. The referent in the target sentence is the vehicle body entity whose details need to be supplemented in the core sentence. The target sentence will contain keywords that reflect a "supplementary relationship," such as "for example," "specific," "especially," and "like." These words indicate that the target sentence is a detailed extension of the core sentence. If the target sentence lacks such keywords, the supplementary relationship will be reflected by describing the "specific detailed characteristics of the entity." For example, if the core sentence mentions "entity function," the target sentence might mention "the entity's functional performance in a certain scenario." Simultaneously, the core sentence must contain a clearly defined vehicle body entity to provide the object for the supplementary explanation.

[0069] Example: The core sentence (level L0) is "poor trunk storage capacity". This sentence mentions the physical "trunk" of the vehicle and points out its "storage capacity" problem. The target sentence is "it cannot fit a 28-inch suitcase". By showing the specific scenario of "cannot fit a 28-inch suitcase", it supplements the details of "poor trunk storage capacity". Here, "it" refers to "trunk" in the core sentence, which is consistent with the logic of supplementing the hypothesis.

[0070] Then construct hypothetical connections 1. Upstream sentence filtering: For the target sentence (including pronouns / omitted subjects), filter sentences within 3 levels upstream and within a fixed publication time. 2. Hypothesis type matching: Based on the identified features, the target sentence and the upstream sentence are matched for hypothesis connection type. For example, if the target sentence contains "still" and the upstream sentence contains "following space", it matches "same topic continuation hypothesis". 3. Candidate entity locking: Based on the hypothesis type, extract all vehicle body entities from the corresponding upstream statements (continuing sentences / reasoning sentences / original sentences / core sentences) to form a candidate entity set.

[0071] For example, this interaction chain includes an upstream statement (level L0) and a target statement (level L1, where L0 contains "the front seats are too far back, and the rear space is small"), and L1 (the target statement) contains "it causes knees to hit the front seats." The referent of "it" is identified through the association between L0 and L1. The subsequent matching is based on the "same topic continuation hypothesis," determined by the fact that L1 does not introduce new vehicle body topics, but only revolves around the "front seats" and "rear space" mentioned in L0, exhibiting implicit topic continuation logic, which conforms to the core feature of the "same topic continuation hypothesis"—"the target statement continues the topic of the upstream statement." The upstream corresponding statement is L0 ("the front seats are too far back, and the rear space is small"), which explicitly contains two vehicle body entities: "front seats" and "rear space," providing direct evidence for subsequently identifying candidate entities. Finally, based on the entities in upstream L0, a candidate entity set is formed as {front seat position, rear space}, containing all vehicle body-related entities mentioned in L0, without any additional redundant or irrelevant entities.

[0072] After obtaining the candidate entity set, this solution uses a "large model for fine-tuning in the vehicle body domain". The large model for fine-tuning in the vehicle body domain can be fine-tuned based on Llama3-8B. For example, using 500,000 vehicle body comment interaction data as corpus, the context statement + hypothetical connection type + candidate entity set are input into the model as a structured prompt, and the model outputs candidate entities + resolution reasons.

[0073] Specifically, the parameters for fine-tuning the large model are as follows: {Task: Solve the problem of referential resolution in vehicle body reviews, and associate pronouns (it / this / that / should) or omitted subjects in the target sentence with the vehicle body entities that are clearly defined in the context.}

[0074] Known information: 1. Contextual Interaction Chain: {List of upstream statements, format: level-statement content, such as L0-"Front seats are too far back, rear legroom is limited"; L1-"It causes knees to hit the front seats"} 2. Assumption connection types: {Same topic continuation assumption / Logical causal assumption / Entity reuse assumption / Supplementary explanation assumption} 3. Candidate entity set: {Entity 1, Entity 2, ..., such as front seat position, rear seat space} 4. Body-related rules: The entity referred to must be a "body part / design parameter / performance-related entity", such as seats, wheelbase, and space, excluding non-body entities such as "vehicle system / power / price".

[0075] Require: 1. Select a unique entity from the candidate entity set as the referent. If none of the candidate entities match, output "No matching entity". 2. Output format: Candidate resolution result - entity name; Resolution reason - Explain why this entity was selected, considering the context, assumed connection type, and vehicle domain rules. The large model then outputs candidate resolution results. For example, the candidate resolution results are shown in Table 2 below: Table 2 - Candidate Dissolution Results

[0076] The following table provides a complete example of a vehicle body design scenario, as shown in Table 3: Table 3 - Examples of Complete Vehicle Body Design Scenarios

[0077] The following provides a detailed description of S3 and S4 in this embodiment.

[0078] Step S3: Collect user expectations and perform semantic reasoning and analysis.

[0079] 1) Text mining: Perform text mining processing on the database of user-expressed expectations and products, including word segmentation, building a thesaurus and assigning categories, etc. 2) Text Analysis: Product performance is selected as key words in the text data. Performance evaluation analysis is performed on the text. The frequency-weighted method is used to denot the number of positive and negative comments on product performance as positive numbers (P) and negative numbers (N), respectively. The total number of evaluations is denoted as T. The weight of each evaluation is represented by w. The specific calculation formula is as follows:

[0080] Where P and T are both positive integers, and N is a negative integer.

[0081] 3) Evaluation quantification using semantic differential method: Semantic differential method is a rating scale method that can be used to measure the semantics or performance of concepts and products. It is used to quantify the semantics of a set of perceptual words, as shown in Table 4 below. This study constructs a semantic evaluation quantification table for design products. The semantic quantification results of each vehicle model can be recorded according to this table. Table 4 Semantic Evaluation Quantification Table

[0082] 4) Design Expectation Mining: Based on human-expressed expectations and language patterns, summarize the aspects that users tend to value more. Divide product performance factors into N priority levels and assign each factor an expected score, denoted as . The score ranges from 0 to 1, with higher scores indicating greater importance of performance factors to the user. If a user does not mention a particular performance characteristic, it indicates low expectations and importance placed on that performance; therefore, the expected score for that aspect can be assigned as 0, making the sum of the semantic weights of the user's expected product performance 1, i.e.:

[0083] Step S4: Quantify customer expectations and vehicle body data using a sensory engineering model.

[0084] The quantitative theoretical methods of sensible engineering are applied to influence vehicle design decisions based on design and subjective image evaluation values. The specific quantitative model is described below: 1) Define a quantitative model relating each vehicle model and its performance characteristics to subjective evaluations. Quantitative theory analyzes a single variable to determine the influence of several categorically independent parameters on the dependent variable. Dummy variables are used to define the categorical parameters, which are called design elements. Therefore, the purpose of applying quantitative theory is to discover the intrinsic relationship between design elements and product performance, maximizing the match with the degree of positive subjective imagery. 2) Based on the emphasis of user input expectations and the purpose and requirements of the design, analyze and determine the weight of each performance aspect of the product that the customer expects; 3) Calculate the corresponding score for each existing model product, which is the correlation coefficient of user expectations and preferences. It is the relationship between the design elements of existing benchmark models and the user's emotional expectation score. 4) The semantic difference quantification table of product design benchmarking is multiplied by the semantic weights of the corresponding customer's expected product performance, and then summed. The vehicle models with higher preference correlation coefficients indicate that the various performance aspects of that vehicle model better match the user's expectations. This can be expressed by the formula.

[0085] Where: i is the number of product categories, i=1,2…m; j is the number of product performance metrics, j=1,2…n; and D is the semantic evaluation quantification result. This represents the expected score.

[0086] In summary, this invention presents a vehicle body selection design method based on Kansei Engineering. It employs a comprehensive analysis combining text analysis and Kansei Engineering principles. Based on the established model, it identifies the product with the highest customer expectation score, derives the target product, determines the optimal design solution, and further defines the original structural data of the designed product, thus integrating theory and practice. Designers can refer to this method during subsequent design processes, aiming to improve the efficiency and reliability of automotive body design.

[0087] In some specific embodiments, taking six expected performance indicators—space, appearance, cost-effectiveness, comfort, power, and fuel consumption—as examples, the results of semantic differential processing of the data for ten different models are listed in Table 5.

[0088] Table 5 Results of the Semantic Dissimilarity Scale

[0089] Using the user expectation "I need a car that is spacious, powerful, fuel-efficient, and well-insulated" as input, the semantic weight ratio of the user's expected product performance is calculated and shown in Table 6.

[0090] Table 6 Semantic Weights of User-Expected Product Performance

[0091] The quantitative calculation results are shown in Table 7. The most suitable model is model 10 with a score of 0.3151. The body data of the model is output to lay the foundation for further design by the designers.

[0092] Table 7 Quantitative Calculation Results

[0093] This invention further provides a vehicle body selection design system based on kinetic engineering, such as... Figure 2 As shown, it includes: Module 1 retrieves the vehicle's body data and comment information from the data that is allowed to be accessed; Database module 2 uses natural language processing technology to parse and organize the acquired comment text data to form a database; Inference module 3 collects user expectations and performs semantic reasoning and analysis on the collected user expectations based on the formed database to obtain the semantic weights of the product performance expected by the user. Matching module 4 uses the Kansei Engineering model to quantitatively match the semantic weights of the product performance expected by the user with the acquired vehicle body data, and determines the vehicle body data corresponding to the model that best matches the user's expectations.

[0094] It is understood that the technical effects of the vehicle body selection design system based on kinetic engineering provided in this application are the same as the technical effects of the method embodiments in the foregoing embodiments, and this application will not elaborate on this.

[0095] This application uses specific embodiments to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the methods and ideas of this application. At the same time, for those skilled in the art, there may be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A vehicle body selection design method based on kinetic engineering, characterized in that, include: Obtain the vehicle's body data and comment information from the data that is granted access; Natural language processing technology is used to parse and organize the acquired comment text data to form a database; User expectations are collected, and semantic reasoning and analysis are performed on the collected user expectations based on the formed database to obtain the semantic weights of product performance expected by users. By using the Kansei Engineering model, the semantic weights of the product performance expected by users are quantitatively matched with the acquired vehicle body data to determine the vehicle body data corresponding to the model that best matches the user's expectations.

2. The vehicle body selection design method based on kinetic engineering according to claim 1, characterized in that, The process of parsing and organizing the acquired comment text data using natural language processing technology to form a database includes: Perform text preprocessing on the comment text data; Based on the text preprocessing results, feature extraction and structural analysis are performed on the text; Based on the results of feature extraction and structural analysis, the text is classified and clustered. Based on the classification and clustering results, association analysis and pattern recognition are performed; Post-processing of the association analysis and pattern recognition results yields parsed and organized text data; The parsed and organized text data is stored in a preset manner to form the database.

3. The vehicle body selection design method based on kinetic engineering according to claim 1, characterized in that, The process of collecting user expectations and performing semantic reasoning and analysis on the collected user expectations based on the formed database to obtain the semantic weights of product performance based on user expectations includes: The user's expressed expectations are subjected to text mining processing against the database, and the processed user expectation text is output. Based on the processed user expectation text, identify key words for product performance, evaluate and analyze product performance based on these key words, and output the performance evaluation and analysis results. Using the semantic differential method, a semantic evaluation quantification table for the designed product is constructed based on the performance evaluation analysis results, and the semantic evaluation quantification table is output. Based on the semantic evaluation quantification table and user language expression, we can mine the user's expected priority of product performance and assign expected scores to obtain the semantic weight of the product performance expected by the user.

4. The vehicle body selection design method based on kinetic engineering according to claim 3, characterized in that, The process of text mining is performed on the user's expressed expectations and the database to output the processed user expectation text, including: The user's expressed expectations are compared with the text data in the database using word segmentation. Based on the word segmentation results, a thesaurus is constructed; Based on the constructed thesaurus, the segmented text data is assigned categories; Integrate the category assignment results and output the processed user-expected text.

5. The vehicle body selection design method based on kinetic engineering according to claim 3, characterized in that, The process involves identifying key words representing product performance based on the processed user expectation text, evaluating and analyzing product performance based on these key words, and outputting performance evaluation and analysis results, including: From the processed user expectation text, filter and identify key words that characterize product performance; Based on the aforementioned key words, the number of positive and negative reviews of the product performance corresponding to each key word is counted. The total number of evaluations of product performance corresponding to the key term is counted; The evaluation weight of the product's performance is determined based on the number of positive reviews, the number of negative reviews, and the total number of evaluations. The performance evaluation weights of all products are integrated, and the performance evaluation analysis results are output.

6. The vehicle body selection design method based on kinetic engineering according to claim 3, characterized in that, The process involves using a semantic evaluation quantification table and user language expression to mine the user's expected priorities regarding product performance and assigning them expected scores, thus obtaining the semantic weights of the user's expected product performance, including: Based on the language users use to express their expectations, summarize the product performance aspects that users focus on. Based on the degree of user concern regarding the performance of each product, the performance aspects of the products are prioritized. Assign a desired score to each priority level for product performance aspects, with the desired score ranging from 0 to 1, and assigning a score of 0 to product performance aspects not mentioned by the user. Adjust the expected scores for each product's performance aspects so that the sum of the expected scores for all product performance aspects is 1; Based on the adjusted expected score, determine and output the semantic weights of the product performance expected by the user.

7. The vehicle body selection design method based on kinetic engineering according to claim 3, characterized in that, The process of using a sensory engineering model to quantitatively match the semantic weights of user-expected product performance with the acquired vehicle body data to determine the vehicle body data corresponding to the model that best matches user expectations, serving as the basis for vehicle body design, includes: Define a quantitative model relating each vehicle model to its various performance subjective evaluations; Based on the emphasis of user input expectations and the purpose requirements of vehicle body design, and combined with the semantic weights of product performance obtained from user expectations, the weights of each product performance are determined. Based on the defined quantitative model and the determined weights of each product performance, the correlation coefficient between existing vehicle models and user expectations is calculated. The preference correlation coefficient is verified by multiplying the quantification results of each vehicle model in the semantic evaluation quantification table with the determined weights of each product performance and then summing the results. Select the car models with the highest preference correlation coefficient, extract the corresponding vehicle body data for each model, and output it.

8. The vehicle body selection design method based on kinetic engineering according to claim 7, characterized in that, The quantitative model defining the relationship between each vehicle model and its various performance subjective evaluations includes: The design elements of the vehicle model are used as classification-independent parameters, and dummy variables are defined to represent these classification-independent parameters. Analyze the degree of influence of each of the classification independent parameters on the vehicle product performance; Based on the degree of influence of each of the aforementioned independent parameters on the vehicle product performance, the intrinsic relationship between design elements and product performance is determined; Based on the aforementioned inherent relationships, a quantitative model is constructed and output regarding the relationship between vehicle models and various subjective performance evaluations.

9. The vehicle body selection design method based on kinetic engineering according to claim 7, characterized in that, Based on the defined quantitative model and the determined weights of each product performance, the correlation coefficient between existing vehicle models and user preferences is calculated, including: From the acquired vehicle body data, extract the design element data of existing vehicle models; The extracted design element data is correlated with the user's emotional expectation score to determine the degree of correlation between the design element data and the user's emotional expectation score. Based on the degree of correlation, calculate the correlation coefficient of user expectation preferences for each vehicle model. Output the correlation coefficients of user expectations and preferences for all vehicle models.

10. A vehicle body selection design system based on kinetic engineering, characterized in that, include: The acquisition module retrieves the vehicle's body data and comment information from the data that is allowed to be accessed. The database module uses natural language processing technology to parse and organize the acquired comment text data to form a database. The reasoning module collects user expectations and performs semantic reasoning and analysis on the collected user expectations based on the formed database to obtain the semantic weights of the product performance expected by the user. The matching module uses a sensory engineering model to quantitatively match the semantic weights of the product performance expected by the user with the acquired vehicle body data, and determines the vehicle body data corresponding to the model that best matches the user's expectations.

Images