An intelligent recommendation system and method for precision shaft design

Abstract

The present application provides an intelligent recommendation system and method for precision shaft design, which relates to the field of intelligent recommendation technology. The inline relationship between each visual image feature in the visual demand space is converted into a demand association feature of the visual modality through the global visual features of the visual image in the demand document; a demand decision tree of the semantic modality is constructed according to the text inline relationship of each semantic demand feature in the context in the semantic demand space, and then the demand association feature of the semantic modality is determined by the structural relationship between the demand decision tree and the semantic demand feature in the semantic demand space; the demand association feature of the semantic modality and the demand association feature of the visual modality are modally fused, and then the features after modal fusion are used as the demand features of precision shaft design, and the demand document is automatically recommended through the demand features. The above scheme can be used to modally fuse the visual and semantic demand features, thereby improving the accuracy of automated recommendations in precision shaft design.

Description

Technical Field

[0001] The present application relates to the field of intelligent recommendation technology, and more specifically, to an intelligent recommendation system and method for precision shaft design. Background Art

[0002] In the design of precision shafts, the precision design of shafts and bearings is crucial and needs to be determined according to specific matching requirements and functional objectives. The dimension chain is a key technology to ensure the overall accuracy of the mechanical system. The dimension chain involves the relationship between parts and the control of cumulative errors. In addition, the matching accuracy of rolling bearings is also an important part of precision shaft design that cannot be ignored, which directly affects the rotation accuracy and reliability of the shaft.

[0003] In traditional design recommendation systems, visual information and semantic information are usually processed by independent modules, which leads to a lack of coordination and unity in the processing. Visual information is usually extracted through computer vision technology, such as the object, color, texture and other features of the image, while semantic information is extracted through natural language processing technology, such as design needs, requirements, constraints, etc. in the text. However, the processing of these two types of information is often unrelated to each other, which may lead to inconsistent and conflicting information in the precision shaft design recommendation strategy. Therefore, how to perform modal fusion of visual and semantic demand features to improve the accuracy of automated recommendations in precision shaft design is a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides an intelligent recommendation system and method for precision shaft design, which can perform modal fusion of visual and semantic demand features, thereby improving the accuracy of automated recommendations in precision shaft design.

[0005] In a first aspect, the present application provides an intelligent recommendation method for precision shaft design, the recommendation method comprising the following steps: Obtaining a requirement document for precision shaft design, performing multimodal feature extraction on the requirement document, and obtaining a visual requirement space of a visual modality and a semantic requirement space of a semantic modality; Converting the inline relationship between each visual image feature in the visual demand space into a demand association feature of a visual modality through the global visual feature of the visual image in the demand document; Constructing a semantic modal demand decision tree according to the text inline relationship of each semantic demand feature in the context in the semantic demand space, and then determining the demand association feature of the semantic modal from the demand decision tree and the structural relationship between the semantic demand features in the semantic demand space; Modality fusion is performed on the demand-related features of the semantic modality and the demand-related features of the visual modality, and the features after modality fusion are used as demand features for precision shaft design, and the demand document is automatically recommended based on the demand features.

[0006] In this embodiment, multimodal feature extraction is performed on the demand document to obtain a visual demand space of a visual modality and a semantic demand space of a semantic modality, specifically including: Extracting visual demand features and semantic demand features of the demand document through a pre-trained document model; Determine the visual demand space of the visual modality according to all visual demand features; The semantic requirement space of the semantic modality is determined according to all the semantic requirement features.

[0007] In this embodiment, converting the inline relationship between the visual image features in the visual demand space into the demand association features of the visual modality through the global visual features of the visual image in the demand document specifically includes: extracting global visual features of the visual image from the requirement document; Constructing a structural relationship model of visual modalities based on the global visual features; The structural relationship model is used to convert the inline relationship between each visual image feature in the visual demand space into a demand-related feature of the visual modality.

[0008] In this embodiment, the structural relationship model is a dependency relationship model based on graph theory.

[0009] In this embodiment, constructing a semantic modal demand decision tree according to the text inline relationship of each semantic demand feature in the context in the semantic demand space specifically includes: Extracting paragraph semantics of each text paragraph from the requirement document; Determine the text inline relationship of the semantic requirement features in the context in the semantic requirement space through all paragraph semantics; A semantic modality demand decision tree is generated based on the text inline relationship.

[0010] In this embodiment, determining the demand association features of the semantic modality based on the structural relationship between the demand decision tree and the semantic demand features in the semantic demand space specifically includes: Determine a structural decision tree of semantic modality through the structural relationship of semantic demand features in the semantic demand space; Decision association is performed on the structure decision tree and the demand decision tree to obtain demand association features of semantic modality.

[0011] In this embodiment, the semantic modality is a modality based on text language information.

[0012] In this embodiment, the visual modality is a modality based on image information.

[0013] In this embodiment, performing modal fusion on the demand-related features of the semantic modality and the demand-related features of the visual modality specifically includes: Obtaining a semantic fusion weight of the semantic modality and a visual fusion weight of the visual modality; The demand-related features of the semantic modality and the demand-related features of the visual modality are fused by using the semantic fusion weight and the visual fusion weight to obtain fused features.

[0014] In a second aspect, the present application provides an intelligent recommendation system for precision shaft design, which is used to perform an intelligent recommendation method for precision shaft design, and the recommendation system includes: A feature extraction module is used to obtain a requirement document for precision shaft design, perform multimodal feature extraction on the requirement document, and obtain a visual requirement space of a visual modality and a semantic requirement space of a semantic modality; A visual modality module, used for converting the inline relationship between each visual image feature in the visual demand space into a demand association feature of the visual modality through the global visual feature of the visual image in the demand document; A semantic modality module, used to construct a semantic modality demand decision tree according to the text inline relationship of each semantic demand feature in the semantic demand space in the context, and then determine the demand association feature of the semantic modality from the demand decision tree and the structural relationship between the semantic demand features in the semantic demand space; The feature fusion module is used to perform modal fusion on the demand-related features of the semantic modality and the demand-related features of the visual modality, and then use the features after modal fusion as the demand features of precision shaft design, and automatically recommend the demand document through the demand features.

[0015] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects: Obtain a requirement document for precision shaft design, perform multimodal feature extraction on the requirement document, and obtain a visual requirement space of the visual modality and a semantic requirement space of the semantic modality; convert the inline relationship between each visual image feature in the visual requirement space into a requirement association feature of the visual modality through the global visual features of the visual image in the requirement document; construct a requirement decision tree of the semantic modality according to the text inline relationship of each semantic requirement feature in the context in the semantic requirement space, and then determine the requirement association feature of the semantic modality from the structural relationship between the requirement decision tree and the semantic requirement feature in the semantic requirement space; perform modal fusion on the requirement association feature of the semantic modality and the requirement association feature of the visual modality, and then use the features after modal fusion as the requirement feature of precision shaft design, and automatically recommend the requirement document through the requirement feature.

[0016] It can be seen that in this application, the demand-related features of the semantic modality and the demand-related features of the visual modality can be modally fused, and then the features after modal fusion are used as the demand features of precision shaft design, and the demand document is automatically recommended through the demand features; first, by extracting the global visual features in the visual image and analyzing its inline relationship, the design elements involved in the image can be fully understood, thereby refining the demand-related features of the visual modality, laying the foundation for modal fusion, ensuring the systematization and accuracy of visual information, and by combining the visual demand features with the features of the semantic modality, a more comprehensive design solution recommendation can be provided, significantly improving the automatic recommendation accuracy of precision shaft design; then, by analyzing the semantic demand features in the text and their contextual inline relationships, the dependencies and priorities between the requirements can be identified, thereby constructing a demand decision tree of the semantic modality, which helps to determine the logical structure and constraints of the design requirements, and provides a clear framework for decision-making. The fusion of the demand-related features of the semantic modality with the features of the visual modality can not only fully capture the design details, but also ensure the comprehensiveness and accuracy of the design requirements, further improving the automatic recommendation accuracy of precision shaft design.

[0017] In summary, the modal fusion of visual and semantic demand features can be achieved, thereby improving the accuracy of automated recommendations in precision shaft design. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 is a flow chart of an intelligent recommendation method for precision shaft design provided by the present application; Figure 2 is an exemplary flow chart of constructing a demand decision tree for semantic modality provided by the present application; Figure 3 is an exemplary flow chart for performing modality fusion according to the present application; Figure 4 This is a module structure diagram of an intelligent recommendation system for precision shaft design provided in this application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] The embodiment of the present application provides an intelligent recommendation system and method for precision shaft design, the core of which is to obtain a requirement document for precision shaft design, perform multimodal feature extraction on the requirement document, and obtain a visual requirement space of the visual modality and a semantic requirement space of the semantic modality; convert the inline relationship between each visual image feature in the visual requirement space into a requirement association feature of the visual modality through the global visual features of the visual image in the requirement document; construct a semantic modality requirement decision tree according to the text inline relationship of each semantic requirement feature in the context in the semantic requirement space, and then determine the semantic modality requirement association feature from the structural relationship between the requirement decision tree and the semantic requirement feature in the semantic requirement space; perform modal fusion on the semantic modality requirement association feature and the visual modality requirement association feature, and then use the modal fusion feature as the requirement feature of precision shaft design, and automatically recommend the requirement document through the requirement feature. The above scheme can be used to perform modal fusion on the visual and semantic requirement features, thereby improving the accuracy of automatic recommendation in precision shaft design.

[0022] Embodiment 1: In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 As shown in FIG. 1 , this figure is an exemplary flow chart of an intelligent recommendation method for precision shaft design according to this embodiment of the present application, and the recommendation method includes the following steps: In step S1, a requirement document for precision shaft design is obtained, and multimodal feature extraction is performed on the requirement document to obtain a visual requirement space of a visual modality and a semantic requirement space of a semantic modality.

[0023] It should be noted that in the present application, the requirement document is a multimodal document used to describe the design of precision shafts, and the requirement document includes image information and text information. The semantic modality is a modality based on text language information; the visual modality is a modality based on image information; in specific implementation, the requirement document for precision shaft design can be obtained from the precision shaft design recommendation system.

[0024] In this embodiment, multimodal feature extraction is performed on the demand document to obtain a visual demand space of a visual modality and a semantic demand space of a semantic modality, specifically in the following manner, namely: Extracting visual demand features and semantic demand features of the demand document through a pre-trained document model; Determine the visual demand space of the visual modality according to all visual demand features; The semantic requirement space of the semantic modality is determined according to all the semantic requirement features.

[0025] It should be noted that in this application, the visual and semantic features in the requirement documents can be extracted through the pre-trained document model, and the image data can be processed by convolutional neural network based on deep learning to extract visual features. At the same time, recurrent neural network or transformer model such as BERT is used to process text data to capture semantic information, so as to integrate visual and text information and provide comprehensive requirement features for precision shaft design.

[0026] In the specific implementation, first, the visual demand features and semantic demand features of the demand document are extracted through a pre-trained document model, the semantic demand features represent the features that can express text demand information, and the visual demand features represent the features that can express image and video demand information; then, the set of all visual demand features can be used as the visual demand space of the visual modality, wherein the visual demand space represents the information space composed of all image and video demand information; finally, the set of all semantic demand features can be used as the semantic demand space of the semantic modality, wherein the semantic demand space represents the information space composed of all text demand information.

[0027] In step S2, the inline relationship between each visual image feature in the visual demand space is converted into a demand association feature of a visual modality through the global visual features of the visual image in the demand document.

[0028] In this embodiment, the inline relationship between the visual image features in the visual demand space is converted into the demand association feature of the visual modality through the global visual features of the visual image in the demand document, which can be specifically implemented in the following manner, namely: extracting global visual features of the visual image from the requirement document; Constructing a structural relationship model of visual modalities based on the global visual features; The structural relationship model is used to convert the inline relationship between each visual image feature in the visual demand space into a demand-related feature of the visual modality.

[0029] It should be noted that in the present application, the requirement-related features of the visual modality represent the interdependence and relationship between the various visual image features in the requirement document; the global visual features reflect the global properties of the visual content; the structural relationship model represents a mathematical model that describes the organization and relationship between the various visual image features under the visual modality, and the structural relationship model is a dependency relationship model based on graph theory.

[0030] In the specific implementation, first, the image analysis model based on deep learning can be used to perform global image processing on the visual images in the demand document, and the result of the global image processing in the image analysis model can be used as the global visual feature of the visual image; then, a structural relationship model based on a directed graph is initialized, and the global visual feature is used as the root node in the structural relationship model to complete the construction of the structural relationship model of the visual modality; finally, the structural relationship model is used to perform feature transformation on the inline relationship between each visual image feature in the visual demand space, and the global visual feature is integrated through the root node. Then, the characteristics of the directed graph are used to perform feature transformation on the inline relationship between each visual image feature in the visual demand space, so that the result of the feature transformation is used as the demand association feature of the visual modality, which helps to capture the hierarchy and dependency relationship between image features, thereby more accurately representing the complex structure of visual information.

[0031] In step S3, a demand decision tree of the semantic modality is constructed according to the text inline relationship of each semantic demand feature in the context in the semantic demand space, and then the demand association characteristics of the semantic modality are determined by the structural relationship between the demand decision tree and the semantic demand features in the semantic demand space.

[0032] Preferably, in this embodiment, reference Figure 2 As shown, this figure is an exemplary flow chart of constructing a demand decision tree for semantic modality in an embodiment of the present application. In this embodiment, the demand decision tree for semantic modality is constructed according to the text inline relationship of each semantic demand feature in the context in the semantic demand space, which can be implemented by the following steps: First, in step S31, the paragraph semantics of each text paragraph is extracted from the demand document; Then, in step S32, the text inline relationship of the semantic requirement feature in the context in the semantic requirement space is determined through all paragraph semantics; Finally, in step S33, a semantic modality demand decision tree is generated based on the text inline relationship.

[0033] It should be noted that in this application, the requirement decision tree is a tree diagram that displays different semantic requirements and their priorities, constraints and decision paths; paragraph semantics represents the core meaning and content of each paragraph in the requirement document; and text inline relationships represent the semantic logical connections between text paragraphs in the requirement document.

[0034] In the specific implementation, first, the requirement document is preprocessed as necessary, including noise removal, punctuation processing, word segmentation, part-of-speech tagging and other operations. The requirement document is segmented into paragraphs, and each text paragraph is processed as an independent semantic unit. Natural language processing technology (such as BERT, GPT, word embedding, etc.) is used to extract the semantic information of each text paragraph as the paragraph semantics of each text paragraph; then, based on the extracted paragraph semantics, the main requirement features involved in the document are identified, such as functional requirements, performance requirements, quality requirements, etc., and the main requirement features are matched to different goals or constraints, and the contextual relationship between paragraphs is determined, that is, by analyzing the association and logical reasoning between paragraphs, it is clear how different requirement features are related to each other in the document. Context-aware models (such as BERT) can be used to analyze the relationship between different paragraphs to determine the requirement features that are inline in the document. and independent or contradictory demand features, and through grammatical dependency analysis, word sense disambiguation, coreference resolution and other technologies, determine the semantic inline relationships between different demand features, such as "dependency", "conditional relationship", "priority", etc., and build a semantic graph. The edges in the semantic graph can be used as the text inline relationships of semantic demand features in the context in the semantic demand space, where the nodes in the semantic graph are semantic demand features and the edges are the relationships between semantic demand features; finally, a multi-level decision tree is generated based on the text inline relationships and the semantic demand space, where each layer of the decision tree corresponds to a decision node, each decision node represents a specific demand feature, and each decision node will be divided according to the demand features involved, and the branches correspond to different decision processes. At each decision node, the optimal decision path is selected according to the demand features and the contextual information in the text, and a semantic modal demand decision tree can be obtained.

[0035] In this embodiment, the demand association features of the semantic modality can be determined by the structural relationship between the demand decision tree and the semantic demand features in the semantic demand space in the following manner, namely: Determine a structural decision tree of semantic modality through the structural relationship of semantic demand features in the semantic demand space; Decision association is performed on the structure decision tree and the demand decision tree to obtain demand association features of semantic modality.

[0036] It should be noted that, in the present application, the demand association feature of the semantic modality represents the mutual relationship and dependency structure between the various semantic demands in the demand document; the structural decision tree is a decision framework generated based on the dependency, priority and constraint relationship between the demand features; the structural relationship of the semantic demand space refers to the structural connection between the various semantic demand features in the demand document, which can be based on the graph representation method (for example: graph convolutional network (GCN), each semantic demand feature can be regarded as a node in the graph, and the edges between the nodes represent the relationship between the semantic demand features (such as dependency, constraint, etc.), and the structural relationship of the semantic demand space can be obtained.

[0037] In specific implementation, first, a hierarchical decision tree algorithm (for example, CART algorithm) can be used to aggregate semantic demand features with direct relationships in the structural relationship into nodes of the same layer. Each node represents a specific demand mode (for example, functional requirements and performance requirements), and the branches between nodes represent sub-requirements or constraints under the mode, so that a structural decision tree of the semantic mode can be obtained. The branches of the decision tree should be selected according to the structural relationship between the semantic demand features, and the selection criteria of the optimization path, such as priority, dependency, etc. For example, in a structural decision tree, the semantic demand feature with the highest priority is used as the root node of the decision, and other semantic demand features that need to rely on the root node decision are used as child nodes; then, the nodes in the structural decision tree are compared with the nodes in the demand decision tree through feature mapping technology (for example, Jaccard similarity), so as to screen out similar and interrelated parts, where the similar parts are manifested as the overlap of the positions of the same semantic demand features in the two decision trees, and the interrelated parts are manifested as the strong correlation between different semantic demand features, so that the collection of similar and interrelated parts is used as the demand association feature of the semantic mode.

[0038] In step S4, the demand-related features of the semantic modality and the demand-related features of the visual modality are modally fused, and the features after modal fusion are used as demand features for precision shaft design, and the demand document is automatically recommended based on the demand features.

[0039] Preferably, in this embodiment, reference Figure 3 As shown, this figure is an exemplary flow chart of modal fusion in an embodiment of the present application. In this embodiment, modal fusion of the demand-related features of the semantic modality and the demand-related features of the visual modality can be specifically implemented by the following steps: First, in step S41, a semantic fusion weight of the semantic modality and a visual fusion weight of the visual modality are obtained; Then, in step S42, the demand-related features of the semantic modality and the demand-related features of the visual modality are fused by using the semantic fusion weight and the visual fusion weight to obtain fused features.

[0040] In specific implementation, first, the semantic fusion weight of the semantic modality and the visual fusion weight of the visual modality can be obtained from the precision shaft design recommendation system. The semantic fusion weight represents the contribution of semantic information in multimodal fusion and determines the influence of text features on decision-making. The visual fusion weight represents the contribution of visual information in multimodal fusion and determines the influence of image features on decision-making. Then, a feature fusion model based on weight fusion is initialized, and the semantic fusion weight is used as the fusion weight of the semantic modality in the feature fusion model, and the visual fusion weight is used as the fusion weight of the visual modality in the feature fusion model. The feature fusion model is used to fuse the demand-related features of the semantic modality and the demand-related features of the visual modality, and the result of the feature fusion can be used as the fused feature.

[0041] In this embodiment, the automatic recommendation of the demand document through the demand characteristics can be specifically carried out in the following manner, namely: matching the demand characteristics with the standards and solutions in the demand characteristic library, and combining content-based recommendation, collaborative filtering and hybrid recommendation methods to provide users with personalized recommendation results based on the demand characteristics.

[0042] It can be seen that in this application, the demand-related features of the semantic modality and the demand-related features of the visual modality can be modally fused, and then the features after modal fusion can be used as the demand features of precision shaft design, and the demand documents can be automatically recommended through the demand features; first, by extracting the global visual features in the visual image and analyzing its inline relationship, the design elements involved in the image can be fully understood, thereby refining the demand-related features of the visual modality, laying the foundation for modal fusion, ensuring the systematization and accuracy of visual information, and by combining the visual demand features with the features of the semantic modality, a more comprehensive design solution recommendation can be provided, significantly improving the accuracy of automatic recommendation for precision shaft design; then, by analyzing the semantic demand features in the text and their contextual inline relationships, the dependencies and priorities between the requirements can be identified, thereby constructing a demand decision tree of the semantic modality, which helps to determine the logical structure and constraints of the design requirements, and provides a clear framework for decision-making. The fusion of the demand-related features of the semantic modality with the features of the visual modality can not only fully capture the design details, but also ensure the comprehensiveness and accuracy of the design requirements, further improving the accuracy of automatic recommendation for precision shaft design.

[0043] In summary, the modal fusion of visual and semantic demand features can be achieved, thereby improving the accuracy of automated recommendations in precision shaft design.

[0044] Embodiment 2: This application provides an intelligent recommendation system for precision shaft design. Figure 4 As shown, this figure is a schematic diagram of a recommendation system according to this embodiment of the present application, and the recommendation system includes: The feature extraction module 100 is used to obtain a requirement document for precision shaft design, perform multimodal feature extraction on the requirement document, and obtain a visual requirement space of a visual modality and a semantic requirement space of a semantic modality; A visual modality module 200, configured to convert the inline relationship between the visual image features in the visual demand space into a demand association feature of a visual modality through the global visual features of the visual image in the demand document; The semantic modality module 300 is used to construct a semantic modality demand decision tree according to the text inline relationship of each semantic demand feature in the semantic demand space in the context, and then determine the demand association feature of the semantic modality from the demand decision tree and the structural relationship between the semantic demand features in the semantic demand space; The feature fusion module 400 is used to perform modal fusion on the demand-related features of the semantic modality and the demand-related features of the visual modality, and then use the features after modal fusion as the demand features of precision shaft design, and automatically recommend the demand document through the demand features.

[0045] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0046] A person skilled in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, the storage medium including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically-erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0047] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. An intelligent recommendation method for precision shaft design, characterized in that: The recommended method comprises the following steps: Obtaining a requirement document for precision shaft design, performing multimodal feature extraction on the requirement document, and obtaining a visual requirement space of a visual modality and a semantic requirement space of a semantic modality; Converting the inline relationship between each visual image feature in the visual demand space into a demand association feature of a visual modality through the global visual feature of the visual image in the demand document; Constructing a semantic modal demand decision tree according to the text inline relationship of each semantic demand feature in the context in the semantic demand space, and then determining the demand association feature of the semantic modal from the demand decision tree and the structural relationship between the semantic demand features in the semantic demand space; Modality fusion is performed on the demand-related features of the semantic modality and the demand-related features of the visual modality, and the features after modality fusion are used as demand features for precision shaft design, and the demand document is automatically recommended based on the demand features.

2. An intelligent recommendation method for precision shaft design according to claim 1, characterized in that: Performing multimodal feature extraction on the demand document to obtain a visual demand space of a visual modality and a semantic demand space of a semantic modality specifically includes: Extracting visual demand features and semantic demand features of the demand document through a pre-trained document model; Determine the visual demand space of the visual modality according to all visual demand features; The semantic requirement space of the semantic modality is determined according to all the semantic requirement features.

3. The intelligent recommendation method for precision shaft design according to claim 1, characterized in that: The method of converting the inline relationship between the visual image features in the visual demand space into the demand association features of the visual modality through the global visual features of the visual image in the demand document specifically includes: extracting global visual features of the visual image from the requirement document; Constructing a structural relationship model of visual modalities based on the global visual features; The structural relationship model is used to convert the inline relationship between each visual image feature in the visual demand space into a demand-related feature of the visual modality.

4. An intelligent recommendation method for precision shaft design as claimed in claim 3, characterized in that: The structural relationship model is a dependency relationship model based on graph theory.

5. The intelligent recommendation method for precision shaft design according to claim 1, characterized in that: Constructing a semantic modal demand decision tree according to the text inline relationship of each semantic demand feature in the semantic demand space in the context specifically includes: Extracting paragraph semantics of each text paragraph from the requirement document; Determine the text inline relationship of the semantic requirement features in the context in the semantic requirement space through all paragraph semantics; A semantic modality demand decision tree is generated based on the text inline relationship.

6. The intelligent recommendation method for precision shaft design according to claim 1, characterized in that: Determining the demand association features of the semantic modality based on the structural relationship between the demand decision tree and the semantic demand features in the semantic demand space specifically includes: Determine a structural decision tree of semantic modality through the structural relationship of semantic demand features in the semantic demand space; Decision association is performed on the structure decision tree and the demand decision tree to obtain demand association features of semantic modality.

7. The intelligent recommendation method for precision shaft design according to claim 1, characterized in that: The semantic modality is a modality based on text language information.

8. The intelligent recommendation method for precision shaft design according to claim 1, characterized in that: The visual modality is a modality based on image information.

9. The intelligent recommendation method for precision shaft design according to claim 1, characterized in that: The modality fusion of the demand-related features of the semantic modality and the demand-related features of the visual modality specifically includes: Obtaining a semantic fusion weight of the semantic modality and a visual fusion weight of the visual modality; The demand-related features of the semantic modality and the demand-related features of the visual modality are fused by using the semantic fusion weight and the visual fusion weight to obtain fused features.

10. An intelligent recommendation system for precision shaft design, used to execute an intelligent recommendation method for precision shaft design as claimed in any one of claims 1 to 9, characterized in that: The recommendation system comprises: A feature extraction module is used to obtain a requirement document for precision shaft design, perform multimodal feature extraction on the requirement document, and obtain a visual requirement space of a visual modality and a semantic requirement space of a semantic modality; A visual modality module, used for converting the inline relationship between each visual image feature in the visual demand space into a demand association feature of the visual modality through the global visual feature of the visual image in the demand document; A semantic modality module, for constructing a semantic modality demand decision tree according to the text inline relationship of each semantic demand feature in the semantic demand space in the context, and then determining the demand association feature of the semantic modality from the demand decision tree and the structural relationship between the semantic demand features in the semantic demand space; The feature fusion module is used to perform modal fusion on the demand-related features of the semantic modality and the demand-related features of the visual modality, and then use the features after modal fusion as the demand features of precision shaft design, and automatically recommend the demand document through the demand features.

Images