An engineering design data intelligent management system and method based on federated learning

Abstract

The application provides an engineering design data intelligent management system and method based on federated learning, comprising: a data module, which is used for acquiring multi-source engineering design data, performing data connection on the multi-source engineering design data, and obtaining an engineering design data connection network; a model module, which is used for training a local model according to the engineering design data connection network, and establishing a global model according to a training result; and a scheme module, which is used for acquiring engineering design requirement data, and generating an engineering scheme according to the global model and the engineering design requirement data. The engineering design data connection network is established and the model is trained through the federated learning technology, so that the outflow of original data can be effectively avoided, and data security is ensured.

Description

Technical Field

[0001] This invention relates to the field of data information security technology, and in particular to an intelligent management system and method for engineering design data based on federated learning. Background Technology

[0002] Engineering design, as a core component of the engineering construction field, serves as a bridge connecting creativity and reality, planning and implementation. It not only concerns the quality and safety of infrastructure such as buildings, transportation, and energy, but also directly impacts the sustainability of urban development. Driven by new urbanization and the digital economy, China's engineering design industry is undergoing a profound transformation from traditional models to intelligent, green, and globalized approaches.

[0003] The current integrated management of engineering design data mainly relies on centralized platforms. All parties involved aggregate their data to a central server, which performs unified data cleaning, integration, and storage. Model training and scheme generation are then carried out based on the centrally stored data. Because engineering design data needs to be shared, data leakage is difficult to avoid, posing a potential risk of harming user privacy. Summary of the Invention

[0004] This invention provides an intelligent management system and method for engineering design data based on federated learning, which solves the problem of insufficient security of engineering design data when stored centrally in the prior art.

[0005] On the one hand, this invention provides an intelligent management system for engineering design data based on federated learning, which includes: The data module is used to acquire multi-source engineering design data, connect the multi-source engineering design data to obtain an engineering design data connection network; the model module is used to train the local model based on the engineering design data connection network, and establish a global model based on the training results; the scheme module is used to acquire engineering design requirement data, and generate engineering schemes based on the global model and the engineering design requirement data.

[0006] Furthermore, the data module performs data connection on multi-source engineering design data to obtain an engineering design data connection network, including: dividing the multi-source engineering design data into structural data to obtain multiple engineering structure data; extracting field features from the engineering structure data and determining the semantic information of the engineering structure data based on the field features; establishing engineering structure data nodes, connecting each engineering structure data node based on the semantic information, and establishing an engineering design data connection network based on the engineering structure data nodes with connection relationships.

[0007] Furthermore, the semantic information of the engineering structure data is determined based on field characteristics, including: determining the field names of the engineering structure data and the field names of adjacent fields within a preset neighborhood based on field characteristics; determining the logical relationship and co-occurrence characteristics between the current field and adjacent fields based on the field names of adjacent fields within the preset neighborhood; determining the adjacent field information of the engineering structure data based on the logical relationship and co-occurrence characteristics; obtaining the hierarchical relationship of the engineering structure data; and combining the field names, adjacent field information, and hierarchical relationship of the engineering structure data to obtain semantic information.

[0008] Furthermore, the engineering structure data nodes are connected based on semantic information, including: clustering the engineering structure data nodes based on semantic information, determining the cluster of each engineering structure data node based on the clustering results; and connecting the engineering structure data nodes in the same cluster to complete the connection of the engineering structure data nodes.

[0009] Furthermore, clustering of engineering structure data nodes based on semantic information includes: establishing a sample dataset of engineering structure data nodes based on their semantic information; randomly selecting k initial cluster centers from the sample dataset; calculating the semantic distance between the engineering structure data nodes in the sample dataset and the initial cluster centers; dividing the engineering structure data nodes into corresponding partitions based on the semantic distance; calculating the mean semantic distance within each partition; recalculating the cluster centers based on the mean semantic distance; and repeating the above steps iteratively until the cluster centers no longer change or the maximum number of iterations is reached, thus completing the clustering of engineering structure data nodes.

[0010] Further, the semantic distance between the engineering structure data nodes in the sample dataset and the initial cluster center is calculated, including: calculating the cosine similarity of the corresponding field names and adjacent field information of the engineering structure data nodes and the initial cluster center based on the semantic information; determining the first semantic distance and the second semantic distance based on the cosine similarity of the field names and adjacent field information; establishing hierarchical path sequences between the engineering structure data nodes and the initial cluster center based on the semantic information; calculating the lowest common ancestor depth between the engineering structure data nodes and the initial cluster center based on the hierarchical path sequences to obtain the third semantic distance; and performing a weighted summation of the first semantic distance, the second semantic distance, and the third semantic distance to obtain the semantic distance between the engineering structure data nodes and the initial cluster center.

[0011] Furthermore, the model module is used to train the local model based on the engineering design data connection network and to establish a global model based on the training results. This includes: establishing a training set based on the engineering structure data corresponding to the engineering design data connection network; training the local model based on the training set; determining the training parameter weights based on the engineering design data connection network; adjusting the training parameters of the local model based on the training parameter weights; and aggregating the adjusted training parameters to obtain the global model.

[0012] Furthermore, the training parameter weights are determined based on the engineering design data connection network, including: calculating the degree of each engineering design data node in the engineering design data connection network; filtering out engineering design data nodes whose degree exceeds a first preset threshold based on the degree of each engineering design data node; setting the engineering design data nodes whose degree exceeds the first preset threshold as center nodes; obtaining the connected components of the center nodes; determining the node score of the corresponding center node based on the connected components; determining the comprehensive node score based on the node scores of all center nodes; and determining the training parameter weights based on the comprehensive node score.

[0013] Further, the node score of the corresponding central node is determined based on the connected components, including: obtaining the degree of each engineering design data node in the connected components, and determining the first node score based on the degree of each engineering design data node; obtaining the cosine similarity between each engineering design data node in the connected components and the corresponding central node, and determining the second node score based on the cosine similarity between each engineering design data node and the corresponding central node; and performing a weighted summation of the first node score and the second node score to obtain the node score of the central node.

[0014] On the other hand, the present invention also provides an intelligent management method for engineering design data based on federated learning, comprising: Acquire multi-source engineering design data, connect the multi-source engineering design data to obtain an engineering design data connection network; train the local model based on the engineering design data connection network, and establish a global model based on the training results; acquire engineering design requirement data, and generate an engineering scheme based on the global model and the engineering design requirement data.

[0015] This invention provides an intelligent management system and method for engineering design data based on federated learning. The system comprises a data module for acquiring multi-source engineering design data and connecting this data to form an engineering design data connection network; a model module for training a local model based on the engineering design data connection network and establishing a global model based on the training results; and a scheme module for acquiring engineering design requirement data and generating engineering schemes based on the global model and the engineering design requirement data. By establishing the engineering design data connection network and training the model using federated learning technology, the outflow of raw data can be effectively prevented, ensuring data security. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the structure of an intelligent management system for engineering design data based on federated learning, provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating an intelligent management method for engineering design data based on federated learning, provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0019] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0020] Figure 1 This is a schematic diagram of the structure of the intelligent management system for engineering design data based on federated learning provided in an embodiment of the present invention, including: The data module is used to acquire multi-source engineering design data, connect the multi-source engineering design data to obtain an engineering design data connection network; the model module is used to train the local model based on the engineering design data connection network, and establish a global model based on the training results; the scheme module is used to acquire engineering design requirement data, and generate engineering schemes based on the global model and the engineering design requirement data.

[0021] In this embodiment, the multi-source engineering design data includes design drawings, structural calculation sheets, material specification lists, construction plans, engineering specifications, and other data of the engineering project. By establishing an engineering design data connection network, the multi-source engineering design data is connected to achieve semantic integration of the data. The local model is trained through the engineering design data connection network to establish a global model. The engineering design requirement data is input into the global model, including the basic parameters, functional requirements, performance indicators, constraints, budget range, and other data of the engineering project. The global model performs inference and calculation based on the input design requirement data and outputs an engineering solution that matches the design requirements.

[0022] In some embodiments of this application, the data module performs data connection on multi-source engineering design data to obtain an engineering design data connection network, including: dividing the multi-source engineering design data into structural parts to obtain multiple engineering structure data; extracting field features from the engineering structure data and determining the semantic information of the engineering structure data based on the field features; establishing engineering structure data nodes; connecting each engineering structure data node based on the semantic information; and establishing an engineering design data connection network based on the engineering structure data nodes with connection relationships.

[0023] In this embodiment, structural analysis is performed on the engineering structural data based on its content organization method to identify the hierarchical structure and constituent units of the data, dividing each piece of engineering design data into several engineering structural data units. For example, a building structural design drawing is divided into several engineering structural data units such as floor slab structural data, beam and column structural data, and foundation structural data. Fields are extracted from the engineering structural data using OCR recognition, image recognition, and other methods to obtain their field features, thereby identifying the semantic information of the engineering structural data. The semantic information is then used to connect the engineering structural data nodes, establishing a network of multiple engineering design data connections.

[0024] In some embodiments of this application, determining the semantic information of engineering structure data based on field features includes: determining the field name of the engineering structure data and the field names of adjacent fields within a preset neighborhood based on the field features; determining the logical relationship and co-occurrence features between the current field and adjacent fields based on the field names of adjacent fields within the preset neighborhood; determining the adjacent field information of the engineering structure data based on the logical relationship and co-occurrence features; obtaining the hierarchical relationship of the engineering structure data; and combining the field names, adjacent field information, and hierarchical relationship of the engineering structure data to obtain semantic information.

[0025] In this embodiment, the logical relationship and co-occurrence characteristics between the current field and adjacent fields are analyzed. Semantic keywords between the co-occurring fields of the current field and adjacent fields are extracted. The information of adjacent fields is determined by the semantic keywords of the co-occurring fields. The hierarchical relationship is, for example, the hierarchical path of project → sub-project → sub-item project → component. The semantic information is obtained by combining the field names, adjacent field information and hierarchical relationship of the engineering structure data. The semantic information is represented by vectorization.

[0026] In some embodiments of this application, connecting the engineering structure data nodes according to semantic information includes: clustering the engineering structure data nodes according to semantic information, determining the cluster of each engineering structure data node according to the clustering results, and connecting the engineering structure data nodes in the same cluster to complete the connection of the engineering structure data nodes.

[0027] In this embodiment, each engineering structure data is abstracted as a node, and the attributes of the node include the semantic information of the engineering structure data. The engineering structure data nodes are clustered based on the semantic information to connect the engineering structure data nodes and form an engineering design data network. This engineering design data connection network is essentially a graph structure containing nodes and edges; nodes represent engineering structure data, and edges represent the semantic relationships between data. In some embodiments of this application, clustering engineering structure data nodes based on semantic information includes: establishing a sample dataset of engineering structure data nodes based on the semantic information of the engineering structure data nodes; randomly selecting k initial cluster centers from the sample dataset; calculating the semantic distance between the engineering structure data nodes in the sample dataset and the initial cluster centers; dividing the engineering structure data nodes into corresponding partitions based on the semantic distance; calculating the mean semantic distance within each partition; recalculating the cluster centers based on the mean semantic distance; and repeating the above steps iteratively until the cluster centers no longer change or the number of iterations reaches the maximum number of iterations, thereby completing the clustering of engineering structure data nodes.

[0028] In this embodiment, the engineering structure data nodes are clustered based on semantic information using the k-means clustering algorithm. The value of k can be set according to the size of the engineering structure data nodes and the expected accuracy.

[0029] In some embodiments of this application, calculating the semantic distance between the engineering structure data nodes and the initial cluster center in the sample dataset includes: calculating the cosine similarity of the corresponding field names and adjacent field information of the engineering structure data nodes and the initial cluster center based on the semantic information; determining the first semantic distance and the second semantic distance based on the cosine similarity of the field names and adjacent field information; establishing hierarchical path sequences between the engineering structure data nodes and the initial cluster center based on the semantic information; calculating the lowest common ancestor depth between the engineering structure data nodes and the initial cluster center based on the hierarchical path sequences to obtain the third semantic distance; and performing a weighted summation of the first semantic distance, the second semantic distance, and the third semantic distance to obtain the semantic distance between the engineering structure data nodes and the initial cluster center.

[0030] In this embodiment, the lowest common ancestor depth between the engineering structure data nodes and the initial cluster centers is calculated based on the hierarchical path sequence to obtain the third semantic distance. Specifically, this is obtained using the third semantic distance calculation formula, which is as follows: , in, The depth of the lowest common ancestor. The hierarchical path depth of the engineering structure data nodes. The hierarchical path depth for the initial cluster centers.

[0031] In some embodiments of this application, the model module is used to train a local model based on an engineering design data connection network and to establish a global model based on the training results, including: establishing a training set based on the engineering structure data corresponding to the engineering design data connection network, and training the local model based on the training set; determining the training parameter weights based on the engineering design data connection network, and adjusting the training parameters of the local model based on the training parameter weights; and aggregating the adjusted training parameters to obtain a global model.

[0032] In this embodiment, a local training set is established based on the engineering structure data corresponding to each node in the engineering design data connection network. This local training set is then used to train a pre-defined local model. The local model can be a deep learning model suitable for engineering design, such as a Graph Neural Network (GNN), or other machine learning models suitable for structured data modeling. During training, the training parameter weights of each training parameter are determined based on the engineering design data connection network, and the training parameters of the local model are adjusted accordingly. After each participant completes local model training and parameter adjustment, the adjusted local training parameters are aggregated to generate a global model. The aggregation method can be a federated averaging algorithm.

[0033] In some embodiments of this application, determining the training parameter weights based on the engineering design data connection network includes: calculating the degree of each engineering design data node in the engineering design data connection network; filtering out engineering design data nodes whose degree exceeds a first preset threshold based on the degree of each engineering design data node; setting the engineering design data nodes whose degree exceeds the first preset threshold as center nodes; obtaining the connected components of the center nodes; determining the node score of the corresponding center node based on the connected components; determining the comprehensive node score based on the node scores of all center nodes; and determining the training parameter weights based on the comprehensive node score.

[0034] In some embodiments of this application, determining the node score of the corresponding central node based on the connected components includes: obtaining the degree of each engineering design data node in the connected components, and determining a first node score based on the degree of each engineering design data node; obtaining the cosine similarity between each engineering design data node in the connected components and the corresponding central node, and determining a second node score based on the cosine similarity between each engineering design data node and the corresponding central node; and performing a weighted summation of the first node score and the second node score to obtain the node score of the central node.

[0035] In this embodiment, a first node score is determined based on the degree of each engineering design data node in the connected component. A higher degree indicates a higher density of connections and a greater information carrying capacity of the node in the network, thus contributing positively to the first node score. Simultaneously, the cosine similarity between each engineering design data node in the connected component and its corresponding central node is obtained. A second node score is determined based on the cosine similarity between each node and the central node. A higher similarity indicates stronger semantic consistency between the node and the central node, thus contributing positively to the second node score. The first and second node scores are weighted and summed to obtain the node score of the central node. A comprehensive node score is determined based on the node scores of all central nodes, which in turn determines the weights of the training parameters, giving higher weights to the training parameters corresponding to central nodes with higher node scores during global aggregation.

[0036] Based on the same general inventive concept, this invention also protects a method for intelligent management of engineering design data based on federated learning, such as... Figure 2 As shown, it includes: S101, acquire multi-source engineering design data, perform data connection on the multi-source engineering design data, and obtain the engineering design data connection network; S102, Train the local model by connecting the network based on the engineering design data, and establish a global model based on the training results; S103: Obtain engineering design requirement data and generate an engineering plan based on the global model and engineering design requirement data.

[0037] It should be noted that all relevant information that may be involved in the various embodiments of the present invention is processed in strict accordance with the requirements of laws and regulations, following the principles of legality, legitimacy, and necessity, based on the reasonable purpose of the business scenario, and is information that users actively provide or generate during the use of the product / service, as well as information obtained with user authorization.

[0038] The information processed by this invention may vary depending on the specific product / service scenario and should be based on the specific scenario in which the user uses the product / service. This may involve user account information, device information, or other related information. This invention will treat the relevant information and its processing with the utmost diligence.

[0039] This invention places great emphasis on the security of relevant information and has adopted reasonable and feasible security protection measures that comply with industry standards to protect user information and prevent unauthorized access, public disclosure, use, modification, damage or loss of relevant information.

[0040] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0041] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent management system for engineering design data based on federated learning, characterized in that, include: The data module is used to acquire multi-source engineering design data, connect the multi-source engineering design data, and obtain an engineering design data connection network. The model module is used to train the local model by connecting to the network based on engineering design data, and to build a global model based on the training results. The solution module is used to acquire engineering design requirement data and generate engineering solutions based on the global model and engineering design requirement data.

2. The intelligent management system for engineering design data based on federated learning according to claim 1, characterized in that, The data module performs data connectivity on multi-source engineering design data to obtain an engineering design data connectivity network, including: The multi-source engineering design data is structurally divided to obtain multiple engineering structural data. Extract field features from engineering structure data, and determine the semantic information of the engineering structure data based on the field features; Establish engineering structure data nodes, connect each engineering structure data node according to semantic information, and establish an engineering design data connection network based on the engineering structure data nodes with existing connections.

3. The intelligent management system for engineering design data based on federated learning according to claim 2, characterized in that, The semantic information of engineering structure data is determined based on field characteristics, including: The field names of the engineering structure data and the field names of adjacent fields in the preset neighborhood are determined based on the field characteristics. The logical relationship and co-occurrence characteristics between the current field and adjacent fields are determined based on the field names of adjacent fields in the preset neighborhood. Based on logical relationships and co-occurrence characteristics, the adjacent field information of the engineering structure data is determined, the hierarchical relationship of the engineering structure data is obtained, and the field names, adjacent field information and hierarchical relationship of the engineering structure data are combined to obtain semantic information.

4. The intelligent management system for engineering design data based on federated learning according to claim 2, characterized in that, Connect the data nodes of each engineering structure based on semantic information, including: Cluster the engineering structure data nodes based on semantic information, and determine the cluster of each engineering structure data node based on the clustering results; Connect the engineering structure data nodes that are in the same cluster to complete the connection of engineering structure data nodes.

5. The intelligent management system for engineering design data based on federated learning according to claim 4, characterized in that, Clustering of engineering structure data nodes based on semantic information includes: A sample dataset of engineering structure data nodes is established based on the semantic information of the engineering structure data nodes, and k initial cluster centers are randomly selected from the sample dataset. Calculate the semantic distance between the semantic information of the engineering structure data nodes in the sample dataset and the initial cluster center, and divide the engineering structure data nodes into the corresponding partitions based on the semantic distance of the semantic information. Calculate the mean semantic distance within each partition, and recalculate the cluster centers based on the mean semantic distance; Repeat the above steps until the cluster centers no longer change or the maximum number of iterations is reached, thus completing the clustering of the engineering structure data nodes.

6. The intelligent management system for engineering design data based on federated learning according to claim 5, characterized in that, Calculate the semantic distance between the engineering structure data nodes in the sample dataset and the initial cluster centers, including: The cosine similarity between the field names and adjacent field information of the engineering structure data nodes and the initial cluster center is calculated based on semantic information. The first semantic distance and the second semantic distance are determined based on the cosine similarity between the field names and adjacent field information. Based on semantic information, hierarchical path sequences are established between engineering structure data nodes and initial cluster centers. The lowest common ancestor depth between engineering structure data nodes and initial cluster centers is calculated based on the hierarchical path sequences to obtain the third semantic distance. The first semantic distance, the second semantic distance, and the third semantic distance are weighted and summed to obtain the semantic distance between the semantic information of the engineering structure data nodes and the initial cluster centers.

7. The intelligent management system for engineering design data based on federated learning according to claim 1, characterized in that, The model module is used to train a local model by connecting to a network based on engineering design data, and to build a global model based on the training results, including: A training set is established by connecting the engineering design data to the corresponding engineering structure data of the network, and the local model is trained based on the training set. The training parameter weights are determined by connecting the network based on the engineering design data, and the training parameters of the local model are adjusted based on the training parameter weights. The adjusted training parameters are aggregated to obtain the global model.

8. The intelligent management system for engineering design data based on federated learning according to claim 7, characterized in that, The training parameter weights are determined by connecting the network based on engineering design data, including: Calculate the degree of each engineering design data node in the engineering design data connection network, and filter out engineering design data nodes whose degree exceeds the first preset threshold based on the degree of each engineering design data node; The engineering design data node whose degree exceeds the first preset threshold is set as the center node, the connected components of the center node are obtained, and the node score of the corresponding center node is determined based on the connected components. The comprehensive node score is determined based on the node scores of all central nodes, and the weights of the training parameters are determined based on the comprehensive node score.

9. The intelligent management system for engineering design data based on federated learning according to claim 8, characterized in that, The node score of the corresponding central node is determined based on the connected components, including: Obtain the degree of each engineering design data node in the connected component, and determine the score of the first node based on the degree of each engineering design data node; Obtain the cosine similarity between each engineering design data node and its corresponding central node in the connected component, and determine the score of the second node based on the cosine similarity between each engineering design data node and its corresponding central node. The node score of the center node is obtained by weighted summation of the scores of the first node and the second node.

10. A method for intelligent management of engineering design data based on federated learning, characterized in that, include: Acquire multi-source engineering design data, connect the multi-source engineering design data, and obtain an engineering design data connection network; The local model is trained by connecting to the network based on engineering design data, and a global model is built based on the training results. Obtain engineering design requirements data, and generate engineering solutions based on the global model and engineering design requirements data.

Images