Industrial design component library management and automatic matching method and system

Abstract

The invention discloses an industrial design component library management and automatic matching method and system, and relates to the technical field of XXX industrial design, and the method comprises the following steps: obtaining the dimension attribute of a component, and carrying out the fusion modeling of the dimension attribute of the component; obtaining design requirements, extracting keywords according to the design requirements, matching the keywords with semantic tags in the component library, and obtaining semantic similarity based on a cosine similarity algorithm; constructing a matching model according to the semantic similarity in combination with the dimension attributes of the components, and comprehensively evaluating the dimension attributes of the components by using a weighted multi-dimensional data fusion technology to obtain matching degree scores of the components and the design requirements; and comparing the matching degree score with a preset matching threshold value, evaluating an automatic matching effect according to a comparison result, and dynamically updating the attribute weight of the component according to the automatic matching effect. According to the method, the problem of separation of semantic understanding and attribute matching is solved, rapid matching of components and design requirements is realized, the design period is remarkably shortened, and the design efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial design, and more specifically, to a method and system for managing and automatically matching an industrial design component library. Background Art

[0002] A component library is a database containing various design components. For effective management, the library usually includes the metadata of the components and the relationships between the components. In the design processes of many industries such as architecture and mechanical manufacturing, designers need to select appropriate components from a vast component library to meet the design requirements. Traditional matching methods mainly rely on manual screening. Designers need to view the detailed information of the components one by one and compare it with the design requirements. This way not only consumes a large amount of time and energy, but also is easily affected by human factors, resulting in inaccurate matching results and greatly reducing the design efficiency.

[0003] Deficiencies of the prior art:

[0004] Some existing automated matching technologies often separately process semantic understanding and component attribute matching. In terms of semantic matching, it simply relies on keyword matching and cannot deeply understand the semantic connotations of design requirements, resulting in low precision of semantic matching. In terms of attribute matching, the mutual relationships between multiple-dimensional attributes are not fully considered, and the components cannot be comprehensively and accurately evaluated, making it difficult to meet the requirements of complex design scenarios. With the continuous changes in design projects and the continuous update of the component library, the original matching model is difficult to adapt to new requirements and data. Existing matching systems lack an effective weight update mechanism and cannot dynamically adjust the model parameters according to the actual matching effect, resulting in a gradual decline in the accuracy and efficiency of matching.

[0005] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0006] To overcome the above defects of the prior art, embodiments of the present invention provide a method and system for managing and automatically matching an industrial design component library. By constructing an automated matching system and through the collaborative work of a fusion module, a matching module, an evaluation module, and an update module, rapid matching of components and design requirements is achieved, significantly shortening the design cycle and improving the design efficiency to solve the problems raised in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] An industrial design component library management and automatic matching method, comprising the following steps:

[0009] Obtain the dimensional attributes of the components and perform fusion modeling on the dimensional attributes of the components;

[0010] Obtain design requirements, extract keywords according to the design requirements, match them with the semantic tags in the component library, and obtain the semantic similarity based on the cosine similarity algorithm;

[0011] Construct a matching model according to the semantic similarity combined with the dimensional attributes of the components, and use the weighted multi-dimensional data fusion technology to comprehensively evaluate the dimensional attributes of the components to obtain the matching degree score between the components and the design requirements;

[0012] Compare the matching degree score with a preset matching threshold, evaluate the automatic matching effect according to the comparison result, and dynamically update the attribute weights of the components according to the automatic matching effect.

[0013] In a preferred embodiment, the process of obtaining the dimensional attributes of the components and performing fusion modeling on the dimensional attributes of the components is as follows:

[0014] Extract the dimensional attribute data of each component from the component library, including size information, material properties, and performance parameters;

[0015] Perform min-max normalization on the dimensional attribute data to scale the data to the range [0, 1] to ensure that the data is on the same scale;

[0016] Combine the dimensional attribute data of each component into a multi-dimensional feature vector V i ={a i1 , a i2 ,...a in}, where a ij is the dimensional attribute of the component.

[0017] In a preferred embodiment, the process of extracting keywords according to the design requirements is as follows:

[0018] According to the design requirement text, use a keyword extraction method based on the attention mechanism to extract keywords through natural language processing technology, and obtain a word sequence {c1, c2,...c n} after word segmentation and part-of-speech tagging; where c j is a word or phrase in the text;

[0019] For each word c j , use the pre-trained Word2Vec word vector model to obtain the word vector h j ;

[0020] If the dimension of the word vector obtained by model training is d, for the j-th word c n in the word sequence {c1, c2,...c j}, the word vector h j ={h j1 , h j2,...h jd};

[0021] Calculate the attention score Q for each word j , compare and analyze the attention score with a preset attention score threshold, and when the attention score is greater than the preset attention score threshold, use the word corresponding to the attention score as a keyword.

[0022] In a preferred embodiment, the process of calculating the attention score for each word is as follows:

[0023] Obtain historical design requirements and label the keywords as the training dataset for the weight matrix w and the bias term B;

[0024] Define a cross-entropy loss function L to measure the difference between the keywords predicted by the model and the actually labeled keywords;

[0025] Use the gradient descent algorithm to minimize the cross-entropy loss function L and continuously iterate to update the weight matrix w and the bias term B;

[0026] Transform the word vector h through a fully connected layer j and calculate the score e by combining the weight matrix w and the bias term B j ;

[0027] According to the score e j Calculate the attention score for each word.

[0028] In a preferred embodiment, the semantic similarity calculation process is as follows:

[0029] Weighted sum the word vectors of each keyword according to the attention score to construct the semantic vector d of the design requirement;

[0030] For the semantic label of the i-th component in the component library, tokenize the semantic label to obtain a word sequence, generate a word vector for each word, calculate the attention score, and weighted sum to obtain the semantic label vector s of the component;

[0031] Use the classical cosine similarity formula to calculate the cosine similarity between the design requirement semantic vector and the component semantic label vector, and combine the attention score Q of the keyword j , to obtain the semantic similarity S i .

[0032] In a preferred embodiment, the matching degree score process between the component and the design requirement is as follows:

[0033] The weight vector representation of the component dimension attribute is w = {w1, w2... w n} where n is the number of component dimension attributes, w jRepresents the weight of the j-th dimensional attribute;

[0034] The vector representation of the component dimensional attribute is V i = {a i1 , a i2 ,... a in}, where a ij represents the value of the j-th dimensional attribute of the i-th component;

[0035] According to the principle of weighted multi-dimensional data fusion, the semantic similarity and the component dimensional attributes are linearly weighted and summed to obtain the matching degree score between the i-th component and the design requirements.

[0036] In a preferred embodiment, the matching degree score is compared with a preset matching threshold, and the process of evaluating the automatic matching effect according to the comparison result is as follows:

[0037] Compare the matching degree score with the preset matching threshold. If the matching degree score is greater than or equal to the preset matching threshold, it is considered that the component is successfully automatically matched with the design requirements;

[0038] If the matching degree score is less than the preset matching threshold, it is considered that the component fails to be automatically matched with the design requirements.

[0039] In a preferred embodiment, the process of dynamically updating the attribute weight of the component according to the automatic matching effect is as follows:

[0040] If the component is successfully matched with the design requirements, let the reward function R be 1;

[0041] If the component fails to be matched with the design requirements, let the reward function R be 0;

[0042] Taking the success or failure of the match as a reward signal, the weight vector w is updated using the Q-learning algorithm according to the reward function R. By continuously iterating and optimizing the weight vector w, the matching effect is continuously improved.

[0043] In a preferred embodiment, the process of updating the weight vector w using the Q-learning algorithm according to the reward function R and continuously iterating and optimizing the weight vector w to continuously improve the matching effect is as follows:

[0044] Take the component dimensional attribute weight vector w = {w1, w2,..., w n} as the state in reinforcement learning. After each round of matching process ends, the state is updated according to the current matching result;

[0045] The action a is defined as the adjustment strategy for the weight vector w;

[0046] Use the reward function R to provide feedback for each state-action pair;

[0047] Create a two-dimensional table Q(w, a) to store the Q-values for taking each action a in each state w. Initially, all Q-values are set to 0 for initialization.

[0048] In the current state w, select an action a according to the ε-greedy policy, randomly select an action with probability ε, and select the action with the maximum Q-value with probability 1 - ε.

[0049] Execute the selected action a and observe the new state w 1 and the obtained reward R, and update the Q-value.

[0050] As the number of iterations increases, the Q-value table will gradually converge, thus finding the optimal weight adjustment strategy.

[0051] In a preferred embodiment, it includes a fusion module, a matching module, an evaluation module, and an update module, and there are connections between the modules:

[0052] The fusion module is used to obtain the dimensional attributes of components and perform fusion modeling on the dimensional attributes of components.

[0053] The matching module is used to obtain design requirements, extract keywords according to the design requirements, and match them with the semantic tags in the component library, and obtain the semantic similarity based on the cosine similarity algorithm.

[0054] The evaluation module is used to construct a matching model according to the semantic similarity combined with the dimensional attributes of components, and comprehensively evaluate the dimensional attributes of components by using the weighted multi-dimensional data fusion technology to obtain the matching degree score between components and design requirements.

[0055] The update module is used to compare the matching degree score with a preset matching threshold, evaluate the automatic matching effect according to the comparison result, and dynamically update the attribute weights of components according to the automatic matching effect.

[0056] The technical effects and advantages of the industrial design component library management and automatic matching method and system of the present invention:

[0057] 1. In the traditional design process, designers need to spend a lot of time manually screening components. This system can quickly process the design requirements and the information in the component library through the collaborative work of the automated fusion module, matching module, evaluation module, and update module. For example, in architectural design, designers only need to input the design requirements, and the system can screen out the components that meet the requirements from a huge architectural component library in a short time. Compared with manual screening, the design cycle is greatly shortened, allowing the design team to invest more time and energy in creative design work. Each module operates in parallel in the system. The fusion module is responsible for quickly obtaining and integrating the dimensional attributes of components, and the matching module simultaneously conducts semantic analysis and matching of design requirements. This parallel processing mechanism further improves the matching speed and meets the requirements of modern design projects for high efficiency.

[0058] 2. Through the matching module with the help of advanced natural language processing technology, this invention can deeply explore the semantic connotations of design requirements, not just limited to keyword matching. Taking mechanical manufacturing design as an example, the system can accurately understand information such as technical terms and function descriptions in design requirements and precisely match them with the semantic tags in the component library, avoiding matching errors caused by semantic understanding deviations and improving the accuracy of semantic matching. The evaluation module, through weighted multi-dimensional data fusion technology, fully considers the mutual relationships between multiple dimensional attributes of components and conducts a comprehensive and accurate evaluation of components. In product design, the system not only considers physical attributes such as the size and shape of components but also comprehensively considers factors such as material properties and costs to ensure that the selected components highly match the design requirements. The update module compares the matching degree score with a preset threshold and dynamically updates the attribute weights of components according to the matching effect. With the changes in design projects and the update of the component library, the system can automatically adjust the parameters of the matching model to maintain high matching accuracy and efficiency. For example, in emerging technology fields where design requirements and component libraries change frequently, this system can quickly adapt to these changes and provide continuous and reliable support for design work. Through continuous feedback and adjustment, the system can achieve self-optimization, reduce manual intervention, and reduce errors caused by human factors. This not only improves the stability and reliability of the system but also provides a more intelligent and efficient design tool for the design team. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic structural diagram of a method for managing and automatically matching an industrial design component library according to the present invention.

[0060] Figure 2 It is a schematic structural diagram of a system for managing and automatically matching an industrial design component library according to the present invention.

[0061] Figure 3 It is a schematic flowchart of a method for managing and automatically matching an industrial design component library according to the present invention. Detailed implementation manners

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] Embodiment 1 Figure 1 A method for managing and automatically matching an industrial design component library of the present invention is given.

[0064] S10. Obtain the dimensional attributes of the components and perform fusion modeling on the dimensional attributes of the components;

[0065] Extract the dimensional attribute data of each component from the component library, including size information, material properties, and performance parameters;

[0066] Perform min-max normalization processing on the dimensional attribute data, scale the data to the interval [0, 1] to ensure that the data is on the same scale, and avoid the influence of certain attributes on the model being too large;

[0067] Combine the dimensional attribute data of each component into a multi-dimensional feature vector V i ={a i1 , a i2 ,...a in}, where a ij are the dimensional attributes of the component, referring to size information, material properties, and performance parameters.

[0068] According to the extracted and normalized dimensional attribute data, determine the composition of the multi-dimensional feature vector of each component. For example, assuming that the size information has 3 features (length, width, height), the material property has 2 features (density, elastic modulus), and the performance parameter has 2 features (load-bearing capacity, seismic grade), then the feature vector of each component will be a 7-dimensional vector.

[0069] S20. Obtain the design requirements, extract keywords according to the design requirements, and match them with the semantic tags in the component library to obtain the semantic similarity based on the cosine similarity algorithm;

[0070] According to the design requirement text, use a keyword extraction method based on the attention mechanism to extract keywords and key semantic information through natural language processing technology, and obtain a word sequence {c1, c2,...c n} after word segmentation and part-of-speech tagging; where c j is a word or phrase in the text;

[0071] Next, for each word c j , the word vector h of each word is usually obtained using a pre-trained Word2Vec word vector model j ;

[0072] If the dimension of the word vector obtained by model training is d, for the j-th word c n in the word sequence {c1, c2,... c j}, the word vector h j = {h j1 , h j2 ,... h jd};

[0073] Obtain historical design requirements, and label keywords and key semantic information as the training dataset for the weight matrix w and the bias term B. Define a cross-entropy loss function L to measure the difference between the keywords predicted by the model and the actually labeled keywords;

[0074] Use the gradient descent algorithm to minimize the cross-entropy loss function L, and continuously update the weight matrix and the bias term formula as follows:

[0075] In the formula, η is the learning rate, which controls the step size of each update;

[0076] Transform the word vector h j through a fully connected layer, and calculate the score e j by combining the weight matrix w and the bias term B. The specific calculation formula is:

[0077] In the formula, w k is the k-th element of the weight matrix w;

[0078] Calculate the attention score Q j of each word according to the score e j . The attention score reflects the importance of the word in the text. The calculation formula is as follows:

[0079] In the formula, e j is the score calculated through the fully connected layer;

[0080] Compare and analyze the attention score with a preset attention score threshold. When the attention score is greater than the preset attention score threshold, the word corresponding to the attention score is used as a keyword;

[0081] Weightedly sum the word vectors of each keyword according to the attention score to construct the semantic vector d of the design requirement. The formula is:

[0082] For the semantic tags of the i-th component in the component library, adopt the same processing method as the design requirement text. Segment the semantic tags to obtain a sequence of words, generate word vectors for each word, calculate the attention scores, and perform weighted summation to obtain the semantic tag vector s of the component;

[0083] Use the classical cosine similarity formula to calculate the cosine similarity between the design requirement semantic vector and the component semantic tag vector, combined with the attention score Q of the keywords j , to obtain the semantic similarity S i , and the calculation formula is:

[0084] Semantic similarity plays a crucial role in the management and automatic matching of industrial design component libraries, specifically reflected in the following aspects:

[0085] During the industrial design process, designers may need to quickly select appropriate parts or modules from the component library. The semantic similarity algorithm can automatically match and recommend the most suitable components by analyzing the similarity between the current design requirements and the existing components in the component library. For example, if a designer requires a certain component to have a specific function, the semantic similarity model can automatically recommend components with similar functions by comparing the function descriptions, performance parameters, etc. of the components.

[0086] Using semantic similarity can improve the search system in the component library, helping designers quickly find the components they need through natural language descriptions. For example, when a designer enters a fuzzy description (such as "lightweight support frame"), the system can understand the intention of the description based on semantic similarity and find the components that meet the functional requirements, rather than just searching based on keyword matching.

[0087] The industrial design field is constantly evolving, with new materials, new technologies, and new design concepts emerging continuously. Semantic similarity can help extract useful components and design knowledge from old design libraries and transfer them to new design tasks. When introducing new components into the component library, the system can automatically calculate the semantic similarity between the new components and the existing components and decide how to integrate or update the new components with the existing components.

[0088] In complex design tasks, multiple components may need to be combined into a system or product. Semantic similarity can help designers identify potential synergies between different components. For example, some components may be functionally similar and can achieve better compatibility and performance when combined. Through semantic similarity analysis, the system can recommend more collaborative component combinations to optimize the design scheme.

[0089] In teamwork, different designers may have different understandings of components. Semantic similarity helps team members have a unified understanding of components by analyzing their functions and uses, reducing misunderstandings and design deviations, thus supporting teamwork and decision-making.

[0090] Semantic similarity technology provides an intelligent way in the management and automatic matching of industrial design component libraries. It not only improves the management efficiency of components in the library but also enables functions such as automatic matching, intelligent recommendation, and optimized selection during the design process. It can help designers quickly find the components that best meet their needs and improve design efficiency and quality through more accurate matching.

[0091] S30, construct a matching model according to semantic similarity combined with the dimensional attributes of components, and use weighted multi-dimensional data fusion technology to comprehensively evaluate the dimensional attributes of components to obtain the matching degree score between components and design requirements.

[0092] The weight vector of the dimensional attributes of components is expressed as w = {w1, w2..., w n}, where n is the number of dimensional attributes of components, and w j represents the weight of the jth dimensional attribute, and the sum of all weights is usually 1.

[0093] The vector representation of the dimensional attributes of components is V i = {a i1 , a i2 ,...a in}, where a ij represents the value of the jth dimensional attribute of the ith component;

[0094] According to the principle of weighted multi-dimensional data fusion, linearly weighted sum of semantic similarity and dimensional attributes of components is performed to obtain the matching degree score between the ith component and design requirements. The calculation formula is as follows:

[0095]

[0096] In the formula, M i is the matching degree score between the ith component and design requirements, S i is the semantic similarity, μ is the weight of semantic similarity, 1 - μ is the weight of dimensional attributes of components, w j represents the weight of the jth dimensional attribute, and a ij represents the value of the jth dimensional attribute of the ith component;

[0097] Determine the weight of semantic similarity: Collect data on design requirements, selected components, and their matching results in past projects to construct a training dataset.

[0098] The dataset should include the design requirement text, semantic tags and dimensional attribute values of components, as well as a flag indicating whether the actual match is successful.

[0099] Construct a simple linear regression model or logistic regression model, with the matching result as the dependent variable and the weighted sum of semantic similarity and component dimensional attributes as the independent variables.

[0100] During model training, the optimization goal is to minimize the error between the predicted result and the actual matching result.

[0101] Determine the semantic similarity weight and the component dimensional attribute weight vector based on the model parameters obtained through training.

[0102] S40, Compare the matching score with a preset matching threshold, evaluate the automatic matching effect according to the comparison result, and dynamically update the attribute weights of the components according to the automatic matching effect.

[0103] Compare the matching score with the preset matching threshold. If the matching score is greater than or equal to the preset matching threshold, it is considered that the component automatically matches successfully with the design requirement;

[0104] If the matching score is less than the preset matching threshold, it is considered that the component automatically matches fails with the design requirement.

[0105] If the component matches successfully with the design requirement, set the reward function R to 1;

[0106] If the component matches fails with the design requirement, set the reward function R to 0;

[0107] Using the success or failure of the match as the reward signal, update the weight vector w using the Q-learning algorithm according to the reward function R. By continuously iterating and optimizing the weight vector w, the matching effect can be continuously improved. The process is as follows:

[0108] Take the component dimensional attribute weight vector w = {w1, w2,..., w n} as the state in reinforcement learning. After each round of matching process ends, the system will update the state according to the current matching result;

[0109] The action a is defined as the adjustment strategy for the weight vector w; for example, a series of discrete actions can be designed, such as increasing or decreasing the weight of a certain dimensional attribute by a certain step size;

[0110] Use the above-defined reward function R to provide feedback for each state-action pair;

[0111] Create a two-dimensional table Q(w,a) to store the Q values for taking each action a in each state w. Initially, the Q values can all be set to 0 for initialization;

[0112] In the current state w, an action a is selected according to the ε-greedy strategy. An action is randomly selected with a probability of ε, and the action with the maximum Q value is selected with a probability of 1 - ε;

[0113] Execute the selected action a and observe the new state w 1 and the obtained reward R, and update the Q value. The formula is:

[0114] In the formula, γ represents the discount factor, and its value range is [0, 1]. The closer it is to 1, the more the algorithm focuses on future rewards. The closer it is to 0, the more the algorithm focuses on current rewards; R is the reward function as the number of iterations increases; a 1 Possible actions in the new state w 1 below;

[0115] As the number of iterations increases, the Q-value table will gradually converge, thus finding the optimal weight adjustment strategy.

[0116] Example 2, Figure 2 This invention provides an industrial design component library management and automatic matching system.

[0117] The fusion module is used to obtain the dimensional attributes of components and perform fusion modeling on the dimensional attributes of components;

[0118] The matching module is used to obtain design requirements, extract keywords according to the design requirements, and match them with the semantic tags in the component library to obtain the semantic similarity based on the cosine similarity algorithm;

[0119] The evaluation module is used to construct a matching model according to the semantic similarity combined with the dimensional attributes of components, and use the weighted multi-dimensional data fusion technology to comprehensively evaluate the dimensional attributes of components to obtain the matching degree score between components and design requirements;

[0120] The update module is used to compare the matching degree score with a preset matching threshold, evaluate the automatic matching effect according to the comparison result, and dynamically update the attribute weights of components according to the automatic matching effect.

[0121] Figure 3 This invention provides a schematic diagram of the process of an industrial design component library management and automatic matching method.

[0122] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0123] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0124] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of the examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician may use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0125] In addition, the functional modules in the various embodiments of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0126] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0127] Finally: The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An industrial design component library management and automatic matching method, characterized in that It includes the following steps: Obtain the dimensional attributes of the components and perform fusion modeling on the dimensional attributes of the components; Obtain the design requirements, extract keywords according to the design requirements, match them with the semantic tags in the component library, and obtain the semantic similarity based on the cosine similarity algorithm; Construct a matching model according to the semantic similarity combined with the dimensional attributes of the components, and use the weighted multi-dimensional data fusion technology to comprehensively evaluate the dimensional attributes of the components to obtain the matching degree score between the components and the design requirements; Compare the matching degree score with the preset matching threshold, evaluate the automatic matching effect according to the comparison result, and dynamically update the attribute weights of the components according to the automatic matching effect.

2. The industrial design component library management and automatic matching method according to claim 1, characterized in that The process of obtaining the dimensional attributes of the components and performing fusion modeling on the dimensional attributes of the components is as follows: Extract the dimensional attribute data of each component from the component library, including size information, material attributes, and performance parameters; Perform min-max normalization processing on the dimensional attribute data to scale the data to the interval [0, 1] to ensure that the data is on the same scale; Combine the dimensional attribute data of each component into a multi-dimensional feature vector V i ={a i1 , a i2 ,... a in}, where a ij is the dimensional attribute of the component.

3. The industrial design component library management and automatic matching method according to claim 2, wherein The process of extracting keywords according to the design requirements is as follows: According to the design requirement text, a keyword extraction method based on the attention mechanism is used to extract keywords through natural language processing technology, and after word segmentation and part-of-speech tagging, a word sequence {c1, c2,... c n} is obtained; where c j is a word or phrase in the text; For each word c j , obtain the word vector h of each word using the pre-trained Word2Vec word vector model j ; If the dimension of the word vectors obtained by model training is d, for the j-th word c n in the word sequence {c1, c2,... c j}, the word vector h j = {h j1 , h j2 ,... h jd}; Calculate the attention score Q for each word j , compare and analyze the attention score with a preset attention score threshold. When the attention score is greater than the preset attention score threshold, use the word corresponding to the attention score as a keyword.

4. An industrial design component library management and automatic matching method according to claim 3, characterized in that The process of calculating the attention score of each word is as follows: Obtain historical design requirements and label the keywords as the training data set for the weight matrix w and the bias term B; Define a cross-entropy loss function L to measure the difference between the keywords predicted by the model and the actually labeled keywords; Use the gradient descent algorithm to minimize the cross-entropy loss function L, and continuously iterate to update the weight matrix w and the bias term B; Transform the word vector h through a fully connected layer j and calculate the score e by combining the weight matrix w and the bias term B j ; According to the fraction e j Calculate the attention score for each word.

5. An industrial design component library management and automatic matching method according to claim 4, characterized in that The process of calculating the semantic similarity is as follows: Weight-sum the word vectors of each keyword according to the attention score to construct the semantic vector d of the design requirements; For the semantic tags of the i-th component in the component library, segment the semantic tags to obtain the word sequence, generate word vectors for each word, calculate the attention score, and weight-sum to obtain the semantic tag vector s of the component; Calculate the cosine similarity between the semantic vector of the design requirement and the semantic label vector of the component using the classical cosine similarity formula, and combine it with the attention score Q of the keyword j to obtain the semantic similarity S i .

6. The industrial design component library management and automatic matching method according to claim 5, characterized in that The process of the matching degree score between the component and the design requirements is as follows: The weight vector of the component dimension attributes is represented as w = {w1, w2,..., w n}, where n is the number of component dimension attributes, and w j represents the weight of the j-th dimension attribute; The vector representation of the component dimension attribute is V i = {a i1 , a i2 ,... a in}, where a ij represents the j-th dimensional attribute value of the i-th component; According to the principle of weighted multi-dimensional data fusion, linearly weight-sum the semantic similarity and the dimensional attributes of the components to obtain the matching degree score between the i-th component and the design requirements.

7. The industrial design component library management and automatic matching method according to claim 6, wherein The process of comparing the matching degree score with the preset matching threshold and evaluating the automatic matching effect according to the comparison result is as follows: Compare the matching degree score with the preset matching threshold. If the matching degree score is greater than or equal to the preset matching threshold, it is considered that the component and the design requirements are automatically matched successfully; If the matching degree score is less than the preset matching threshold, it is considered that the component and the design requirements are automatically matched failed.

8. An industrial design component library management and automatic matching method according to claim 7, characterized in that, The process of dynamically updating the attribute weights of the components according to the automatic matching effect is as follows: If the component and the design requirements are matched successfully, set the reward function R to 1; If the component and the design requirements are matched failed, set the reward function R to 0; Use the matching success or failure as the reward signal, and update the weight vector w using the Q-learning algorithm according to the reward function R. By continuously iterating and optimizing the weight vector w, the matching effect is continuously improved.

9. The industrial design component library management and automatic matching method according to claim 8, characterized in that The process of updating the weight vector w using the Q-learning algorithm according to the reward function R and continuously iterating and optimizing the weight vector w to continuously improve the matching effect is as follows: Take the component dimension attribute weight vector w = {w1, w2,..., w n} as the state in reinforcement learning. After each round of matching process ends, update the state according to the current matching result; The action a is defined as the adjustment strategy for the weight vector w; Provide feedback for each state-action pair using the reward function R; Create a two-dimensional table Q(w,a) to store the Q-values for taking each action a in each state w. Initially, all Q-values are set to 0 and initialized; In the current state w, select an action a according to the ε-greedy policy, randomly select an action with probability ε, and select the action with the maximum Q-value with probability 1-ε; Execute the selected action a, observe the new state w 1 and the obtained reward R, and update the Q value; As the number of iterations increases, the Q-value table will gradually converge to find the optimal weight adjustment strategy.

10. An industrial design component library management and automatic matching system, characterized in that, Including a fusion module, a matching module, an evaluation module, and an update module, there are connections between the modules: The fusion module is used to obtain the dimensional attributes of components and perform fusion modeling on the dimensional attributes of components; The matching module is used to obtain design requirements, extract keywords according to the design requirements, and match them with the semantic labels in the component library to obtain the semantic similarity based on the cosine similarity algorithm; The evaluation module is used to construct a matching model according to the semantic similarity combined with the dimensional attributes of components, and comprehensively evaluate the dimensional attributes of components using the weighted multi-dimensional data fusion technology to obtain the matching degree score between components and design requirements; The update module is used to compare the matching degree score with a preset matching threshold, evaluate the automatic matching effect according to the comparison result, and dynamically update the attribute weights of components according to the automatic matching effect.

Images