A method, device, and medium for storing component data of a digital twin model.

Abstract

This application discloses a method, device, and medium for storing component data of a digital twin model, relating to the field of data storage. The method includes: acquiring component data of the digital twin model; establishing a graph structure dataset based on the component data; determining the feature representation and category of each component using a pre-trained deep learning model based on the graph structure dataset; the deep learning model includes a graph convolutional neural network, a fully connected layer, and a classifier connected sequentially, where the output of the graph convolutional neural network is the feature representation, and the output of the classifier is the category; generating a personalized hash code corresponding to each component based on its feature vector, feature representation, and category, and storing it in a database. This application improves the query efficiency of digital twin model components and reduces the physical space required to store component data, thus reducing resource consumption.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data storage, and in particular to a digital twin model component data storage method, device and medium. BACKGROUND

[0002] In the intelligent pump station management, Building Information Modeling (BIM) is an important part of the data platform, which is crucial for efficient data query and pump station information visualization. The rapid retrieval and optimization of BIM not only concerns the system performance, but also directly affects the decision efficiency, operation and maintenance effect and user experience.

[0003] BIM contains highly complex geometric shapes, engineering properties, spatial relationships and other multi-dimensional information. The correlation and hierarchy between these data require efficient indexing and search strategies. Traditional BIM component data is stored in a structured way in the database. This structured storage method provides high management and query capabilities for objects and their attributes, and is suitable for rule-based applications and traditional data analysis tasks. However, as BIM component data becomes more complex and contains a large amount of data, this traditional storage method also exposes many technical challenges and limitations, and its technical shortcomings are:

[0004] 1. Indexing difficulty: The query efficiency of traditional databases on a large amount of data is limited by its indexing strategy and algorithm. When dealing with large BIMs (containing thousands or even tens of thousands of components), searching for specific component information directly in the original data can become time-consuming and inefficient, especially when precise matching or similarity retrieval is required.

[0005] 2. Data redundancy and conflict: Current storage methods can lead to a large number of duplicate data records, especially when dealing with similar but not identical BIM components. This not only increases the demand for storage space, but also increases the possibility of error rate, and can make subsequent queries and management more complex.

[0006] 3. Low query efficiency: Under the existing data organization structure, complex searches in large databases often rely on full table scans or index construction, which can be greatly limited by performance bottlenecks.

[0007] 4. Resource consumption: As the complexity and size of BIMs grow, the demand for storing and processing these data also increases. This not only involves hardware costs (such as servers, storage devices, etc.), but also involves operation and management costs and power consumption issues.

[0008] 5. Limited flexibility: Traditional databases often have strict restrictions on data types and structures, which may not be suitable for the flexible and changing BIM requirements, such as adding new component types or modifying existing component properties. Poor adaptability may lead to the need for frequent data migration or reconstruction operations, increasing the cost of time and resources. SUMMARY

[0009] The purpose of the present application is to provide a digital twin model component data storage method, device and medium, which can improve the query efficiency of digital twin model components and reduce the physical space required for storing component data, and reduce resource consumption.

[0010] To achieve the above purpose, the present application provides the following solutions:

[0011] In a first aspect, the present application provides a digital twin model component data storage method, comprising:

[0012] Obtaining component data of a digital twin model; the component data includes feature vectors of each component and relationships between components;

[0013] According to the component data, a graph structure data set is established; the graph structure data set includes node features and graph structure features, the node features include each component and the feature vector of each component, and the graph structure features include the relationship between the components;

[0014] According to the graph structure data set, a pre-trained deep learning model is used to determine the feature representation and category of each component; wherein the deep learning model includes a graph convolutional neural network, a fully connected layer and a classifier connected in turn, the output of the graph convolutional neural network is the feature representation, and the output of the classifier is the category;

[0015] According to the feature vector of each component, the feature representation of each component and the category of each component, a personalized hash code corresponding to each component is generated and stored in the database.

[0016] In a second aspect, the present application provides a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned digital twin model component data storage method.

[0017] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the above-mentioned digital twin model component data storage method.

[0018] According to the specific embodiments provided by the present application, the present application has the following technical effects:

[0019] The application provides a storage method and device of digital twin model component data and a medium. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 An application environment diagram of a storage method of digital twin model component data in an embodiment of the present application;

[0022] Figure 2 A flowchart of a storage method of digital twin model component data provided in an embodiment of the present application;

[0023] Figure 3 An overall framework diagram of a storage method of digital twin model component data provided in an embodiment of the present application;

[0024] Figure 4 A schematic diagram of a deep learning model in an embodiment of the present application;

[0025] Figure 5 A schematic diagram of a graph structure data set in an embodiment of the present application. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0027] First, some technical terms involved in the embodiments of the present application are introduced.

[0028] BIM component: BIM refers to the whole process management of design, construction and operation using building information model. BIM component data refers to a set of data used to describe a specific construction element in the BIM system, including but not limited to geometric characteristics (such as shape, size), physical properties (such as material type, weight), environmental impact factors (such as energy consumption estimation) and related connection relationships and other associated information.

[0029] Graph Convolutional Networks (GCN): A deep learning architecture for processing graph data, which can perform convolution operations directly on nodes, edges and entire graph structures to capture complex local and global relationships in data, and can well understand and generate feature representations with high value for complex structures.

[0030] Loss function: A measure used in machine learning and deep learning to evaluate the performance of a model. It measures the difference between the model's output and the actual result, and is used to guide the adjustment of weights in the training process to optimize prediction accuracy.

[0031] Hash algorithm: A process that converts input data of arbitrary size into fixed-size output (usually called "hash" or "hash code"). It is widely used in information retrieval, database indexing, secure storage, etc., especially in handling large amounts of data to provide fast lookup and matching capabilities.

[0032] The present application aims to solve the problem of low retrieval efficiency of BIM components in current intelligent pump stations, and slow loading speed when components change or associated real-time data changes. A storage method for digital twin model component data is provided, which combines deep learning, feature extraction and hash coding process to generate personalized hash codes. The personalized hash codes of each component are stored in the database in advance. When searching on the current terminal, the corresponding component can be efficiently found from the massive data. When new BIM components are stored, the trained deep learning model can be used for rapid storage, facilitating subsequent retrieval, thereby realizing effective filtering and rapid positioning of large-scale BIM component data, and significantly reducing query time.

[0033] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0034] The storage method for digital twin model component data provided by the embodiments of the present application can be applied to, for example Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server. Terminal 102 can send component data of the digital twin model to server 104. Server 104 generates a personalized hash code for each component based on the received component data and stores it in a database. Furthermore, in some embodiments, the method for storing digital twin model component data can also be implemented separately by server 104 or terminal 102.

[0035] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0036] In one exemplary embodiment, such as Figure 2 and Figure 3 As shown, a method for storing digital twin model component data is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 201 to 204.

[0037] Step 201: Obtain component data for the digital twin model. The component data includes the feature vector of each component and the relationships between components. The digital twin model is a building information model of a digital twin intelligent pumping station.

[0038] In a specific application example, step 201 includes steps 11 to 13.

[0039] Step 11: Obtain the attribute information of each component in the digital twin model.

[0040] Specifically, attribute information reflecting the similarity or difference between components is extracted from multi-dimensional BIM. Based on this attribute information, the components are then classified into simple categories. For each component in the BIM, its geometric information has already been associated with relevant attribute features during the modeling process. This attribute information is exported, including geometric attributes (such as volume and shape), physical attributes (such as material type and weight), and environmental influencing factors (lighting, temperature), and is organized into k categories, such as [door, table, chair, generator, wall, gate, transformer, ...]. 1*k .

[0041] Step 12: Determine the relationships between components based on the attribute information of each component.

[0042] Step 13: Normalize the attribute information of each component to obtain the feature vector of each component.

[0043] By normalizing the attribute information in each category, the feature vectors of each component are made to have the same dimension. Specifically, smaller feature vectors are padded with zero vectors to make each feature vector have a matrix data format of the same size. Assuming a door consists of four points and a table consists of six points, the door is padded with two zero vectors to have the same number of points as the table.

[0044] The final component data is Φ = {Φ1, Φ2, Φ3, ..., Φ} n}; where Φ is the component data, Φ n Let n be the feature vector of the nth component and its relationship with other components, where n is the number of components.

[0045] Step 202: Based on the component data, establish a graph structure dataset. The graph structure dataset includes node features and graph structure features. The node features include each component and its feature vector, and the graph structure features include the relationships between components.

[0046] Step 203: Based on the graph structure dataset, use a pre-trained deep learning model to determine the feature representation and category of each component. For example, Figure 4 As shown, the deep learning model includes a graph convolutional neural network, a fully connected layer, and a classifier connected in sequence. The output of the graph convolutional neural network is a feature representation, and the output of the classifier is a category.

[0047] This application constructs a deep learning model capable of processing graph-structured data to extract and understand complex component relationships from BIM. The deep learning model is pre-trained using a training sample set. Each training sample in the set includes a graph-structured dataset sample and the true category of each component within it. Each category of the graph-structured dataset sample contains at least 10 components. For example, if the graph-structured dataset sample includes Table 1, Table 2, Chair 1, Chair 2, Transformer 1, Transformer 2, and Transformer 3, then Table 1 and Table 2 belong to the same category. The loss function used during deep learning model training can be a hash loss function. The loss function value is calculated by measuring the error between the classifier's output category and the true category. If the loss function value does not reach the target range, training and optimization continue until the loss function value is minimized and performs well on the test set.

[0048] like Figure 5 As shown, for a graph-structured dataset G, there are N nodes ( Figure 4 In a graph convolutional neural network (Graph Convolutional Neural Network), each node (R, S, T, U, V) has its own features. A node represents a component, and its features represent the feature vector of that component. Let these nodes form an N*D matrix, i.e., the node features X, where D represents the feature dimension of each node, i.e., the dimension of the feature vector. Furthermore, the relationships between nodes can also be extracted as an N*N relation matrix, i.e., the graph structure features A, also known as the adjacency matrix. The node features X and graph structure features A are the inputs to the graph convolutional neural network. The graph convolutional neural network outputs the probability that each component belongs to each category (R1 has a probability of 1.1 for the first category, R2 has a probability of 3.2 for the second category, R3 has a probability of 4.2 for the third category, and so on).

[0049] The propagation mechanism between layers in an L-layer graph convolutional neural network based on hierarchical propagation rules is as follows:

[0050]

[0051] Among them, H l+1 This is the activation matrix of the (l+1)th layer. Let W be the activation matrix of the l-th layer. (l) Let be the weight matrix of the l-th layer, and δ(·) denote the nonlinear activation function. I N It is an N-dimensional identity matrix. Since the diagonals of the graph structure feature A are all 0, it is related to H. (l) When multiplying, the characteristics of the node itself are ignored, and an identity matrix I is added to A. N This will make the diagonal elements become 1. for The standardized matrix, for The data in the p-th row and q-th column, where Q is... Total number of columns, for The standardized data in the p-th row and q-th column, due to It is an unnormalized matrix, if compared with H (l) Direct multiplication alters the original feature distribution because different nodes have different numbers and weights of edges. Some nodes are connected by multiple edges, causing the aggregated feature values ​​of nodes with more edges to be much larger than those of nodes with fewer edges. Therefore, it is necessary to... Standardization processing is carried out to make Each row adds up to 1. It is a symmetric and normalized matrix.

[0052] In this application, the graph convolutional neural network has 4 propagation layers. After the input data is calculated by the above propagation layer formula, it enters the fully connected layer. After fully connected processing, it enters the softmax classifier and outputs k results, including the probability of each component belonging to each category.

[0053] Step 204: Generate a personalized hash code for each component based on its feature vector, feature representation, and category, and store it in the database.

[0054] This application generates a personalized hash code for each component. This maintains category differentiation while being space-efficient, reducing hash collisions and improving component retrieval efficiency.

[0055] In a specific application example, step 204 includes steps 41 and 42.

[0056] Step 41: For the a-th generation process of any component, based on the generation number and the component's feature information, a primary hash function and an auxiliary hash function are used to obtain the preliminary hash code of the component generated in the a-th generation. The feature information includes feature vector, feature representation, and category; a ≥ 0.

[0057] Specifically, based on the component's characteristic information, a primary hash function is used to determine the component's primary hash code. Based on the generation count and the component's characteristic information, an auxiliary hash function is used to determine the component's auxiliary hash code. Based on the primary hash code and the auxiliary hash code, the initial hash code of the component generated in the a-th generation is determined.

[0058] NewHashValue=H1(data)+a×H2(data);

[0059] Wherein, NewHashValue is the initial hash code of the component generated in the a-th generation, data is the feature information of the component, H1() is the main hash function, H2() is the auxiliary hash function, and a is the generation number.

[0060] The process of determining the main hash code of a component using a main hash function based on the component's feature information includes: performing a hash calculation on the component's feature information to obtain the component's first hash code; performing a linear transformation on the component's first hash code to obtain the component's second hash code; and using the HashingTrick algorithm to map the component's second hash code to a fixed-length binary hash code to obtain the component's main hash code.

[0061] The main hash code of the i-th component is determined using the following formula:

[0062] h i '=(g(h i ))mod2 m ;

[0063] Among them, h i ' is the main hash code of the i-th component, h i Let m be the first hash code of the i-th component, m be the length of the binary hash code, mod represents the modulo operation, g() represents the linear transformation, and the softplus activation function is used here to provide a smooth approximation of positive real numbers and map real values ​​to non-negative values. These non-negative values ​​are then mapped to fixed-length binary hash codes through the HashingTrick algorithm.

[0064] The basic idea for determining the master hash code is to perform a modulo operation on each real number using one or more hash functions, and then extract the corresponding bits as part of the hash code based on the result. Assuming we need to generate an m-bit binary hash code, we first perform a modulo operation on each element g(h... i Multiple different hash functions are used, each of which randomly maps to a set of integers. Ideally, each h... i ' is a uniform distribution between 0 and 2 m A random number of -1.

[0065] Step 42: Determine whether the preliminary hash code of the component generated in the a-th generation is the same as the personalized hash code of any target component. If yes, proceed with the (a+1)-th generation process of the component. If no, use the preliminary hash code of the component generated in the a-th generation as the personalized hash code of the component. The target component is any component other than the component mentioned above.

[0066] Furthermore, when generating personalized hash codes for components, it is also necessary to consider the similarity between hash codes of components of the same category and the diversity between hash codes of components of different categories, thereby reducing hash collisions. Specifically, formula L is used. hash (h i ”,h' j ')=W·|d(h i ”,h' j ')| p The difference between two personalized hash codes is calculated to increase the similarity between hash codes of components of the same category and decrease the similarity between hash codes of components of different categories. Where h... i " is the personalized hash code of the i-th component, h' j ' is the personalized hash code of the j-th component, L hash (h i ”,h' j ') is h i "and h' j The difference between ', W is the weighting coefficient, d(h) i ”,h' j ') is h i "and h" j The distance between them, p is a positive integer used to increase the sensitivity of the loss function.

[0067] The personalized hash codes generated in this application not only consider the internal feature information of each BIM component but also its correlation information in the geometric structure. This enables more effective representation learning and feature extraction capabilities for BIM of massive components, thereby generating more accurate and robust hash codes. By handling hash collisions, the similarity problem of different components in the hash space is resolved, improving retrieval efficiency and accuracy.

[0068] In hash table design, the goal is to map the feature information of each component to a unique position in the table using a hash function. However, in real-world applications, the feature information of BIM components often far exceeds the number of possible hash codes, leading to a probability that different components might map to the same hash code. To ensure that the generated hash codes are independent at each position (i.e., different binary bits corresponding to different hash functions) and do not cause collisions, this application employs a "rehashing method" for calculation. If a component finds that its position is already occupied after the first hash code generation (i.e., a hash collision has occurred), by introducing a second (or more) hash function, new and different hash values ​​can be generated for known hash collisions, thereby reducing the probability of collisions and improving the efficiency of the hash table.

[0069] To better understand the process of generating a personalized hash code for each component in step 204 of this application, the steps are described in detail below:

[0070] (1) Choose two hash functions: H1() and H2(). H1() is the main hash function used for initial hash processing, and H2() is the auxiliary hash function used to handle collisions.

[0071] (2) Calculate the master hash value: For each component, use the master hash function H1() to obtain a master hash code and determine where the component should be placed.

[0072] (3) If the position is already occupied, a new hash code is calculated using the auxiliary hash function H2() and the known master hash code as parameters.

[0073] (4) Loop processing: If the position of the newly calculated hash code is still occupied, continue to use the same H2() and thus increase the number of generation a to find another available position until an empty position is found.

[0074] This application provides an efficient, accurate, and highly scalable method for processing and indexing large amounts of data for BIM components. It can not only effectively reduce the conflict rate, but also adapt to the specific needs of different categories. When it is necessary to retrieve a specific type of BIM component, the component can be quickly located through a hash table, thereby improving query efficiency and system performance in practical applications.

[0075] In summary, the beneficial effects of this application include at least the following:

[0076] 1. Improved Query Efficiency: Utilizing GCN to process BIM component data enables more efficient data indexing and retrieval. By training GCN to understand the complex relationships between components, it can intelligently parse and match user queries, reducing unnecessary full table scans or complex calculations, thereby significantly improving query performance. By generating personalized hash codes, it can quickly find related or similar objects in a predefined hash space, significantly reducing query time complexity. It supports not only exact matching but also fuzzy queries, improving search efficiency in large databases.

[0077] 2. Reduced data redundancy and conflicts: Data processed by GCN can more accurately represent the relationships and attributes between components, and the hashing process eliminates or reduces the possibility of duplicate records. This not only saves storage space, but also reduces maintenance costs and query errors caused by mismatches, and improves data consistency and reliability.

[0078] 3. Reduced resource consumption: By storing personalized hash codes, the physical space required to store data is reduced, and the demand for hardware resources when processing large amounts of data is lowered.

[0079] 4. Enhanced flexibility and adaptability: By learning the dynamic relationships between BIM components, GCN can respond more flexibly to changes in user needs or the addition of new components. It can adaptively adjust prediction and matching strategies without frequent data reconstruction or index updates, thus becoming more efficient in handling rapidly changing requirements. Furthermore, it can better integrate different data formats and types, increasing overall compatibility and scalability.

[0080] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0081] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0082] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0083] In this application, all actions to acquire signals, information, or data are carried out in compliance with the relevant data protection laws and policies of the country where the location is situated, and with the authorization granted by the owner of the relevant device.

[0084] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0085] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited thereto.

[0086] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0087] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A storage method of digital twin model component data, characterized by, The storage method of the digital twin model component data comprises: obtaining component data of a digital twin model; the component data comprises a feature vector of each component and a relationship between components; establishing a graph structure data set according to the component data; the graph structure data set comprises node features and graph structure features, the node features comprising each component and a feature vector of each component, and the graph structure features comprising a relationship between components; determining a feature representation and a category of each component by using a pre-trained deep learning model according to the graph structure data set; the deep learning model comprises a graph convolutional neural network, a full connection layer and a classifier connected in sequence, the output of the graph convolutional neural network being the feature representation, and the output of the classifier being the category; generating a personalized hash code corresponding to each component according to the feature vector of each component, the feature representation of each component and the category of each component respectively, and storing the personalized hash code in a database; wherein the personalized hash code corresponding to each component is generated according to the feature vector of each component, the feature representation of each component and the category of each component respectively, and specifically comprises: for the a-th generation process of any component, a primary hash code of the component generated in the a-th time is obtained by using a primary hash function and an auxiliary hash function according to the generation number and feature information of the component; the feature information comprises a feature vector, a feature representation and a category; a≥0; it is judged whether the primary hash code of the component generated in the a-th time is the same as the personalized hash code of any target component, if yes, the a+1-th generation process of the component is performed, and if not, the primary hash code of the component generated in the a-th time is taken as the personalized hash code of the component; the target component is the remaining component except the component; the process of generating the personalized hash code for each component specifically comprises: selecting two hash functions: H1() and H2(); H1() is a primary hash function for initial hash processing, and H2() is an auxiliary hash function for processing conflicts; calculating a primary hash value: for each component, a primary hash code is obtained by using the primary hash function H1(), to determine the position where the component should be placed; if the position is occupied, a new hash code is calculated by using the auxiliary hash function H2() and the known primary hash code as parameters; loop processing: if the position of the newly calculated hash code is still occupied, another available position is searched by using the same H2() and the generation number a which is incremented, until an empty position is found. 2.The method of claim 1, wherein, The component data of the digital twin model is obtained, specifically comprising: obtaining attribute information of each component in the digital twin model; determining the relationship between components according to the attribute information of each component; normalizing the attribute information of each component to obtain a feature vector of each component. 3.The method of claim 1, wherein, The digital twin model is a building information model of a digital twin intelligent pump station. 4.The method of claim 1, wherein, The primary hash code of the component generated in the a-th time is obtained by using the primary hash function and the auxiliary hash function according to the generation number and the feature information of the component, specifically comprising: According to the feature information of the component, a main hash code of the component is determined by using a main hash function; According to the generation number and the feature information of the component, an auxiliary hash code of the component is determined by using an auxiliary hash function; According to the main hash code and the auxiliary hash code, a preliminary hash code of the component generated for the a-th time is determined.

5. The storage method of digital twin model component data according to claim 4, characterized in that, According to the feature information of the component, a main hash code of the component is determined by using a main hash function, specifically comprising: Hash calculation is performed on the feature information of the component to obtain a first hash code of the component; Linear transformation is performed on the first hash code of the component to obtain a second hash code of the component; The second hash code of the component is mapped to a fixed-length binary hash code by using a Hashing Trick algorithm to obtain the main hash code of the component.

6. The storage method of digital twin model component data according to claim 5, characterized in that, The main hash code of the i-th component is determined by using the following formula: h i ' = (g(h i )) mod 2 m ; where h i is the main hash code of the ith component, h i is the first hash code of the ith component, m is the length of the binary hash code, g() represents a linear transformation, and mod represents a modulo operation. 7.The method of claim 1, wherein, The preliminary hash code of the component generated for the a-th time is obtained by using the following formula: NewHashValue = H1(data) + a x H2(data); Wherein, NewHashValue is the preliminary hash code of the component generated for the a-th time, data is the feature information of the component, H1() is the main hash function, H2() is the auxiliary hash function, and a is the generation number.

8. A computer device comprising: A memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor executes the computer program to implement the storage method of the component data of the digital twin model according to any one of claims 1-7.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the storage method of the component data of the digital twin model according to any one of claims 1-7.

Images