New energy automobile model design method and system, electronic equipment and storage medium

Abstract

The invention discloses a new energy vehicle modeling design method and system, electronic equipment and a storage medium, and belongs to the technical field of vehicle industry design, and the method comprises the steps: carrying out the weight calculation and sorting of user perceptual vocabularies obtained through an evaluation construction map based on an interval type 2 trapezoidal fuzzy Kano model, so as to recognize key perceptual demands; establishing a nonlinear mapping relation between the key perceptual vocabularies and the form design characteristics of the new energy automobile based on a limit gradient lifting model optimized by a He-Marx optimization algorithm; predicting and generating an optimal morphological feature combination with the highest perceptual evaluation value by using the mapping relation; and performing subjective and objective comprehensive evaluation on the generated design scheme in combination with an eye movement tracking technology and a discrete information data fluctuation weighting method, and screening out a final design scheme. According to the method, subjective and fuzzy perceptual requirements of the user can be accurately quantified and sorted, a nonlinear mapping relation between the perceptual requirements of the user and specific product design characteristics is established, and the modeling design efficiency of the new energy automobile is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicle industrial design, in particular to a new energy vehicle modeling design method and system based on user emotional needs analysis and artificial intelligence technology, an electronic device and a storage medium. BACKGROUND

[0002] With the increasing emphasis on environmental protection and sustainable development worldwide, new energy vehicles (NEVs) have become the core development direction of the automotive industry. In the fiercely competitive market, consumers' car purchase decisions are no longer limited to hard indicators such as vehicle performance, range, etc. The emotional resonance and charm perception triggered by the appearance of the vehicle, i.e. "charming form", is playing an increasingly important role. How to accurately convert consumers' fuzzy and subjective emotional needs (Kansei) into specific and objective design elements is a major challenge currently faced by the automotive design field.

[0003] Traditional automotive design processes largely rely on designers' personal experience and subjective judgment, lacking a systematic and scientific method to quantify and analyze users' emotional preferences. The Kano model is introduced for user demand classification, but the traditional Kano model uses discrete scales (such as "like", "as it should be") when processing evaluation information, making it difficult to effectively deal with the fuzziness, uncertainty and individual differences of user feelings. In addition, when establishing the relationship between emotional needs and product design features, linear models or simple statistical methods are usually used, making it difficult to capture the complex, nonlinear mapping relationship between the two, resulting in design solutions that may deviate from users' true preferences.

[0004] Therefore, there is an urgent need for an innovative method that can effectively handle user emotional uncertainty, accurately establish the mapping relationship between emotions and design features, and scientifically evaluate design solutions to improve the efficiency and success rate of new energy vehicle charm form design. SUMMARY

[0005] Based on this, the present application provides a new energy vehicle modeling design method, system, electronic device and storage medium, aiming to accurately quantify and sort users' subjective and fuzzy emotional needs, establish a nonlinear mapping relationship between users' emotional needs and specific product design features, and finally design a scientific and comprehensive evaluation system.

[0006] In a first aspect, the present application provides a new energy vehicle modeling design method, comprising the following steps: Based on a pre-set new energy vehicle sample set and user Kano questionnaire survey data for emotional vocabulary, an interval 2-type trapezoidal fuzzy Kano model is used to calculate and determine the comprehensive importance weight of multiple emotional vocabulary, and a pre-set number of key emotional vocabulary with the highest weight are selected; The morphological characteristics of the new energy vehicle sample set are deconstructed to obtain multi-dimensional morphological characteristic data; based on the multi-dimensional morphological characteristic data and the user evaluation value of the key Kansei vocabulary, an extreme gradient boosting model optimized by a hippo optimization algorithm is trained to establish a mapping model between the morphological characteristics and the key Kansei vocabulary; Using the mapping model, the perceptual evaluation value of all morphological characteristic combinations within a predetermined range is predicted to determine the optimal morphological characteristic combination corresponding to the key Kansei vocabulary, and at least one candidate design scheme is generated according to the optimal morphological characteristic combination; The visual perception data of the user on the candidate design scheme is obtained by using eye tracking technology, and the candidate design scheme is objectively quantitatively analyzed by using the discrete information data fluctuation weighting method, and the final new energy vehicle modeling design scheme is determined from the candidate design scheme based on the visual perception data and the objective quantitative analysis result.

[0007] As an optional implementation mode of the first aspect of the application, the step of calculating and determining the comprehensive importance weight of the plurality of Kansei vocabularies comprises: based on the improved CRITIC weight method, using the Gini coefficient instead of the standard deviation to calculate the distribution difference of each Kansei vocabulary under different Kano categories to determine its objective importance; based on the least deviation method, combining the fuzzy evaluation of experts on the importance of different Kano categories to determine the satisfaction importance of each Kansei vocabulary; the objective importance and the satisfaction importance are weighted and integrated to obtain the comprehensive importance weight.

[0008] As an optional implementation mode of the first aspect of the application, the step of using the interval 2-type trapezoidal fuzzy Kano model comprises: quantifying the language scale set in the Kano questionnaire into interval 2-type trapezoidal fuzzy numbers; the user uses the interval 2-type trapezoidal fuzzy numbers to answer the positive and negative questions of the Kansei vocabulary to obtain fuzzy evaluation values; by calculating the distance between the fuzzy evaluation values and the fuzzy numbers corresponding to each language scale, the membership degree of adjacent language scales is determined, and then a Kano category membership degree vector is generated to classify and weight the Kansei vocabulary.

[0009] As an optional implementation mode of the first aspect of the application, the step of deconstructing the morphological characteristics of the new energy vehicle sample set comprises: using morphological analysis method, the new energy vehicle shape is divided into a plurality of design parts, the design parts include windows, lamps, air intake grilles, hubs and rearview mirrors; a plurality of types of characteristics are defined for each design part, and the type characteristics are coded to form multi-dimensional morphological characteristic data.

[0010] As an optional implementation of the first aspect of the application, the step of training an extreme gradient boosting model optimized by a hippo optimization algorithm includes: taking hyperparameters of an XGBoost model, including the number of trees, the depth of trees and the learning rate, as individuals in a hippo population; iteratively updating the combination of hyperparameters by simulating the swimming stage and the walking stage of the hippo population, wherein the swimming stage is used for global exploration and the walking stage is used for local development; taking the mean square error as the fitness function to evaluate the performance of the XGBoost model corresponding to each combination of hyperparameters until a preset number of iterations is reached or the model performance converges, thereby determining the optimal combination of hyperparameters.

[0011] As an optional implementation of the first aspect of the application, the step of obtaining visual perception data of the user on the candidate design scheme by using eye tracking technology includes: recording the eye movement trajectory of the user when observing the candidate design scheme, and counting eye movement indicators including total fixation time, average fixation time, fixation count and first fixation time; determining the visual perception priority order of the user on each candidate design scheme based on the standardized mean value of the eye movement indicators.

[0012] As an optional implementation of the first aspect of the application, the step of objectively quantitatively analyzing the candidate design scheme by using the discrete information data fluctuation weighting method includes: collecting scores of multiple evaluators on each candidate design scheme for the key perceptual words; calculating the deviation coefficient and the independent information ratio of each candidate design scheme under the score of each key perceptual word; combining the deviation coefficient and the independent information ratio to calculate the pure information quantity of each candidate design scheme, and normalizing it to obtain the objective weight of each candidate design scheme; sorting each candidate design scheme based on the objective weight.

[0013] In a second aspect, the embodiments of the application provide a new energy vehicle modeling design system, which comprises: A perceptual demand quantification module is configured to use an interval 2-type trapezoidal fuzzy Kano model to calculate and determine the comprehensive importance weight of multiple perceptual words based on a preset new energy vehicle sample set and user Kano questionnaire survey data for perceptual words, and to filter out a preset number of key perceptual words with the highest weight; A mapping model construction module is configured to deconstruct the form features of the new energy vehicle sample set to obtain multi-dimensional form feature data, and to train an extreme gradient boosting model optimized by a hippo optimization algorithm based on the multi-dimensional form feature data and user evaluation values for the key perceptual words to establish a mapping model between the form features and the key perceptual words. An optimal scheme generation module is configured to utilize the mapping model to predict perceptual evaluation values of all form feature combinations within a preset range, determine an optimal form feature combination corresponding to the key perceptual words, and generate at least one candidate design scheme according to the optimal form feature combination. A comprehensive evaluation screening module is configured to acquire visual perception data of the candidate design schemes of a user by using an eye tracking technology, and perform objective quantitative analysis on the candidate design schemes by using a discrete information data fluctuation weighting method, and determine a final new energy vehicle modeling design scheme from the candidate design schemes in combination with the visual perception data and the objective quantitative analysis result.

[0014] In a third aspect, an electronic device is provided, which includes a processor, a memory, and a program or instructions stored on the memory and executable on the processor, and the program or instructions are executed by the processor to implement the steps of the method in the first aspect.

[0015] In a fourth aspect, a readable storage medium is provided, which stores a program or instructions, and the program or instructions are executed by a processor to implement the steps of the method in the first aspect.

[0016] Compared with the prior art, the present application has the following beneficial effects: 1. Improved accuracy of demand analysis: IT2Tr-FKM is applied to NEV form design for the first time, and the fuzziness and uncertainty in user perceptual evaluation are effectively handled by interval 2-type trapezoidal fuzzy numbers, so that the weight ranking of perceptual demand is more scientific and reliable.

[0017] 2. Intelligent mining of design elements: The HO-XGBoost model is used to successfully establish a complex nonlinear mapping relationship between user emotions and specific design features. The introduction of the HO algorithm significantly improves the prediction accuracy and generalization ability of the XGBoost model, which can efficiently and accurately mine the optimal design feature combination.

[0018] 3. Enhanced scientificity of scheme evaluation: The eye tracking technology and DIDF weighting method are innovatively combined. Eye tracking captures the user's subconscious visual preferences, and the DIDF method provides an objective weight for the user's subjective score, and the combination of the two establishes a multi-dimensional evaluation system that combines subjective and objective factors, making the selection of the final scheme more comprehensive and reliable.

[0019] 4. Improved overall design efficiency and quality: The entire framework transforms the original subjective and scattered design process into a data-driven and intelligent decision-making systematic process, significantly shortening the design cycle and producing new energy vehicle design that better meets the target user's emotional preferences and has stronger market competitiveness. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flow chart of a new energy vehicle modeling design method according to an embodiment of the present application; Figure 2 is a technical roadmap of a new energy vehicle modeling design method according to an embodiment of the present application; Figure 3 is a schematic diagram of new energy vehicle form disassembly according to an embodiment of the present application; Figure 4 is a structural schematic diagram of a new energy vehicle modeling design system according to an embodiment of the present application.

[0021] The following specific embodiments will further illustrate the present application in conjunction with the above drawings. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0023] The terms "first", "second", and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein. In addition, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / ", generally indicates that the front and rear associated objects are in a "or" relationship. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise explicitly specified.

[0024] Embodiment 1 Please refer to Figure 1 is a flow chart of a new energy vehicle modeling design method according to an embodiment of the present application, and refer to Figure 2 is a technical roadmap of a new energy vehicle modeling design method according to an embodiment of the present application. The method can include the following steps: S1: Based on the preset new energy vehicle sample set and the user Kano questionnaire survey data for the perceptual vocabulary, an interval 2 type trapezoidal fuzzy Kano model (IT2Tr-FKM) is used to calculate and determine the comprehensive importance weight of multiple perceptual vocabulary, and a preset number of key perceptual vocabulary with the highest weight is selected.

[0025] Specifically, in order to solve the defects of the existing fuzzy Kano model (FKM) in dealing with the subjectivity, uncertainty and individual differences of evaluation information in user demand classification, this step uses interval type-2 trapezoidal fuzzy number (IT2TrFN) to process the uncertainty of evaluation information in the Kansei evaluation process, and constructs an IT2Tr-FKM integrated model. The process includes two main stages of classification and weighting of Kansei vocabulary.

[0026] 1. Definition and application of interval type-2 trapezoidal fuzzy number (IT2TrFN) Interval type-2 fuzzy set (IT2FS) introduces the uncertainty of the membership function itself on the basis of traditional fuzzy set, allowing decision makers to consider the uncertainty of membership degree and shape / position of membership function. IT2TrFN describes the evaluation information through primary membership degree and secondary membership degree, as shown in formula (1).

[0027] Definition 1: In the formula, represents an interval type-2 trapezoidal fuzzy set, represents the lower boundary trapezoidal fuzzy number of , represents the upper boundary trapezoidal fuzzy number of , , is the starting point of the upper and lower boundary trapezoids; , , , is the inflection point of the upper and lower boundary trapezoids; , is the end point of the upper and lower boundary trapezoids; , represents the height of the upper and lower boundary trapezoids of . The ten parameters can be freely adjusted according to the actual situation, with very high flexibility.

[0028] Definition 2: Distance calculation formula Let set and set be two non-negative IT2TrFNs on the domain X as formula (2)-(3).

[0029] represents the height of the upper boundary trapezoid of set , represents the height of the lower boundary trapezoid of set ; To measure the difference between two IT2TrFNs, the distance between fuzzy sets and is calculated The formula is defined as equation (4): Where: , represents the height of the lower bound trapezoid, represents the height of the upper bound trapezoid.

[0030] 2. Classification of Kansei Vocabularies First, through interviews and KJ methods, etc., 5 Kansei vocabularies are extracted N E i ( i = 1, 2,..., N ) 。 Then, the language scale set of the Kano questionnaire (such as dissatisfied, acceptable, indifferent, as it should be, satisfied) is quantified as IT2TrFN, denoted by standard language scale C.

[0031] The user's answers to the positive and negative questions for each Kansei vocabulary are represented by IT2TrFN. The distance between the user's fuzzy evaluation value and each standard language scale C is calculated by equation (4), and then the membership degrees of the evaluation value to the left and right adjacent language scales are calculated according to equation (5).

[0032] Where: represents the fuzzy evaluation value x , the membership function of the fuzzy evaluation value to the left adjacent standard language scale x , the membership function of the fuzzy evaluation value to the right adjacent standard language scale , the distance function for calculating , in the classification of Kansei vocabularies, the standard language scale C={C1,C2,C3,C4,C5} is quantified as interval type 2 trapezoidal fuzzy numbers, i represents the quantized value of the i-th standard language scale. represents the i-th scale in the Kano questionnaire, . The sum of the membership degrees of the evaluation value to the adjacent left and right scales is , and the membership degree values to other scales are all 0. d(˙) represents the function for calculating the membership degree, and represents the membership degree set of the question. ​​​

[0033] The membership sets of the positive and inverse problems are calculated by formula (5) and represented by P i and U i respectively. Then, a 5x5 fuzzy evaluation matrix L i is generated according to the preference information given by the questionnaire, as shown in formula (6): The above steps are repeated to count the Kano category membership vectors of the i-th sensibility word by the j-th participant, and the overall category membership vector of the i-th sensibility word is H as shown in formula (7): wherein denotes the Kano category membership vector of the i-th sensibility word by the j-th participant, h and denote the Kano ratios corresponding to the six different Kano categories of the i-th sensibility word, respectively. i

[0034] The Kano category corresponding to the highest membership value is the category to which the sensibility word belongs, and the category distribution matrix of the sensibility words is obtained by collecting the category membership vectors, as shown in formula (8): D wherein N denotes the total number of sensibility words, and K denotes the total number of Kano categories.

[0035] 3. Weighting of Sensibility Words First, objective importance calculation: improved CRITIC weight method is adopted. Gini coefficient is used instead of standard deviation to measure the distribution difference (contrast) of sensibility words in each Kano category, as shown in formula (9). Combined with the conflict between words, the objective importance of the i-th sensibility word is calculated by formula (10) .

[0036] wherein Gini coefficient is used instead of standard deviation to measure the distribution of category membership information, and denotes: wherein denotes the membership of the i-th sensibility word to the j-th Kano category, i and p denotes the membership of the i-th sensibility word to the j-th Kano category, ​​​​The closer the calculated value of the Gini coefficient of the emotional vocabulary is to 1, the more unbalanced the information distribution is, and the more information is contained.

[0037] The objective importance of the mth emotional vocabulary can be calculated by formula (10) i In the formula, represents the information carrying capacity of the mth emotional vocabulary, i

[0038] Thus, the objective importance of the emotional vocabulary is

[0039] Secondly, satisfaction importance calculation: in view of the different contributions of different categories of emotional vocabulary to customer satisfaction, experts are invited to make fuzzy evaluation on the importance of Kano categories themselves. Through the least square method, combined with the Kano category membership degree vector of the emotional vocabulary, the satisfaction importance of the emotional vocabulary is calculated according to formulas (11)-(12)

[0040] The Kano category importance is represented by vector , an expert fuzzy evaluation matrix is established to determine the importance of Kano categories, and then the satisfaction importance of the emotional vocabulary is calculated through the least square method.

[0041] According to the Kano category membership degree vector of the emotional vocabulary , the absolute satisfaction importance of the emotional vocabulary is obtained : The absolute satisfaction importance of the emotional vocabulary is normalized, as formula (12): Thus, the satisfaction importance of the emotional vocabulary is , represents the satisfaction importance of the mth emotional vocabulary. i

[0042] Finally, comprehensive importance calculation: the objective importance and the satisfaction importance are weighted and integrated, and the comprehensive importance is obtained according to formula (13) .

[0043] The comprehensive importance vector of the emotional vocabulary is represented as .

[0044] ​​​​​​​​Finally, all the emotional words are ranked according to the comprehensive importance, and the highest weight words are selected as the key emotional words.

[0045] Exemplarily, to construct a representative new energy vehicle (NEV) sample set, 221 NEV pictures of different brands and models were collected from mainstream automobile websites, professional automobile magazines and other multimedia channels. To ensure data quality, all pictures were screened, and pictures with low resolution, poor angle or seriously disturbed by complex light environment were removed, and finally 110 high-quality, multi-angle NEV sample pictures were retained. Using Adobe Photoshop software, the 110 pictures were standardized, including uniform size (15cm x 15cm), background removal, color balance adjustment, and were made into standardized sample cards. Then, to dig into the deep emotional needs of users, 10 subjects with different backgrounds (including designers, students and ordinary consumers) were invited to participate in a depth interview based on the evaluation construction graph (EGM) method. In the interview, the subjects were shown the NEV sample cards, and were guided to freely express their preferences, dislikes and original reasons for producing these feelings by comparing different vehicle forms. After the interview, the KJ method (affinity diagram method) was used to systematically organize the large amount of original corpus collected. Similar or related descriptions were inducted and merged, and finally an EGM evaluation construction graph containing three levels was constructed: 52 evaluation items (i.e. emotional words) in the upper layer, 16 evaluation items (original reasons) in the middle layer, and 200 evaluation items (specific form conditions) in the lower layer. After further refinement and screening, 14 most representative upper emotional words were finally determined, including “simple”, “unique”, “dynamic”, “technological”, “high-level”, “introverted” and the like, which constitute the basis for subsequent quantification of emotional needs.

[0046] S2: decompose the form features of the new energy vehicle sample set to obtain multi-dimensional form feature data; based on the multi-dimensional form feature data and the user evaluation value for the key emotional words, train an extreme gradient boosting model (XGBoost) optimized by a hippo optimization algorithm (HO) to establish a mapping model between the form features and the key emotional words.

[0047] Specifically, the HO-XGBoost model is used to establish a nonlinear mapping relationship between the design features and the user emotional evaluation.

[0048] 1. XGBoost model XGBoost predicts by integrating multiple decision trees. For input samples , the predicted value is the cumulative output of all L decision trees , then the The predicted value for each sample is given by formula (14): XGBoost's optimization objective includes a loss function and a regularization term, and is approximated using a second-order Taylor expansion. In the... In the next iteration, the objective function is optimized. for: in The first derivative of the loss function. The loss function is expressed as follows: The first-order partial derivative, It is a loss function that represents the difference between the actual value and the predicted value. Indicates the first t Tree samples x i The predicted value, The second derivative of the loss function. The loss function is expressed as follows: The second-order partial derivatives of . It is the first The number of leaf nodes in a tree. The score is for the leaf nodes. and It is the regularization coefficient that controls the complexity of the model.

[0049] 2. Hippo Optimization Algorithm (HO) The HO algorithm is used to optimize the hyperparameters (such as the number of trees, depth, learning rate, etc.) of the XGBoost model. It finds the optimal combination of hyperparameters by simulating the "swimming phase" (global exploration) and "walking phase" (local development) of a hippopotamus population.

[0050] Swimming phase: Simulate hippo group behavior to generate candidate hyperparameter combinations, as shown in formula (16). Individuals learn from the globally optimal combination, while introducing randomness to explore new spaces.

[0051] Initial generation was achieved by simulating hippopotamus group behavior. Group of hyperparameter combinations This ensures that each dimension of each hyperparameter set is within a preset range. Each hyperparameter combination is analogous to an individual hippopotamus; the current hyperparameter combination... Refer to the globally optimal combination ,formula This reflects the behavior of individuals learning from successful individuals and approaching the optimal solution, while Then, the random factor is introduced to simulate the hippopotamus autonomous exploration of new directions, making the algorithm search the hyperparameter space with both directionality and explorability, preventing it from falling into local optimum, and then generating candidate hyperparameter combinations by simulating the behavior of hippopotamus groups, and calculating as formula (16).

[0052] wherein represents the new candidate value of the j th hyperparameter, represents the current value of the j th hyperparameter, represents the global optimal value of the j th hyperparameter.

[0053] a and b The value range is generally between [0, 1], and the value is taken to balance the exploration and utilization ability of the algorithm to control the search direction coefficient. The experience value is taken a = 0.5, b = 0.3. , is a uniformly distributed random number.

[0054] Model evaluation: use mean squared error (MSE) as the fitness function to evaluate the performance of each hyperparameter combination, as formula (17).

[0055] For the generated candidate hyperparameter combination , build an XGBoost model and train it on the training set, and then calculate the mean squared error (MSE) through the validation set to evaluate the model performance. Assuming that there are samples in the validation set, the sample true value is , and the predicted value is , then the mean squared error calculation is as formula (17): If the new candidate hyperparameter combination is better, , then update the global optimal solution.

[0056] Walking phase: fine-tune around the current global optimal solution for fine search, as formula (18).

[0057] In the walking phase, the global optimal solution is fine-tuned according to the following formula to generate a new hyperparameter combination .

[0058] wherein, represents the value of the th hyperparameter in the updated hyperparameter combination. represents the value of the jth hyperparameter in the current global optimal hyperparameter combination. is a random variable uniformly distributed in is the jth value of the random variable uniformly distributed in Step size As a dynamic parameter, the value range is usually in (0, 1]. U represents the uniform distribution, represents the continuous uniform distribution in the interval [-1, 1]. It decays with the number of iterations, giving a larger value in the early iterations, which can make the algorithm quickly search the solution space with a large step size to speed up the convergence process. As the iteration progresses, the value gradually decreases, which is convenient for fine-tuning the hyperparameters to improve the accuracy of the solution. For example, where is the total number of iterations, and the initial value .

[0059] Such a loop iteration, constantly repeating the swimming phase of the HO algorithm, model evaluation, walking phase, until the maximum number of iterations or convergence is reached. The optimal hyperparameter combination is finally determined, and a high-precision XGBoost prediction model is trained.

[0060] Exemplarily: IT2Tr-FKM questionnaire design and data collection: Based on the 14 emotional words selected in the previous step, a questionnaire based on the interval type 2 trapezoidal fuzzy Kano model (IT2Tr-FKM) was designed. Unlike traditional Kano questionnaires, this questionnaire quantifies the user's five subjective feeling language scales, namely "dissatisfied (D)", "acceptable (W)", "indifferent (N)", "as it should be (M)", and "satisfied (L)", into interval type 2 trapezoidal fuzzy numbers (IT2TrFN). These fuzzy numbers can capture the degree of hesitation and uncertainty of the evaluation. The questionnaire includes positive questions for each emotional word ("If the car has XX feature, how do you feel?") and negative questions ("If the car does not have XX feature, how do you feel?"). The questionnaire was distributed to 60 graduate students majoring in industrial design, who have a deeper understanding of car design. Finally, 49 valid questionnaires were collected.

[0061] Kano attribute classification: The data from the 49 collected questionnaires were processed. Take the Kano term "simple" as an example. If a respondent's answer to the positive question corresponds to the fuzzy number [(7, 7.8, 7.8, 9.2; 0.53), (5.2, 7.4, 8.6, 9.8; 1)], and the answer to the reverse question is [(0.2, 1.8, 1.8, 3.2; 0.53), (0, 1.2, 2.6, 4.8; 1)]. First, the distance between the answer fuzzy number and the five standard language scale fuzzy numbers is calculated according to formulas (4) and (5), thereby determining the membership of the answer to the adjacent language scale. Then, the membership is converted into a 5x5 fuzzy evaluation matrix according to formula (6). Comparing this matrix with the pre-set Kano category evaluation table, the membership of the user's "simple" term to the charm attribute (A), one-dimensional attribute (O), essential attribute (M), indifferent attribute (I), and reverse attribute (R) can be calculated, forming a membership vector. The membership vectors of all 49 respondents are summed and averaged, and the overall Kano category membership vector of the "simple" term is obtained according to formula (7). Repeat this process to finally obtain the distribution matrix of the 14 Kano terms in A, O, M, I, R, and Q (Q is a suspicious result). According to the principle of the highest membership, the main Kano attribute of each term is determined, for example, "introversive" is classified as an essential attribute, and "delicate" is classified as a one-dimensional attribute.

[0062] Synthetic importance weighting of Kano terms: Objective importance calculation: To measure the importance of each Kano term in the user's demand structure, an improved CRITIC method (ICRITIC) is used. Based on the above Kano category distribution matrix, the Gini coefficient is used to replace the standard deviation in the traditional CRITIC method according to formula (9), to more accurately measure the imbalance (i.e., contrast) of the distribution of each Kano term in the six Kano categories. Then, the conflict between terms is calculated according to formula (10), and the contrast and conflict are combined to calculate the objective importance vector Wo of the 14 Kano terms.

[0063] Satisfaction importance calculation: To incorporate the influence of user satisfaction, 49 experts were invited to evaluate the importance of the five Kano categories, namely charm (A), one-dimensional (O), essential (M), indifferent (I), and reverse (R). Then, according to the minimum deviation method defined in formulas (11) and (12), the experts' fuzzy evaluation is combined with the Kano category membership of each Kano term to calculate the satisfaction importance vector Ws of the 14 Kano terms. S .

[0064] Comprehensive importance calculation: Finally, according to formula (13), the objective importance Wo and the satisfaction importance Ws are integrated by weighting to obtain the final comprehensive importance vector W S The fusion is performed to obtain the final comprehensive importance vector W. The calculation results show that the comprehensive importance scores of the three Kansei words "introversive (K7)", "delicate (K 11 ) "and "high-level (K6)" are the highest. Therefore, these three words are determined as the key Kansei words for subsequent design and modeling.

[0065] S3: Using the mapping model, predict the Kansei evaluation value of all shape feature combinations within the preset range, determine the optimal shape feature combination corresponding to the key Kansei words, and generate at least one candidate design scheme according to the optimal shape feature combination.

[0066] Exemplarily: Shape feature deconstruction and training data preparation: Referring to Figure 3 , the appearance of the NEV is systematically disassembled by using the shape analysis method. The whole vehicle shape is divided into five key design parts: X1 - vehicle window, X2 - vehicle lamp, X3 - air intake grille, X4 - vehicle hub, and X5 - rearview mirror. A variety of specific type features are defined for each part, and they are digitally coded. For example, there are five types of vehicle windows, which are coded as 1 to 5. For the 110 NEV samples in step one, the features of the five parts are analyzed and coded one by one. At the same time, 100 subjects are invited to score the 110 sample pictures on the three key Kansei words "introversive", "delicate" and "high-level" respectively using a seven-point Likert scale. The average of all scores is taken to finally construct a training data set containing 110 records. The input of each record is a 5-dimensional shape feature code (such as [3, 2, 4, 1, 5]), and the output is a 3-dimensional Kansei evaluation value (such as [5.8, 6.2, 6.5]).

[0067] HO-XGBoost mapping model training: To capture the complex nonlinear relationship between morphological features and perceptual evaluation, this embodiment uses the limit gradient boosting model (XGBoost) optimized by the Hippocampus Optimization (HO) algorithm. The dataset constructed in the previous step is divided into training set and test set in the ratio of 8:2. The population size of the Hippocampus Optimization algorithm is set to 5, and the maximum number of iterations is set to 100. The optimization goal of the HO algorithm is to find a combination of hyperparameters that can minimize the mean square error (MSE) of the XGBoost model prediction. These hyperparameters include the number of trees, the maximum depth of the tree, and the learning rate. During the iteration process, the HO algorithm simulates the "swimming phase" (global exploration, searching for new possible solution space) and "walking phase" (local development, fine search around the current optimal solution) of the Hippocampus population, efficiently updating the hyperparameter combination. After training is completed, an optimized, high-precision HO-XGBoost prediction model for each key perceptual word is obtained.

[0068] Optimal scheme generation: Using the three trained HO-XGBoost models, a comprehensive perceptual score prediction is performed for all possible morphological feature combinations. Since there are 5 types for each part, there are a total of 5 ^5 =3125 possible combinations. These 3125 combinations are used as input to predict their scores in the "intimate", "delicate", and "high-level" dimensions. The prediction results determine the morphological feature combination with the highest score in each dimension, for example: The combination with the highest score in the "intimate" dimension is [5, 5, 1, 3, 2].

[0069] The combination with the highest score in the "delicate" dimension is [5, 5, 3, 4, 5].

[0070] The combination with the highest score in the "high-level" dimension is [5, 2, 5, 1, 4].

[0071] Based on these three data-driven optimal feature combinations, the designer creatively draws sketches of multiple candidate design schemes, and quickly renders high-quality visual candidate design schemes through generative AI tools such as Midjourney.

[0072] S4: Use eye tracking technology to obtain users' visual perception data of the candidate design schemes, and use the Discrete Information Data Fluctuation Weighting Method (DIDF) to objectively quantify the candidate design schemes. Based on the visual perception data and the objective quantitative analysis results, the final new energy vehicle styling design scheme is determined from the candidate design schemes.

[0073] Specifically, this step combines subjective and objective evaluation methods to achieve a comprehensive and scientific assessment of candidate solutions.

[0074] 1. Eye-tracking technology (subjective unconscious evaluation) This study uses eye-tracking experiments to verify whether the product meets the customer's emotional preferences and to identify the focus areas of user visual attention. By recording and analyzing user eye movement indicators (such as total fixation duration, first fixation duration, etc.) when observing design solutions, we can understand the user's "first visual reaction" and determine which design solutions can quickly attract user attention. However, high visual perception sequence does not necessarily mean better design, as it may be influenced by non-core design factors such as color and location. To address this shortcoming, the objective DIDF weighting method is introduced.

[0075] 2. Discrete Information Data Fluctuation Weighting Method (DIDF) (objective quantitative analysis) DIDF is an objective weighting method that combines index independence and data volatility to process expert or user rating data for candidate solutions. The core steps are as follows: Step 1: Calculate the dispersion coefficient of each index The dispersion coefficient reflects the consistency or volatility of the evaluation data.

[0076] In the formula: is the standard deviation, is the mean value.

[0077] Step 2: Calculate the independent information ratio of each index.

[0078] Calculate the independent information ratio of each index Perform multiple linear regression with a certain index as the dependent variable and other indexes as independent variables, and get the determination coefficient of the index Then: Here, represents the goodness of fit of the th index. In the extreme case, if the goodness of fit is 0, it means that the index can provide complete information; if is 1, it means that the index cannot provide any independent information and can be deleted.

[0079] Step 3: Calculate the pure information quantity of each index .

[0080] Standardize , to get the standardized , Then the pure information quantity is calculated .

[0081] wherein is the deviation coefficient of the first index before standardization, is the independent information ratio of the first index before standardization. By standardization, the differences in dimensions of different indexes are eliminated, and then the two are multiplied to obtain the pure information quantity.

[0082] Fourth step: calculating the weight of each index .

[0083] Fifth step: evaluation index standardization and weighted summary.

[0084] The evaluation indexes are standardized, and then the weight obtained in the fourth step is used for weighted summary, so as to obtain the final evaluation result.

[0085] Exemplarily: Eye movement experiment (subjective subconscious evaluation): from the numerous schemes generated in the previous step, 12 design schemes (labeled as A1 to A 12 ) ranked in the top 4 in the comprehensive performance of three key perceptual vocabulary dimensions were selected. Five users with NEV driving experience were recruited to participate in the eye tracking experiment. In the experiment, the renderings of the 12 schemes were integrated into one large picture, and the users were required to observe freely within 3 minutes. The Tobii Pro Glasses 2 wearable eye tracker was used to record the eye movement trajectory of the user in real time. After the experiment, the eye movement data was analyzed by Tobii Pro Lab software, and 7 key eye movement indexes were extracted, including total fixation duration, average fixation duration, fixation count, first fixation duration, etc. After standardizing the 7 index data of each scheme, the average score was calculated to quantify the subconscious visual attention of the user to each scheme. The results show that the visual perception priority of the user is: A3>A 10 >A 12 >A9>…, indicating that schemes A3 and A 10 can attract the visual attention of the user most.

[0086] DIDF empowerment (objective quantitative evaluation): In order to obtain a more objective evaluation, 32 designers with industrial design professional background were invited to score the "degree of reserve", "degree of exquisite" and "degree of high-class" of 12 candidate schemes using a seven-point Likert scale. After collecting all the scoring data, the discrete information data fluctuation empowerment method (DIDF) was used for processing. According to formulas (19)-(24), the scoring data of each scheme under each kansei vocabulary was used to calculate its dispersion coefficient (reflecting the consistency of evaluation) and independent information ratio (reflecting the uniqueness of evaluation data) in turn. Combined with these two indicators, the pure information quantity of each scheme was calculated and normalized to obtain the objective weight of each scheme under three kansei dimensions. The weights of the three dimensions were added up to get the comprehensive evaluation score and ranking of the 12 schemes. The results show that the comprehensive ranking is: A3>A 10 >A9>A 12 >…, indicating that schemes A3 and A 10 have the highest comprehensive quality in the eyes of experts.

[0087] Final scheme determination: The results of the above two evaluation methods are as follows: the eye movement experiment reveals the user's subconscious preference, and the DIDF empowerment method provides an objective quantitative ranking of expert evaluation. Both methods consistently rank schemes A3 and A 10 in the top two. After the final discussion and trade-off of the design team, the scheme A3 with the best overall performance is selected as the final new energy vehicle charm form design scheme. Finally, the professional three-dimensional modeling software such as Rhino 3D is used to accurately digitize the scheme A3, laying a foundation for subsequent engineering development.

[0088] To verify the superiority of the HO-XGBoost model used in the present application, its performance is compared with four other commonly used artificial intelligence prediction models, including GRU-Attention, CNN, RBF and LSTM. All models use the same training set and test set. The comparison results show that, whether in the training set or in the test set, the HO-XGBoost model has the highest determination coefficient (R²) (close to 0.97-0.99) and the lowest root mean square error (RMSE) when predicting the "Reserved", "Exquisite" and "High-class" kansei evaluation values. This result powerfully proves that the HO-XGBoost model has the highest prediction accuracy and the best generalization ability, and can most effectively establish the complex mapping relationship between the NEV form features and the user's kansei needs.

[0089] Example 2 Please refer to Figure 4, as shown in the structure schematic diagram of a new energy vehicle modeling design system proposed in the second embodiment of the present application, the system comprises the following key modules: The perceptual demand quantification module 100 is configured to calculate and determine the comprehensive importance weight of a plurality of perceptual words based on a preset new energy vehicle sample set and user Kano questionnaire survey data for perceptual words, and to filter out a preset number of key perceptual words with the highest weight by using an interval 2-type trapezoidal fuzzy Kano model. The mapping model construction module 200 is configured to deconstruct the morphological characteristics of the new energy vehicle sample set to obtain multi-dimensional morphological characteristic data, and to train an extreme gradient boosting model optimized by a river horse optimization algorithm based on the multi-dimensional morphological characteristic data and user evaluation values for the key perceptual words to establish a mapping model between morphological characteristics and key perceptual words. The optimal scheme generation module 300 is configured to use the mapping model to predict the perceptual evaluation values of all morphological characteristic combinations within a preset range, determine the optimal morphological characteristic combination corresponding to the key perceptual words, and generate at least one candidate design scheme according to the optimal morphological characteristic combination. The comprehensive evaluation and screening module 400 is configured to use eye tracking technology to obtain visual perception data of users for the candidate design scheme, and to use a discrete information data fluctuation weighting method to objectively and quantitatively analyze the candidate design scheme, and to determine the final new energy vehicle modeling design scheme from the candidate design scheme in combination with the visual perception data and the objective and quantitative analysis results.

[0090] The new energy vehicle modeling design system in the embodiments of the present application can be a device, a component in a terminal, an integrated circuit, or a chip. The device can be a mobile electronic device or a non-mobile electronic device. Illustratively, the mobile electronic device can be a mobile phone, a tablet computer, a notebook computer, a palm computer, a vehicle-mounted electronic device, a wearable device, an Ultra-mobile Personal Computer (UMPC), a netbook, or a Personal Digital Assistant (PDA), etc., and the non-mobile electronic device can be a server, a Network Attached Storage (NAS), a Personal Computer (PC), etc., and the embodiments of the present application are not limited in this regard.

[0091] The new energy vehicle modeling design system in the embodiments of the present application can be a device with an operating system. The operating system can be an Android operating system, an IOS operating system, or other possible operating systems, and the embodiments of the present application are not limited in this regard.

[0092] The new energy vehicle modeling design system provided by the embodiment of the application can realize Figure 1 The method embodiment of the new energy vehicle modeling design method realizes each process, and to avoid repetition, each process will not be described here.

[0093] Optionally, the embodiment of the application further provides an electronic device, including a processor, a memory, a program or instructions stored on the memory and executable on the processor, which, when executed by the processor, realizes each process of the above-mentioned new energy vehicle modeling design method embodiment and achieves the same technical effect. To avoid repetition, each process will not be described here.

[0094] The embodiment of the application further provides a readable storage medium, which stores a program or instructions, which, when executed by a processor, realizes each process of the above-mentioned new energy vehicle modeling design method embodiment and achieves the same technical effect. To avoid repetition, each process will not be described here.

[0095] The processor is the processor of the electronic device in the above-mentioned embodiment. The readable storage medium includes a computer readable storage medium, such as a computer readable memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, etc.

[0096] It should be noted that in this paper, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of another identical element in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the present application is not limited to the order of functions shown or discussed, but also includes functions performed in a substantially simultaneous manner or in the opposite order, for example, the described method can be performed in an order different from the described order, and various steps can also be added, omitted or combined. In addition, the features described with reference to some examples can be combined in other examples.

[0097] Through the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned example methods can be realized by means of software and a necessary general hardware platform, and of course, can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product in essence or in the form of a part that contributes to the prior art. The computer software product is stored in a storage medium (such as a ROM / RAM, a magnetic disk, or an optical disk) and includes a plurality of instructions for causing a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in the various embodiments of the present application.

[0098] The embodiments of the present application are described above in combination with the drawings, but the present application is not limited to the above-mentioned specific embodiments, and the above-mentioned specific embodiments are only illustrative and not restrictive. Those skilled in the art can make many forms without departing from the scope of the present application under the inspiration of the present application, and all of them belong to the protection of the present application.

Claims

1. A new energy vehicle modeling design method, characterized in that, The method comprises the following steps: Based on the preset new energy vehicle sample set and the user Kano questionnaire survey data for the emotional vocabulary, an interval 2-type trapezoidal fuzzy Kano model is used to calculate and determine the comprehensive importance weight of multiple emotional vocabulary, and a preset number of key emotional vocabulary with the highest weight is screened out; The form features of the new energy vehicle sample set are deconstructed to obtain multi-dimensional form feature data; Based on the multi-dimensional form feature data and the user evaluation value for the key emotional vocabulary, an extreme gradient boosting model optimized by a hippo optimization algorithm is trained to establish a mapping model between the form features and the key emotional vocabulary; The mapping model is used to predict the emotional evaluation value of all form feature combinations within a preset range, determine the optimal form feature combination corresponding to the key emotional vocabulary, and generate at least one candidate design scheme according to the optimal form feature combination; Eye tracking technology is used to obtain visual perception data of the user on the candidate design scheme, and a discrete information data fluctuation weighting method is used to objectively and quantitatively analyze the candidate design scheme, and the visual perception data and the objective and quantitative analysis result are combined to determine the final new energy vehicle modeling design scheme from the candidate design scheme.

2. The method of claim 1, wherein, The step of calculating and determining the comprehensive importance weight of multiple emotional vocabulary comprises: Based on the improved CRITIC weight method, the Gini coefficient is used to replace the standard deviation to calculate the distribution difference of each emotional vocabulary under different Kano categories to determine its objective importance; Based on the least squares method, the fuzzy evaluation of experts on the importance of different Kano categories is combined to determine the satisfaction importance of each emotional vocabulary; The objective importance and the satisfaction importance are weighted and integrated to obtain the comprehensive importance weight.

3. The method according to claim 1 or 2, characterized in that, The step of using the interval 2-type trapezoidal fuzzy Kano model comprises: The language scale in the Kano questionnaire is quantified as an interval 2-type trapezoidal fuzzy number; The user uses the interval 2-type trapezoidal fuzzy number to answer the positive and negative questions of the emotional vocabulary to obtain fuzzy evaluation values; By calculating the distance between the fuzzy evaluation values and the fuzzy numbers corresponding to each language scale, the membership degree of adjacent language scales is determined, and a Kano category membership degree vector is generated to classify and calculate the weight of the emotional vocabulary.

4. The method of claim 1, wherein, The step of deconstructing the form features of the new energy vehicle sample set comprises: Using the form analysis method, the form of the new energy vehicle is divided into multiple design parts, and the design parts include the window, the lamp, the air intake grille, the hub and the rearview mirror; Multiple types of features are defined for each design part, and the type features are coded to form multi-dimensional form feature data.

5. The method of claim 1, wherein, The step of training an extreme gradient boosting model optimized by a hippo optimization algorithm comprises: The hyperparameters of the XGBoost model, including the number of trees, the depth of the tree and the learning rate, are used as individuals in the hippo population; The hyperparameter combination is iteratively updated by simulating the swimming stage and the walking stage of the hippo population, wherein the swimming stage is used for global exploration and the walking stage is used for local development; The mean square error is used as the fitness function to evaluate the performance of the XGBoost model corresponding to each hyperparameter combination until a preset iteration number is reached or the model performance converges, thereby determining the optimal hyperparameter combination.

6. The method of claim 1, wherein, The step of obtaining visual perception data of the candidate design scheme by using eye tracking technology comprises: Recording the eye movement trajectory of the user when observing the candidate design scheme, and counting the eye movement indicators including total fixation duration, average fixation duration, fixation count and first fixation duration; Based on the standardized average of the eye movement indicators, the visual perception priority order of the user for each candidate design scheme is determined.

7. The method of claim 1, wherein, The step of objectively quantitatively analyzing the candidate design scheme by using the discrete information data fluctuation weighting method comprises: Collecting the scores of multiple evaluators on each candidate design scheme for the key emotional words; Calculating the dispersion coefficient and independent information ratio of each candidate design scheme under the score of each key emotional word; Combining the dispersion coefficient and independent information ratio to calculate the pure information quantity of each candidate design scheme, and normalizing it to obtain the objective weight of each candidate design scheme; Based on the objective weight, the candidate design schemes are sorted.

8. A new energy vehicle modeling design system, characterized in that, Comprise: The emotional demand quantification module is configured to use the interval 2-type trapezoidal fuzzy Kano model to calculate and determine the comprehensive importance weight of multiple emotional words based on a preset new energy vehicle sample set and user Kano questionnaire survey data for emotional words, and to select a preset number of key emotional words with the highest weight; The mapping model construction module is configured to deconstruct the form features of the new energy vehicle sample set to obtain multi-dimensional form feature data; Based on the multi-dimensional form feature data and the user evaluation value for the key emotional words, an extreme gradient boosting model optimized by the Rhino optimization algorithm is trained to establish a mapping model between the form features and the key emotional words; The optimal scheme generation module is configured to use the mapping model to predict the emotional evaluation value of all form feature combinations within a preset range, determine the optimal form feature combination corresponding to the key emotional words, and generate at least one candidate design scheme according to the optimal form feature combination; The comprehensive evaluation and screening module is configured to obtain visual perception data of the candidate design scheme by using eye tracking technology, and objectively quantitatively analyze the candidate design scheme by using the discrete information data fluctuation weighting method, and determine the final new energy vehicle styling design scheme from the candidate design scheme based on the visual perception data and the objective quantitative analysis result.

9. An electronic device, comprising: Comprise a processor, a memory and a program or instruction stored on the memory and executable on the processor, when the program or instruction is executed by the processor, the steps of a new energy vehicle styling design method as claimed in any one of claims 1-7 are realized.

10. A readable storage medium, characterized by, The program or instruction is stored on the readable storage medium, and when the program or instruction is executed by the processor, the steps of a new energy vehicle styling design method as claimed in any one of claims 1-7 are realized.

Images