Product modeling design method and system

Abstract

The application provides a product modeling design method and system, comprising the following steps: obtaining a research object, establishing a product perceptual vocabulary library and a modeling sample library; performing preliminary screening on the product perceptual vocabulary library based on a word vector; scoring product modeling samples by using a semantic difference method; screening an advantage perception intention vocabulary by using a factor analysis method; obtaining product modeling characteristic morphological elements and coding by using a morphological analysis method; establishing a mapping relationship between the characteristic morphological elements and the product perceptual intention modeling elements based on a three-layer BP neural network model, and performing finite element analysis based on the screened modeling elements; performing topological optimization on force distribution characteristics of product modeling, coupling a topologically optimized structure model and product perceptual intention modeling elements; outputting a product modeling scheme; evaluating the product modeling scheme based on an eye tracking experiment, and generating a product modeling design scheme when the evaluation is passed.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of product design, and particularly relates to a product modeling design optimization method and system. BACKGROUND

[0002] Modeling design is an important part of product design and a main link in the product life cycle, which directly determines the performance of the product. With the development of science and technology and the complication of design objects, higher requirements are put forward for product design, so that the product modeling design cannot be evaluated by intuition and experience, but should be comprehensively and scientifically evaluated by using advanced theories and methods.

[0003] In extracting the perceptual image words of product modeling, the existing researches mostly use questionnaire survey method to extract, and the simple evaluation method has the disadvantages of poor precision and small objective proportion, so a reasonable processing and quantization method should be used to visualize the information.

[0004] In the field of industrial design, there is no reasonable and comprehensive evaluation method for the complex mapping problem of the appearance features such as product modeling and spatial layout and the perceptual cognition of users, most of the researches are subjective, which will affect the product design scheme, lead to the product designed to be unable to meet the emotional needs of users, and further lead to the modeling design and structural design of the product to be separated from each other. In the actual production stage of the product, it will lead to redundant use of materials and more consumption of materials. This will increase the production cycle and production cost of the product for the enterprise, and cannot adapt to the actual demand to improve the efficiency and quality of product design. SUMMARY

[0005] In view of the deficiencies in the prior art, the purpose of the present application is to provide a product modeling design method and system.

[0006] In order to achieve the above purpose, the present application is realized by the following technical scheme:

[0007] In the first aspect, the embodiment of the present application provides a product modeling design method, comprising the following steps:

[0008] Step 1: determining a target product, establishing a target product perceptual word library and a modeling sample library;

[0009] Step 2: preliminarily classifying and screening the product perceptual word library based on word vectors;

[0010] Step 3: scoring the product modeling sample by using semantic differential method based on the modeling sample library in step 1;

[0011] Step 4: screening the advantage perceptual intention words by using factor analysis method based on the product perceptual word library in step 2;

[0012] Step 5 obtains product modeling feature morphological elements and encodes based on the modeling sample library described in step 2 using morphological analysis method;

[0013] Step 6 establishes a mapping relationship between the encoded feature morphological elements and the perceptual intention modeling elements based on a three-layer BP neural network model, obtains screened modeling elements through the mapping relationship, and performs finite element analysis based on the screened modeling elements;

[0014] Step 7 performs topology optimization on the stress sub-feature of product modeling based on the finite element analysis of step 6, and couples the topology-optimized structure model with the product perceptual intention modeling elements;

[0015] Step 8 outputs a product modeling scheme;

[0016] Step 9 evaluates the product modeling scheme based on an eye tracking experiment, and generates a product modeling design scheme if the evaluation is passed.

[0017] In an implementation manner of the present application, step 1 specifically includes: collecting comment text data of a target product based on plane media, network media and investigation, etc., and performing word segmentation on the comment text data to obtain evaluation words; extracting high-frequency words in the evaluation words for evaluating appearance as center words, extracting adjectives in the evaluation words to obtain adjective words, collecting perceptual words and modeling samples of the target product, and establishing a material library.

[0018] In an implementation manner of the present application, step 2 specifically includes: converting the evaluation words into word vectors, calculating the similarity of each adjective word and the center word based on the word vectors, and extracting corresponding adjective words as initial perceptual image words according to the similarity.

[0019] In an implementation manner of the present application, step 3 specifically includes formulating a seven-point quantification table of semantic difference based on the preliminary screened perceptual intention words, and letting the subjects score the modeling samples through the seven-point quantification table.

[0020] In an implementation manner of the present application, step 4 specifically includes: analyzing the average value of the scores of the seven-point quantification table of semantic difference as data basis through a factor analysis method in SPSS software, extracting main components, and finally obtaining dominant perceptual intention words.

[0021] In an implementation manner of the present application, step 5 specifically includes: disassembling product components in a product modeling feature library according to product functions and morphological features, obtaining product modeling component morphological elements, and adopting binary encoding for a typical modeling of each component.

[0022] In an implementation form of the present application, the step 6 specifically comprises: taking product modeling component coding as an input layer, taking advantage perception intention vocabulary as an output layer, taking a Log-sigmoid function as an implicit layer, taking a purelin transfer function as an output layer, training samples, and establishing a mapping relationship between product modeling elements and emotional intention; constructing a geometric model of the screened modeling elements through a three-dimensional modeling software, setting material properties of the geometric model through an Ansys software, then performing mesh division and quality detection on the model, and performing finite element load stress analysis after the detection is passed.

[0023] In an implementation form of the present application, the step 7 specifically comprises: according to product modeling stress sub-feature, first determining an optimization design region to prevent model boundaries and key connecting components from being optimized out; secondly defining a constraint condition to determine an optimal product modeling optimization result; coupling the optimized product modeling elements with the modeling elements conforming to the user emotional intention obtained above to design multiple product modeling schemes conforming to the user emotion.

[0024] In an implementation form of the present application, in the step 8, the product modeling scheme is evaluated based on an eye movement tracking experiment, specifically comprising: first obtaining multiple eye movement index data through an eye movement experiment, and secondly establishing a relationship model between the eye movement data and the evaluation value through an SVM mathematical model to score the product modeling scheme, and obtaining an optimal product model scheme based on the score result.

[0025] In a second aspect, the embodiments of the present application further provide a product modeling design system, comprising the following modules.

[0026] The obtaining module is configured to obtain a research object, establish a product emotional vocabulary library and a modeling sample library.

[0027] The preliminary screening module is configured to preliminarily screen the product emotional vocabulary library based on a word vector.

[0028] The scoring module is configured to score the product modeling sample by using a semantic difference method.

[0029] The screening module is configured to screen advantage perception intention vocabulary based on a factor analysis method.

[0030] The feature form element obtaining and coding module is configured to obtain product modeling feature form elements and code based on the product modeling sample library by using a form analysis method.

[0031] The mapping relationship establishing and analyzing module is configured to establish a mapping relationship between product modeling elements and emotional intention based on a three-layer BP neural network model, and perform finite element analysis based on the screened modeling elements.

[0032] The topological optimization and coupling module is configured to perform topological optimization based on force distribution characteristics of the product modeling, and to perform coupling based on a structure model after topological optimization and product emotional intention modeling elements.

[0033] The output module is configured to output the product modeling scheme.

[0034] The generation module is configured to evaluate the product modeling scheme based on the eye tracking experiment, and to generate a product modeling design scheme if the evaluation is passed.

[0035] The beneficial effects of the above embodiments of the present application are as follows:

[0036] The product modeling design optimization method proposed in the present application uses BP neural network and topological optimization for product modeling design, realizes the change of design method from emotion to reason, reduces the influence of designer's subjective factors on product design scheme in the design process, and effectively improves the rationality of product structure design, ensuring the accuracy and efficiency of product modeling design. Coupling based on the structure model after topological optimization and the product emotional intention modeling elements effectively solves the problem of mutual separation of product modeling design and structure design, and reduces the problem of redundant material use and high material consumption in modeling design. The eye tracking experiment is used to evaluate the product modeling design scheme, and the SVM is used to establish a mathematical model between eye movement data and product modeling design evaluation, which objectively evaluates the pros and cons of product modeling design, and better promotes the change of product modeling design oriented by user demand. The method proposed in the present application effectively reduces the production cycle and production cost of products, improves the speed of product modeling design scheme, and provides a new method for product modeling design oriented by consumer spiritual needs. BRIEF DESCRIPTION OF DRAWINGS

[0037] The drawings accompanying the specification of the present application serve to provide a further understanding of the present application, and the illustrative embodiments of the present application and their descriptions serve to explain the present application, and do not constitute an improper limitation on the present application.

[0038] Figure 1 is the overall design flowchart of the present application;

[0039] Figure 2 is a processing process schematic diagram of step 9 of the present application;

[0040] Figure 3 is a gravel map;

[0041] Figure 4 Network training results;

[0042] Figure 5 Goodness-of-fit simulation;

[0043] Figure 6 Front headlamp design area

[0044] Figure 7(a) Car light modeling

[0045] Figure 7(b) Front cover design white film

[0046] Figure 8(a), Figure 8(b) Front cover design white model

[0047] Figure 9(a), Figure 9(b), Figure 9(c) Cab design white model

[0048] Figure 10 Roof light design white model

[0049] Figure 11 Figure 12 Final tractor model DETAILED DESCRIPTION

[0050] It should be noted that the following detailed description is illustrative only and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0051] It should be noted that the terms used herein are only intended to describe specific embodiments and are not intended to limit exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should also be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of a feature, step, operation, device, component and / or combination thereof;

[0052] As introduced in the background, there are deficiencies in the prior art, in order to solve the above technical problems, the present application proposes a product modeling design method and system.

[0053] In order to better meet the emotional needs of users in product modeling design, the embodiment needs to obtain the relationship between accurate emotional intention words and product modeling mapping. For this purpose, the embodiment is based on the preliminary screening of product emotional word library based on word vector, and uses semantic difference method to score, and then performs secondary screening based on factor analysis method to obtain representative advantage emotional intention words. Effectively solve the problem of lack of representativeness and strong subjective speculation when only relying on questionnaire survey method to obtain emotional intention words. Provide visual and quantifiable data support.

[0054] ​In order to improve the prediction accuracy of product emotional intention, the mapping relationship between product modeling elements and emotional intention is quantitatively converted, and the best quantitative combination of product modeling elements is realized. In view of this, on the basis of obtaining product modeling component form elements from the product modeling feature library and coding, the mapping relationship between product modeling elements and emotional intention vocabulary is established by using a three-layer BP neural network model, the product modeling image prediction is realized, and objective data support is provided for selecting and reasonably combining the modeling form of each product component. The problems that the relationship between the two mapping parties cannot be accurately described and the subjectivity is serious in the evaluation process are effectively solved.

[0055] In order to improve the structural strength of the modeling element, reduce the excessive use of materials in modeling design, and solve the problem of too miscellaneous use of materials, in view of this, on the basis of finite element analysis of the screened modeling elements, the topological optimization of the stress distribution characteristics of product modeling is carried out, so that the design position of product modeling elements can be reasonably divided according to the stress characteristics of product structure.

[0056] In order to avoid the disconnection between product modeling design and structural design, in view of this, the emotional image product modeling element is coupled with the optimized product structure, so that a product modeling design scheme with reasonable structure and meeting the emotional needs of users can be efficiently obtained, and the production cycle and production cost of the product are effectively reduced.

[0057] After obtaining the product modeling scheme, in order to solve the problem of strong subjectivity and lack of objective theoretical support in the process of obtaining user emotional cognitive image, the bottom support of cognitive origin is provided for design, in view of this, on the basis of obtaining multiple eye movement index data through eye movement experiment, the relationship model between eye movement data and evaluation value is established by using SVM mathematical model, the product modeling scheme is scored, and the optimal product model scheme is obtained based on the score result. The design method is changed from emotion to reason, the accuracy and efficiency of product modeling design are ensured, and the product design can better adapt to customer demand and market demand.

[0058] Specifically, the product modeling design optimization method provided by the embodiment of the present application comprises the following steps:

[0059] Step 1: obtaining a research object, establishing a product emotional vocabulary library and a modeling sample library;

[0060] Step 2: preliminarily screening the product emotional vocabulary library based on a word vector;

[0061] Step 3: scoring and classifying the product modeling sample by using a semantic differential method (SD analysis method);

[0062] Step 4: screening the advantage perception intention vocabulary based on the factor analysis method;

[0063] Step 5: obtaining the product modeling feature form elements and coding based on the product modeling sample library and using the form analysis method;

[0064] Step 6: establishing the mapping relationship between the product modeling elements and the perceptual intention based on the three-layer BP neural network model; and performing finite element analysis based on the screened modeling elements;

[0065] Step 7: performing topology optimization based on the stress sub-feature of the product modeling; and coupling the structure model after topology optimization with the product perceptual intention modeling elements;

[0066] Step 8: outputting the product modeling scheme;

[0067] Step 9: evaluating the product modeling scheme based on the eye tracking experiment, and generating the product modeling design scheme if the evaluation is passed.

[0068] The above determining the research object, establishing the product perceptual vocabulary library and the modeling sample library, specifically includes: collecting the perceptual vocabulary and modeling sample of the target product based on the plane media, network media and investigation, and establishing the material library.

[0069] The above preliminary screening of the product perceptual vocabulary library and the modeling sample library based on the word vector, specifically includes: converting the evaluation vocabulary into a word vector, calculating the similarity of each adjective vocabulary and the center vocabulary based on the word vector, and extracting the corresponding adjective vocabulary as the initial perceptual image vocabulary according to the similarity.

[0070] The above scoring the product modeling sample by using the semantic differential method (SD analysis method), specifically includes: formulating a semantic differential seven-point quantification table based on the preliminary screened perceptual intention vocabulary, and letting the subjects score the modeling sample through the seven-point quantification table.

[0071] The above screening the advantage perception intention vocabulary based on the factor analysis method, specifically includes: analyzing the average value of the semantic differential seven-point quantification table score as data basis by using the factor analysis method in the SPSS software, extracting the main components, and finally obtaining the advantage perception intention vocabulary.

[0072] The above obtaining the product modeling feature form elements and coding based on the product modeling sample library and using the form analysis method, specifically includes: disassembling the product components in the product modeling feature library according to the product function and form characteristics, obtaining the product modeling component form elements, and using binary coding for the typical modeling of each component.

[0073] The mapping relationship between the product modeling elements and the perceptual intention is established based on the three-layer BP neural network model, and specifically includes: a product modeling component code as an input layer, an advantage perception intention vocabulary as an output layer, an implicit layer using a Log-sigmoid function, and an output layer using a purelin transfer function. The mapping relationship between the product modeling elements and the perceptual intention is established by training the samples.

[0074] The finite element analysis based on the screened modeling elements includes: constructing a geometric model of the screened modeling elements through a three-dimensional modeling software, setting material properties of the geometric model through Ansys software, then dividing the model into grids and performing quality detection, and performing finite element load stress analysis after the detection is passed.

[0075] The topology optimization based on the stress division features of the product modeling includes: determining the optimization design area according to the stress division features of the product modeling to prevent the model boundary and key connecting components from being optimized out, and then defining the constraint conditions to determine the optimal product modeling optimization result.

[0076] The coupling between the structure model after the topology optimization and the product perceptual intention modeling elements includes: coupling the optimized product modeling elements with the modeling elements obtained above that meet the user's perceptual intention to design multiple product modeling schemes that meet the user's emotions.

[0077] The product modeling scheme is evaluated based on the eye tracking experiment, and specifically includes: first, obtaining multiple eye movement index data through an eye movement experiment, and then establishing a relationship model between the eye movement data and the evaluation value by using an SVM mathematical model to score the product modeling scheme, and obtaining the optimal product model scheme based on the score result.

[0078] Further, the embodiment also provides a product modeling design system, including the following modules:

[0079] The acquisition module is configured to acquire a research object, and establish a product perceptual vocabulary library and a modeling sample library.

[0080] Specifically, the module is configured to establish a product perceptual vocabulary library and a modeling sample library, and specifically includes: collecting perceptual vocabulary and modeling samples of target products based on plane media, network media, and investigation, and establishing a material library.

[0081] The preliminary screening module is configured to preliminarily screen the product emotional vocabulary library and the modeling sample library based on a questionnaire method; specifically, the module is configured to convert the evaluation vocabulary into a word vector, calculate the similarity of each adjective vocabulary and the center vocabulary based on the word vector, and extract the corresponding adjective vocabulary as an initial emotional image vocabulary according to the similarity.

[0082] The scoring module is configured to score the product modeling sample by using a semantic difference method; specifically, the module is configured to develop a semantic difference seven-point quantification table based on the emotional intention vocabulary after preliminary screening, and let the subjects score the modeling sample through the seven-point quantification table.

[0083] The screening module is configured to screen the advantage perception intention vocabulary based on a factor analysis method; specifically, the module is configured to analyze the average value of the semantic difference seven-point quantification table score as data basis by using the factor analysis method in the SPSS software, extract the main constituent components, and finally obtain the advantage perception intention vocabulary.

[0084] The characteristic morphological element acquisition and coding module is configured to acquire and code the product modeling characteristic morphological elements based on the product modeling sample library by using a morphological analysis method; specifically, the module is configured to disassemble the product components in the product modeling characteristic library according to the product function and morphological characteristics, obtain the product modeling component morphological elements, and use binary coding for the typical modeling of each component.

[0085] The mapping relationship establishment and analysis module is configured to establish the mapping relationship between the product modeling elements and the emotional intention based on a three-layer BP neural network model, and perform finite element analysis based on the screened modeling elements; specifically, the module is configured to use the product modeling component code as the input layer, the advantage perception intention vocabulary as the output layer, use the Log-sigmoid function for the hidden layer, and use the purelin transfer function for the output layer, train the sample, establish the mapping relationship between the product modeling elements and the emotional intention, construct the geometric model of the screened modeling elements by using a three-dimensional modeling software, set the material properties of the geometric model by using the Ansys software, then divide the model into grids and perform quality detection, and perform finite element load stress analysis after the detection is passed.

[0086] The topological optimization and coupling module is configured to perform topological optimization based on stress distribution characteristics of the product shape, and to perform coupling based on a structure model after topological optimization and product emotional intention shape elements; specifically, the module is configured to first determine an optimization design region according to stress distribution characteristics of the product shape, to prevent model boundaries and key connection components from being optimized out; second, to define constraint conditions and determine an optimal product shape optimization result; and to perform coupling of the optimized product shape elements and the obtained shape elements conforming to the user's emotional intention to design multiple product shape schemes conforming to the user's emotion.

[0087] The output module is configured to output the product shape scheme.

[0088] The generation module is configured to evaluate the product shape scheme based on an eye tracking experiment, and to generate a product shape design scheme if the evaluation passes; specifically, the module is configured to first obtain multiple eye movement index data through an eye tracking experiment, second, to establish a relationship model between eye movement data and evaluation values by an SVM mathematical model, to score the product shape scheme, and to obtain an optimal product model scheme based on the score result, as shown in Figure 2 .

[0089] The above only describes preferred embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various modifications and changes to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0090] Example 1

[0091] The specific steps and processes of the product shape design method and system disclosed in the embodiment are described in detail by taking a tractor shape as a target product.

[0092] 1. Establish a target product emotional vocabulary library and a shape sample library, and the specific steps are as follows:

[0093] Taking "tractor" as a keyword, product review data related to the tractor shape is collected by using Python tools to grab the data through relevant data models, papers, journals, tractor sales websites, etc. Adjectives related to the tractor shape and tractor shape samples are collected to establish a product emotional vocabulary corpus and a shape sample library. As shown in Tables 1 and 2.

[0094] Table 1 Tractor Emotional Vocabulary Library

[0095]

[0096] Table 2 Tractor Shape Sample Library

[0097]

[0098] 2. The product emotional vocabulary library is preliminarily classified and screened based on a clustering analysis method, and the specific steps are as follows:

[0099] In order to narrow the range of emotional image, the cosine similarity between the word vector corresponding to the center vocabulary and the word vector corresponding to each adjective vocabulary is calculated, the calculation result is taken as the similarity of the center vocabulary and the adjective vocabulary, the adjective vocabulary with a similarity exceeding a preset similarity threshold is extracted as a relevant vocabulary, and the word frequency of each relevant vocabulary in the evaluation vocabulary is obtained; the relevant vocabularies corresponding to each center vocabulary are merged, and the relevant vocabulary with a word frequency exceeding a preset word frequency threshold is extracted, and an initial emotional image vocabulary is obtained.

[0100] 3. The product modeling sample is scored by using a semantic difference method (SD analysis method), and the specific steps are as follows:

[0101] (1) Emotional vocabulary re-screening

[0102] The re-screening of the emotional vocabulary is the premise of formulating the seven-point quantification table. Therefore, the emotional vocabulary is screened from a professional point of view by taking product designers and researchers as the main interviewees. According to previous experience, the number of core emotional vocabularies is concentrated in the range of 5 to 10, which is convenient for the subsequent formulation of the semantic difference seven-point quantification table. First, 21 graduate students with design background and 9 relevant professional practitioners are selected in the campus as the main body for investigation and research. Each interviewee can select 7 groups of vocabularies, and the data obtained are compared with the frequency of selection of each emotional vocabulary and the concentrated discussion of the interviewees. We finally select seven groups of core emotional vocabularies, namely, "technological-traditional", "high-end-low-end", "powerful-mild", "light-heavy", "massive-individual", "dynamic-static", and "beautiful-ugly".

[0103] (2) Formulation of seven-point quantification table and scoring

[0104] Finally, the seven groups of emotional vocabulary groups are drawn into the semantic difference seven-point quantification table. In the table, the emotional vocabularies of opposite meanings in the same group are arranged on both sides, and seven levels of -3, -2, -1, 0, 1, 2, and 3 are set to score the emotional image of the sample modeling, which is an important step to quantify the emotional image of the product. Through the questionnaire survey of 30 testers with design background such as design professional teachers and students in the school and enterprise modeling designers, the semantic difference seven-point quantification table is used to score the electric tractor modeling samples (as shown in Table 3). In order to make the scores of the interviewees more intuitive, the scores of the 30 modeling samples are processed by average value (as shown in Table 4), and finally the scoring data of the 30 modeling samples in the modeling sample library are obtained.

[0105] Table 3 Semantic difference seven-point quantification table

[0106]

[0107] Table 4 Average score of semantic differential evaluation

[0108]

[0109] (3) Classification of modeling samples

[0110] You also classify the 30 samples into 7 groups according to the scores of the 7-point quantification table. According to the data in Table 4, the highest scoring perceptual word in each modeling sample is used as the classification basis for the classification of the modeling samples in Table 2. The specific classification is shown in Table 5:

[0111] Table 5 Classification of modeling samples

[0112]

[0113] 4. The advantage perception intention word is screened based on the factor analysis method, and the specific steps are as follows:

[0114] The average value of the scores of the semantic differential 7-point quantification table is analyzed by applying the factor analysis method in the SPSS software to extract the main constituent components. Finally, the advantage perception intention word that can best represent the perceptual image of the tractor is obtained.

[0115] Firstly, the KMO test and Bartlett's sphericity test are used to test the data model, and the results are shown in Table 6. After the results show that the data model meets the requirements of factor analysis, the data is analyzed by factor analysis method to obtain the gravel map of the perceptual intention information of the tractor (Table 7), total variance explanation (Table 7) and component matrix after rotation (Table 8). Figure 3

[0116] Table 6 KMO and Bartlett test

[0117]

[0118] Table 7 Total variance explanation

[0119]

[0120] Table 8 Component matrix after rotation

[0121]

[0122] ​In the rotated component matrix (as shown in Table 8), the closer the loading coefficient of a component to 1, the more it represents the perceptual image of the electric tractor. The loading coefficients of components 1, 2, and 3 are 0.864, 0.952, and 0.969, respectively, which are the largest among all the loading coefficients, indicating that these three components can represent the perceptual image of the tractor. Therefore, the dominant perceptual words that conform to the perceptual image of the tractor are "light, powerful, and dynamic".

[0123] 5. Based on the product modeling sample library, the product modeling feature form elements are obtained by using the form analysis method and are coded. The specific steps are as follows:

[0124] Based on the form analysis method, the product form elements in the tractor product modeling sample library are divided. According to the product function and form characteristics, the tractor is disassembled into 5 parts, each modeling part belongs to no more than 7 design features and no less than 4 design features. At the same time, the binary coding is used for the typical modeling of each part, as shown in Table 9.

[0125] Table 9 Modeling feature library

[0126]

[0127] 6. Based on the three-layer BP neural network model, the mapping relationship between the coded modeling feature library and the perceptual intention modeling elements is established. The screened modeling elements are obtained through the mapping relationship. Based on the screened modeling elements, the finite element analysis is carried out. The specific steps are as follows:

[0128] Combined with the modeling feature library, the Delphi method is used to select the modeling samples multiple times to determine the final dominant modeling sample library, which is coded. Through the 1-7 Likert scale, 20 non-design personnel and 7 design professionals are invited to conduct a second image evaluation experiment. The evaluation object is the screened dominant tractor samples. The evaluation results are processed by averaging, and the preference degree of users to the dominant samples is analyzed. Table 10 shows part of the data of samples 1-10.

[0129] Table 10 Part of sample evaluation experiment and sample coding

[0130]

[0131] According to the above analysis, the network structure is constructed based on Matlab R2018a software platform for simulation test. After pre-test of historical data for many times, the training effect of 3-layer neural network structure is better, 15-bit 0 or 1 code input, i.e. 15 nodes in the input layer, 3 dominant perception intention outputs, i.e. 3 nodes in the output layer, 11 hidden layer nodes are determined by formula (1). The output layer adopts purelin transfer function, formula (2), the network learning times are set to 10000, the step is 0.7, the additional momentum factor is 0.2, the error target value is 0.003, and the training output error is measured by mean square error. The first 20 sample data in table 10 are fitted and trained, and the training results are shown in Figure 4 、 Figure 5 , which verifies the effectiveness of the network training model. The design feature codes of the 8 samples numbered 41~48 in table 10 are imported into the network input layer, and the performance detection is evaluated by mean square error analysis function MSE, expression (3)

[0132] (1)

[0133] Wherein: G is the hidden layer neuron node, N is the input layer neuron node, H is the output layer neuron node, and a is usually an adjustment constant between 1-10.

[0134] (2)

[0135] (3)

[0136] Wherein: y r is the network output value, y ※ r is the target evaluation value.

[0137] In order to make the design output image modeling meet the emotional needs of consumers, based on the neural network algorithm of Matlab program, the maximum value of image evaluation is calculated by traversal. The calculation results are shown in table 11.

[0138] Table 11 Network test results

[0139]

[0140] Therefore, the optimal combination of tractor image modeling "powerful intention" elements can be deduced, as shown in table 12.

[0141] Table 12 Optimal combination of tractor image modeling "powerful intention" elements

[0142]

[0143] Based on the BP neural network selected modeling elements, at the same time for the subsequent structure optimization design, in the three-dimensional software to the selected modeling of simplified parametric modeling;

[0144] Before the finite element analysis of the parametric model, the material properties of the cab and the front hood are set. According to the characteristics of the working environment of the tractor, the front hood and the cab are required to have certain structural strength, so the structural steel with a yield strength of 235 MPa is selected in this paper. The specific material parameters are shown in Table 12, and the material properties are set in Ansys;

[0145] Table 12 Material properties of structural steel

[0146]

[0147] After the model is established, in order to improve the calculation efficiency and adaptability of the subsequent stress analysis of the model, tetrahedral mesh is used as the division unit to divide the mesh of the model.

[0148] After the mesh of the model is divided, the mesh is inspected based on the functions provided in Ansys software. After the inspection is qualified, combined with the actual use of the tractor, three limit conditions are set to analyze the structure of the front hood and the cab by finite element method. The three limit conditions are: sudden braking condition in driving state (ignoring the occurrence time), sudden turning condition (ignoring the occurrence time) and static state with heavy load on the top of the model.

[0149] The surface load of the model under the three limit conditions is set; after the parameters are set, the finite element mechanics analysis of the model under the three limit conditions is carried out based on Ansys software, and the stress distribution characteristics are obtained.

[0150] 7. Based on the finite element analysis of step 6, the stress distribution characteristics of the product modeling are optimized, and the structure model after topology optimization is coupled with the product perceptual intention modeling elements, which are as follows:

[0151] Based on the perceptual intention words obtained by factor analysis and the topology optimization results, the cab is modeled in detail;

[0152] Based on the structure optimization model obtained by finite element analysis and topology optimization of the front hood, the design area of the car lamp is reasonably divided (such as Figure 6 ), and the car lamp model is coupled with the front hood based on the car lamp elements obtained by three-layer BP neural network as shown in Fig. 7(a) and Fig. 7(b).

[0153] Similarly, based on the structural optimization model obtained from the finite element analysis and topology optimization of the front hood and cab, the grille, windows, and roof lights are designed and coupled to obtain the final effect model, as shown in Figures 8(a), 8(b), 9(a), 9(b), and 9(c). Figure 10 As shown.

[0154] Based on the coupling results of model elements and combined with the design of secondary elements, the optimized design of the tractor shape is completed, with the following effect: Figure 11 , Figure 12 As shown.

Claims

1. A product styling design method characterized by comprising: The method comprises the following steps: Step 1: determining a target product, establishing a target product perceptual vocabulary library and a modeling sample library; Step 2: performing preliminary classification and screening on the product perceptual vocabulary library based on word vectors; Step 3: scoring product modeling samples based on the modeling sample library in step 1 by using a semantic difference method; Step 4: screening advantage perception intention vocabulary by using a factor analysis method based on the product perceptual vocabulary library in step 2; Step 5: obtaining product modeling feature morphological elements and encoding by using a morphological analysis method based on the modeling sample library in step 1; Step 6: establishing a mapping relationship between the encoded feature morphological elements and perceptual intention modeling elements based on a three-layer BP neural network model, obtaining screened modeling elements through the mapping relationship, and performing finite element analysis based on the screened modeling elements; The step 6 specifically comprises: taking the coding of product modeling components as an input layer, taking the advantage perception intention vocabulary as an output layer, taking a Log-sigmoid function as an implicit layer, taking a purelin transfer function as an output layer, training samples, and establishing a mapping relationship between product modeling elements and perceptual intentions; constructing a geometric model of the screened modeling elements through a three-dimensional modeling software, setting material properties of the geometric model through an Ansys software, then performing mesh division and quality detection on the model, and performing finite element load stress analysis after the detection is passed; Step 7: based on the finite element analysis in step 6, performing topology optimization on the stress sub-feature of product modeling, and coupling the topology-optimized structure model with the product perceptual intention modeling elements; The step 7 specifically comprises: determining an optimization design region according to the stress sub-feature of product modeling, to prevent the model boundary and key connecting components from being optimized out; defining constraint conditions to determine an optimal product modeling optimization result; coupling the optimized product modeling elements with the modeling elements meeting the user perceptual intention, to design multiple product modeling schemes meeting the user emotion; Step 8: outputting product modeling schemes; Step 9: evaluating the product modeling schemes based on an eye tracking experiment, and generating a product modeling design scheme in the case of passing the evaluation.

2. The product design method according to claim 1, wherein Step 1 specifically comprises: collecting comment text data of a target product based on a plane medium, a network medium and a survey method, performing word segmentation on the comment text data to obtain evaluation vocabulary, extracting high-frequency vocabulary in the evaluation vocabulary for evaluating appearance as central vocabulary, extracting adjectives in the evaluation vocabulary to obtain adjective vocabulary, collecting perceptual vocabulary and modeling samples of the target product, and establishing a material library.

3. The product design method according to claim 2, wherein Step 2 specifically comprises: converting the evaluation vocabulary into word vectors, calculating the similarity of each adjective vocabulary and the central vocabulary based on the word vectors, and extracting corresponding adjective vocabulary as initial perceptual image vocabulary according to the similarity.

4. The product design method according to claim 1, wherein Step 3 specifically comprises: formulating a seven-point quantification table based on the perceptual intention vocabulary after preliminary screening, and letting subjects score the modeling samples through the seven-point quantification table.

5. The product design method according to claim 1, wherein The step 4 specifically comprises: taking the average value of the scores of the semantic differential seven-point quantification table as data basis to analyze by factor analysis method, extracting the main constituent components, and finally obtaining the dominant perception intention vocabulary.

6. The product design method according to claim 1, wherein The step 5 specifically comprises: in the product modeling feature library, the product components are disassembled according to the product function and shape characteristics, the product modeling component shape elements are obtained, and binary coding is used for the typical modeling of each component.

7. The product design method according to Claim 1, wherein In the step 8, the product modeling scheme is evaluated based on the eye tracking experiment, specifically comprising: first, obtaining multiple eye movement index data through the eye movement experiment, and second, establishing a relationship model between the eye movement data and the evaluation value by the SVM mathematical model, scoring the product modeling scheme, and obtaining the optimal product model scheme based on the score result.

8. A system for implementing the product design method according to any one of claims 1 to 7, characterized in that, The method comprises the following modules: An acquisition module configured to acquire research objects, establish a product emotional vocabulary library and a modeling sample library; A preliminary screening module configured to preliminarily screen the product emotional vocabulary library and the modeling sample library based on a questionnaire survey method; A scoring module configured to score the product modeling sample by using a semantic differential method; A screening module configured to screen the dominant perception intention vocabulary based on a factor analysis method; A feature shape element acquisition and coding module configured to acquire product modeling feature shape elements and code them based on the product modeling sample library by using a shape analysis method; A mapping relationship establishment and analysis module configured to establish a mapping relationship between the product modeling elements and the emotional intention based on a three-layer BP neural network model, and perform finite element analysis based on the screened modeling elements; A topology optimization and coupling module configured to perform topology optimization based on the stress distribution characteristics of the product modeling, and perform coupling based on the structure model after topology optimization and the product emotional intention modeling elements; An output module configured to output a product modeling scheme; A generation module configured to evaluate the product modeling scheme based on an eye tracking experiment, and generate a product modeling design scheme if the evaluation is passed.

Images