A digital twin construction method, device, equipment and computer storage medium

Abstract

Embodiments of the present application provide a digital twin construction method, device and equipment and a computer storage medium. The method comprises extracting a feature vector of first attribute information of a target entity to obtain a corresponding multi-dimensional feature vector; performing correlation degree quantization processing on the multi-dimensional feature vector and sorting feature values from large to small, selecting the first N feature values and the corresponding correlation degree feature vectors, and performing weighted summation processing to obtain a multi-dimensional feature optimization twin coefficient; when the multi-dimensional feature optimization twin coefficient is not less than a standard multi-dimensional feature optimization twin coefficient, instantiating a standard digital twin template combined with the first attribute information to obtain a target digital twin; and when the multi-dimensional feature optimization twin coefficient is less than the standard multi-dimensional feature optimization twin coefficient, constructing the target digital twin using the first attribute information. Embodiments of the present application use the first attribute information to directly construct the target digital twin without template matching, thereby expanding the application scenarios.

Description

Technical Field

[0001] This application belongs to the field of Internet of Things (IoT) technology, and in particular relates to a method, apparatus, device and computer storage medium for constructing a digital twin. Background Technology

[0002] With the development of IoT digital twin technology, it has been applied to various fields. Taking the fire protection field as an example, existing technologies embed pre-designed fire linkage calculation rules into building digital delivery software. After the equipment and facilities that require corresponding digital twins are connected, calculations are performed based on these pre-designed rules to generate fire linkage data, helping building management to digitally replicate and control equipment and facilities. However, these pre-designed fire linkage calculation rules are mainly based on general fire protection standards, and the generated fire linkage data cannot be effectively adapted to actual application scenarios. Furthermore, for equipment and facilities without pre-designed rules, it is impossible to build corresponding digital twins. On the other hand, building a specific digital twin for each piece of equipment and facility for a specific scenario increases the construction difficulty, limiting the widespread application of digital twin technology.

[0003] Therefore, how to provide a method that can improve the scenario adaptability of the constructed digital twin, be applied to the construction of digital twins for new equipment and facilities, and at the same time take into account the low construction difficulty is a problem that needs to be solved by those skilled in the art. Summary of the Invention

[0004] This application provides a method, apparatus, device, and computer storage medium for constructing digital twins, which can solve the problems that existing generated fire linkage data cannot be effectively adapted to actual application scenarios, and that digital twins cannot be constructed for equipment and facilities without pre-defined rules; while constructing a digital twin for each specific scenario for each piece of equipment and facility increases the construction difficulty and limits the widespread application of digital twin technology.

[0005] In a first aspect, embodiments of this application provide a method for constructing a digital twin, the method comprising: Obtain the first attribute information of the target entity; the first attribute information is information used to characterize the features of the target entity. Extract the feature vector of the first attribute information to obtain the corresponding multidimensional feature vector; The correlation degree of the multidimensional feature vector is quantized to obtain the correlation degree feature vector and the corresponding feature value. The correlation degree feature vector represents the correlation combination of multidimensional features, and the feature value represents the correlation degree quantization value of the corresponding correlation degree feature vector. Based on the descending order of the feature values, select the first N feature values ​​and the corresponding correlation feature vectors; The first N eigenvalues ​​and their corresponding correlation feature vectors are weighted and summed to obtain the multidimensional feature optimized twinning coefficients; the multidimensional feature optimized twinning coefficients are used to characterize the importance of the correlation between multidimensional features; When the multidimensional feature optimization twin coefficient is not less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity, the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficient is instantiated by combining the first attribute information to obtain the target digital twin of the target entity. When the twin coefficient of the multidimensional feature optimization is less than the standard twin coefficient of the target entity, the target digital twin of the target entity is constructed using the first attribute information.

[0006] In one feasible implementation, the above-described construction of a target digital twin of the target entity using the first attribute information may include: Construct a specific digital twin template for the target entity based on the first attribute information; The target digital twin of the target entity is obtained by instantiating based on the first attribute information and a specific digital twin template.

[0007] In one feasible implementation, after constructing a specific digital twin template of the target entity based on the first attribute information, the following may also be included: Obtain the product type attribute information of the target entity; Input the product type attribute information and the first attribute information into the initialization model to obtain the initialization model weight values; The first attribute information and the configuration information of the standard digital twin template are both input into the similarity weight index model and the optimization model, respectively, and the similarity weight index and optimization correction coefficient are obtained respectively. The sedimentation factor value is obtained based on the initial model weight value, similarity weight index, and optimization correction coefficient; If the precipitation factor value is greater than or equal to the preset precipitation factor value, the specific digital twin template will be updated to the latest standard digital twin template of the target entity.

[0008] In one feasible implementation, the method may further include: If the precipitation factor value is less than the preset precipitation factor value, the second attribute information is added to the first attribute information to obtain the supplemented first attribute information. Return to the steps of inputting product type attribute information and first attribute information into the initialization model to obtain the initialization model weight values.

[0009] In one feasible implementation, the above-described correlation quantification processing of the multidimensional feature vector to obtain the correlation feature vector and the corresponding feature value may include: Constructing the covariance matrix using multidimensional eigenvectors; The covariance matrix is ​​subjected to eigenvalue decomposition to obtain the correlation degree eigenvector and eigenvalue.

[0010] In one feasible implementation, the weighted summation of the first N feature values ​​and their corresponding correlation feature vectors to obtain the multidimensional feature-optimized twinning coefficients can include: Obtain the normalized eigenvalues ​​of the first N eigenvalues, and the ratio of the magnitude of each correlation eigenvector to the magnitude of the maximum correlation eigenvector in the correlation eigenvectors corresponding to the first N eigenvalues; The normalized eigenvalues ​​and ratios are weighted and summed to obtain the multidimensional feature-optimized twin coefficients.

[0011] In one feasible implementation, the method may further include: Obtain sample attribute information from entity sample data; sample attribute information is information used to characterize the features of entity sample data. The sample attribute information is processed by feature vector extraction to obtain the corresponding multidimensional feature vector of the sample; The mean of the multidimensional feature vector of the sample is normalized to zero to obtain the zeroed multidimensional feature vector of the sample. The sample covariance matrix is ​​constructed based on the zero-normalized sample multidimensional feature vector, and the sample covariance matrix is ​​subjected to eigenvalue decomposition to obtain the sample correlation feature vector and sample feature values. Based on the sorting order of sample feature values ​​from largest to smallest, select the first N sample feature values ​​and the corresponding sample correlation feature vectors; Normalize the sample feature values ​​to obtain normalized sample feature values; The standard multidimensional feature optimization twin coefficient is obtained by weighted summing of the sample correlation feature vectors corresponding to the first N sample feature values, the sample ratio of the feature vector magnitude of the sample correlation feature vector to the feature vector magnitude of the maximum sample correlation feature vector, and the sample normalized feature values.

[0012] Secondly, embodiments of this application provide a digital twin construction apparatus, which may include: The acquisition module is used to acquire the first attribute information of the target entity; the first attribute information is information used to characterize the features of the target entity. The feature vector extraction module is used to extract the feature vector of the first attribute information to obtain the corresponding multi-dimensional feature vector. The correlation degree quantification module is used to quantify the correlation degree of multi-dimensional feature vectors to obtain correlation degree feature vectors and corresponding feature values. The correlation degree feature vectors represent the correlation combination of multi-dimensional features, and the feature values ​​represent the correlation degree quantification value of the corresponding correlation degree feature vectors. The selection module is used to select the top N feature values ​​and their corresponding correlation feature vectors based on the descending order of feature values. The weighted summation module is used to perform weighted summation on the first N feature values ​​and their corresponding correlation feature vectors to obtain the multidimensional feature optimized twin coefficients; the multidimensional feature optimized twin coefficients are used to characterize the importance of the correlation between multidimensional features; The instantiation module is used to instantiate the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficient, combined with the first attribute information, when the multidimensional feature optimization twin coefficient is not less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity, so as to obtain the target digital twin of the target entity. The construction module is used to construct a target digital twin of the target entity using the first attribute information when the twin coefficient of the multidimensional feature optimization is less than the standard twin coefficient of the target entity.

[0013] Thirdly, embodiments of this application provide a digital twin construction device, which may include: a processor and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the digital twin construction method as described above.

[0014] Fourthly, embodiments of this application provide a computer storage medium storing computer program instructions, which, when executed by a processor, implement the digital twin construction method described above.

[0015] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements any of the methods for constructing a digital twin in the above embodiments.

[0016] The digital twin construction method, apparatus, device, and computer storage medium of this application extract feature vectors from first attribute information to obtain multi-dimensional feature vectors, thereby transforming the original first attribute information into feature parameters that can be used for subsequent quantization processing. Then, by performing correlation degree quantization processing on the multi-dimensional feature vectors, correlation degree feature vectors representing the correlation combination of multi-dimensional features are obtained, along with feature values ​​representing the quantized correlation degree values ​​of the corresponding correlation degree feature vectors. This yields the correlation patterns and corresponding degree information between multi-dimensional features. Finally, following a selection order from largest to smallest, several feature values ​​and their corresponding correlation degree feature vectors are selected and weighted summed to filter out highly correlated features. The correlation degree feature vector and its corresponding feature value are weighted and summed to reduce the amount of data processed, simplify the processing complexity, and reduce the impact of signal-to-noise ratio on the results. When the multidimensional feature optimization twin coefficient is greater than or equal to the standard multidimensional feature optimization twin coefficient, the target digital twin is instantiated using the standard digital twin template combined with the first attribute information. The standard digital twin template can be directly called without manual configuration for the scenario, which improves the efficiency of digital twin construction. When the multidimensional feature optimization twin coefficient is less than the standard multidimensional feature optimization twin coefficient, the target digital twin is directly constructed using the first attribute information. Digital twins can also be deployed even without template matching, which broadens the application scenarios of this method. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an architecture diagram of a digital twin construction model provided in an embodiment of this application; Figure 2 This is a flowchart illustrating a digital twin construction method provided in an embodiment of this application; Figure 3 This is a schematic diagram illustrating the process of converting a specific digital twin template into a standard digital twin template, as provided in an embodiment of this application. Figure 4 This is a flowchart illustrating a method for calculating standard multidimensional feature optimization twin coefficients provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a digital twin construction device provided in an embodiment of this application; Figure 6 This is a schematic diagram of the hardware structure of a digital twin construction device provided in an embodiment of this application; The annotations in the attached figures are explained as follows: 101 - Equipment and facility attribute information acquisition node, 102 - Equipment and facility classification and identification node, 103 - Standard digital twin instantiation node, 104 - Specific digital twin construction node. Detailed Implementation

[0019] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0020] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0021] Taking the fire protection field as an example: With the development of IoT digital twin technology and the increasing demand for fire safety, digital twin technology has been applied to fire protection scenarios. This digital twin technology can collect real data through the Internet of Things and quickly build a virtual model that maps to the physical fire protection scenario. It has the ability to quickly build an accurate mapping between the twin world and the physical world, which is the key to timely response and handling of emergency events. It can enhance the management capabilities of the fire scene and can be used for emergency command, drills and other scenarios.

[0022] There is an existing method for automatically calculating fire linkage relationships. This method converts professional knowledge of fire linkage into computer-readable rules, matches these rules to calculate the fire linkage relationships of equipment and facilities, and then stores these relationships on a twin data platform. The data is then transmitted to the client via an interface. However, the converted rules, based on the professional knowledge of fire linkage, are essentially pre-defined lists and logics—pre-made rules. These rules cannot be updated in real-time to address complex real-world situations. For example, if only the linkage logic between smoke sensors and sprinkler heads is pre-set, but not for newly added equipment such as smart fire doors or temperature detectors in special areas, the smart fire door and temperature detector are not in the pre-defined list of identifiable rules. The system cannot confirm the relationships between the smart fire door and temperature detector, or with other equipment and facilities (e.g., it cannot determine which detector the smart fire door should be linked to for closure). Therefore, it cannot replicate the state of the smart fire door and temperature sensor, nor can it control them, and it cannot adapt to equipment and facilities without corresponding pre-defined calculation rules. Furthermore, the prefabricated rules require manual revision and cannot be automatically updated, resulting in the method of calculating fire linkage relationships requiring manual revision of the prefabricated rules and having poor real-time performance.

[0023] There is another existing method for constructing digital twin models. This method involves a series of steps, including receiving requests, building a 3D model, static mapping correction, dynamic interaction, simulation, fault prediction, real-time monitoring, and model correction, to construct the digital twin model. This method has advantages in real-time data transmission, sharing, and model performance optimization. However, it is highly dependent on preliminary work, especially the need to pre-collect a large amount of target point cloud data and perform tedious and high-quality optimization processing on this target point cloud data. This increases the complexity and time cost of model construction, limiting the widespread application and rapid iteration capability of this method.

[0024] Therefore, how to provide a method that can improve the scenario adaptability of the constructed digital twin, be applied to scenarios where digital twins are constructed using newly added equipment and facilities, and at the same time take into account the low complexity of digital twin construction, is a problem that needs to be solved by those skilled in the art.

[0025] To address the aforementioned technical problems, embodiments of this application provide a method, apparatus, device, and computer storage medium for constructing a digital twin.

[0026] Specifically, this application first illustrates the process of the above-mentioned digital twin construction method through an application scenario example. (See also...) Figure 1 , Figure 1This is an architecture diagram of a digital twin construction model provided in an embodiment of this application.

[0027] The architecture of the digital twin construction model proposed in this application may include: a device / facility attribute information acquisition node 101, a device / facility classification and identification node 102, a standard digital twin instantiation node 103, and a specific digital twin construction node 104. The device / facility attribute information acquisition node 101 is connected to the device / facility classification and identification node 102 and is used to send the acquired device / facility attribute information to the device / facility classification and identification node 102. This application embodiment efficiently integrates IoT data and 3D spatial data by constructing a mapping between an IoT digital twin model and a 3D entity model, ensuring the real-time nature and accuracy of information. Based on the rendering capabilities of WebGPU (Web Graphics Processing Unit) technology, it can recreate real-world scenes through a web browser.

[0028] The aforementioned equipment and facility classification and identification node 102 is connected to the standard digital twin instantiation node 103 and the specific digital twin construction node 104, respectively. Upon receiving equipment and facility attribute information, the equipment and facility classification and identification node 102 identifies whether a standard digital twin template corresponding to the equipment and facility attribute information exists in the database. Specifically, through data processing, original feature vector extraction, feature vector filtering, and weighted summation of the equipment and facility attribute information, a multi-dimensional feature-optimized twin coefficient M1 is obtained for the equipment and facility. Based on the M1 value of the equipment and facility, it is compared with the M value corresponding to the standard digital twin template of the same or similar type of equipment and facility. It can be set that if the M1 value is greater than or equal to the M value, the standard digital twin template is considered to be the standard digital twin template corresponding to the aforementioned equipment and facility, meaning that the standard digital twin template can be reused; it can be set that if the M1 value is less than the M value, the standard digital twin template does not exist in the database, meaning that the standard digital twin template cannot be reused.

[0029] When a standard digital twin template corresponding to the aforementioned equipment and facility attribute information exists in the database, the equipment and facility classification and identification node 102 sends the aforementioned equipment and facility attribute information to the standard digital twin instantiation node 103. The standard digital twin instantiation node 103 receives the corresponding standard digital twin template, combines the aforementioned standard digital twin template with the equipment and facility attribute information, and instantiates it in a specific scenario to obtain a digital twin of the equipment and facility. This eliminates the need for manual configuration of relevant information and solves the problem of low mapping efficiency in existing technologies.

[0030] If no standard digital twin template corresponding to the aforementioned equipment and facility attribute information exists in the database, the equipment and facility classification and identification node 102 sends the aforementioned equipment and facility attribute information to a specific digital twin construction node 104. The specific digital twin construction node 104 receives the aforementioned equipment and facility attribute information and, based on the received specific digital twin rules and the aforementioned equipment and facility attribute information, constructs a digital twin of the equipment and facility. Specifically, in this embodiment, multi-angle view information of the equipment and facility can be analyzed and processed to obtain equipment and facility attribute information corresponding to the corresponding category, and a digital twin of the equipment and facility can be constructed based on this equipment and facility attribute information.

[0031] Furthermore, to prevent low mapping efficiency caused by the need to construct a digital twin based on the attribute information of a specific device or facility each time it is connected to the model (that is, a device or facility for which there is no corresponding standard digital twin template in the database), a specific digital twin template can be constructed based on the attribute information of the specific device or facility after it is first connected to the model. Then, using the attribute information of the specific device or facility, the product type information corresponding to the specific device or facility, and the standard digital twin template with the highest matching degree in the database (this standard digital twin template with the highest matching degree can come from the same type of device or facility or from a similar type of device or facility), a sedimentation factor P value is calculated. The P value is used to characterize the proportion of the characteristic association information of the specific device or facility that the currently selected attribute information can carry in the total characteristic association information of the specific device or facility. The larger the P value, the more capable the currently selected attribute information is of representing the main attribute information of the specific device or facility. Conversely, if the P value is less than a preset value, it means that the currently selected attribute information cannot be used as the main attribute information of the specific device or facility.

[0032] In this embodiment, if the P value is greater than or equal to the preset value, the specific digital twin template can be updated to the standard digital twin template of the specific equipment and facility. Subsequently, when a device or facility that is compatible with the updated standard digital twin template is accessed, the updated standard digital twin template can be directly matched, as described above in the case of "when a standard digital twin template corresponding to the attribute information of the aforementioned equipment and facility exists in the database," to construct a digital twin. If the P value is less than the preset value, additional attribute information of the specific equipment and facility can be added to the original attribute information of the specific equipment and facility. The corresponding P value can then be recalculated using the additional attribute information of the specific equipment and facility until the P value calculated using the additional attribute information of the specific equipment and facility is greater than or equal to the preset value. At this point, the additional attribute information of the specific equipment and facility is taken as the final attribute information of the specific equipment and facility, and the specific digital twin template constructed using the final attribute information of the specific equipment and facility is updated to the standard digital twin template of the specific equipment and facility.

[0033] The method described above for constructing digital twins of equipment and facilities not only solves the problem of incomplete standard digital twin templates in existing technologies, which prevents the construction of digital twins for some equipment and facilities, but also allows the construction of specific digital twin templates that meet the conditions to serve as new standard digital twin templates corresponding to the specific equipment and facilities. This eliminates the need to construct digital twins from scratch each time, ensuring both comprehensive mapping and efficiency. The model instantiates the attribute information of the accessed equipment and facilities in the scenario, enabling users to build a reliable and convenient fire-fighting linkage data twin platform.

[0034] The following section first introduces the digital twin construction method provided in the embodiments of this application.

[0035] Figure 2 A flowchart illustrating a digital twin construction method according to an embodiment of this application is shown. Figure 2 As shown, the method may include the following steps: S201: Obtain the first attribute information of the target entity; the first attribute information is information used to characterize the features of the target entity; S202: Extract the feature vector of the first attribute information to obtain the corresponding multi-dimensional feature vector; S203: Perform correlation degree quantization processing on the multidimensional feature vectors to obtain correlation degree feature vectors and corresponding feature values; the correlation degree feature vectors represent the correlation combination methods of multidimensional features, and the feature values ​​represent the correlation degree quantization values ​​of the corresponding correlation degree feature vectors; S204: Based on the descending order of feature values, select the first N feature values ​​and the corresponding correlation feature vectors; S205: The first N eigenvalues ​​and their corresponding correlation feature vectors are weighted and summed to obtain the multidimensional feature optimization twin coefficients; the multidimensional feature optimization twin coefficients are used to characterize the importance of the correlation between multidimensional features; S206: When the multidimensional feature optimization twin coefficient is not less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity, the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficient is instantiated by combining the first attribute information to obtain the target digital twin of the target entity. S207: When the twin coefficient of the multidimensional feature optimization is less than the standard twin coefficient of the target entity, construct the target digital twin of the target entity using the first attribute information.

[0036] The following is a detailed explanation of each step: S201: Obtain the first attribute information of the target entity; the first attribute information is information used to characterize the features of the target entity.

[0037] The digital twin constructed in this embodiment can be commonly applied to building fire protection scenarios, or other scenarios. The aforementioned digital twin is a virtual representation or model corresponding to a physical object, process, or system. It creates a virtual copy that can map the physical entity in real time by capturing and integrating real-time data from sensors, devices, and other sources, combined with analysis, processing, and visualization technologies. This digital twin is not merely a static digital model, but a dynamic model that continuously evolves and adapts based on received real-time data about the corresponding physical entity, enabling real-time monitoring, analysis, and optimization of the mapping performance. The target entity is the entity to which the digital twin needs to be constructed, and the first attribute information is the attribute information required for constructing the digital twin, i.e., information used to characterize the features of the target entity. Furthermore, it should be noted that this embodiment does not specifically limit the aforementioned first attribute information; the specific types of first attribute information to be collected can be determined according to the actual application scenario. The first attribute information may specifically include product type information of the target entity (for example, the product type may include sensor type, fire truck type, etc.), location information, name information, model information, manufacturer information, status information, image information, and other attribute data information. The sensor types in the above product type information can be further divided into smoke sensor types, temperature sensor types, water pump types, fire hydrant pump types, and sprinkler pump types. This embodiment can acquire multiple first attribute information for a single target entity. This embodiment does not limit the specific acquisition method. For example, for a single target entity, the step of acquiring first attribute information can be performed once at multiple time sequences to obtain multiple first attribute information corresponding to the same target entity; or, at the same time, first attribute information can be acquired once under multiple set conditions, also obtaining multiple first attribute information corresponding to the same target entity. Here, acquiring first attribute information once under multiple set conditions requires further explanation that these set conditions need to be determined based on the specific type of first attribute information being acquired. For example, when the acquired first attribute information includes image information, image information of the target entity can be acquired from multiple different viewpoints.

[0038] Further elaboration on the aforementioned practical application scenarios is needed. The classification of practical application scenarios is based on the corresponding categories of the target entities. Using the building fire protection field as an example, the practical application scenarios can be specifically divided into high-rise building fire application scenarios, commercial building fire application scenarios, and industrial fire application scenarios. In this embodiment, the aforementioned first attribute information can be collected on the IoT platform using a pre-set page integration tool or a pre-set API (Application Programming) interface (specifically, a standardized API interface using JSON format data, which is a lightweight text data exchange format). In this embodiment, the corresponding information can be filled in on the page integration tool, including project name, data source name, username, user ID, organization path, access key, Kafka listening URL, signature algorithm version information, signature method, expiration time, etc. Based on the method of collecting first attribute information using the page integration tool, we take collecting the product type information of the target entity as an example. Then, by selecting the corresponding target entity on the page integration tool, the mapping of the corresponding digital twin can be completed.

[0039] In this embodiment, obtaining the first attribute information of the target entity can be done either in real-time (i.e., immediately after completing one acquisition operation, the next acquisition operation is performed) or at preset acquisition time intervals. This application does not limit the value of the preset acquisition time interval; for example, it can be set to 1 minute, 2 minutes, or 5 minutes. This application does not limit the basis for setting the preset acquisition time interval; it can be customized by the operator or set according to the priority of the current business. Specifically, the higher the priority of the current business, the smaller the corresponding preset acquisition time interval.

[0040] S202: Extract the feature vector of the first attribute information to obtain the corresponding multidimensional feature vector.

[0041] Based on the aforementioned first attribute information, feature vectors corresponding to multiple categories are extracted, namely the aforementioned multidimensional feature vectors. Each element in the extracted feature vector is a quantized value, used to characterize the features of the corresponding first attribute information. For ease of understanding, an example is given: taking the first attribute information as including product type information, location information, status information, and image information, and specifically the product type being a fire truck, the elements in the extracted feature vector can include the water pump pressure feature element and the fire truck ambient temperature feature element corresponding to this fire truck type; the direction feature element and the position feature element corresponding to the location information; the speed feature element, acceleration feature element, engine speed feature element, engine temperature feature element, and atmospheric pressure feature element corresponding to the status information; and the texture feature element and the image feature element corresponding to the image information. All the aforementioned elements can constitute a multidimensional feature vector that quantifies and characterizes the target entity. In a feasible embodiment, when multiple first attribute information is obtained for a single target entity, a corresponding multidimensional feature vector is extracted for each first attribute information. Thus, a multidimensional feature vector set is extracted for each target entity, and this set includes multiple of the aforementioned multidimensional feature vectors.

[0042] S203: Perform correlation degree quantification processing on the multidimensional feature vectors to obtain correlation degree feature vectors and corresponding feature values; the correlation degree feature vectors represent the correlation combination method of multidimensional features, and the feature values ​​represent the correlation degree quantification value of the corresponding correlation degree feature vectors.

[0043] This step focuses on quantifying the relationships between elements in a multidimensional feature vector to determine whether a relationship exists between each element and other elements, and the strength of that relationship. Several methods can be used to quantify the correlation degree of the multidimensional feature vector. Specifically, a mutual information matrix (MI) approach can be used; alternatively, a covariance matrix approach can be used; or correlation coefficient matrices, Mahalanobis distance matrices, or GNN (Graph Neural Network) methods can be employed. It is crucial that any specific method used to quantify the correlation degree of the multidimensional feature vector outputs the aforementioned correlation feature vector and its corresponding feature values. It should be noted here that the output correlation feature vector and its corresponding feature values ​​are not the same as the multidimensional feature vector mentioned above. The elements in the multidimensional feature vector are used to represent the various features of the target entity, while the elements in this correlation feature vector are used to represent the correlation relationship between multiple features of the target entity.

[0044] In one feasible embodiment, to improve the accuracy of the mapping between the constructed digital twin and the target entity, the aforementioned correlation quantification processing of the multi-dimensional feature vector to obtain the correlation feature vector and the corresponding feature value may include: Constructing the covariance matrix using multidimensional eigenvectors; The covariance matrix is ​​subjected to eigenvalue decomposition to obtain the correlation degree eigenvector and eigenvalue.

[0045] This application specifically utilizes multi-dimensional feature vectors to construct a covariance matrix and performs eigenvalue decomposition on the constructed covariance matrix. This method accurately captures and quantifies the correlation between elements in the multi-dimensional feature vectors, ensuring the accuracy of the subsequently constructed digital twin. The essence of the aforementioned covariance matrix is ​​to quantify the degree of linear correlation between elements in any two dimensions of the multi-dimensional feature vectors. For example, the (i, j)-th element in the covariance matrix represents the covariance value between the i-th and j-th elements. A positive covariance value indicates a positive correlation between the two elements; a negative covariance value indicates a negative correlation; and a zero covariance value indicates no linear correlation between the two elements. This embodiment utilizes multidimensional feature vectors to construct a covariance matrix. Specifically, it uses multidimensional feature vectors from multiple samples to construct the covariance matrix. Each sample includes a multidimensional feature vector representing the target entity. The elements in the multidimensional feature vector exhibit fluctuations across different samples and are not entirely identical, ensuring that the covariance matrix constructed using the multidimensional feature vectors from multiple input samples is meaningful in actual calculation. For example, in a feasible embodiment, view images of the target entity from multiple perspectives can be acquired, such as a front view image, a top view image, and other view images different from the front and top views. Specifically, this embodiment can be configured to process the acquired view images using a convolutional neural network. Processing the view images allows for the analysis of the target entity's geometric shape, surface texture, color distribution, size, structure, texture, and mapping features. When subsequently constructing a digital twin, the digital twin can be built based on the aforementioned geometric shape features, surface texture features, mapping features, and color distribution features. This embodiment does not limit the specific method of acquiring view images of the target entity from multiple perspectives; it can be obtained directly from the manufacturer or through devices such as drones and cameras.

[0046] In one feasible embodiment, before constructing the covariance matrix using the multidimensional feature vectors, the elements of different categories in the multidimensional feature vectors can be standardized to eliminate the problem that some elements have excessively large values ​​relative to other elements, affecting the accuracy of the final digital twin mapping. Specifically, a mean-zeroing method can be used, replacing the original elements in the multidimensional feature vectors with the difference between the original element and the corresponding sample mean, where the sample mean is the mean calculated using elements of the same type from all samples.

[0047] In one specific embodiment, to avoid the influence of the absolute values ​​of the data in the multidimensional feature vectors of the extracted samples on the results, and to avoid the interference of the mean differences between different features on the accuracy of the correlation degree quantification results, the relative values ​​of the data can be used instead of the absolute values ​​of the data for processing, thereby achieving the above effect. In this embodiment, after extracting the multidimensional feature vectors of multiple samples, the covariance matrix can be constructed based on the elements of each category in the multiple samples. The calculation of the eigenvalues ​​in the covariance matrix can refer to formula (1): (1); in, Let be the covariance value between the i-th element and the j-th element in the above covariance matrix; The number of samples in the input multidimensional feature vector; n is the number of elements in the feature vector set, which is a positive integer, and i and j are both taken from the range of n, for example, i, j∈{1,2,…,n}; The value of the i-th element in the k-th sample after standardization (in this embodiment, the standardization is the mean-zeroing process mentioned above); Let be the sample mean of the i-th element after standardization across multiple samples. Let j be the value of the j-th element in the k-th sample after standardization. is the sample mean of the j-th element after standardization in multiple samples.

[0048] After obtaining the above covariance values, we select to perform eigenvalue decomposition on the covariance matrix based on the above covariance values ​​to obtain the above correlation degree eigenvectors and corresponding eigenvalues.

[0049] S204: Based on the descending order of feature values, select the top N feature values ​​and their corresponding correlation feature vectors.

[0050] Based on the magnitude of the eigenvalues, the eigenvectors with the largest eigenvalues ​​in the covariance matrix are selected, which means the eigenvalues ​​with high correlation and the corresponding correlation degree eigenvectors are selected.

[0051] S205: The first N eigenvalues ​​and their corresponding correlation feature vectors are weighted and summed to obtain the multidimensional feature optimization twin coefficients; the multidimensional feature optimization twin coefficients are used to characterize the importance of the correlation between multidimensional features.

[0052] In this embodiment, the importance of the relationships between elements in the multidimensional feature vector is obtained through weighted summation. This multidimensional feature-optimized twin coefficient can also reflect the influence weights of the features corresponding to different elements in the multidimensional feature vector on the digital twin, which can improve the construction speed when building the digital twin.

[0053] In one feasible embodiment, to improve the accuracy of the output multidimensional feature-optimized twin coefficients, the above-mentioned weighted summation of the first N feature values ​​and the corresponding correlation feature vectors to obtain the multidimensional feature-optimized twin coefficients may include: Obtain the normalized eigenvalues ​​of the first N eigenvalues, and the ratio of the magnitude of each correlation eigenvector to the magnitude of the maximum correlation eigenvector in the correlation eigenvectors corresponding to the first N eigenvalues; The normalized eigenvalues ​​and ratios are weighted and summed to obtain the multidimensional feature-optimized twin coefficients.

[0054] In one specific embodiment, the formula for calculating the twin coefficients of multidimensional feature optimization is shown in formula (2).

[0055] (2); in, The number of highly correlated eigenvectors selected after eigenvalue decomposition of the covariance matrix, which is the aforementioned N value; The weights of the i-th correlation feature vector; Let be the eigenvalue corresponding to the i-th correlation feature vector; The sum of all eigenvalues ​​is used to normalize the eigenvalues, such that... This indicates the relative importance of the i-th eigenvalue; Let be the magnitude of the i-th correlation feature vector; The magnitude of all correlation feature vectors. The importance of the feature vector used to characterize the degree of association at this point.

[0056] In this embodiment, after obtaining the multidimensional feature-optimized twin coefficients, these coefficients can be compared with the standard multidimensional feature-optimized twin coefficients corresponding to the target entity. Based on the comparison results, corresponding operations are performed to construct a digital twin. In this embodiment, the standard multidimensional feature-optimized twin coefficients corresponding to the target entity can be either the same type as the target entity or the type most closely related to the target entity.

[0057] S206: When the multidimensional feature optimization twin coefficient is not less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity, the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficient is instantiated by combining the first attribute information to obtain the target digital twin of the target entity.

[0058] In this embodiment, when the multidimensional feature optimization twin coefficient is greater than or equal to the standard multidimensional feature optimization twin coefficient corresponding to the target entity, the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficient can be used for instantiation. At this point, the standard digital twin template can be directly used to quickly model the target entity's corresponding target digital twin. Simultaneously, the first attribute information of the target entity is mapped onto the target digital twin using a scale, enabling real-time data acquisition and scene linkage. In this embodiment, the corresponding target entity information can be selected through a page integration tool. In the instantiation step, the target entity's location information, name information, product type information, model information, manufacturer information, and status information can all be bound to the corresponding target digital twin, and it can be instantiated into a preset twin scene.

[0059] S207: When the twin coefficient of the multidimensional feature optimization is less than the standard twin coefficient of the target entity, construct the target digital twin of the target entity using the first attribute information.

[0060] In this embodiment, when the multidimensional feature-optimized twin coefficient is less than the standard multidimensional feature-optimized twin coefficient corresponding to the target entity, a target digital twin can be constructed by selecting the target entity through the page integration tool. Simultaneously, when it is necessary to instantiate a multimodal target digital twin (i.e., a collection of multiple target digital twins), information from multiple types of target entities can be selected and instantiated simultaneously using the page integration tool. This selection and instantiation of multiple types of target entity information could, for example, simultaneously instantiate fire-fighting aircraft and fire-fighting robots, and these multiple types of target entities can interconnect and collaborate through the Internet of Things to share data information.

[0061] Based on the above steps, the embodiments of this application can quickly construct corresponding digital twins. Digital twins in the same scenario can be interconnected and communicate with each other through Internet of Things (IoT) technology to share data and form an intelligent collaborative system. For example, by instantiating a standard digital twin template, a digital twin is obtained, such as the digital twin corresponding to a fire truck. In the corresponding scenario, the command center can track the location, status, and equipment information of the fire truck in real time. When constructing a digital twin for a fire-fighting aircraft, a target digital twin is constructed based on the first attribute information of the fire-fighting aircraft. At this time, the constructed target digital twin can conduct aerial reconnaissance and provide fire information to fire trucks on the ground. The constructed digital twin corresponding to a smoke sensor can monitor the environmental parameters of the fire scene in real time and adjust the fire-fighting strategy. This series of digital twins can provide accurate information for remote command and dispatch in scenario applications.

[0062] In one feasible embodiment, to improve the flexibility and scalability of digital twin construction and to achieve reusability of the digital twin of the target entity, the above-mentioned construction of the target digital twin of the target entity using the first attribute information may include: Construct a specific digital twin template for the target entity based on the first attribute information; The target digital twin of the target entity is obtained by instantiating based on the first attribute information and a specific digital twin template.

[0063] This application embodiment illustrates a method for constructing a target digital twin when the multidimensional feature optimization twin coefficient is less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity. In this embodiment, a specific digital twin template corresponding to the target entity is first constructed using first attribute information. The construction process of this specific digital twin template can refer to the construction method of existing standard digital twin templates in the database. After constructing this specific digital twin template, a method for constructing a target digital twin of the target entity when the multidimensional feature optimization twin coefficient is greater than or equal to the standard multidimensional feature optimization twin coefficient corresponding to the target entity can be referenced. The first attribute information is combined with the specific digital twin template for instantiation to obtain the target digital twin. By first constructing the specific digital twin template, the target digital twin can be constructed using the same specific digital twin template when constructing the same entity.

[0064] In one feasible embodiment, to improve the efficiency of constructing a digital twin for a target entity that can adapt to the aforementioned specific digital twin template but does not subsequently have a matching standard digital twin template in the database, and to improve the efficiency of subsequently constructing a digital twin for the target entity again when constructing a specific digital twin template based on the first attribute information of the target entity, the following steps may be included after constructing the specific digital twin template of the target entity based on the first attribute information. (See reference...) Figure 3 , Figure 3 This is a schematic diagram illustrating the process of converting a specific digital twin template into a standard digital twin template, as provided in an embodiment of this application.

[0065] S301: Obtain the product type attribute information of the target entity.

[0066] The product type attribute information can be consistent with the product type information described in S201 above. For example, the product type attribute information of the target entity obtained at this time can include sensor type, fire truck type, etc., indicating which type of equipment and facilities the target entity belongs to.

[0067] S302: Input the product type attribute information and the first attribute information into the initialization model to obtain the initialization model weight values.

[0068] In this embodiment, the first attribute information may include image information. Specifically, the image of the target entity can be obtained first, and then the image can be processed by a convolutional neural network to obtain the first attribute information corresponding to the target entity.

[0069] S303: Input the first attribute information and the configuration information of the standard digital twin template into the similarity weight index model and the optimization model respectively, and obtain the similarity weight index and optimization correction coefficient respectively.

[0070] The standard digital twin template here refers to the template corresponding to the target entity. It can be a standard digital twin template in the database that belongs to the same device type as the target entity, or it can be a standard digital twin template in the database that is closest to the device type of the target entity. In specific processing, the database can be searched for the closest standard digital twin template based on the target entity.

[0071] S304: Based on the initial model weight values, similarity weight index, and optimization correction coefficients, the sedimentation factor value is obtained.

[0072] The sedimentation factor value ranges from [0,1], and based on this sedimentation factor value, it can be determined whether a specific digital twin template corresponding to the target entity can be used as the standard digital twin template for the target entity. For example, when the sedimentation factor of the specific digital twin template corresponding to the fire rescue robot product is 0.5, and the standard digital twin constructed from the standard digital twin template corresponding to the fire rescue robot has a sedimentation factor of 0.8, it can be determined that the specific digital twin template constructed using the first attribute information of the currently collected target entity is insufficient to be sedimented into a new standard digital twin template corresponding to the aforementioned fire rescue robot product. However, when the sedimentation factor of the specific digital twin template reaches 0.8, the corresponding key feature attribute data can be sedimented into standard IoT digital twin data, facilitating users to quickly construct the target digital twin effect based on the fire rescue robot product.

[0073] The formula for calculating this precipitation factor value can be found in Formula 3: (3); in, This refers to the precipitation factor value; The function initializes the model weights and outputs scalar weight values. It is a similarity weight exponential function; For the tuning function, output the scalar tuning coefficients; The feature vector information corresponding to the standard digital twin template, For product type parameter vectors, For the twin model parameter tensor; As a standard twin template, For multidimensional feature vectors, and The stability constant can be specifically set as described above. and They are 0.01 and 0.99 respectively; This is the feature vector of the current image.

[0074] S305: If the precipitation factor value is greater than or equal to the preset precipitation factor value, update the specific digital twin template to the latest standard digital twin template of the target entity.

[0075] In this embodiment, the preset precipitation factor value can be calculated based on the standard digital twin template corresponding to the target entity, construct a standard digital twin, and then use the above method to calculate the precipitation factor value of the standard digital twin as the preset precipitation factor value.

[0076] In one feasible embodiment, in order to achieve the conversion of a specific digital twin template to a standard digital twin template even in complex scenarios and to improve the success rate of converting a specific digital twin template to a standard digital twin template, the method may further include: If the precipitation factor value is less than the preset precipitation factor value, the second attribute information is added to the first attribute information to obtain the supplemented first attribute information. Return to the steps of inputting product type attribute information and first attribute information into the initialization model to obtain the initialization model weight values.

[0077] In this embodiment, when the precipitation factor value of the target entity is less than the preset precipitation factor value, the specific digital twin template constructed based on the target entity cannot be directly used as the standard digital twin template. At this time, more first attribute information for the target entity will be collected, and the precipitation factor value of the target entity will be iteratively calculated until the precipitation factor value is greater than or equal to the preset precipitation factor value.

[0078] In one feasible embodiment, to improve the accuracy of determining the twinning coefficients for standard multidimensional feature optimization, the above method may further include the following steps. For details, please refer to... Figure 4 , Figure 4 This is a flowchart illustrating a method for calculating standard multidimensional feature optimization twin coefficients provided in an embodiment of this application.

[0079] S401: Obtain sample attribute information of entity sample data; sample attribute information is information used to characterize the features of entity sample data.

[0080] S402: Perform feature vector extraction processing on the sample attribute information to obtain the corresponding multidimensional feature vector of the sample.

[0081] S403: Perform mean-zeroing processing on the multidimensional feature vectors of the samples to obtain zeroed multidimensional feature vectors of the samples.

[0082] S404: Construct a sample covariance matrix based on the zeroed sample multidimensional feature vector, and perform eigenvalue decomposition on the sample covariance matrix to obtain the sample correlation feature vector and sample feature values.

[0083] S405: Based on the descending order of sample feature values, select the top N sample feature values ​​and the corresponding sample correlation feature vectors.

[0084] S406: Normalize the sample feature values ​​to obtain normalized feature values.

[0085] S407: The sample ratio of the magnitude of the sample association feature vector to the magnitude of the maximum sample association feature vector, and the sample normalized feature values, are weighted and summed to obtain the standard multidimensional feature optimization twin coefficient.

[0086] In this embodiment, the steps for calculating the standard multidimensional feature optimization twin coefficient can refer to the above-described method for calculating the multidimensional feature optimization twin coefficient of the target entity. The difference lies in that, here, sample attribute information of the entity sample data is obtained and processed, and multiple entity sample data exist for the same type of entity. This entity sample data can be obtained from the historical data of the corresponding entity and obtained by data filtering.

[0087] The digital twin construction method provided in this application includes: S201: obtaining first attribute information of the target entity; the first attribute information is information used to characterize the features of the target entity; S202: extracting feature vectors from the first attribute information to obtain corresponding multi-dimensional feature vectors; S203: performing correlation degree quantization processing on the multi-dimensional feature vectors to obtain correlation degree feature vectors and corresponding feature values; the correlation degree feature vectors characterize the correlation combination method of multi-dimensional features, and the feature values ​​characterize the correlation degree quantization value of the corresponding correlation degree feature vectors; S204: selecting the first N feature values ​​and corresponding correlation degree feature vectors based on the descending order of feature values; S205: The first N feature values ​​and their corresponding correlation feature vectors are weighted and summed to obtain the multidimensional feature optimization twin coefficients. The multidimensional feature optimization twin coefficients are used to characterize the importance of the correlation between multidimensional features. S206: If the multidimensional feature optimization twin coefficients are not less than the standard multidimensional feature optimization twin coefficients corresponding to the target entity, the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficients is instantiated by combining the first attribute information to obtain the target digital twin of the target entity. S207: If the multidimensional feature optimization twin coefficients are less than the standard multidimensional feature optimization twin coefficients corresponding to the target entity, the target digital twin of the target entity is constructed using the first attribute information.

[0088] This embodiment extracts the feature vector of the first attribute information to obtain a multi-dimensional feature vector, thus transforming the original first attribute information into feature parameters that can be used for subsequent quantization processing. Then, by performing correlation quantization processing on this multi-dimensional feature vector, correlation degree feature vectors are obtained to characterize the correlation combination of multi-dimensional features, and feature values ​​are obtained to characterize the correlation degree quantization values ​​of the corresponding correlation degree feature vectors. This yields the correlation patterns and corresponding degree information between multi-dimensional features. Then, following a selection order from largest to smallest, several feature values ​​and their corresponding correlation degree feature vectors are selected and weighted and summed to select the correlation degree feature vectors with high correlation and their corresponding features. The eigenvalues ​​are weighted and summed to reduce the amount of data processed, simplify the processing complexity, and reduce the impact of the signal-to-noise ratio on the results. When the twin coefficient optimized by the multidimensional feature is greater than or equal to the twin coefficient optimized by the standard multidimensional feature, the target digital twin is instantiated using the standard digital twin template combined with the first attribute information. The standard digital twin template can be directly called without manual configuration for the scenario, which improves the efficiency of digital twin construction. When the twin coefficient optimized by the multidimensional feature is less than the twin coefficient optimized by the standard multidimensional feature, the target digital twin is directly constructed using the first attribute information. Digital twins can also be deployed even without template matching, which broadens the application scenarios of this method.

[0089] Furthermore, this application embodiment improves the accuracy of the mapping between the constructed digital twin and the target entity by constructing a covariance matrix using multidimensional feature vectors to quantify the correlation degree of these multidimensional feature vectors. Normalizing the data before weighted summation improves the accuracy of the output multidimensional feature-optimized twin coefficients. Constructing a specific digital twin template for the target entity based on the first attribute information when the multidimensional feature-optimized twin coefficient is less than the standard multidimensional feature-optimized twin coefficient corresponding to the target entity enhances the flexibility and scalability of digital twin construction and the reusability of constructing digital twins for the target entity. Introducing a sedimentation factor value, and updating the specific digital twin template to the latest standard digital twin template of the target entity when the sedimentation factor value is greater than or equal to a preset sedimentation factor value, improves the accuracy of subsequent mappings in the database. For target entities that do not have a suitable standard digital twin template but can adapt to the aforementioned specific digital twin template, the efficiency of constructing a digital twin is improved, as well as the efficiency of subsequently constructing a digital twin for the target entity when constructing a specific digital twin template based on the first attribute information of the target entity. By adding the second attribute information to the first attribute information when the sedimentation factor value is less than the preset sedimentation factor value, the supplemented first attribute information is obtained, and then it is determined again whether the currently constructed specific digital twin template can be used as a standard digital twin template. This enables the conversion of specific digital twin templates to standard digital twin templates even in complex scenarios, improving the success rate of converting specific digital twin templates to standard digital twin templates. By using entity sample data to calculate the standard multidimensional feature optimization twin coefficient, the accuracy of determining the standard multidimensional feature optimization twin coefficient is improved.

[0090] Figure 5 This is a schematic diagram of a device structure provided in an embodiment of this application. Figure 5 As shown, the device may include: The acquisition module 501 is used to acquire the first attribute information of the target entity; the first attribute information is information used to characterize the features of the target entity. The feature vector extraction module 502 is used to extract the feature vector of the first attribute information to obtain the corresponding multi-dimensional feature vector. The correlation degree quantification processing module 503 is used to perform correlation degree quantification processing on multi-dimensional feature vectors to obtain correlation degree feature vectors and corresponding feature values; the correlation degree feature vectors represent the correlation combination method of multi-dimensional features, and the feature values ​​represent the correlation degree quantification value of the corresponding correlation degree feature vectors; Module 504 is used to select the first N feature values ​​and their corresponding correlation feature vectors based on the descending order of feature values. The weighted summation processing module 505 is used to perform weighted summation processing on the first N feature values ​​and the corresponding correlation feature vectors to obtain the multidimensional feature optimized twin coefficients; the multidimensional feature optimized twin coefficients are used to characterize the importance of the correlation between multidimensional features; The instantiation module 506 is used to instantiate the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficient, combined with the first attribute information, when the multidimensional feature optimization twin coefficient is not less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity, so as to obtain the target digital twin of the target entity. The construction module 507 is used to construct a target digital twin of the target entity using the first attribute information when the multidimensional feature optimization twin coefficient is less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity.

[0091] The network resource allocation device described above is explained in detail below: In one embodiment, the step of constructing a target digital twin of the target entity using the first attribute information in the construction module 507 can be configured as follows: constructing a specific digital twin template of the target entity based on the first attribute information; and instantiating the target digital twin of the target entity based on the first attribute information and the specific digital twin template.

[0092] In one embodiment, the above-described apparatus may further include: The product type attribute information acquisition module is used to acquire the product type attribute information of the target entity. The initialization model weight value output module is used to input product type attribute information and first attribute information into the initialization model to obtain the initialization model weight value; The similarity weight index and optimization correction coefficient output module is used to input the first attribute information and the configuration information of the standard digital twin template into the similarity weight index model and the optimization model respectively, and obtain the similarity weight index and optimization correction coefficient respectively. The sedimentation factor value output module is used to obtain the sedimentation factor value based on the initial model weight value, similarity weight index and optimization correction coefficient; The update module is used to update a specific digital twin template to the latest standard digital twin template of the target entity when the precipitation factor value is greater than or equal to the preset precipitation factor value.

[0093] In one embodiment, the above-described apparatus may further include: The supplementary module is used to add the second attribute information to the first attribute information when the precipitation factor value is less than the preset precipitation factor value, so as to obtain the supplemented first attribute information. The loop module is used to return the steps of inputting product type attribute information and first attribute information into the initialization model to obtain the initialization model weight values.

[0094] In one embodiment, the aforementioned correlation degree quantification processing module 503 may include: Covariance matrix construction unit, used to construct a covariance matrix using multidimensional eigenvectors; The covariance matrix decomposition unit is used to perform eigenvalue decomposition on the covariance matrix to obtain the correlation degree eigenvector and eigenvalue.

[0095] In one embodiment, the weighted summation processing module 505 described above may include: The ratio determination unit is used to obtain the normalized feature values ​​of the first N feature values, and the ratio of the magnitude of each correlation feature vector to the magnitude of the maximum correlation feature vector in the correlation feature vectors corresponding to the first N feature values. The weighted summation processing unit is used to perform weighted summation processing on the normalized eigenvalues ​​and ratios to obtain the multidimensional feature-optimized twin coefficients.

[0096] In one embodiment, the above-described apparatus may further include: The sample attribute information acquisition module is used to acquire sample attribute information of entity sample data; sample attribute information is information used to characterize the features of entity sample data. The sample multidimensional feature vector extraction module is used to extract feature vectors from sample attribute information to obtain the corresponding sample multidimensional feature vectors. The mean-zeroing processing module is used to perform mean-zeroing processing on the multi-dimensional feature vector of the samples to obtain the zeroed sample multi-dimensional feature vector; The module for calculating the sample correlation feature vector and sample feature value is used to construct the sample covariance matrix based on the zeroed sample multidimensional feature vector, and to perform eigenvalue decomposition on the sample covariance matrix to obtain the sample correlation feature vector and sample feature value. The sample feature value selection module is used to select the top N sample feature values ​​and the corresponding sample correlation feature vectors based on the sorting order of sample feature values ​​from largest to smallest. The sample feature value normalization processing module is used to normalize the sample feature values ​​to obtain the sample normalized feature values. The standard multidimensional feature optimization twin coefficient calculation module is used to perform weighted summation on the sample correlation feature vector corresponding to the first N sample feature values, the sample ratio of the sample correlation feature vector magnitude to the maximum sample correlation feature vector magnitude, and the sample normalized feature values ​​to obtain the standard multidimensional feature optimization twin coefficient.

[0097] The digital twin construction apparatus provided in this application includes an acquisition module 501 for acquiring first attribute information of a target entity; the first attribute information is information used to characterize the features of the target entity; a feature vector extraction module 502 for extracting feature vectors from the first attribute information to obtain corresponding multi-dimensional feature vectors; a correlation degree quantification processing module 503 for performing correlation degree quantification processing on the multi-dimensional feature vectors to obtain correlation degree feature vectors and corresponding feature values; the correlation degree feature vectors characterize the correlation combination method of multi-dimensional features, and the feature values ​​characterize the correlation degree quantification value of the corresponding correlation degree feature vectors; and a selection module 504 for selecting the first N feature values ​​and corresponding correlation degree feature vectors based on the descending order of feature values. The weighted summation processing module 505 is used to perform weighted summation processing on the first N feature values ​​and the corresponding correlation feature vectors to obtain the multidimensional feature optimized twin coefficients; the multidimensional feature optimized twin coefficients are used to characterize the importance of the correlation between multidimensional features; the instantiation module 506 is used to instantiate the standard digital twin template corresponding to the standard multidimensional feature optimized twin coefficients, combined with the first attribute information, when the multidimensional feature optimized twin coefficients are not less than the standard multidimensional feature optimized twin coefficients corresponding to the target entity, to obtain the target digital twin of the target entity; the construction module 607 is used to construct the target digital twin of the target entity using the first attribute information when the multidimensional feature optimized twin coefficients are less than the standard multidimensional feature optimized twin coefficients corresponding to the target entity.

[0098] This application embodiment extracts the feature vector of the first attribute information to obtain a multi-dimensional feature vector, thereby transforming the original first attribute information into feature parameters that can be used for subsequent quantization processing. Then, by performing correlation degree quantization processing on this multi-dimensional feature vector, correlation degree feature vectors representing the correlation combination of multi-dimensional features are obtained, along with feature values ​​representing the correlation degree quantization values ​​of the corresponding correlation degree feature vectors. This yields the correlation patterns and corresponding degree information between multi-dimensional features. Then, following a selection order from largest to smallest, several feature values ​​and their corresponding correlation degree feature vectors are selected and weighted summed to filter out highly correlated correlation degree feature vectors and their corresponding... The feature values ​​are weighted and summed to reduce the amount of data processed, simplify the processing complexity, and reduce the impact of the signal-to-noise ratio on the results. When the twin coefficient of the multidimensional feature optimization is greater than or equal to the standard multidimensional feature optimization twin coefficient, the target digital twin is instantiated using the standard digital twin template combined with the first attribute information. The standard digital twin template can be directly called without manual configuration for the scenario, which improves the efficiency of digital twin construction. When the twin coefficient of the multidimensional feature optimization is less than the standard multidimensional feature optimization twin coefficient, the target digital twin is directly constructed using the first attribute information. Digital twins can also be deployed even without template matching, which broadens the application scenarios of this method.

[0099] Figure 6 A schematic diagram of the hardware structure of the digital twin construction device provided in an embodiment of this application is shown.

[0100] The digital twin building device may include a processor 601 and a memory 602 storing computer program instructions.

[0101] Specifically, the processor 601 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0102] Memory 602 may include mass storage for data or instructions. For example, and not limitingly, memory 602 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one instance, memory 602 may include removable or non-removable (or fixed) media, or memory 602 may be non-volatile solid-state memory. Memory 602 may be internal or external to the integrated gateway disaster recovery device.

[0103] Memory 602 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this disclosure.

[0104] The processor 601 reads and executes computer program instructions stored in the memory 602 to achieve... Figure 2 The digital twin construction method in the illustrated embodiment.

[0105] In one example, the digital twin building device may also include a communication interface 603 and a bus 604. Wherein, as Figure 6 As shown, the processor 601, memory 602, and communication interface 603 are connected through bus 604 and complete communication with each other.

[0106] The communication interface 603 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0107] Bus 604 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 304 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0108] Furthermore, in conjunction with the digital twin construction methods described in the above embodiments, this application embodiment can provide a computer storage medium for implementation. This computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the digital twin construction methods described in the above embodiments.

[0109] This application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the methods for constructing a digital twin in the above embodiments.

[0110] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0111] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, read-only memory (ROM), flash memory, erasable read-only memory (EROM), floppy disks, compact disc read-only memory (CD-ROM), optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0112] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0113] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0114] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for constructing a digital twin, characterized in that, include: Obtain the first attribute information of the target entity; The first attribute information is information used to characterize the features of the target entity; Extract the feature vector of the first attribute information to obtain the corresponding multidimensional feature vector; The multidimensional feature vectors are subjected to correlation degree quantization processing to obtain correlation degree feature vectors and corresponding feature values; the correlation degree feature vectors represent the correlation combination method of multidimensional features, and the feature values ​​represent the correlation degree quantization value of the corresponding correlation degree feature vectors; Based on the descending order of the feature values, select the first N feature values ​​and the corresponding correlation feature vectors; The first N feature values ​​and the corresponding correlation feature vectors are weighted and summed to obtain the multidimensional feature-optimized twinning coefficients. The optimized twinning coefficients of the multidimensional features are used to characterize the importance of the correlation between the multidimensional features; If the multidimensional feature optimization twin coefficient is not less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity, the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficient is instantiated in combination with the first attribute information to obtain the target digital twin of the target entity. If the multidimensional feature-optimized twin coefficient is less than the standard multidimensional feature-optimized twin coefficient corresponding to the target entity, a target digital twin of the target entity is constructed using the first attribute information.

2. The digital twin construction method according to claim 1, characterized in that, The step of constructing the target digital twin of the target entity using the first attribute information includes: Construct a specific digital twin template of the target entity based on the first attribute information; Based on the first attribute information and the specific digital twin template, an instantiation is performed to obtain the target digital twin of the target entity.

3. The digital twin construction method according to claim 2, characterized in that, After constructing a specific digital twin template of the target entity based on the first attribute information, the method further includes: Obtain the product type attribute information of the target entity; The product type attribute information and the first attribute information are input into the initialization model to obtain the initialization model weight values; The first attribute information and the configuration information of the standard digital twin template are both input into the similarity weight index model and the optimization model, respectively, and the similarity weight index and optimization correction coefficient are obtained respectively. Based on the initial model weight values, the similarity weight index, and the optimization correction coefficients, the sedimentation factor value is obtained; If the precipitation factor value is greater than or equal to the preset precipitation factor value, the specific digital twin template is updated to the latest standard digital twin template of the target entity.

4. The digital twin construction method according to claim 3, characterized in that, If the precipitation factor value is less than the preset precipitation factor value, the second attribute information is added to the first attribute information to obtain the supplemented first attribute information. Return to the step of inputting the product type attribute information and the first attribute information into the initialization model to obtain the initialization model weight value.

5. The method for constructing a digital twin according to claim 1, characterized in that, The step of performing correlation quantization on the multidimensional feature vector to obtain the correlation feature vector and the corresponding feature value includes: Construct a covariance matrix using the multidimensional feature vectors; The covariance matrix is ​​subjected to eigenvalue decomposition to obtain the correlation degree eigenvector and the eigenvalue.

6. The method for constructing a digital twin according to claim 1, characterized in that, The step of performing a weighted summation of the first N feature values ​​and the corresponding correlation feature vectors to obtain the multidimensional feature-optimized twinning coefficients includes: Obtain the normalized feature values ​​of the first N feature values, and the ratio of the magnitude of each correlation feature vector to the magnitude of the maximum correlation feature vector in the correlation feature vector corresponding to the first N feature values; The normalized eigenvalues ​​and the ratios are weighted and summed to obtain the multidimensional feature-optimized twin coefficients.

7. The method for constructing a digital twin according to claim 1, characterized in that, The method further includes: Obtain sample attribute information of entity sample data; the sample attribute information is information used to characterize the features of the entity sample data. The sample attribute information is processed by feature vector extraction to obtain the corresponding multidimensional feature vector of the sample; The mean of the sample multidimensional feature vector is normalized to zero to obtain the zeroed sample multidimensional feature vector; Based on the zeroed sample multidimensional feature vector, a sample covariance matrix is ​​constructed, and the sample covariance matrix is ​​subjected to eigenvalue decomposition to obtain the sample correlation feature vector and sample feature values. Based on the descending order of the sample feature values, the first N sample feature values ​​and the corresponding sample correlation feature vectors are selected. The sample feature values ​​are normalized to obtain sample normalized feature values; The standard multidimensional feature optimization twin coefficient is obtained by weighted summing of the sample correlation feature vectors corresponding to the first N sample feature values, the sample correlation feature vector magnitude to the maximum sample correlation feature vector magnitude, and the sample normalized feature values.

8. A digital twin construction device, characterized in that, The device includes: The acquisition module is used to acquire first attribute information of the target entity; the first attribute information is information used to characterize the features of the target entity. The feature vector extraction module is used to extract the feature vector of the first attribute information to obtain the corresponding multidimensional feature vector. The correlation degree quantification processing module is used to perform correlation degree quantification processing on the multi-dimensional feature vector to obtain the correlation degree feature vector and the corresponding feature value; the correlation degree feature vector represents the correlation combination method of the multi-dimensional features, and the feature value represents the correlation degree quantification value of the corresponding correlation degree feature vector; The selection module is used to select the first N feature values ​​and the corresponding correlation feature vectors based on the descending order of the feature values. The weighted summation processing module is used to perform weighted summation processing on the first N feature values ​​and the corresponding correlation degree feature vectors to obtain the multidimensional feature optimized twin coefficients; the multidimensional feature optimized twin coefficients are used to characterize the importance of the correlation between the multidimensional features; The instantiation module is used to instantiate the standard digital twin template corresponding to the standard multidimensional feature optimization twin coefficient, in combination with the first attribute information, when the multidimensional feature optimization twin coefficient is not less than the standard multidimensional feature optimization twin coefficient corresponding to the target entity, so as to obtain the target digital twin of the target entity. A construction module is used to construct a target digital twin of the target entity using the first attribute information when the multidimensional feature-optimized twin coefficient is less than the standard multidimensional feature-optimized twin coefficient corresponding to the target entity.

9. A digital twin construction device, characterized in that, The device includes: a processor and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the digital twin construction method as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores computer program instructions, which, when executed by a processor, implement the digital twin construction method as described in any one of claims 1-7.

Images