Product model generation method and apparatus, electronic device, and readable storage medium

Abstract

The application relates to a product model generation method and device, electronic equipment and a readable storage medium. The method comprises the following steps: obtaining product key elements from acquired product demand information, and constructing structured product demand data according to the product key elements; matching a target reference model from a preset reference model database based on the structured product demand data; matching target reference parameter instructions from a preset reference instruction database according to the target reference model and historical design data corresponding to the target reference model; performing parameter mapping based on a structured product demand data, a target reference model and a reference parameter instruction set according to a preset parameter recommendation neural network, and generating a target design instruction set; and generating a product model based on a parameter driving technology and the target design instruction set, thereby obtaining a first target product model. The scheme provided in the application can automatically drive product model generation in a whole process, improve design efficiency and result consistency, and further meet the demand of rapid market response.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of product design and modeling, and particularly relates to a product model generation method and device, electronic equipment and a readable storage medium. BACKGROUND

[0002] In the field of non-standard mechanical equipment, parameterized design technology has been widely used, which significantly improves the design efficiency and consistency of specific product series.

[0003] In related technologies, engineers usually need to manually analyze user requirements, set or adjust design parameters according to experience, and drive the system to iterate and optimize. However, the above method is time-consuming and laborious, and the efficiency is low when facing the increasingly diversified and personalized user order requirements. In addition, the design quality is closely related to the personal experience and state of the engineer, which is difficult to realize standardization and scaling, and it is difficult to meet the needs of rapid market response. SUMMARY

[0004] To solve or partially solve the problems in the related art, the present application provides a product model generation method, device, electronic equipment and readable storage medium, which can automatically drive the product model generation in the whole process, thereby effectively shortening the product design cycle, improving the design efficiency and result consistency, and further meeting the needs of rapid market response in the product model generation process.

[0005] The first aspect of the present application provides a product model generation method, comprising

[0006] obtaining product key elements from the obtained product requirement information, and constructing structured product requirement data according to the product key elements;

[0007] matching a target reference model from a preset reference model database based on the structured product requirement data;

[0008] matching a target reference parameter instruction from a preset reference instruction database according to the target reference model and historical design data corresponding to the target reference model;

[0009] performing parameter mapping based on the structured product requirement data, the target reference model and the reference parameter instruction set according to a preset parameter recommendation neural network, and generating a target design instruction set;

[0010] generating a product model based on parameter-driven technology and the target design instruction set to obtain a first target product model.

[0011] In some embodiments, the product key elements are obtained from the obtained product requirement information, and the structured product requirement data is constructed according to the product key elements, comprising:

[0012] perform named entity recognition on the obtained product requirement information based on the semantic analysis model to obtain entity elements within a preset range;

[0013] establish a logical relationship tree among the entity elements based on a semantic role labeling rule according to the entity elements, and generate structured product requirement data according to the logical relationship tree.

[0014] In some embodiments, the matching of the target reference model from the preset reference model database based on the structured product requirement data comprises:

[0015] performing product series similarity matching based on the structured product requirement data by calling a preset reference model database to obtain a matching result of the product series similarity matching;

[0016] determining the target reference model from the preset reference model database based on the matching result of the product series similarity matching and a preset similarity threshold.

[0017] In some embodiments, the determination of the target reference model from the preset reference model database based on the matching result of the product series similarity matching and the preset similarity threshold comprises:

[0018] determining a plurality of recommended models from the preset reference model database based on the matching result of the product series similarity matching and the preset similarity threshold;

[0019] determining the target reference model from the plurality of recommended models according to a received determination instruction.

[0020] In some embodiments, the generation of the product model based on the parameter-driven technology and the target design instruction set to obtain a first target product model comprises:

[0021] obtaining basic model data associated with the target reference model from a preset basic template database;

[0022] performing adaptive modification on the basic model data based on a parameter-driven technology and the target design instruction set to obtain target model data;

[0023] generating a model according to the target model data to obtain the first target product model.

[0024] In some embodiments, the method further comprises:

[0025] performing iterative optimization on the first target product model based on received structured feedback data to obtain a second target product model.

[0026] A second aspect of this application provides a product model generation apparatus, comprising:

[0027] The product requirement extraction module is used to extract key product elements from the acquired product requirement information and construct structured product requirement data based on the key product elements.

[0028] The benchmark model matching module is used to match a target benchmark model from a preset benchmark model database based on the structured product requirement data.

[0029] The benchmark parameter instruction matching module is used to match the target benchmark parameter instruction from a preset benchmark instruction database based on the target benchmark model and the first historical design data corresponding to the target benchmark model.

[0030] The design instruction set acquisition module is used to generate a target design instruction set by performing parameter mapping on the structured product requirement data, the target benchmark model, and the benchmark parameter instruction set based on preset parameters and a recommended neural network.

[0031] The model generation module is used to generate a product model based on parameter-driven technology and the target design instruction set, thus obtaining the first target product model.

[0032] In some embodiments, the device further includes:

[0033] The model optimization module is used to iteratively optimize the first target product model based on the received structured feedback data to obtain the second target product model.

[0034] A third aspect of this application provides an electronic device, comprising:

[0035] Processor; and

[0036] A memory that stores executable code, which, when executed by the processor, causes the processor to perform the method described above.

[0037] A fourth aspect of this application provides a computer-readable storage medium having executable code stored thereon, which, when executed by a processor of an electronic device, causes the processor to perform the method described above.

[0038] The technical solution provided in this application may include the following beneficial effects:

[0039] The technical scheme of the present application comprises the following steps: after obtaining product demand information, automatically extracting key elements from the product demand information in advance and constructing structured product demand data, matching target reference models and related historical design data based on the structured data, performing precise parameter mapping to generate target design instruction set by using a parameter recommendation neural network, and finally realizing automatic generation of product models based on parameter driving technology. Through the above product model generation method, the product model is generated through full-process automation driving, thereby effectively shortening the product design cycle, effectively reducing the dependence on individual experience and the subjective bias of manual screening, improving the design efficiency and result consistency, and thus making the product model generation process meet the needs of rapid market response.

[0040] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0041] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings, in which like reference characters designate like elements throughout the several views.

[0042] Figure 1 is a flowchart of a product model generation method according to an embodiment of the present application;

[0043] Figure 2 is another flowchart of a product model generation method according to an embodiment of the present application;

[0044] Figure 3 is a structural diagram of a product model generation device according to an embodiment of the present application;

[0045] Figure 4 is another structural diagram of a product model generation device according to an embodiment of the present application;

[0046] Figure 5 is a structural diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0047] The embodiments of the present application will be described in detail with reference to the accompanying drawings. Although the embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0048] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting. As used in this application and the appended claims, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0049] It should be understood that although the terms "first," "second," "third," etc. can be used in this application to describe various information, the information should not be limited to these terms. These terms are only used to distinguish one type of information from another type of information. For example, without departing from the scope of the application, first information can also be referred to as second information, and similarly, second information can also be referred to as first information. Therefore, the features defined with "first," "second" can explicitly or implicitly include one or more of the features. In the description of the application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0050] In the conventional existing parameterized design process of non-standard mechanical equipment, the customized requirements submitted by the user usually contain unstructured natural language descriptions, and engineers need to manually analyze the requirement text to extract key elements such as functions, performances and application scenarios. Due to the lack of automatic requirement structuring mechanism, the manual analysis process is easily affected by subjective experience, resulting in insufficient completeness and accuracy of requirement element extraction. When matching historical benchmark models, relying on manual experience to screen similar product series, it is difficult to quantitatively evaluate the matching degree of the model and the requirement, and it is easy to produce matching deviation. In the parameter instruction generation link, engineers need to manually adjust the design parameters based on historical data, and the relevance of parameter adjustment strategy to the current requirement scene depends on individual experience accumulation, resulting in low efficiency of design instruction set generation and the risk of parameter redundancy or absence.

[0051] To solve the above problems, the product model generation method provided by the embodiments of the application can automatically drive the product model generation in the whole process, thereby effectively shortening the product design cycle, improving the design efficiency and result consistency, and further enabling the product model generation process to meet the needs of rapid market response.

[0052] The technical solutions of the embodiments of the application are described in detail below with reference to the drawings.

[0053] Figure 1 FIG. 1 is a flowchart of a product model generation method according to an embodiment of the application.

[0054] Referring to Figure 1 The product model generation method according to the application comprises:

[0055] S110, obtaining product key elements from the obtained product demand information, and constructing structured product demand data according to the product key elements.

[0056] In this step, product key elements are extracted from product demand information received from the front end, and structured product demand data matching user demand is constructed according to the extracted product key elements.

[0057] Among them, the product demand information can be product demand description text information input by the user through the graphical user interface.

[0058] Among them, the product key elements can refer to core entity elements such as functions, performances, and application scenarios extracted from the product demand information. For example, a semantic analysis model can be used for named entity recognition and semantic role labeling of product demand information, so as to convert unstructured product demand information into machine-identifiable structured product demand data.

[0059] S120, matching a target reference model from a preset reference model database based on the structured product demand data.

[0060] In this step, after obtaining the structured product demand data, the target reference model that matches the structured product demand data is matched by the pre-stored model data in the preset reference model data.

[0061] It should be understood that the target reference model can be used to show the user at the front end, and when the target reference model is determined and meets the user demand, the related parameters of the subsequent target product model are further designed according to the same or similar user demand as the target reference model, which effectively improves the fit degree of the subsequent target product model and the user demand.

[0062] Among them, the target reference model can refer to a preset product model that matches the user demand corresponding to the structured product demand data.

[0063] Among them, the target reference model can be filtered and matched from the reference model database through similarity calculation.

[0064] S130, matching a target reference parameter instruction from a preset reference instruction database according to the target reference model and historical design data corresponding to the target reference model.

[0065] In this step, according to the obtained target reference model, the pre-stored historical design data corresponding to the target reference model is obtained, and the associated target reference parameter instruction is matched from the preset reference instruction database.

[0066] The historical design data can be a set of historical design data associated with the target reference model. The data types of the historical design data can include, but are not limited to, historical design cases, industry standards, material properties.

[0067] In this step, the parameter recommendation neural network is used to perform parameter mapping based on the structured product requirement data, the target reference model, and the reference parameter instruction set, to generate a target design instruction set.

[0068] In this step, the parameter recommendation neural network is used to perform parameter mapping based on the structured product requirement data, the target reference model, and the reference parameter instruction set, to generate a target design instruction set.

[0069] The preset parameter recommendation neural network can be a machine learning model used to generate design instructions and obtained through pre-construction and pre-training. The historical design data can be used to train the neural network to obtain the preset parameter recommendation neural network, so that the preset parameter recommendation neural network can automatically generate a design instruction set that meets the current requirement through parameter mapping based on the input structured product requirement data, the target reference model, and the reference parameter instruction set.

[0070] The target design instruction set can be a set of predefined rules and algorithms used to control the relationship between parameters. The target design instruction set can be automatically executed by a modeling system or software.

[0071] For example, the target design instruction set can include the following contents:

[0072] Geometric constraints: "cylinder height = diameter x 2";

[0073] Performance constraints: "material thickness >= load calculation value x safety factor";

[0074] Logical rules: "if product type = outdoor equipment, then protection level >= IP67".

[0075] S150, based on parameter-driven technology and target design instruction set to generate product model, get the first target product model.

[0076] In this step, based on the parameter-driven technology and the obtained target design instruction set, a product model that meets the user's demand is generated, and a first target product model is obtained.

[0077] It should be understood that the first target product model can include, but is not limited to, at least one of product BOM structure data, product modeling data, and engineering drawing data. The first target product model incorporates the rules or algorithms corresponding to the target design instruction set.

[0078] The parameter driving technology can refer to a technology of automatically modifying model parameters based on a design instruction set. For example, adaptive adjustment can be performed on a basic model by calling a parameterized modeling tool.

[0079] In this embodiment, the product model generation method of the present application automatically extracts key elements from product demand information and constructs structured product demand data after obtaining product demand information. The target reference model and related historical design data are matched based on the structured data, and the target design instruction set is generated by precise parameter mapping using a parameter recommendation neural network. Finally, the product model is automatically generated based on the parameter driving technology. Through the above product model generation method, the product model is generated through full-process automation driving, thereby effectively shortening the product design cycle and effectively reducing the dependence on individual experience and the subjective bias of manual screening, improving design efficiency and result consistency, and thus making the product model generation process meet the needs of rapid market response.

[0080] Figure 2 FIG. 2 is another flowchart of the product model generation method according to an embodiment of the present application.

[0081] Referring to Figure 2 The product model generation method of the present application comprises:

[0082] In S210, the obtained product demand information is subjected to named entity recognition based on a semantic analysis model, to obtain entity elements within a preset range.

[0083] In this step, the obtained product demand information is subjected to named entity recognition by a pre-trained semantic analysis model, to obtain entity elements within a preset range.

[0084] The semantic analysis model can be a natural language processing model based on deep learning. The natural language processing model can process the input product demand text, thereby identifying the entity elements within the preset range by understanding the context semantics of the text. Of course, the semantic analysis model can also use an AI model in related technologies. For example, the semantic analysis model can use a BERT (Bidirectional Encoder Representations from Transformers) model or a GPT (Generative Pre-trained Transformer) model, which is not limited herein.

[0085] The preset range of the entity element can at least include function, performance, and application scenario. The function can refer to the core role or task assumed by the product, such as supporting load bearing, material handling, material cutting, etc. The performance can refer to the quantitative or qualitative level achieved by the product in realizing its function, such as load bearing capacity, running speed, cutting accuracy, etc. The application scenario can refer to the specific use environment, industry, or working condition to which the product is directed, such as industrial automation production line, aerospace field, automobile manufacturing, etc.

[0086] As an example, for product demand information corresponding to an industrial robot, the semantic analysis model can identify "welding" as a function entity, "cutting accuracy 0.05mm" as a performance entity, and "automobile manufacturing" as an application scenario entity.

[0087] S220, according to the entity element, a logical relationship tree between the entity elements is established based on the semantic role labeling rule, and structured product demand data is generated according to the logical relationship tree.

[0088] In this step, according to the obtained entity element and based on the pre-set semantic role labeling rule, a logical relationship tree corresponding to the preset relationship type between the entity elements is established, and structured product demand data that can be directly processed by a computer is generated according to the logical relationship tree.

[0089] The semantic role labeling rule can be configured to identify the syntactic relationship and semantic role between the entity elements.

[0090] For example, in the demand "cutting in a high-temperature environment to achieve a cutting accuracy of 5 microns", it can be identified that "cutting" is the action subject, "5 micron cutting accuracy" is the limiting condition corresponding to the aforementioned action subject, and "high-temperature environment" is the application scenario corresponding to the action subject and the limiting condition.

[0091] The logical relationship tree can be a preset relationship between entities stored by a tree data structure. The root node of the logical relationship tree can be a core function, the child nodes of the logical relationship tree can be performance parameters and application scenarios, and the edges of the logical relationship tree represent the logical association between the core function, the performance parameters, and the application scenarios.

[0092] The structured product demand data can be one of JSON or XML, a standardized format that can be directly used for computer processing.

[0093] S230, based on the structured product demand data, a preset reference model database is called to perform product series similarity matching, and a matching result of the product series similarity matching is obtained.

[0094] In this step, after obtaining the structured product demand data, a preset benchmark model database pre-stored with historical model data is called, and historical model data corresponding to the structured product demand data is matched from the preset benchmark model database as a matching result based on a similarity algorithm.

[0095] The product series similarity matching process can include the following steps:

[0096] The functional elements, performance elements, and application scenarios in the structured product demand data are converted into multi-dimensional vectors, and the obtained multi-dimensional vectors are compared with corresponding feature vectors in the historical model data in the preset benchmark model database through a cosine similarity algorithm to output a similarity score.

[0097] The similarity score can be in numerical or percentage format.

[0098] The hierarchical weight distribution basis for similarity matching can be provided according to the logical relationship tree in the structured product demand data. For example, the functional element weight accounts for 60%, the performance element accounts for 30%, and the application scenario accounts for 10%. In this way, the relevance of the matching criteria in the matching process and the product technical characteristics can be ensured.

[0099] S240, determining a target benchmark model from the preset benchmark model database based on the matching result of the product series similarity matching and a preset similarity threshold.

[0100] In this step, the target benchmark model is determined from the preset benchmark model database according to the matching result of the product series similarity matching and the preset similarity threshold.

[0101] The matching result of the product series similarity matching can correspond to all models in the preset benchmark model database. That is, the matching result of the product series similarity matching can have a similarity matching score corresponding to all models in the preset benchmark model database.

[0102] For example, the preset similarity threshold can be set to 0.8. After the matching result of the product series similarity matching, the benchmark models with a similarity score greater than or equal to 0.8 are selected as candidate target benchmark models, and the model with the highest similarity is finally selected from these candidate models as the final target benchmark model.

[0103] In some embodiments, the matching process for obtaining the target benchmark model can include:

[0104] S241, determining a plurality of recommended models from the preset benchmark model database based on a preset similarity threshold.

[0105] In this step, based on the obtained product series similarity matching matching result and the preset similarity threshold, a preset number or a plurality of recommended models meeting a preset condition are determined from the preset reference model database.

[0106] The plurality of recommended models can meet a preset association condition. The preset association condition can include at least one of similar functions, similar target user groups, similar core technologies, and the same product category. It should be understood that each of the similar functions, the similar target user groups, the similar core technologies, and the same product category can be applied alone or in combination. Of course, the preset association condition can also include other adaptive setting conditions, which are not limited here.

[0107] The similar function condition can be determined by comparing the function keywords extracted by the semantic matching algorithm, such as calculating the similarity between “precision transmission” and “power transmission”. The similar target user group condition can be determined by clustering analysis based on user portrait labels, such as distinguishing between medical device user groups and industrial device user groups. The similar core technology condition can be determined by calculating the cosine similarity of the technology feature vector, such as comparing the technical parameters of hydraulic drive and electric drive. The same product category condition can be determined by matching the pre-set classification code, such as classifying “industrial robots” and “service robots” into different product categories.

[0108] It should be understood that the plurality of recommended models can be product models of the same series or similar series. The same series or similar series meet the pre-associated condition, and the plurality of recommended models are determined from the product models of the same series or similar series that meet the pre-associated condition under the premise of meeting the preset similarity threshold. For example, different stroke specifications of variant products in the same series of hydraulic cylinders.

[0109] As an example, the preset similarity threshold is set to 0.8, and the model number is 5. Based on the preset similarity threshold and the preset association condition, 5 recommended models are selected from the preset reference model database, wherein the preset association condition includes similar functions, similar target user groups, similar core technologies, and the same product category, and corresponds to weights of 0.4, 0.3, 0.2, and 0.1, respectively. The selection process is as follows:

[0110] First, the models meeting the similar function condition are selected by comparing the matching degree of the function description keywords. For example, if the target product requirement includes the “automatic production line” function, the models with the same or similar function description are selected.

[0111] Second, on the basis of the first step, the same target user group condition model is screened by comparing the user portrait features. For example, if the target product is for the "manufacturing enterprise" user group, the model suitable for the same or similar user group is screened out.

[0112] Third, on the basis of the second step, the model meeting the similar core technology condition is screened out by comparing the technical field classification. For example, if the target product involves "machine vision" technology, the model applying the same or similar core technology is screened out.

[0113] Fourth, on the basis of the third step, the model meeting the same product category condition is screened out by comparing the product categories. For example, if the target product belongs to the "industrial robot" category, the model belonging to the same category is screened out.

[0114] Fifth, on the basis of the fourth step, the comprehensive matching score of the screened model is calculated according to the preset corresponding association condition weight, and the model with a comprehensive matching score exceeding the preset similarity threshold value 0.8 and a number of 5 is selected as the recommended model.

[0115] It can be known that by setting the preset association condition and the preset similarity threshold value, the accuracy and efficiency of the target reference model matching can be further improved, and a more reliable and applicable basis is provided for subsequent product model generation.

[0116] Among them, multiple recommended models can be displayed in the same interactive interface at the same time, so that the user can obtain the information of multiple recommended models from the same interactive interface.

[0117] S242, determining the target reference model from the multiple recommended models according to the received determination instruction.

[0118] In this step, the target reference model is determined from the multiple recommended models according to the determination instruction received through the interactive interface.

[0119] It should be understood that the determination instruction can be a check operation of the user on the candidate model list or a click confirmation action on any model from the interactive interface.

[0120] As an example, when the user demand corresponds to a 50-ton hydraulic machine, according to the corresponding structured product demand data, the reference model database is preliminarily screened by the preset similarity threshold value, five same series reference models with a working pressure range of 45-55 tons are screened out, and then the recommended model set is pushed to the interactive interface of the user terminal for visual display. After receiving the selection operation of the user on a specific model, the model corresponding to the selection operation is determined as the target reference model. Through the above-mentioned way, the user decision-making link is introduced, and the candidate model is finally determined by the artificial confirmation instruction, which can effectively make up for the possible semantic understanding deviation of pure algorithm matching.

[0121] S250, matching the target reference parameter instruction from the preset reference instruction database according to the target reference model and the historical design data corresponding to the target reference model.

[0122] In this step, according to the obtained target reference model, the pre-stored historical design data corresponding to the target reference model is obtained, and the associated target reference parameter instruction is matched from the preset reference instruction database.

[0123] S260, parameter mapping is performed by the preset parameter recommendation neural network based on the structured product requirement data, the target reference model and the reference parameter instruction set to generate a target design instruction set.

[0124] In this step, the preset parameter recommendation neural network is used to perform parameter mapping based on the structured product requirement data, the target reference model and the reference parameter instruction set to generate a target design instruction set that meets the current user's requirements.

[0125] In some embodiments, the preset parameter recommendation neural network can be obtained by pre-training the preset recommendation neural network according to all reference models in the preset reference model database, historical design data corresponding to all reference models and the preset reference instruction database to obtain a trained preset recommendation neural network.

[0126] It can be understood that through the pre-training process of multi-dimensional training data, the preset recommendation neural network obtained can complete the establishment of the end-to-end mapping relationship of product requirement features to parameter instruction sets. For example, when inputting structured requirement data containing "high torque, low speed", the preset recommendation neural network can automatically generate a matching gear modulus and reduction ratio parameter combination and recommend the optimal value. Therefore, the instruction set generated by the preset parameter recommendation neural network can adapt to the optimal parameter recommendation requirements in different scenarios.

[0127] Among them, the plurality of reference models in the reference model database can cover different product series, for example, containing at least three core models of mechanical structures, each model corresponding to a different functional parameter range.

[0128] Among them, the historical design data corresponding to all reference models can contain parameter adjustment records in a preset number of actual design cases corresponding to all reference models, for example, the corresponding relationship between material strength parameters and load conditions. It should be understood that through the historical design data corresponding to all reference models, the user's preference for parameter adjustment in actual application can be analyzed.

[0129] Among them, the instruction set in the preset reference instruction database can be stored according to the industry standard format. For example, all instruction sets in the reference instruction database are stored through an instruction template containing an ISO standard code.

[0130] In the pre-training process, the neural network optimizes the weight parameters through a back propagation algorithm. For example, the training period can be set to 500 rounds, and the loss function adopts a combination of mean square error and cross entropy for training optimization.

[0131] S270, obtaining the basic model data associated with the target reference model from the preset basic template database.

[0132] In this step, when the target reference model is determined, the basic model data with the same topological structure as the target reference model is matched from the preset basic template database.

[0133] The preset basic template database can contain basic model data of multiple product series, wherein each basic model data is associated with one or more target reference models.

[0134] The preset basic template database can establish the association between the basic model data and the target reference model through a hash index or metadata tag. In this way, the matching degree between the model feature vector and the reference model parameter can be calculated by a cosine similarity algorithm, so as to realize the matching between the basic model data and the target reference model.

[0135] All basic model data in the preset basic template database can be pre-associated with the target reference model, and the corresponding basic model data can be directly obtained by querying the identifier of the target reference model in the subsequent matching process.

[0136] For example, the identifier of the target reference model is used as a query condition to obtain the corresponding basic model data in the preset basic template database through a query operation.

[0137] S280, based on the parameter driving technology and the target design instruction set, the basic model data is adaptively modified to obtain the target model data.

[0138] In this step, the target design instruction set obtained based on the parameter driving technology is disassembled into multiple parameter groups and a parameter mapping table is established. Based on the parameter mapping table, the corresponding parameters of the basic model data are one-to-one corresponding and adaptively modified.

[0139] The adaptive modification process can adopt a hierarchical optimization strategy, which preferentially adjusts the key parameters affecting the function implementation of the model, and then optimizes the secondary parameters. For example, in the hydraulic system design, the cylinder stroke parameter is preferentially adjusted, and then the sealing material parameter is optimized.

[0140] In the adaptive modification process, the parameters corresponding to the basic model data can be adaptively adjusted according to the pre-set priority types. For example, the priority types include core feature parameters and non-core feature parameters. In the adaptive modification process, the core feature parameters of the basic model data can be retained, and only the non-core parameters of the basic model data are dynamically adjusted according to the target design instruction set. For example, in the design of the transmission mechanism of the non-standard mechanical equipment, the core parameters such as gear modulus and center distance are retained, and the non-core parameters such as tooth width coefficient and lubrication mode are adjusted. Further, the adaptive modification process can include but is not limited to updating, adding, and deleting operations on the parameters of the basic model data.

[0141] S290, model generation is performed according to the target model data, and a first target product model is obtained.

[0142] In this step, the target model data obtained by the adaptive modification based on the basic model data and the target design instruction set is used for model generation, and a first target product model is obtained.

[0143] In the model generation process, the application program interface of the three-dimensional modeling software can be used to convert the target model data into a three-dimensional geometric model. It can be understood that the generated first target product model can include complete data associated with the product model, such as geometric information, material properties, and assembly relationship.

[0144] In some embodiments, after obtaining the first target product model, the product model generation method of the present application can further include the following steps:

[0145] S2100, based on the received structured feedback data, the first target product model is iteratively optimized to obtain a second target product model.

[0146] In this step, after obtaining the first target product model, the structured feedback data fed back by the user or the designer is received from the pre-set data interface, and the first target product model is iteratively optimized based on the structured feedback data.

[0147] It should be understood that the first target product model as the initially generated product model can not fully meet the actual demand or have local optimization space. If only relying on the initial design process, the product model can still not meet the actual demand of the user. By generating the first target product model, based on the structured feedback data fed back by the user or the designer, the model is further optimized and adjusted based on the first target product model, so that the finally obtained product model can better meet the actual demand of the user and have better market adaptability.

[0148] The structured feedback data can include parameter adjustment instructions. The parameter adjustment instructions are used to indicate that the corresponding adjustment operations are performed on the corresponding parameters of the first target product model.

[0149] The iterative optimization process can adopt a reinforcement learning algorithm, input the structured feedback data as a reward function, and dynamically adjust the model parameter update direction for iterative optimization of the first target product model. The iterative optimization process can be implemented using related technologies, which will not be described here.

[0150] In the iterative optimization of the first target product model, it can also be determined whether the parameter adjustment instruction in the received structured feedback data exceeds the preset template parameter range of the first target product model. When it is determined that the parameter adjustment instruction in the received structured feedback data exceeds the preset template parameter range of the first target product model, the model iterative optimization is performed on the basis model data corresponding to the first target product model to obtain a third target product model.

[0151] It is not difficult to understand that when the iterative optimization of the first target product model may cause the design parameters to deviate from the reasonable constraint range of the benchmark model, if the iterative optimization is still performed, the stability of the original product model design logic may be damaged. Therefore, when it is determined that the iterative optimization process of the current model may cause the design parameters to deviate from the reasonable constraint range of the benchmark model, the model iterative optimization is performed again based on the basis model data, which can effectively ensure the stability of the product model design.

[0152] The third target product model obtained in the above manner not only retains the user's personalized demand for the product model, but also avoids the imbalance of the model performance caused by the parameter out-of-range in the final product model generation effect by returning to the basic design framework, effectively avoiding the impact of parameter adjustment out of the preset range on the stability of the product model.

[0153] In this embodiment, the product model generation method of the present application realizes automatic construction of the logical relationship tree between entity elements based on product demand information by performing named entity recognition and semantic role labeling based on a semantic analysis model. The target benchmark model is matched according to the logical relationship tree for user selection, which can effectively improve the understanding accuracy of structured product demand data, thereby improving the fit degree of the obtained target benchmark model and the actual demand of the user. In addition, the target benchmark model is screened by relying on the product series similarity matching algorithm combined with the preset threshold, which can effectively quantify the objectivity of the matching process, eliminate the interference of artificial experience on model selection, and ensure the accuracy and repeatability of benchmark model matching. In addition, the process of automatically iterative optimization of the model using structured feedback data is set, forming a closed-loop mechanism of design-feedback-optimization, which can further improve the model quality while reducing manual intervention, and further enhance the adaptability and market response agility of product design.

[0154] Corresponding to the foregoing application function implementation method embodiments, the application further provides a product model generation apparatus, an electronic device, and corresponding embodiments.

[0155] Figure 3 FIG. 1 is a structural schematic diagram of a product model generation apparatus according to an embodiment of the application.

[0156] Referring to Figure 3 The product model generation apparatus 300 according to the application includes a product demand extraction module 310, a reference model matching module 320, a reference parameter instruction matching module 330, a design instruction set acquisition module 340, and a model generation module 350.

[0157] The product demand extraction module 310 is configured to acquire product key elements from acquired product demand information and construct structured product demand data according to the product key elements.

[0158] In some embodiments, the product demand extraction module 310 can further perform named entity recognition on the acquired product demand information based on a semantic analysis model to obtain entity elements within a preset range; establish a logical relationship tree between each entity element based on a semantic role labeling rule, and generate structured product demand data according to the logical relationship tree.

[0159] The reference model matching module 320 is configured to match a target reference model from a preset reference model database based on the structured product demand data.

[0160] In some embodiments, the reference model matching module 320 can further perform product series similarity matching by calling the preset reference model database based on the structured product demand data to obtain a matching result of the product series similarity matching; and determine the target reference model from the preset reference model database based on the matching result of the product series similarity matching and based on a preset similarity threshold.

[0161] In some embodiments, the reference model matching module 320 can further determine a plurality of recommended models from the preset reference model database based on the matching result of the product series similarity matching and based on the preset similarity threshold; and determine the target reference model from the plurality of recommended models according to a received determination instruction.

[0162] The reference parameter instruction matching module 330 is configured to match target reference parameter instructions from a preset reference instruction database according to the target reference model and historical design data corresponding to the target reference model.

[0163] The design instruction set acquisition module 340 is configured to perform parameter mapping based on the structured product demand data, the target reference model, and the reference parameter instruction set by a preset parameter recommendation neural network to generate a target design instruction set.

[0164] The model generation module 350 is configured to generate a product model based on a parameter-driven technology and a target design instruction set, to obtain a first target product model.

[0165] In some embodiments, the model generation module 350 can further obtain, from a preset basic template database, basic model data associated with the target reference model; perform adaptive modification on the basic model data based on the parameter-driven technology and the target design instruction set, to obtain target model data; and perform model generation according to the target model data, to obtain the first target product model.

[0166] Figure 4 FIG. 10 is another structural schematic diagram of a product model generation apparatus according to an embodiment of the present application.

[0167] Referring to FIG. 10, Figure 4 The product model generation apparatus 300 according to the present application comprises a product demand extraction module 310, a reference model matching module 320, a reference parameter instruction matching module 330, a design instruction set acquisition module 340, a model generation module 350, and a model iteration module 360.

[0168] The model iteration module 360 is configured to perform iterative optimization on the first target product model based on the received structured feedback data, to obtain a second target product model.

[0169] In this embodiment, the product model generation method according to the present application comprises the following steps: after obtaining product demand information, automatically extracting key elements from the product demand information and constructing structured product demand data; matching a target reference model and related historical design data based on the structured data; performing precise parameter mapping to generate a target design instruction set by using a parameter recommendation neural network; and finally, automatically generating a product model based on a parameter-driven technology. Through the above product model generation method, the product model generation is driven by full-process automation, thereby effectively shortening the product design cycle, effectively reducing the dependence on individual experience and the subjective bias of manual screening, improving the design efficiency and result consistency, and further enabling the product model generation process to meet the needs of rapid market response.

[0170] As to the apparatus in the above embodiments, the specific manners in which various modules perform operations have been described in detail in the embodiments of the method, and will not be described herein in detail.

[0171] Figure 5 FIG. 10 is another structural schematic diagram of a product model generation apparatus according to an embodiment of the present application.

[0172] Referring to FIG. 10, Figure 5 The electronic device 1000 comprises a memory 1010 and a processor 1020.

[0173] The processor 1020 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gates or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can be any conventional processor.

[0174] The memory 1010 can include various types of storage units, such as a system memory, a read-only memory (ROM), and a permanent storage device. Among them, the ROM can store static data or instructions required by the processor 1020 or other modules of the computer. The permanent storage device can be a readable and writable storage device. The permanent storage device can be a non-volatile storage device that does not lose stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a mass storage device (such as a magnetic or optical disk, a flash memory) as a permanent storage device. In some other embodiments, the permanent storage device can be a removable storage device (such as a floppy disk, an optical drive). The system memory can be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory can store some or all instructions and data required by the processor during runtime. In addition, the memory 1010 can include a combination of any computer readable storage media, including various types of semiconductor memory chips (such as DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), magnetic disks and / or optical disks. In some embodiments, the memory 1010 can include a readable and / or writable removable storage device, such as a compact disc (CD), a read-only digital versatile disc (such as DVD-ROM, double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (such as an SD card, a min SD card, a Micro-SD card, etc.), a magnetic floppy disk, etc. The computer readable storage medium does not include a carrier wave and a transient electronic signal transmitted through wireless or wired transmission.

[0175] The memory 1010 stores executable code, which, when processed by the processor 1020, can cause the processor 1020 to perform part or all of the above-mentioned methods.

[0176] In addition, the method according to the present application can also be implemented as a computer program or computer program product, which comprises computer program code instructions for executing part or all of the steps in the above method according to the present application.

[0177] Alternatively, the present application can also be implemented as a computer readable storage medium (or non-transitory machine readable storage medium or machine readable storage medium) having stored executable code (or computer program or computer instruction code) which, when executed by a processor of an electronic device (or a server, etc.), causes the processor to execute part or all of the steps of the above method according to the present application.

[0178] The above has described the embodiments of the present application, the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes are obvious to those skilled in the art without departing from the scope and spirit of the described embodiments. The selection of the terms used herein is intended to best explain the principles of the embodiments, practical application or improvement of the technology in the market, or to enable other ordinary skilled in the art to understand the embodiments disclosed herein.

Claims

1. A product model generation method characterized by, comprising obtaining product key elements from the obtained product demand information, and constructing structured product demand data according to the product key elements; wherein, including: performing named entity recognition on the obtained product demand information based on a semantic analysis model to obtain entity elements within a preset range; establishing a logical relationship tree between each entity element based on a semantic role labeling rule according to the entity elements, and generating structured product demand data according to the logical relationship tree; matching a target reference model from a preset reference model database based on the structured product demand data; wherein, including: performing product series similarity matching by calling the preset reference model database based on the structured product demand data to obtain a matching result of the product series similarity matching; determining a plurality of recommended models from the preset reference model database based on the matching result of the product series similarity matching and based on a preset similarity threshold; determining a target reference model from the plurality of recommended models according to a received determination instruction; wherein, the plurality of recommended models are product models of the same series or similar series that meet a preset association condition, and the preset association condition includes at least one of similar functions, similar target user groups, similar core technologies, and the same product category; matching a target reference parameter instruction from a preset reference instruction database according to the target reference model and historical design data corresponding to the target reference model; performing parameter mapping based on the structured product demand data, the target reference model, and the reference parameter instruction set by a preset parameter recommendation neural network, automatically generating a design instruction set adapted to current demand according to the input structured product demand data, the target reference model, and the reference parameter instruction set, and generating a target design instruction set; generating a product model based on parameter-driven technology and the target design instruction set to obtain a first target product model.

2. The method of claim 1, wherein, The generating of the product model based on the parameter-driven technology and the target design instruction set to obtain the first target product model includes: obtaining basic model data associated with the target reference model from a preset basic template database; performing self-adaptive modification on the basic model data based on parameter-driven technology and the target design instruction set to obtain target model data; generating a model according to the target model data to obtain a first target product model.

3. The method according to any one of claims 1 to 2, characterized in that, The method further includes: performing iterative optimization on the first target product model based on received structured feedback data to obtain a second target product model.

4. A product model generation apparatus characterized by comprising: comprising: a product demand extraction module configured to obtain product key elements from the obtained product demand information, and construct structured product demand data according to the product key elements; wherein, including: performing named entity recognition on the obtained product demand information based on a semantic analysis model to obtain entity elements within a preset range; establishing a logical relationship tree between each entity element based on a semantic role labeling rule according to the entity elements, and generating structured product demand data according to the logical relationship tree; The benchmark model matching module is configured to match a target benchmark model from a preset benchmark model database based on the structured product demand data. The benchmark model matching module includes: performing product series similarity matching based on the structured product demand data and the preset benchmark model database to obtain a matching result of the product series similarity matching; determining a plurality of recommended models from the preset benchmark model database based on the matching result of the product series similarity matching and a preset similarity threshold; and determining the target benchmark model from the plurality of recommended models according to a received determination instruction. The plurality of recommended models are product models of the same series or similar series that meet a preset association condition. The preset association condition includes at least one of the following: similar functions, similar target user groups, similar core technologies, and the same product category. The benchmark parameter instruction matching module is configured to match a target benchmark parameter instruction from a preset benchmark instruction database based on the target benchmark model and first historical design data corresponding to the target benchmark model. The design instruction set acquisition module is configured to perform parameter mapping based on the structured product demand data, the target benchmark model, and the benchmark parameter instruction set by using a preset parameter recommendation neural network, to automatically generate a design instruction set that is adapted to current demand based on the input structured product demand data, the target benchmark model, and the benchmark parameter instruction set, and to generate a target design instruction set. The model generation module is configured to generate a product model based on a parameter driving technology and the target design instruction set, to obtain a first target product model.

5. The apparatus of claim 4, wherein, The device further includes: The model optimization module is configured to iteratively optimize the first target product model based on received structured feedback data, to obtain a second target product model.

6. An electronic device, comprising: The device includes: a processor; and a memory having stored thereon executable code that, when executed by the processor, causes the processor to perform the method of any one of claims 1-2.

7. A computer-readable storage medium having stored thereon executable code that, when executed by a processor of an electronic device, causes the processor to perform the method of any one of claims 1-2. ​ ​

Images