Digital twin modeling-oriented intention-driven key attribute selection method and device

Abstract

The invention relates to the technical field of digital twinning, in particular to an intention-driven key attribute selection method and device for digital twinning modeling. The method comprises the following steps of: aiming at an industrial digital twinning model construction problem and an attribute selection and updating strategy based on time-varying entropy and uncertainty optimization, dynamically adjusting the updating frequency of a high-entropy (remarkable change) attribute and reducing the updating frequency of a low-entropy (small change) attribute, so that the calculation burden of model construction is reduced, and the modeling efficiency is improved. By calculating the conditional mutual information of each attribute to the target variable, the attribute having a significant contribution to reduction of conditional entropy is selected, the redundant attribute is reduced, and the modeling accuracy is improved. According to the scheme, the key problems that in a traditional model building method, the attributes adopt the fixed updating frequency and modeling is conducted through all the attributes are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twins, and particularly to an intention-driven key attribute selection method and device for digital twin modeling. Background Art

[0002] Digital Twin (DT) is a technology that takes the integration and fusion of data and models as the basis and core. By constructing an accurate digital mapping of a physical object in real time in the digital space, and based on data integration, analysis, and prediction, it simulates, validates, predicts, and controls the entire life cycle process of a physical entity, and finally forms an intelligent decision-making optimization closed loop. This method associates and maps the real-time collected data to the digital twin model, thereby identifying, tracking, and monitoring physical entities, and improving the accuracy of model construction. At the same time, through the digital twin model, it predicts and analyzes the behavior of the simulation object, diagnoses and warns about faults, locates and records problems, thereby improving system efficiency, reducing computational overhead, and achieving data-driven optimization decisions. Digital twins are continuously penetrating the industrial field and expanding to vertical industries such as transportation, health care, etc., realizing application values such as mechanism description, anomaly diagnosis, risk prediction, and decision assistance, and are expected to become a general-purpose technology for the digital transformation of the economic and social industries in the future.

[0003] The rapid development of the integration of digital twins and industry can significantly improve system efficiency and accuracy, reduce computational overhead, and provide strong support for industrial digital transformation. DT provides a platform for the testing and analysis of complex systems, where the digital / virtual counterparts of real-world systems can provide real-time monitoring, analysis, evaluation, and prediction [S. Mihai et al., "Digital Twins: A Survey on Enabling Technologies, Challenges, Trends and Future Prospects," in IEEE Communications Surveys & Tutorials, vol. 24, no. 4, pp. 2255-2291, Fourth quarter 2022.]. The digital twin model is an important part of digital twins and an important prerequisite for realizing digital twin functions. It can not only accurately simulate the behavior and state of physical entities, but also provide optimization suggestions, conduct fault prediction, and improve equipment health management by analyzing virtual models. With the help of digital twin models, industrial enterprises can achieve intelligent management, improve system efficiency, and enhance the accuracy and real-time nature of decision-making.

[0004] However, due to the development of sensing technology and intelligent technology, as well as the dynamic changes in the operating environment of complex equipment, the amount of real-time monitoring data of equipment has doubled, presenting typical characteristics of industrial big data such as high speed, multi-source heterogeneity, and volatility [T. Mortlock, D. Muthirayan, S. -Y. Yu, P. P. Khargonekar and M. Abdullah AlFaruque, "Graph Learning for Cognitive Digital Twins in ManufacturingSystems," in IEEE Transactions on Emerging Topics in Computing, vol. 10, no.1, pp. 34-45, 1 Jan.-March 2022.]. In the process of building a digital twin model, the above situation leads to the vastness and complexity of real-time data, making real-time data acquisition a huge challenge. If one attempts to build a model for all attributes of the modeling object, it will not only consume a large amount of computing resources but also may cause a sharp increase in the complexity of the model, thus reducing efficiency and response speed. Especially in the case of inconsistent data quality or transmission delay, how to efficiently and accurately process this vast amount of data to ensure the efficiency and accuracy of the model has become a key challenge in digital twin applications.

[0005] To address the challenge of difficult real-time data acquisition caused by vast and complex data, the present invention proposes a method for intention-driven key attribute selection. Specifically, for each modeling object, based on the evaluation of uncertainty and attribute contribution, attributes that significantly contribute to reducing conditional entropy are selected, and then targeted modeling is performed on these attributes, which is beneficial to improving the efficiency and accuracy of the model. This method of attribute selection usually realizes the deep feature extraction and modeling of system data by using intelligent algorithms combined with some technical methods in the fields of machine learning and artificial intelligence [Z. Zheng, J. Oh, M. Hessel, Z. Xu, M. Kroiss, H. Van Hasselt, D. Silver, and S. Singh, “What can learnedintrinsic rewards capture?” in Proc. International Conference on MachineLearning (ICML), 2020, p. 11436–11446.], and adopts a multi-scale and multi-model method to analyze the sensing data at multiple levels and scales, mine and learn the relevant relationships, logical relationships, and main features contained therein, so as to realize the representation and modeling of the super-real state of the system.

[0006] However, these feature extraction methods, such as Principal Component Analysis (PCA) or Convolutional Neural Network (CNN), may require large computational resources. Especially when dealing with large-scale data, they may cause computational bottlenecks and reduce the model construction efficiency. Moreover, for complex tasks, automatic feature extraction requires more data and time for training and tuning, increasing the complexity of industrial production systems. In addition, the prior art does not fully consider the characteristics of the modeled object's attributes changing over time. Usually, a fixed update frequency is adopted for attribute updates, and the dynamic adjustment of the change amplitude of time-varying attributes at different time points is not carried out. For high-entropy attributes, frequent updates can better capture the dynamic behavior of the system, while for low-entropy attributes, the update frequency can be reduced to save computational resources. The existing methods fail to consider the different requirements of different attributes, which may lead to low computational efficiency.

[0007] In recent years, the continuous development of digital twins has been widely applied in various fields, effectively addressing the computational pressure on digital twin models brought about by massive data. How to reduce the computational overhead and improve the efficiency of the model while ensuring high accuracy and high real-time performance remains an important challenge in the field of digital twins. The literature [H. Chai, H. Wang, T. Li and Z. Wang, "Generative AI-Driven Digital Twin for Mobile Networks," in IEEE Network, vol. 38, no. 5, pp. 84-92, Sept. 2024.] proposed a digital twin paradigm for mobile networks driven by Generative Artificial Intelligence (GAI). As a key enabling factor for generating DT data, GAI can implicitly learn the complex distribution of network data to obtain high-fidelity data. This system can generate high-fidelity digital twin data and provide practical network optimization solutions, enhancing the flexibility and accuracy of network design. The literature [B. Zhang, M. Zhang, T. Dong, M. Lu and H. Li, "Design of Digital Twin System for DC Contactor Condition Monitoring," in IEEE Transactions on Industry Applications, vol. 59, no. 4, pp. 3904-3909, July-Aug. 2023.] proposed a design scheme for a digital twin system for DC contactors, which describes and defines the model through three aspects: geometry, mechanics, and data. This high-fidelity modeling method for digital twins can effectively monitor the device status and evaluate its lifespan.The literature [D. Gautam, G. Thakur, P. Kumar, A. K. Das and Y. Park, "Blockchain Assisted Intra-Twin and Inter-Twin Authentication Scheme for Vehicular Digital Twin System," in IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 10, pp. 15002-15015, Oct. 2024.] proposed a blockchain-based authentication framework to enhance the security of communication and the verifiability of data in the vehicular digital twin network. By integrating blockchain technology, this framework effectively improves the computational and communication efficiency of the system while ensuring data security and privacy protection.

[0008] Existing solutions adopt a fixed update frequency for modeling attributes, ignoring the variation amplitude of time-varying attributes at different time points. For high-entropy attributes, frequent updates may better reflect the dynamic behavior of the system, while low-entropy attributes do not require frequent updates. The existing methods do not consider the requirements of different attributes, which may lead to low computational efficiency.

[0009] During the model construction process, modeling all attributes will increase the computational and storage overhead, and result in an extended real-time model construction time and reduced accuracy. Summary of the Invention

[0010] To solve the technical problems that existing solutions adopt a fixed update frequency for modeling attributes, ignoring the variation amplitude of time-varying attributes at different time points, leading to low computational efficiency, and modeling all attributes will increase the computational and storage overhead, and result in an extended real-time model construction time and reduced accuracy, the embodiments of the present invention provide an intention-driven key attribute selection method and device for digital twin modeling. The technical solutions are as follows:

[0011] On the one hand, an intention-driven key attribute selection method for digital twin modeling is provided, characterized in that the method includes:

[0012] S1. Design different update strategies according to the change of time-varying entropy;

[0013] S2. Calculate the uncertainty of each modeling object within a time slot;

[0014] S3. Evaluate the contribution of the attributes of each modeling object to the target variable based on the uncertainty;

[0015] S4. Optimize the attribute selection criteria and uncertainty;

[0016] S5. Preset resource and modeling cost limitations, select an attribute set that meets the requirements, and complete the intention-driven key attribute selection for digital twin modeling.

[0017] Optionally, in S1, design different update strategies according to the change of time-varying entropy, including:

[0018] For time-varying attributes with high entropy, adopt a frequent update strategy; among them, high-entropy attributes are attributes with significant changes;

[0019] For time-varying attributes with low entropy, adopt a sparse update strategy; among them, low-entropy attributes are stable or less-changing attributes.

[0020] Optionally, in S2, calculate the uncertainty and the number of attribute contributions of each modeling object within a time slot, including:

[0021] Calculate each modeling object according to the following formula (1) within the time slot uncertainty:

[0022] (3)

[0023] where is the entropy of the th modeling object, is a stochastic process representing its multi-dimensional attributes, is the probability distribution.

[0024] Optionally, in S3, evaluate the contribution of the attributes of each modeling object to the target variable based on uncertainty, including:

[0025] Calculate the conditional mutual information between each attribute and the target variable according to the following formula (2):

[0026]

[0027] where is the conditional mutual information;

[0028] Select attributes that make a significant contribution to reducing the conditional entropy;

[0029] Determine the importance of attributes according to the magnitude of the conditional mutual information, and reduce redundant attributes.

[0030] Optionally, in S3, optimize the attribute selection criteria and uncertainty, including:

[0031] Set the optimization goal to minimize the conditional entropy. According to the calculated uncertainty and the number of attribute contributions of each modeling object within a time slot, the optimization problem is as shown in the following formula (3):

[0032] (3)

[0033] where, is the entropy of the -th modeling object, is the set of selected attributes, denoted by ; in the constraint conditions, represents the modeling cost related to the -th attribute of the -th modeling object. The total modeling cost cannot exceed the maximum allowed energy consumption or resource limit, denoted as .

[0034] Optionally, in S5, preset the resource and modeling cost limits, and select the set of attributes that meet the requirements, including:

[0035] Initialize the candidate attribute set as the complete attribute set of the modeling object;

[0036] According to the maximum allowed modeling cost, check one by one whether the attributes meet the cost limit;

[0037] Select the attribute with the highest conditional mutual information and add it to the selected attribute set;

[0038] Update the total modeling cost until all eligible attributes are selected.

[0039] On the other hand, an intention-driven key attribute selection device for digital twin modeling is provided. The device is applied to the intention-driven key attribute selection method for digital twin modeling. The device includes:

[0040] An update strategy design module, configured to design different update strategies according to the change of time-varying entropy;

[0041] An uncertainty calculation module, configured to calculate the uncertainty of each modeling object within a time slot;

[0042] An evaluation module, configured to evaluate the contribution of the attributes of each modeling object to the target variable based on the uncertainty;

[0043] An optimization module, configured to optimize the attribute selection criteria and uncertainty;

[0044] An attribute set selection module, configured to preset the resource and modeling cost limits, select the set of attributes that meet the requirements, and complete the intention-driven key attribute selection for digital twin modeling.

[0045] Optionally, the update strategy design module is further configured to adopt a frequent update strategy for time-varying attributes with high entropy, where the high-entropy attributes are attributes with significant changes.

[0046] For time-varying attributes with low entropy, a sparse update strategy is adopted, where the low-entropy attributes are attributes that are stable or have small changes.

[0047] On the other hand, there is provided an intention-driven key attribute selection device for digital twin modeling. The intention-driven key attribute selection device for digital twin modeling includes: a processor; a memory, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor, any of the methods in the above-mentioned intention-driven key attribute selection method for digital twin modeling is implemented.

[0048] On the other hand, there is provided a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is loaded and executed by a processor to implement any of the methods in the above-mentioned intention-driven key attribute selection method for digital twin modeling.

[0049] The beneficial effects brought by the technical solutions provided in the embodiments of the present invention at least include:

[0050] In the embodiments of the present invention, for the problem of model construction for industrial digital twins, based on the attribute selection and update strategy optimized by time-varying entropy and uncertainty, on the one hand, by dynamically adjusting the update frequency of high-entropy (significantly changing) attributes, the computational burden of low-entropy (small-changing) attributes is reduced, thereby improving the modeling efficiency. On the other hand, by calculating the conditional mutual information of each attribute with respect to the target variable, attributes that significantly contribute to reducing the conditional entropy are selected to optimize the model accuracy. At the same time, considering the modeling cost and resource limitations, it is ensured that the model maintains a reasonable balance in terms of computation and energy consumption. This strategy not only improves the flexibility and prediction accuracy of the model, but also effectively avoids excessive computation and optimizes resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0052] Figure 1 It is a flowchart showing a method for intention-driven key attribute selection for digital twin modeling provided by an embodiment of the present invention;

[0053] Figure 2Framework diagram of the intention-driven key attribute selection method for industrial-oriented digital twin model construction provided by an example of the present invention;

[0054] Figure 3 Algorithm program flow chart of the intention-driven key attribute selection method for industrial-oriented digital twin model construction provided by an example of the present invention;

[0055] Figure 4 Block diagram of the intention-driven key attribute selection device for digital twin modeling provided by an embodiment of the present invention;

[0056] Figure 5 Structure schematic diagram of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0057] Next, the technical solutions in the present invention will be described with reference to the accompanying drawings.

[0058] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0059] In the embodiments of the present invention, sometimes subscripts such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0060] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0061] The embodiments of the present invention provide an intention-driven key attribute selection method for digital twin modeling. This method can be implemented by an intention-driven key attribute selection device for digital twin modeling, and the intention-driven key attribute selection device for digital twin modeling can be a terminal or a server. As Figure 1 shown in the flow chart of the intention-driven key attribute selection method for digital twin modeling, as Figure 1 shown, the intention-driven key attribute selection method proposed by the present invention, the processing flow of this method can include the following steps:

[0062] S1. Design different update strategies according to the change of time-varying entropy;

[0063] In a feasible implementation manner, in S1, designing different update strategies according to the change of time-varying entropy includes:

[0064] For time-varying attributes with high entropy, a frequent update strategy is adopted; among them, high-entropy attributes are attributes with significant changes.

[0065] For time-varying attributes with low entropy, a sparse update strategy is adopted; among them, low-entropy attributes are stable or slightly changing attributes.

[0066] In a feasible implementation, as Figure 2 shown, different update strategies are designed according to the change of time-varying entropy: ① Frequent update of high-entropy attributes: For time-varying attributes with high entropy, that is, attributes with significant changes, they should be updated more frequently. These features contain more information and are crucial for accurately modeling system behavior. ② Sparse update of low-entropy attributes: For time-varying attributes with low entropy, that is, stable or slightly changing attributes, the update frequency can be reduced. These attributes have less impact on the model and do not need to be updated frequently, which helps to reduce computational overhead. Figure 1 is a framework diagram of an intention-driven key attribute selection method for industrial digital twin model construction, which includes an update strategy based on time-varying entropy and the following content of key attribute selection.

[0067] In a feasible implementation, assume that for a digital twin service request for regional traffic control, the selected modeling object set is , including the specific quantitative information of road networks, vehicle flows, traffic signal systems, environmental factors, and traffic events, as well as their historical data. Since there may be correlations and redundancies between attributes, the next step is to select modeling attributes to improve the efficiency of these modeling objects. Assume .

[0068] S2. Calculate the uncertainty of each modeling object within a time slot;

[0069] In a feasible implementation, in S2, calculate the uncertainty and the number of attribute contributions of each modeling object, including;

[0070] Calculate the uncertainty of each modeling object within the time slot according to the following formula (1):

[0071]

[0072] where, is the entropy of the th modeling object, is a stochastic process representing its multi-dimensional attributes, is the probability distribution.

[0073] S3. Evaluate the contribution of the attributes of each modeling object to the target variable based on the uncertainty;

[0074] In a feasible implementation, in S3, based on the uncertainty, evaluate the contribution of the attributes of each modeling object to the target variable, including:

[0075] Calculate the conditional mutual information between each attribute and the target variable according to the following formula (2):

[0076]

[0077] Where is the conditional mutual information, which measures the additional information contributed by the attribute to the target variable given the attribute . In other words, it quantifies how much the uncertainty about is reduced given the knowledge of the other attributes after adding .

[0078] Select the attributes that significantly contribute to reducing the conditional entropy;

[0079] Determine the importance of the attributes according to the magnitude of the conditional mutual information and reduce the redundant attributes.

[0080] S4. Optimize the attribute selection criteria and uncertainty;

[0081] In a feasible implementation, in S4, optimizing the attribute selection criteria and uncertainty includes:

[0082] During the attribute selection process, the goal of this project is to select the attributes that significantly contribute to reducing the conditional entropy , which represents the uncertainty in predicting the target variable given the selected attribute set . Set the optimization goal to minimize the conditional entropy. According to the calculated uncertainty of each modeling object within the time slot and the number of attribute contributions, the optimization problem is as shown in the following formula (3):

[0083] (3)

[0084] Where is the entropy of the th modeling object, is the set of selected attributes, denoted by ; in the constraint condition, represents the modeling cost related to the th attribute of the th modeling object. The total modeling cost cannot exceed the maximum allowable energy consumption or resource limit, denoted as 。

[0085] S5, Preset resources and modeling cost constraints, select the attribute set that meets the requirements, and complete the intent-driven key attribute selection for digital twin modeling. The preset resources depend on the latency and accuracy requirements of service requests, system hardware capabilities, application scenario priorities, data time-varying entropy characteristics, budget constraints, and external environmental factors. The modeling cost is subject to the energy consumption in processes such as computing, transmission, and storage, and the preset resources.

[0086] In a feasible implementation, S5, preset resources and modeling cost constraints, select the attribute set that meets the requirements, including:

[0087] Initialize the candidate attribute set as the complete attribute set of the modeling object;

[0088] According to the maximum allowable modeling cost, check one by one whether the attributes meet the cost constraints;

[0089] Select the attribute with the highest conditional mutual information and add it to the selected attribute set;

[0090] Update the total modeling cost until all eligible attributes are selected.

[0091] In a feasible implementation, in the attribute selection process during modeling, first accept two inputs: the number of attributes of the th modeling object and the maximum allowable modeling cost (step 1). This process initializes the candidate attribute set as the complete attribute set of the modeling object , the selected attribute set is empty, and the current modeling cost and are respectively initialized to zero and an empty set (step 3.a). Then, check the attributes one by one from the candidate set to determine whether their inclusion will exceed the maximum cost . If so, the attribute is removed from ; otherwise, according to equation (2), calculate its conditional mutual information with the target state (step 3.b.c-1). Then, the attribute with the highest conditional mutual information can be found and added to the selected set , and at the same time removed from (steps 4-5). The total modeling cost is updated accordingly (step 6). This process continues until all eligible attributes are selected, and finally the selected attribute set is returned (step 7), and the flowchart is asFigure 3 as shown

[0092] The algorithmic program pseudocode process of the key attribute selection method driven by intention for constructing a digital twin model for industry is as follows:

[0093] 1) Input: the number of attributes of the th modeling object and the maximum allowable modeling cost ;

[0094] 2) Initialization: the candidate attribute set is the complete attribute set of the modeling object , the selected attribute set is empty, and the current modeling cost is zero;

[0095] 3) For i = 1 to execute the loop

[0096] a) Initialization: the temporary variables and are initialized to zero and an empty set respectively;

[0097] b) For to execute the loop

[0098] a) Check the attribute one by one from the candidate set ;

[0099] b) If , then the attribute is removed from ; continue;

[0100] c) Otherwise

[0101] c-1) Calculate the conditional mutual information between and the target state according to formula (2);

[0102] c-2) If , assign to , assign to ;

[0103] 4) Add to the selected set ;

[0104] 5) Remove from ;

[0105] 6) Update the current total modeling cost to ;

[0106] Output: Attribute set .

[0107] In the embodiments of the present invention, a new framework of intention-driven key attributes for digital twin modeling is proposed. For the problem of model construction for industrial digital twins, based on the attribute selection and update strategy optimized by time-varying entropy and uncertainty, by dynamically adjusting the update frequency of high-entropy (significantly changing) attributes and reducing the update frequency of low-entropy (little-changing) attributes, the computational burden of model construction is reduced and the modeling efficiency is improved. By calculating the conditional mutual information of each attribute with respect to the target variable, attributes that significantly contribute to reducing the conditional entropy are selected, redundant attributes are reduced, and the modeling accuracy is improved. The proposed solution solves the key problems in traditional model construction methods, where attributes use a fixed update frequency and all attributes are modeled.

[0108] For the problem of model construction for industrial digital twins, based on the attribute selection and update strategy optimized by time-varying entropy and uncertainty, on the one hand, by dynamically adjusting the update frequency of high-entropy (significantly changing) attributes, the computational burden of low-entropy (little-changing) attributes is reduced, thereby improving the modeling efficiency. On the other hand, by calculating the conditional mutual information of each attribute with respect to the target variable, attributes that significantly contribute to reducing the conditional entropy are selected to optimize the model accuracy. At the same time, considering the modeling cost and resource constraints, it is ensured that the model maintains a reasonable balance in terms of computation and energy consumption. This strategy not only improves the flexibility and prediction accuracy of the model, but also effectively avoids over-computation and optimizes resource utilization.

[0109] Figure 4 is a block diagram of an intention-driven key attribute selection device 300 for digital twin modeling shown according to an exemplary embodiment. The device 300 is used for the intention-driven key attribute selection method for digital twin modeling. Referring to Figure 4 , the device includes an update strategy design module 310, an uncertainty calculation module 320, an evaluation module 330, an optimization module 340, and an attribute set selection module 350. Among them:

[0110] The update strategy design module 310 is used to design different update strategies according to the change of time-varying entropy;

[0111] The uncertainty calculation module 320 is used to calculate the uncertainty of each modeling object within a time slot;

[0112] The evaluation module 330 is used to evaluate the contribution of the attributes of each modeling object to the target variable based on the uncertainty;

[0113] The optimization module 340 is used to optimize the attribute selection criteria and uncertainty;

[0114] The attribute set selection module 350 is used to preset resources and modeling cost limits, select an attribute set that meets the requirements, and complete the intention-driven key attribute selection for digital twin modeling.

[0115] Optionally, the update strategy design module 310 is further used to adopt a frequent update strategy for time-varying attributes with high entropy; among them, high-entropy attributes are attributes with significant changes;

[0116] For time-varying attributes with low entropy, a sparse update strategy is adopted; among them, low-entropy attributes are stable or less-changing attributes.

[0117] Optionally, the uncertainty calculation module 320 is used to calculate the uncertainty of each modeling object in the time slot according to the following formula (1):

[0118]

[0119] where is the entropy of the th modeling object, is a stochastic process representing its multi-dimensional attributes, is the probability distribution.

[0120] Optionally, the evaluation module 330 is used to calculate the conditional mutual information between each attribute and the target variable according to the following formula (2):

[0121]

[0122] where is the conditional mutual information;

[0123] Select attributes that make a significant contribution to reducing the conditional entropy;

[0124] Determine the importance of attributes according to the magnitude of the conditional mutual information, and reduce redundant attributes.

[0125] Optionally, the optimization module 340 is used to set the optimization goal to minimize the conditional entropy. According to the calculated uncertainty of each modeling object in the time slot and the number of attribute contributions, the optimization problem is as shown in the following formula (3):

[0126] (3)

[0127] where is the entropy of the th modeling object, is the set of selected attributes, denoted by Indicates; in the constraints, Indicates The first modeling object The total modeling cost cannot exceed the maximum allowed energy consumption or resource limit, denoted as .

[0128] Optionally, the attribute set selection module 350 is used to initialize the candidate attribute set to be a complete attribute set of the modeling object;

[0129] According to the maximum allowed modeling cost, check whether the attributes meet the cost constraints one by one;

[0130] Select the attribute with the highest conditional mutual information and add it to the set of selected attributes;

[0131] Updates the total modeling cost until all eligible attributes are selected.

[0132] In an embodiment of the present invention, a new framework of intent-driven key attributes for digital twin modeling is proposed. Aiming at the model building problem of industrial digital twins, an attribute selection and update strategy based on time-varying entropy and uncertainty optimization is used to dynamically adjust the update frequency of high entropy (significant changes) attributes and reduce the update frequency of low entropy (small changes) attributes, thereby reducing the computational burden of model building and improving modeling efficiency. By calculating the conditional mutual information of each attribute to the target variable, attributes that contribute significantly to reducing conditional entropy are selected, redundant attributes are reduced, and modeling accuracy is improved. The proposed solution solves the key problems of fixed update frequency and modeling of all attributes in traditional model building methods.

[0133] For the model building problem of industrial digital twins, the attribute selection and update strategy based on time-varying entropy and uncertainty optimization, on the one hand, by dynamically adjusting the update frequency of high entropy (significant changes) attributes, the computational burden of low entropy (small changes) attributes is reduced, thereby improving modeling efficiency. On the other hand, by calculating the conditional mutual information of each attribute on the target variable, the attributes that have significantly contributed to reducing the conditional entropy are selected to optimize the model accuracy. At the same time, considering the modeling cost and resource constraints, ensure that the model maintains a reasonable balance in computing and energy consumption. This strategy not only improves the flexibility and prediction accuracy of the model, but also effectively avoids over-computation and optimizes resource utilization.

[0134] Figure 5 is a structural diagram of an intent-driven key attribute selection device for digital twin modeling provided by an embodiment of the present invention, such as Figure 5 As shown, the intent-driven key attribute selection device for digital twin modeling can include the above Figure 4The intention-driven key attribute selection device for digital twin modeling as shown. Optionally, the intention-driven key attribute selection device 410 for digital twin modeling may include a first processor 2001.

[0135] Optionally, the intention-driven key attribute selection device 410 for digital twin modeling may further include a memory 2002 and a transceiver 2003.

[0136] Among them, the first processor 2001 is connected to the memory 2002 and the transceiver 2003, for example, through a communication bus.

[0137] Next, in combination with Figure 5 Specific introductions will be made to each component of the intention-driven key attribute selection device 410 for digital twin modeling:

[0138] Among them, the first processor 2001 is the control center of the intention-driven key attribute selection device 410 for digital twin modeling, which can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or can be an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention, such as: one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0139] Optionally, the first processor 2001 can execute various functions of the intention-driven key attribute selection device 410 for digital twin modeling by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0140] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 5 the CPU0 and CPU1 shown in

[0141] In a specific implementation, as an embodiment, the intention-driven key attribute selection device 410 for digital twin modeling may also include multiple processors, such as Figure 5The first processor 2001 and the second processor 2004 shown in []. Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here can refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0142] Among them, the memory 2002 is used to store the software program for implementing the solution of the present invention and is controlled by the first processor 2001 for execution. The specific implementation method can refer to the above method embodiment and will not be elaborated here.

[0143] Optionally, the memory 2002 can be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic storage media or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 can be integrated with the first processor 2001 or exist independently and is coupled to the first processor 2001 through an interface circuit of the intent-driven key attribute selection device 410 for digital twin modeling ( Figure 5 not shown in []). The embodiments of the present invention do not make specific limitations on this.

[0144] The transceiver 2003 is used to communicate with a network device or with a terminal device.

[0145] Optionally, the transceiver 2003 can include a receiver and a transmitter ( Figure 5 not shown separately in []). Among them, the receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0146] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or exist independently and is coupled to the first processor 2001 through an interface circuit of the intent-driven key attribute selection device 410 for digital twin modeling ( Figure 5 not shown in []). The embodiments of the present invention do not make specific limitations on this.

[0147] It should be noted that Figure 5 The structure of the intention-driven key attribute selection device 410 for digital twin modeling shown in does not limit the router. The actual knowledge structure recognition device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0148] In addition, for the technical effects of the intention-driven key attribute selection device 410 for digital twin modeling, reference may be made to the technical effects of the intention-driven key attribute selection method for digital twin modeling described in the above method embodiments, which will not be elaborated here.

[0149] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0150] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0151] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable sensors. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be solid-state drives.

[0152] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context.

[0153] It should be understood that in various embodiments of the present invention, the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0154] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this document can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0155] The unit described as a separation component may or may not be physically separated. The component presented as a unit may or may not be a physical unit, that is, it may be located in one place or may be distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the objectives of the solution of this embodiment.

[0156] In addition, each functional unit in various embodiments of the present invention may be integrated into a processing unit, may exist as individual physical units, or two or more units may be integrated into one unit.

[0157] If the described functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention.

[0158] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claimed rights.

Claims

1. An intention-driven key attribute selection method for digital twin modeling, characterized in that, The method includes: S1. Design different update strategies according to the change of time-varying entropy; S2. Calculate the uncertainty of each modeling object within a time slot; S3. Evaluate the contribution of the attributes of each modeling object to the target variable based on the uncertainty; S4. Optimize the attribute selection criteria and uncertainty; S5. Preset resource and modeling cost limitations, select an attribute set that meets the requirements, and complete the intention-driven key attribute selection for digital twin modeling.

2. The intent-driven key attribute selection method for digital twin modeling according to claim 1, characterized in that In S1, designing different update strategies according to the change of time-varying entropy includes: For time-varying attributes with high entropy, adopt a frequent update strategy; among them, high-entropy attributes are attributes with significant changes; For time-varying attributes with low entropy, adopt a sparse update strategy; among them, low-entropy attributes are stable or slightly changing attributes.

3. The intent-driven key attribute selection method for digital twin modeling according to claim 1, characterized in that In S2, calculating the uncertainty and the number of attribute contributions of each modeling object within a time slot includes: Calculate each modeling object according to the following formula (1) In the time slot The uncertainty within: ； Among them, is the entropy of the th modeling object, is a random process representing its multidimensional attributes, is the probability distribution.

4. The intent-driven key attribute selection method for digital twin modeling according to claim 2, characterized in that In S3, evaluating the contribution of the attributes of each modeling object to the target variable based on the uncertainty includes: Calculate the conditional mutual information between each attribute and the target variable according to the following formula (2): ； Among them, is conditional mutual information; Select attributes that make a significant contribution to reducing the conditional entropy; Determine the importance of attributes according to the magnitude of the conditional mutual information, and reduce redundant attributes.

5. The method for selecting key attributes driven by intention for digital twin modeling according to claim 2, characterized in that In S3, optimizing the attribute selection criteria and uncertainty includes: Set the optimization goal to minimize the conditional entropy. According to the calculated uncertainty and the number of attribute contributions of each modeling object within a time slot, the optimization problem is shown in the following formula (3): （3） Among them, is the entropy of the th modeling object, is the set of selected attributes, denoted by ; in the constraint conditions, represents the modeling cost related to the th attribute of the th modeling object. The total modeling cost cannot exceed the maximum allowable energy consumption or resource limit, denoted as .

6. The method for selecting key attributes driven by intention for digital twin modeling according to claim 4, characterized in that S5. Preset resource and modeling cost limitations, select an attribute set that meets the requirements, including: Initialize the candidate attribute set as the complete attribute set of the modeling object; According to the maximum allowable modeling cost, check whether each attribute meets the cost limitation one by one; Select the attribute with the highest conditional mutual information and add it to the selected attribute set; Update the total modeling cost until all eligible attributes are selected.

7. An intent-driven key attribute selection device for digital twin modeling, where the intent-driven key attribute selection device for digital twin modeling is used to implement the intent-driven key attribute selection method for digital twin modeling according to any one of claims 1-6, characterized in that The device includes: An update strategy design module for designing different update strategies according to the change of time-varying entropy; An uncertainty calculation module for calculating the uncertainty of each modeling object within a time slot; An evaluation module for evaluating the contribution of the attributes of each modeling object to the target variable based on the uncertainty; An optimization module for optimizing the attribute selection criteria and uncertainty; An attribute set selection module for presetting resource and modeling cost limitations, selecting an attribute set that meets the requirements, and completing the intention-driven key attribute selection for digital twin modeling.

8. The intent-driven key attribute selection device for digital twin modeling according to claim 7, characterized in that The update strategy design module is further configured to adopt a frequent update strategy for time-varying attributes with high entropy; among them, high-entropy attributes are attributes with significant changes; For time-varying attributes with low entropy, adopt a sparse update strategy; among them, low-entropy attributes are stable or slightly changing attributes.

9. An intention-driven key attribute selection device for digital twin modeling, the intention-driven key attribute selection device for digital twin modeling includes: A processor; A memory, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by the processor, any one of the methods in the intention-driven key attribute selection method for digital twin modeling according to any one of claims 1-6 is implemented.

10. A computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is loaded and executed by a processor to implement any one of the methods in the intention-driven key attribute selection method for digital twin modeling according to any one of claims 1-6.

Images