Enterprise data asset intelligent management system for server research, development and manufacturing

Abstract

The invention relates to the technical field of server research, development and manufacturing, and discloses an enterprise data asset intelligent management system for server research, development and manufacturing. The system comprises a data asset access module used for acquiring initial information of enterprise data assets to be managed; the multi-dimensional feature extraction module is used for extracting content, structure and semantic features of the data assets; the intelligent classification processing module performs multi-dimensional classification processing to obtain a classification result; the dynamic storage management module generates a storage strategy; and the application service output module executes application service output and displays a result. According to the system, efficient access, comprehensive feature extraction, scientific classification, dynamic storage management and convenient application service output of enterprise data assets in the research, development and manufacturing process of the server are realized, and the management level and the utilization value of the enterprise data assets are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of server R&D and manufacturing, and in particular to an enterprise data asset intelligent management system for server R&D and manufacturing. Background Art

[0002] During the server R&D and manufacturing process, companies generate a large amount of data assets, which are of great value to all aspects of a company's R&D, production, and management. However, as server R&D and manufacturing continue to grow, the scale and complexity of these data assets are also increasing, posing numerous challenges to traditional data asset management systems.

[0003] The sources of data assets are diverse and complex. Server R&D and manufacturing involve multiple stages, including design, R&D, production, and testing. Each stage generates a vast amount of data. This data may originate from different devices, systems, and departments, and initial information such as data source identification, format type, storage location, and associated object identification varies. This makes accessing and managing data assets difficult. Traditional access methods struggle to efficiently handle such complex data sources, making it difficult to quickly and accurately acquire and manage these data assets.

[0004] The feature extraction of data assets is not comprehensive and in-depth enough. Traditional feature extraction methods often focus only on surface features of data, such as text content and numerical values, while ignoring the data's structural and semantic characteristics. For example, structural features such as data organization, storage hierarchy, and associations, as well as semantic features such as keywords, contextual relationships, and business meaning, are crucial for the classification and management of data assets. Without comprehensive and in-depth extraction of these features, accurate classification and management of data assets is impossible, which impacts the efficiency of data asset utilization.

[0005] The classification of data assets is not scientific or rational enough. Traditional classification methods typically base their classification on a single dimension, such as data type or usage, and fail to meet the needs of multi-dimensional classification. During server R&D and manufacturing, data assets possess diverse characteristics and attributes, requiring classification across multiple dimensions, including content, structure, and semantics, to achieve accurate results. Traditional classification methods are unable to achieve multi-dimensional classification, resulting in inaccurate data asset classification results and an inability to effectively support data asset storage and application.

[0006] Furthermore, data asset storage management lacks dynamism and intelligence. Traditional storage management policies are typically fixed and unable to dynamically adjust based on data asset classification and actual needs. As data assets continually change and update, fixed storage policies may not meet their storage needs, leading to wasted or insufficient storage resources. Furthermore, traditional storage management policies lack intelligent decision-making capabilities and are unable to automatically generate optimal storage policies based on data asset characteristics and application scenarios, impacting data asset storage efficiency and availability.

[0007] Finally, the application service output of data assets is not efficient and convenient. Traditional application service output methods often require manual intervention. Data asset storage policies cannot be automatically input into the managed enterprise data assets to execute application service output. Furthermore, the output results are not intuitive or clear. This leads to inefficient data asset application and a failure to provide timely and effective support for enterprise decision-making and management. Summary of the Invention

[0008] The purpose of the present invention is to provide an enterprise data asset intelligent management system for server R&D and manufacturing to solve the problems raised in the above background technology.

[0009] To achieve the above objectives, the present invention provides an enterprise data asset intelligent management system for server R&D and manufacturing, the system comprising:

[0010] The data asset access module is used to obtain the initial information of the enterprise data assets to be managed generated during the server R&D and manufacturing process. The initial information includes the data source identifier, format type, storage location and associated object identifier;

[0011] A multi-dimensional feature extraction module, configured to extract content features, structural features, and semantic features of the data assets based on the initial information of the enterprise data assets to be managed;

[0012] An intelligent classification processing module is used to perform multi-dimensional classification processing on the content features, structural features and semantic features to obtain data asset classification results;

[0013] A dynamic storage management module, configured to generate a data asset storage strategy based on the data asset classification result;

[0014] The application service output module is used to input the data asset storage policy into the enterprise data asset to be managed to execute application service output, and transmit the output result to the display terminal for display.

[0015] Preferably, the data asset access module includes:

[0016] The access protocol adaptation unit is used to extract the version information, compatible parameters and transmission rules of the data asset access protocol from the interface configuration library;

[0017] a data type identification unit, configured to extract the function description of each access protocol in the interface configuration library to obtain a set of data type descriptions;

[0018] a feature pre-extraction unit, configured to perform feature extraction on the initial information of the enterprise data asset to be managed and each data type description in the set of data type descriptions to obtain a set of initial information feature vectors and data type feature vectors;

[0019] an association relationship parsing unit, configured to perform association relationship parsing on the set of the initial information feature vector and the data type feature vector to obtain a data-type association optimization vector;

[0020] An access rule generating unit is configured to generate an access rule for the data asset based on the data-type association optimization vector.

[0021] Preferably, the association relationship analysis unit includes:

[0022] a data type feature aggregation subunit, configured to perform feature aggregation processing based on hierarchical mapping on the set of data type feature vectors to obtain a guiding template;

[0023] The cross-domain collaborative parsing subunit is used to perform cross-domain collaborative parsing on the set of the initial information feature vector and the data type feature vector based on the guiding template to obtain the data-type association optimization vector.

[0024] Preferably, the data type feature aggregation subunit includes:

[0025] A data type feature hierarchical mapping secondary subunit is used to perform hierarchical mapping processing on each data type feature vector in the set of data type feature vectors using a mapping matrix to obtain a set of hierarchically transformed data type feature vectors;

[0026] A data type feature matrix arrangement secondary subunit is used to perform matrix arrangement on the set of data type feature vectors after the hierarchical transformation to obtain a data type feature matrix;

[0027] The data type feature matrix extreme value extraction secondary sub-unit is used to extract the extreme values ​​of the data type feature vectors after each level transformation in the data type feature matrix to obtain the data type feature matrix key vector as the guiding template.

[0028] Preferably, the data type feature hierarchical mapping secondary sub-unit includes:

[0029] The data type feature vector is point-multiplied by the mapping matrix and then added to the mapping bias vector in a positional manner to obtain the data type feature vector after hierarchical transformation.

[0030] Preferably, the cross-domain collaborative resolution subunit includes:

[0031] an initial information hierarchical mapping secondary subunit, configured to perform hierarchical mapping processing on the initial information feature vector using a query matrix and a value matrix to obtain an initial information query vector and an initial information value vector;

[0032] a template-guided heterogeneous conversion parsing secondary subunit, configured to input the initial information query vector, the initial information value vector, the transformed data type feature vectors at each level in the data type feature matrix, and the guiding template into a template-guided heterogeneous conversion structure to obtain a sequence of data-type cross-domain collaborative parsing vectors;

[0033] The position mean calculation secondary subunit is used to calculate the position mean vector of the sequence of the data-type cross-domain collaborative resolution vectors to obtain the data-type association optimization vector.

[0034] Preferably, the template-guided isomerization conversion and resolution secondary subunit comprises:

[0035] After calculating the product between the initial information query vector and the transposed vector of the data type feature vector after the hierarchical transformation, the obtained data-type association feature matrix is ​​divided by the modulus length of the key vector of the data type feature matrix by position to obtain a data-type association weight matrix;

[0036] Inputting the data-type association weight matrix into a normalization function for processing to obtain a data-type association normalized weight matrix;

[0037] After multiplying the data-type association normalized weight matrix with the key vector of the data type feature matrix, the obtained feature vector is multiplied by position with the initial information value vector to obtain the data-type cross-domain collaborative resolution vector.

[0038] Preferably, the access rule generating unit includes:

[0039] Inputting the data-type association optimization vector into an access rule recommendation module based on a decision maker to obtain an access rule for the data asset;

[0040] Based on the access rule, an access scope of the data asset is determined.

[0041] Preferably, the multidimensional feature extraction module includes:

[0042] A content feature extraction unit, configured to extract features of text content, numerical content, and multimedia content from the initial information of the enterprise data assets to be managed;

[0043] a structural feature extraction unit, configured to extract features of data organization, storage hierarchy, and association relationships from the initial information;

[0044] A semantic feature extraction unit for analyzing the features of keywords, contextual relationships, and business meanings in the initial information;

[0045] The feature fusion unit is used to fuse the content features, structural features and semantic features to obtain a comprehensive feature vector.

[0046] Preferably, the feature fusion unit includes:

[0047] A content feature weight allocation subunit, configured to allocate weight coefficients to the content features, structural features, and semantic features respectively;

[0048] The feature weighting calculation subunit is used to multiply the content feature, structural feature, and semantic feature by the corresponding weight coefficients respectively and then add them together to obtain the comprehensive feature vector.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] In terms of data asset access, the data asset access module extracts the version information, compatible parameters and transmission rules of the data asset access protocol from the interface configuration library through the access protocol adaptation unit, ensuring the accuracy and compatibility of the access protocol; the data type identification unit extracts the functional descriptions of each access protocol in the interface configuration library to obtain a set of data type descriptions, providing a basis for data type identification; the feature pre-extraction unit extracts features from the initial information and data type descriptions to obtain feature vectors, laying the foundation for subsequent association relationship analysis; the association relationship analysis unit obtains the data-type association optimization vector through hierarchical mapping feature aggregation processing and cross-domain collaborative analysis, improving the accuracy of data and type association; the access rule generation unit generates access rules based on the association optimization vector and determines the access scope, realizing efficient and accurate access to data assets, and solving the access difficulty problem caused by the complex data sources in traditional access methods.

[0051] During feature extraction, the multidimensional feature extraction module's content feature extraction unit extracts features from text, numerical values, and multimedia content. The structural feature extraction unit extracts features related to data organization, storage hierarchy, and relationships. The semantic feature extraction unit analyzes features related to keywords, contextual relationships, and business meaning. The feature fusion unit then fuses these three types of features to create a comprehensive feature vector. This comprehensive and in-depth feature extraction approach, compared to traditional methods that focus solely on surface features, more fully captures the essential characteristics of data assets, providing richer and more accurate feature information for subsequent classification processing.

[0052] The intelligent classification processing module performs multi-dimensional classification processing on content, structural, and semantic features to obtain accurate data asset classification results. Compared with traditional single-dimensional classification methods, multi-dimensional classification can more comprehensively consider the various characteristics and attributes of data assets, making the classification results more scientific and reasonable, and providing more effective support for the storage and application of data assets.

[0053] The dynamic storage management module generates data asset storage policies based on classification results, making storage management dynamic and intelligent. These policies can be adjusted based on data asset classification results and actual needs, avoiding the waste or insufficiency of storage resources caused by traditional fixed storage policies and improving storage resource utilization. Intelligent policy generation also enhances the efficiency and usability of storage management.

[0054] The application service output module inputs storage policies to the enterprise data assets to be managed, executes application service output, and transmits the results to the display terminal for display. This process automates and facilitates application service output, eliminating the need for manual intervention and improving the efficiency of data asset utilization. The intuitive display of output results also provides timely and effective support for enterprise decision-making and management.

[0055] In addition, the entire system forms a complete intelligent data asset management process through the collaborative work between various modules. From data access, feature extraction, classification processing, storage management to application service output, each link is closely connected and supports each other, realizing the comprehensive, efficient and intelligent management of enterprise data assets in the server R&D and manufacturing process, greatly improving the management level and utilization value of enterprise data assets, and providing strong technical support for the development of enterprises in the field of server R&D and manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a working principle diagram of the enterprise data asset intelligent management system for server R&D and manufacturing according to the present invention;

[0057] Figure 2 Workflow diagram for data asset access module;

[0058] Figure 3 Workflow diagram for the data type feature aggregation subunit;

[0059] Figure 4 This is the workflow diagram of the cross-domain collaborative parsing sub-unit. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0061] See also Figure 1-Figure 4 The present invention provides an intelligent enterprise data asset management system for server R&D and manufacturing. The system includes: a data asset access module, a multi-dimensional feature extraction module, an intelligent classification processing module, a dynamic storage management module, and an application service output module. The specific implementation steps are as follows:

[0062] The data asset access module is used to obtain the initial information of the enterprise data assets to be managed generated during the server R&D and manufacturing process. The initial information includes data source identification, format type, storage location and associated object identification.

[0063] The multi-dimensional feature extraction module extracts the content features, structural features and semantic features of the data assets based on the initial information of the enterprise data assets to be managed.

[0064] The intelligent classification processing module performs multi-dimensional classification processing on content features, structural features and semantic features to obtain data asset classification results.

[0065] The dynamic storage management module generates a data asset storage strategy based on the data asset classification results.

[0066] The application service output module inputs the data asset storage strategy into the enterprise data assets to be managed to execute the application service output, and transmits the output results to the display terminal for display.

[0067] Example 1:

[0068] The data asset access module consists of an access protocol adapter unit, a data type identification unit, a feature pre-extraction unit, an association analysis unit, and an access rule generation unit. The access protocol adapter unit extracts the data asset access protocol version information, compatibility parameters, and transmission rules from the interface configuration repository. The interface configuration repository stores information related to various access protocols. This information is extracted to enable subsequent data asset access according to the appropriate protocol.

[0069] The data type identification unit extracts the functional descriptions of each access protocol from the interface configuration library, resulting in a collection of data type descriptions. Each access protocol has a corresponding functional description that details the characteristics and requirements of the data types it can handle. Collecting these descriptions into a collection helps accurately determine the type of data assets.

[0070] The feature pre-extraction unit extracts features from the initial information of the enterprise's data assets and each data type description in the data type description set, ultimately generating an initial information feature vector and a set of data type feature vectors. The initial information, which includes information such as the data source identifier, format type, storage location, and associated object identifiers, is converted into an initial information feature vector using a specific feature extraction method. Similarly, for each data type description in the data type description set, a corresponding extraction method is applied to generate multiple data type feature vectors, ultimately forming a set of data type feature vectors.

[0071] The task of the association analysis unit is to analyze the association between the initial information feature vector and the set of data type feature vectors to obtain the data-type association optimization vector. This unit specifically includes the data type feature aggregation subunit and the cross-domain collaborative analysis subunit.

[0072] The data type feature aggregation subunit performs hierarchical mapping-based feature aggregation on the set of data type feature vectors, ultimately obtaining a guiding template. It further comprises the data type feature hierarchical mapping subunit, the data type feature matrix arrangement subunit, and the data type feature matrix extreme value extraction subunit. The data type feature hierarchical mapping subunit uses a mapping matrix to perform hierarchical mapping on each data type feature vector in the set of data type feature vectors. Specifically, it performs a dot product of the data type feature vector with the mapping matrix and then positionally adds the data type feature vector to the mapping bias vector. This yields a hierarchically transformed data type feature vector. All transformed vectors constitute the set of hierarchically transformed data type feature vectors. The data type feature matrix arrangement subunit performs a matrix arrangement on the set of hierarchically transformed data type feature vectors to obtain a data type feature matrix. The data type feature matrix extreme value extraction subunit extracts the extreme values ​​of each hierarchically transformed data type feature vector from the data type feature matrix. These extreme values ​​form the key vectors of the data type feature matrix, which serve as the guiding template.

[0073] The cross-domain collaborative parsing subunit performs cross-domain collaborative parsing on a set of initial information feature vectors and data type feature vectors based on a guided template to obtain a data-type association optimization vector. It includes a secondary subunit for initial information hierarchical mapping, a secondary subunit for template-guided heterogeneous transformation parsing, and a secondary subunit for positional mean calculation. The secondary subunit for initial information hierarchical mapping uses a query matrix and a value matrix to perform hierarchical mapping processing on the initial information feature vector, thereby obtaining an initial information query vector and an initial information value vector. The secondary subunit for template-guided heterogeneous transformation parsing inputs the initial information query vector, the initial information value vector, the data type feature vectors after each level transformation in the data type feature matrix, and the guiding template into a template-guided heterogeneous transformation structure. After processing, a sequence of data-type cross-domain collaborative parsing vectors is obtained. The secondary subunit for positional mean calculation calculates the positional mean vector of this sequence, ultimately obtaining a data-type association optimization vector.

[0074] The access rule generation unit generates access rules for data assets based on the data-type association optimization vector. Specifically, the data-type association optimization vector is input into the decision-maker-based access rule recommendation module. This module recommends appropriate access rules based on the input vector information, thereby generating access rules for the data asset. Simultaneously, based on the generated access rules, the access scope of the data asset is determined, clarifying the specific scope and restrictions of the data asset when accessing the system.

[0075] Example 2:

[0076] The cross-domain collaborative analysis subunit is an important component of the correlation analysis unit, and internally includes three specific processing units: an initial information hierarchical mapping secondary subunit, a template-guided heterogeneous conversion analysis secondary subunit, and a position mean calculation secondary subunit. Each unit undertakes different processing tasks, and through orderly collaboration, achieves cross-domain collaborative analysis of data, and ultimately obtains a data-type correlation optimization vector.

[0077] The initial information hierarchical mapping secondary subunit is responsible for hierarchical mapping processing of the initial information feature vector using a query matrix and a value matrix. The initial information feature vector is extracted from the initial information of the enterprise data asset to be managed, and includes features corresponding to data source identification, format type, storage location, and correlation object identification. The query matrix and the value matrix are matrices with specific dimensions and parameters that are pre-set, and play a key mapping role in the processing. The specific processing process is to operate the initial information feature vector with the query matrix to obtain an initial information query vector, and to operate the initial information feature vector with the value matrix to obtain an initial information value vector. These two vectors are important input data for subsequent processing, and they convert and represent the initial information features from different angles.

[0078] The template-guided heterogeneous conversion analysis secondary subunit needs to input the initial information query vector, the initial information value vector, the data type feature vector after each level transformation in the data type feature matrix, and the guide template into the heterogeneous conversion structure based on template guidance. The data type feature matrix is obtained by hierarchical mapping processing and matrix arrangement of the data type feature vector, and each level-transformed data type feature vector contains feature information of the corresponding data type. The guide template is obtained by feature aggregation processing of the set of data type feature vectors, and carries key feature information of the data type.

[0079] In the heterogeneous conversion structure, the specific processing steps are as follows: first, calculate the product of the transpose vector of the initial information query vector and the level-transformed data type feature vector, which will obtain a data-type correlation feature matrix. This matrix reflects the correlation degree between the initial information query vector and each level-transformed data type feature vector. Next, divide the obtained data-type correlation feature matrix by the modulus of the key vector of the data type feature matrix according to the position. The data type feature matrix key vector is composed of extreme values extracted from the data type feature matrix, and represents the key features of the data type. The modulus is used for normalization processing of the correlation feature matrix to obtain a data-type correlation weight matrix. The weight matrix reflects the correlation weight between different data type features and the initial information.

[0080] The data-type association weight matrix is ​​then input into a normalization function for processing. The normalization function maps the values ​​in the weight matrix to a specific range, making the individual weight values ​​comparable. After processing, the data-type association normalized weight matrix is ​​obtained. Finally, the data-type association normalized weight matrix is ​​multiplied by the key vector of the data type feature matrix to obtain a feature vector. This feature vector is then positionally multiplied with the initial information value vector to obtain the data-type cross-domain collaborative resolution vector. This vector comprehensively considers the association between the initial information and the data type features, enabling collaborative resolution of cross-domain information.

[0081] The task of the positional mean calculation secondary subunit is to process the sequence of data-type cross-domain collaborative resolution vectors. In the template-guided heterogeneous transformation and resolution secondary subunit, multiple data-type cross-domain collaborative resolution vectors are generated for the data type feature vectors after different hierarchical transformations, thus forming a vector sequence. The positional mean calculation secondary subunit needs to calculate the positional mean vector of this sequence. Specifically, for the value at each position in the vector sequence, the average value of all vectors at that position is calculated, and finally a new vector is obtained. This vector is the data-type association optimization vector.

[0082] After obtaining the data-type association optimization vector, the access rule generation unit inputs it into the decision-maker-based access rule recommendation module. This decision-maker-based access rule recommendation module contains pre-set decision logic and a rule library. Based on the input data-type association optimization vector, it analyzes the characteristics and type of the data asset to recommend appropriate access rules, ultimately generating the access rules for the data asset.

[0083] After obtaining the access rules, the access rule generation unit also needs to determine the access scope of the data asset based on the access rules. This determination requires considering various parameters and restrictions in the access rules, such as data source restrictions, format type requirements, and storage location regulations. By analyzing and processing these conditions, the specific scope of data assets when accessing the system is clarified, including restrictions and requirements on the data types that can be accessed, data sources, and data processing methods.

[0084] Example 3:

[0085] The multi-dimensional feature extraction module is an important component for realizing intelligent management of enterprise data assets. It includes content feature extraction unit, structural feature extraction unit, semantic feature extraction unit and feature fusion unit. Each unit realizes the extraction and fusion of multi-dimensional features of data assets through a specific processing flow, providing a comprehensive feature basis for subsequent intelligent classification processing.

[0086] The function of the content feature extraction unit is to extract features of textual content, numerical content, and multimedia content from the initial information of the enterprise data assets to be managed. During the server R&D and manufacturing process, the content of data assets varies. Textual content may include technical specifications in R&D documents and process specifications in the manufacturing process. Natural language processing technology is used to extract features such as keywords and sentence structure from these texts. Numerical content, such as the performance parameters of server components and cost data in the manufacturing process, can be analyzed statistically to extract features such as mean, variance, and extreme values. Multimedia content includes design drawings during the R&D process and surveillance videos from the manufacturing site. For images, features such as color, texture, and shape can be extracted, while for videos, features such as keyframes and action sequences can be extracted. Feature extraction of these different types of content can fully reflect the content attributes of data assets.

[0087] The structural feature extraction unit is responsible for extracting the data organization, storage hierarchy, and relationship characteristics from the initial information. The data organization reflects the logical arrangement of the data, such as the row and column structure of tabular data and the hierarchical relationship of tree-shaped data. By analyzing the data organization, the corresponding structural features can be extracted. The storage hierarchy refers to the physical distribution level of data in the storage medium, such as the hierarchical division of different folders in a server storage system. Extracting storage hierarchy features helps understand the data storage architecture. The relationship characteristics focus on the mutual connection between data, such as the reference relationship between different R&D documents, the correspondence between component parameters and product models, etc. These relationship characteristics can be extracted through methods such as graph structure analysis. The extraction of these structural features can clearly present the organization and storage mode of data assets.

[0088] The semantic feature extraction unit is tasked with analyzing the keywords, contextual relationships, and business implications of the initial information. Keywords are the words that embody the core content of the data. Key terms can be extracted from the text using algorithms such as word frequency statistics and TF-IDF. Contextual relationships reflect the interactions between keywords within the context. Semantic associations between keywords can be analyzed using techniques such as semantic networks and syntactic analysis. Business implications require understanding the actual business significance of the data, taking into account the business context of server R&D and manufacturing. For example, the specific impact of a performance parameter on server cooling design can be explored using tools such as domain knowledge graphs. Extracting semantic features enables the system to understand the inherent meaning of data assets from a business perspective.

[0089] The function of the feature fusion unit is to fuse content features, structural features, and semantic features to obtain a comprehensive feature vector. This unit contains a content feature weight assignment subunit and a feature weight calculation subunit. The content feature weight assignment subunit assigns weight coefficients to content features, structural features, and semantic features respectively. The determination of the weight coefficients needs to consider the importance of different features in the classification of data assets. For example, in some scenarios, semantic features of business meaning may have a greater impact on the classification results, while in other scenarios, the structural features of the data may be more critical. Weight assignment can be achieved through expert experience setting, machine learning algorithm training, etc., to ensure that the weight coefficients can reasonably reflect the importance of each feature.

[0090] The feature weighting calculation subunit multiplies the content features, structural features, and semantic features by their corresponding weight coefficients and then adds them together to obtain a comprehensive feature vector. Specifically, assuming that the content features are represented by vector C, the structural features are represented by vector S, and the semantic features are represented by vector Y, and the corresponding weight coefficients are w1, w2, and w3, respectively, then the comprehensive feature vector F is calculated as F = w1×C+w2×S+w3×Y. Through this weighted summation method, features of different dimensions are integrated into a unified vector. This vector integrates the content, structure, and semantic information of the data asset, and can more comprehensively describe the characteristics of the data asset.

[0091] In actual applications, the workflow of the multidimensional feature extraction module is as follows: first, the data asset access module obtains the initial information of the enterprise data assets to be managed and transmits it to the multidimensional feature extraction module; then, the content feature extraction unit, the structural feature extraction unit, and the semantic feature extraction unit process the initial information respectively and extract the corresponding feature vectors; then, the feature fusion unit performs weight allocation and weighted calculation on these three types of feature vectors to generate a comprehensive feature vector; finally, the comprehensive feature vector is passed to the intelligent classification processing module as the basis for data asset classification.

[0092] The various units in this embodiment cooperate with each other to extract and fuse features of data assets from different angles, ensuring that the extracted features can fully and accurately reflect the essential attributes of the data assets. The content feature extraction unit covers the specific content form of the data, the structural feature extraction unit reveals the organization and storage method of the data, the semantic feature extraction unit explores the business connotation of the data, and the feature fusion unit organically combines these multi-dimensional features to form a unified comprehensive feature vector. This multi-dimensional feature extraction and fusion method provides a solid foundation for the intelligent classification and subsequent management of data assets, enabling the system to classify and process data assets more accurately, and then generate reasonable storage strategies to achieve intelligent management of data assets. Through such a processing flow, the multi-dimensional feature extraction module plays a key role in the management of data assets of server R&D and manufacturing enterprises, ensuring that data assets can be effectively utilized and managed.

[0093] Example 4:

[0094] The access protocol adapter unit of the data asset access module extracts the version information, compatibility parameters and transmission rules of the data asset access protocol from the interface configuration library. For example, in the server R&D and manufacturing process, a certain type of data asset comes from a specific model of test equipment, and the interface configuration library stores the access protocol corresponding to the equipment. The access protocol adapter unit extracts the version information of the protocol, such as V2.3, to clarify the update and iteration status of the protocol; extracts compatibility parameters, such as the supported operating system versions of Windows Server 2019 and above, Linux CentOS 7.6 and above, to ensure that the equipment and software environment of the access system meet the requirements; extracts transmission rules, such as data transmission using TCP / IP protocol, port number 8080, and the transmission rate must be maintained above 100Mbps to ensure the stability and efficiency of data transmission.

[0095] The data type identification unit extracts the functional description of each access protocol in the interface configuration library to obtain a collection of data type descriptions. For example, the interface configuration library contains access protocols for different types of data, such as R&D design documents, manufacturing process data, and quality inspection reports. For the access protocol for R&D design documents, its functional description will elaborate on the application of the protocol to data types such as CAD drawings and PDF technical documents, and clarify the format requirements, content structure, etc. of these data; for the access protocol for manufacturing process data, the functional description will indicate that it is applicable to types such as production process parameters and equipment operation data, and explain the data collection frequency, unit specifications, etc. These functional descriptions are extracted to form a collection containing descriptions of various data types, providing a basis for the subsequent accurate identification of the type of data assets.

[0096] The feature pre-extraction unit extracts features from the initial information of the enterprise data assets to be managed and each data type description in the data type description set, obtaining a set of initial information feature vectors and data type feature vectors. Taking a BOM (bill of materials) document generated during the development of a server motherboard as an example, its initial information includes a data source identifier of "Motherboard R&D Team", a format type of Excel, a storage location of the server storage path " / R&D Data / Motherboard / BOM / 202506", and an associated object identifier of "Motherboard Model A-001". The feature pre-extraction unit processes this initial information, converting the data source identifier into a corresponding encoding vector, the Excel format type into a specific format feature vector, the storage location into a path feature vector, and the associated object identifier into a model feature vector, thereby obtaining an initial information feature vector. At the same time, for the data type description of the Excel format BOM document in the data type description set, the feature pre-extraction unit extracts the column name information, data format specifications, required fields, and other features contained therein, and converts them into data type feature vectors. Multiple such vectors constitute a set of data type feature vectors.

[0097] The association relationship analysis unit performs association relationship analysis on the set of initial information feature vectors and data type feature vectors to obtain a data-type association optimization vector. The data type feature aggregation subunit of this unit performs feature aggregation processing based on hierarchical mapping on the set of data type feature vectors to obtain a guidance template. Continuing with the BOM document as an example, the set of data type feature vectors contains feature vectors of BOM documents in different Excel formats. The data type feature hierarchical mapping secondary subunit uses a mapping matrix to perform hierarchical mapping processing on these vectors, multiplying each vector by the mapping matrix and then adding the mapping bias vector by position to obtain the data type feature vector after hierarchical transformation; the data type feature matrix arrangement secondary subunit arranges these transformed vectors into a matrix to form a data type feature matrix; the data type feature matrix extreme value extraction secondary subunit extracts the extreme values ​​of the data type feature vectors after each hierarchical transformation in the matrix to obtain the data type feature matrix key vector as a guidance template.

[0098] The cross-domain collaborative parsing subunit performs cross-domain collaborative parsing on the set of initial information feature vectors and data type feature vectors based on the guided template. The initial information hierarchical mapping subunit uses the query matrix and value matrix to perform hierarchical mapping processing on the initial information feature vectors of the BOM document, obtaining the initial information query vector and the initial information value vector. The template-guided heterogeneous transformation parsing subunit inputs these two vectors, the data type feature vectors after each level transformation in the data type feature matrix, and the guiding template into the template-guided heterogeneous transformation structure, calculates the product of the initial information query vector and the transposed vector of the hierarchically transformed data type feature vector to obtain the data-type association feature matrix, divides it by the modulus of the key vector of the data type feature matrix by position, and obtains the data-type association weight matrix. This matrix is ​​input into the normalization function to obtain the normalized weight matrix, which is then multiplied by the key vector of the data type feature matrix. The obtained feature vector is then dot-multiplied by the initial information value vector by position to obtain the data-type cross-domain collaborative parsing vector. The position mean calculation subunit calculates the position mean vector of the vector sequence to obtain the data-type association optimization vector.

[0099] The access rule generation unit generates access rules for data assets based on the data-type association optimization vector and determines the access scope. The data-type association optimization vector is input into the decision-maker-based access rule recommendation module. For BOM documents, an access rule is recommended, such as "Allow the motherboard R&D team to upload data through a specified Excel template. The uploaded file size must not exceed 50MB and must include required fields such as material code, name, specification, quantity, etc." Based on this rule, the access scope is determined to be limited to Excel format BOM documents uploaded by the motherboard R&D team that meet the specified template and format requirements. The storage location is limited to the " / R&D Data / Motherboard / BOM / " path, and the associated object identifier is related data for a specific motherboard model.

[0100] This embodiment takes the BOM document for server motherboard development as an example to illustrate in detail the specific operations of each unit of the data asset access module. From the access protocol adaptation unit to extract protocol information, to the data type identification unit to form a data type description set, to the feature pre-extraction unit to obtain the feature vector, and then through a series of processing by the association relationship analysis unit to obtain the association optimization vector, finally the access rule generation unit generates the access rules and determines the access scope. Each step is closely centered around a specific data asset instance, clearly presenting the complete process of the data asset access system, ensuring that data assets can be accurately accessed according to rules that match their own characteristics, and providing a reliable foundation for subsequent data processing and management.

[0101] Example 5:

[0102] Take a set of data assets generated in the development process of a server power module as an example, and elaborate the specific implementation of embodiment 5 in detail. The set of data assets includes power design drawings, performance test reports, material procurement lists, and production process flowcharts. The initial information is obtained by the data asset access module, including data source identification as “power research and development department”, “test center”, “procurement department”, “production department”, format type as DWG, PDF, Excel, Visio, storage location as server path “ / power module / research and development data / 202506”, and associated object identification as “power model P-002”.

[0103] The multi-dimensional feature extraction module processes the set of data assets. The content feature extraction unit extracts line, color, layer, and other graphical features from the power design drawings, extracts output voltage, current, efficiency, and other numerical features from the performance test reports, extracts material name, specification, quantity, and other text features from the material procurement list, and extracts process node, arrow direction, and other graphical features from the production process flowchart. The structure feature extraction unit analyzes the layer hierarchy of the design drawings, the chapter structure of the test report, the table row and column of the procurement list, and the node association of the process flowchart, and extracts the respective data organization method, storage hierarchy, and association relationship features. The semantic feature extraction unit combines the business background of power research and development and manufacturing, identifies professional keywords such as “load regulation” and “ripple voltage” from the test report, analyzes their business meaning in power performance evaluation, and understands the context relationship and process requirements of nodes such as “welding process” and “aging test” from the process flowchart.

[0104] The feature fusion unit assigns weight coefficients to the content features, structure features, and semantic features. Considering the importance of numerical features in power performance test reports for classification, a higher weight is assigned to the content features; the structure features of design drawings (such as layer organization) have a greater impact on data management, and a moderate weight is assigned to the structure features; the semantic features of process flowcharts (such as process logic) are closely related to production processes, and a corresponding weight is assigned to the semantic features. Through the feature weighting calculation subunit, the features of various types are multiplied by the corresponding weights and added to obtain a comprehensive feature vector, such as the comprehensive feature vector of the power design drawings, which integrates graphical content features, layer structure features, and design business semantic features.

[0105] The intelligent classification processing module performs multi-dimensional classification of data assets based on comprehensive feature vectors. For power supply design drawings, based on their graphical features, layer structure, and design semantics, they are classified as "R&D Design - Drawing Documents." Performance test reports are classified as "Test and Verification - Performance Reports" based on numerical features, report structure, and performance evaluation semantics. Material purchase lists are classified as "Supply Chain - Procurement Documents" based on text features, table structure, and procurement business semantics. Production process flow charts are classified as "Manufacturing Process - Process Documents" based on graphical features, node structure, and process semantics, resulting in data asset classification results.

[0106] The dynamic storage management module generates data asset storage strategies based on the classification results. For power supply design drawings in the "R&D Design - Drawing Documents" category, given their high version control requirements, the generated storage strategy is: store them in the " / Power Module / R&D Data / Design Drawing / " path, enable version management, automatically generate a new copy for each modification, and log the change. Access permissions are set to read and write for members of the Power Supply R&D department, and read-only for other departments. For test reports in the "Test and Verification - Performance Reports" category, due to their timeliness and relevance, the storage strategy is: store them in the " / Power Module / R&D Data / Test Report / " path, index them by test date and power supply model, link them to the design drawings, and allow them to be edited by members of the test center, but viewable by the R&D and Quality departments. Purchase lists in the "Supply Chain - Procurement Documents" category, because they involve supplier information and procurement processes, are stored in the " / Power Module / Supply Chain Data / Purchase List / " path. Sensitive information, such as supplier contact information, is encrypted, and access is restricted to personnel in the Procurement and Finance departments. The process flow chart in the "Manufacturing Process - Process Document" category is stored in the " / Power Module / Manufacturing Data / Process File / " path to meet the needs of on-site reference. This path is synchronized to the production workshop terminal devices and set to read-only mode to prevent accidental modifications during production.

[0107] The application service output module inputs the storage policy into the data assets to be managed to execute the application service output. Taking the power supply design drawing as an example, the system stores it in the specified path according to the storage policy and automatically enables the version management function. When the R&D personnel modify the drawing, the system generates a new version and records the modification time, modifier and modification content; at the same time, the output results are transmitted to the display terminal for display. The R&D personnel can view the version history, access rights and other information of the drawing on the display terminal. After the test report is stored, the system automatically establishes an associated link with the design drawing. When the test personnel view the test report of a certain model of power supply on the display terminal, they can jump to the corresponding design drawing with one click; after the purchase list is encrypted and stored, the purchasing department personnel can log in to the system to view the complete list, and sensitive information will be displayed as asterisks when other departments access it; after the process flow chart is synchronized to the production workshop terminal, the operator can check the production process flow of the current power supply model at any time on the terminal to ensure that the production operation meets the process requirements.

[0108] This embodiment takes the four types of data assets of the server power module as an example, and fully demonstrates the entire process from feature extraction, classification processing to storage strategy generation and application service output. The multi-dimensional feature extraction module extracts data features from the three dimensions of content, structure, and semantics. The intelligent classification processing module completes accurate classification based on comprehensive features. The dynamic storage management module generates differentiated storage strategies for different categories. The application service output module ensures the execution of strategies and visualizes the results. The modules work together to achieve intelligent management of data assets. For example, through classified storage and permission control, the security and traceability of power supply R&D and manufacturing data are guaranteed. Through associated links, it facilitates R&D personnel to review design and test data. Through version management, the modification process of design drawings is ensured to be traceable, providing a specific implementation method for efficient management of data assets for server R&D and manufacturing companies.

[0109] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0110] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An enterprise data asset intelligent management system for server R&D and manufacturing, characterized by: include: The data asset access module is used to obtain the initial information of the enterprise data assets to be managed generated during the server R&D and manufacturing process. The initial information includes the data source identifier, format type, storage location and associated object identifier; A multi-dimensional feature extraction module, configured to extract content features, structural features, and semantic features of the data assets based on the initial information of the enterprise data assets to be managed; An intelligent classification processing module is used to perform multi-dimensional classification processing on the content features, structural features and semantic features to obtain data asset classification results; A dynamic storage management module, configured to generate a data asset storage strategy based on the data asset classification result; The application service output module is used to input the data asset storage policy into the enterprise data asset to be managed to execute application service output, and transmit the output result to the display terminal for display.

2. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 1 is characterized in that: The data asset access module includes: The access protocol adaptation unit is used to extract the version information, compatible parameters and transmission rules of the data asset access protocol from the interface configuration library; a data type identification unit, configured to extract the function description of each access protocol in the interface configuration library to obtain a set of data type descriptions; a feature pre-extraction unit, configured to perform feature extraction on the initial information of the enterprise data asset to be managed and each data type description in the set of data type descriptions to obtain a set of initial information feature vectors and data type feature vectors; an association relationship parsing unit, configured to perform association relationship parsing on the set of the initial information feature vector and the data type feature vector to obtain a data-type association optimization vector; An access rule generating unit is configured to generate an access rule for the data asset based on the data-type association optimization vector.

3. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 2 is characterized in that: The association relationship analysis unit includes: a data type feature aggregation subunit, configured to perform feature aggregation processing based on hierarchical mapping on the set of data type feature vectors to obtain a guiding template; The cross-domain collaborative parsing subunit is used to perform cross-domain collaborative parsing on the set of the initial information feature vector and the data type feature vector based on the guiding template to obtain the data-type association optimization vector.

4. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 3 is characterized in that: The data type feature aggregation subunit includes: A data type feature hierarchical mapping secondary subunit is used to perform hierarchical mapping processing on each data type feature vector in the set of data type feature vectors using a mapping matrix to obtain a set of hierarchically transformed data type feature vectors; A data type feature matrix arrangement secondary subunit is used to perform matrix arrangement on the set of data type feature vectors after the hierarchical transformation to obtain a data type feature matrix; The data type feature matrix extreme value extraction secondary sub-unit is used to extract the extreme values ​​of the data type feature vectors after each level transformation in the data type feature matrix to obtain the data type feature matrix key vector as the guiding template.

5. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 4 is characterized in that: The data type feature hierarchical mapping secondary sub-unit includes: The data type feature vector is point-multiplied by the mapping matrix and then added to the mapping bias vector in a positional manner to obtain the data type feature vector after hierarchical transformation.

6. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 5 is characterized in that: The cross-domain collaborative parsing subunit includes: an initial information hierarchical mapping secondary subunit, configured to perform hierarchical mapping processing on the initial information feature vector using a query matrix and a value matrix to obtain an initial information query vector and an initial information value vector; a template-guided heterogeneous conversion parsing secondary subunit, configured to input the initial information query vector, the initial information value vector, the transformed data type feature vectors at each level in the data type feature matrix, and the guiding template into a template-guided heterogeneous conversion structure to obtain a sequence of data-type cross-domain collaborative parsing vectors; The position mean calculation secondary subunit is used to calculate the position mean vector of the sequence of the data-type cross-domain collaborative resolution vectors to obtain the data-type association optimization vector.

7. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 6 is characterized in that: The template-guided heterogeneous conversion parsing secondary subunit includes: After calculating the product between the initial information query vector and the transposed vector of the data type feature vector after the hierarchical transformation, the obtained data-type association feature matrix is ​​divided by the modulus length of the key vector of the data type feature matrix by position to obtain a data-type association weight matrix; Inputting the data-type association weight matrix into a normalization function for processing to obtain a data-type association normalized weight matrix; After multiplying the data-type association normalized weight matrix with the key vector of the data type feature matrix, the obtained feature vector is multiplied by position with the initial information value vector to obtain the data-type cross-domain collaborative resolution vector.

8. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 7 is characterized in that: The access rule generating unit includes: Inputting the data-type association optimization vector into an access rule recommendation module based on a decision maker to obtain an access rule for the data asset; Based on the access rule, an access scope of the data asset is determined.

9. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 1 is characterized in that: The multidimensional feature extraction module includes: A content feature extraction unit, configured to extract features of text content, numerical content, and multimedia content from the initial information of the enterprise data assets to be managed; a structural feature extraction unit, configured to extract features of data organization, storage hierarchy, and association relationships from the initial information; A semantic feature extraction unit for analyzing the features of keywords, contextual relationships, and business meanings in the initial information; The feature fusion unit is used to fuse the content features, structural features and semantic features to obtain a comprehensive feature vector.

10. The enterprise data asset intelligent management system for server R&D and manufacturing according to claim 9 is characterized in that: The feature fusion unit, include: A content feature weight allocation subunit, configured to allocate weight coefficients to the content features, structural features, and semantic features respectively; The feature weighting calculation subunit is used to multiply the content feature, structural feature, and semantic feature by the corresponding weight coefficients respectively and then add them together to obtain the comprehensive feature vector.

Images