Modeling method and device of digital twin model

Abstract

The invention discloses a modeling method and device for a digital twin model, and is applied to the technical field of digital twin models, and the method comprises the steps: obtaining the operation characteristic data of a plurality of heterogeneous terminals; according to the operation characteristic data of each heterogeneous terminal, selecting a plurality of heterogeneous terminals participating in local training from the plurality of heterogeneous terminals as training terminals; controlling each training terminal to carry out local training, and carrying out asynchronous global updating or synchronous global updating on global model parameters of the digital twin model during the local training period; and if all the training terminals complete local training, constructing the digital twinborn model according to the global model parameters of the digital twinborn model. According to the method, the modeling efficiency and the modeling precision of the digital twin model can be effectively improved, and the method can be widely applied to industrial intelligent manufacturing scenes.

Description

Technical Field

[0001] The present application relates to the technical field of digital twin models, and in particular to a modeling method and device for digital twin models. Background Art

[0002] A digital twin model refers to a virtual model created by simulating equipment, systems or processes in the physical world, which can comprehensively and accurately reflect the various properties and characteristics of the physical entity.

[0003] In related technologies, data-driven technologies such as synchronous federated learning and asynchronous federated learning are used to build digital twin models.

[0004] However, the relevant technology has not fully taken into account the heterogeneity of various industrial heterogeneous terminals, and is unable to select an industrial heterogeneous terminal that is more suitable for model modeling from several industrial heterogeneous terminals. The modeling efficiency of its digital twin model needs to be improved; in addition, the relevant technology still has inherent lag effects and model obsolescence problems, which will lead to unbalanced training costs and model losses. The modeling accuracy of its digital twin model needs to be improved. Summary of the invention

[0005] The embodiments of the present application provide a modeling method and device for a digital twin model, which are used to improve the modeling efficiency and modeling accuracy of the digital twin model.

[0006] On the one hand, an embodiment of the present application provides a modeling method of a digital twin model, the method comprising the following steps: Obtaining operational characteristic data of several heterogeneous terminals; According to the operation characteristic data of each of the heterogeneous terminals, selecting a plurality of the heterogeneous terminals participating in the local training as training terminals from among the plurality of the heterogeneous terminals; Controlling each of the training terminals to perform local training, and performing asynchronous global update or synchronous global update on the global model parameters of the digital twin model during the local training; If all the training terminals have completed local training, the digital twin model is constructed according to the global model parameters of the digital twin model.

[0007] On the other hand, an embodiment of the present application provides a modeling device for a digital twin model, the device comprising: An acquisition module, used to acquire operation characteristic data of a number of heterogeneous terminals; A first processing module, configured to select, from among a plurality of the heterogeneous terminals, a plurality of the heterogeneous terminals participating in the local training as training terminals according to the operation characteristic data of each of the heterogeneous terminals; A second processing module is used to control each of the training terminals to perform local training, and to perform asynchronous global update or synchronous global update on the global model parameters of the digital twin model during the local training; The third processing module is used to construct the digital twin model according to the global model parameters of the digital twin model if all the training terminals have completed local training.

[0008] According to the modeling method and device of the digital twin model of the present application, the heterogeneity of each heterogeneous terminal is characterized by the operating characteristic data of each heterogeneous terminal. While fully considering the heterogeneity of each heterogeneous terminal, a heterogeneous terminal that is more suitable for digital twin model modeling is selected from several heterogeneous terminals to participate in local training. In this way, the degree of adaptation between the heterogeneous terminals and the digital twin model modeling is improved, and the modeling efficiency of the digital twin model is improved. In addition, during the local training of each training terminal, the global model parameters of the digital twin model are asynchronously updated globally or synchronously updated globally to balance the training cost and model loss of the digital twin model. In this way, it can be ensured that the digital twin model can respond quickly to a dynamically changeable environment, enhance the data consistency of the virtual and real space, and improve the modeling accuracy of the digital twin model.

[0009] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a flow chart of the modeling method of the digital twin model provided in this application; Figure 2 It is a schematic diagram of the modeling method of the digital twin model provided in this application; Figure 3 It is a structural diagram of the modeling device of the digital twin model provided in this application. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0012] The present application is further described below in conjunction with the accompanying drawings and specific embodiments. The described embodiments should not be regarded as limiting the present application, and all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.

[0013] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0014] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0015] In the industrial intelligent manufacturing scenario, each industrial heterogeneous terminal is connected to the Industrial Internet of Things (IIoT). The IIoT can collect the operating data of the industrial heterogeneous terminals in real time, and use the operating data of the industrial heterogeneous terminals in combination with the Digital Twin (DT) technology to build a digital twin model of the industrial heterogeneous terminal. The digital twin model can reflect the operating status of the industrial heterogeneous terminal in real time, and can be extended to applications such as real-time monitoring and fault detection. For example, the operating status of the industrial heterogeneous terminal can be obtained through the digital twin model, and then the operating status of the industrial heterogeneous terminal can be fault-detected using a preset fault detection model to obtain a result that characterizes whether the industrial heterogeneous terminal is faulty. This is conducive to reducing the failure risk of the industrial heterogeneous terminal and facilitating preventive equipment maintenance.

[0016] It should be noted that industrial heterogeneous terminals refer to industrial terminals with different structures or functions, and industrial heterogeneous terminals can be flexibly set according to actual industrial intelligent manufacturing. Exemplarily, industrial heterogeneous terminals may include CNC machine tool equipment, industrial robot equipment, intelligent sensing detection equipment, intelligent logistics equipment, intelligent casting equipment, and intelligent stamping equipment, but are not limited to this. Among them, the intelligent sensing detection equipment is equipped with multiple sensors, which can collect various data in the production process in real time and transmit them to other terminals through a wireless network to achieve precise control of industrial intelligent manufacturing; the intelligent logistics equipment is equipped with multiple wireless radio frequency identification (RFID) modules, multiple automatic guided vehicles (AGV) and controllers, which can control the transportation of each automatic guided vehicle through the controller combined with multiple wireless radio frequency modules to achieve rapid sorting, storage and management of goods; intelligent casting equipment is used for automated industrial casting, and intelligent stamping equipment is used for automated industrial stamping.

[0017] It is understandable that the industrial Internet of Things technology uses core Internet of Things technologies such as communication technology, cloud computing, and big data analysis to connect industrial equipment, sensors, control systems, data acquisition equipment, etc. through the Internet or local area network, thereby realizing the interconnection and intelligent management of equipment. In addition, the Digital Twin (DT) model accurately simulates the equipment, system or process in the physical world through digital means, creating a virtual model that completely corresponds to it, which can fully and accurately reflect the various attributes and characteristics of the physical entity.

[0018] At present, the digital twin model of industrial heterogeneous terminals is usually modeled by physical principle-based technology, data-driven technology, or a combination of the two. However, in complex industrial intelligent manufacturing scenarios, the digital twin model is often required to accurately represent the multi-dimensional characteristics of industrial heterogeneous terminals, and the global model parameters of the digital twin model of industrial heterogeneous terminals are required to be updated in real time to ensure the data consistency of the virtual and real space of industrial intelligent manufacturing. Therefore, data-driven technology is more suitable for the construction of digital twin models of industrial heterogeneous terminals.

[0019] When using digital drive technology to model the digital twin model of industrial heterogeneous terminals, due to the limited number of model parameters of the edge server in the industrial Internet of Things, a large number of industrial heterogeneous terminals are often required to participate in the model modeling to improve the modeling efficiency. In this regard, federated learning in data-driven technology is widely used in the construction of digital twin models of industrial heterogeneous terminals.

[0020] In the related technologies, synchronous federated learning or asynchronous federated learning is used to model the digital twin model of industrial heterogeneous terminals. However, on the one hand, the related technologies do not fully consider the heterogeneity of various industrial heterogeneous terminals, and are unable to select industrial heterogeneous terminals that are more suitable for modeling from a number of industrial heterogeneous terminals. The modeling efficiency of the digital twin model of its industrial heterogeneous terminals needs to be improved; on the other hand, the related technologies still have inherent lagging effects and model obsolescence problems, which will lead to unbalanced training costs and model losses. The modeling accuracy of the digital twin model of its industrial heterogeneous terminals needs to be improved.

[0021] In view of this, the embodiments of the present application provide a modeling method and device for a digital twin model, aiming to improve the modeling accuracy and modeling efficiency of the digital twin model of industrial heterogeneous terminals.

[0022] First, the implementation steps of the modeling method of the digital twin model provided by this application will be explained in detail below with reference to the accompanying drawings.

[0023] The modeling method of the digital twin model provided in the embodiment of the present application can be applied to a terminal, a server, or software running in a terminal or a server. The terminal can be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The server can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms. In addition, the server can also be a node server in a blockchain network, but is not limited thereto. Among them, blockchain is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.

[0024] Reference Figure 1 , Figure 1 : is a flow chart of a modeling method of a digital twin model provided in the present application, and the modeling method of the digital twin model mainly includes the following steps S101-S104: S101, obtaining operation characteristic data of a plurality of heterogeneous terminals; S102, selecting a plurality of heterogeneous terminals participating in local training as training terminals from among a plurality of heterogeneous terminals according to the operation characteristic data of each heterogeneous terminal; S103, controlling each training terminal to perform local training, and performing asynchronous global update or synchronous global update on the global model parameters of the digital twin model during the local training; S104: If all training terminals have completed local training, a digital twin model is constructed according to the global model parameters of the digital twin model.

[0025] In an embodiment of the present application, first, operating characteristic data of several heterogeneous terminals are obtained. Heterogeneous terminals refer to industrial heterogeneous terminals in industrial intelligent manufacturing scenarios. The operating characteristic data of each heterogeneous terminal can characterize the heterogeneity of each heterogeneous terminal; then, based on the operating characteristic data of each heterogeneous terminal, multiple heterogeneous terminals participating in local training are selected from several heterogeneous terminals as training terminals; thereafter, each training terminal is controlled to perform local training, and during the local training, the global model parameters of the digital twin model are asynchronously or synchronously updated globally; finally, when all training terminals have completed local training, a digital twin model is constructed according to the global model parameters of the digital twin model. The digital twin model refers to the digital twin model of the industrial heterogeneous terminal in the industrial intelligent manufacturing scenario.

[0026] It can be seen that the embodiment of the present application characterizes the heterogeneity of each industrial heterogeneous terminal through the operating characteristic data of each industrial heterogeneous terminal in the industrial intelligent manufacturing scenario. While fully considering the heterogeneity of each industrial heterogeneous terminal, an industrial heterogeneous terminal that is more suitable for digital twin modeling is selected from a number of industrial heterogeneous terminals to participate in local training, thereby improving the degree of adaptation between the industrial heterogeneous terminals and the digital twin model modeling, thereby improving the modeling efficiency of the digital twin model of the industrial heterogeneous terminal; in addition, during the local training of each training terminal, the global model parameters of the digital twin model are asynchronously updated globally or synchronously updated globally to balance the training cost and model loss of the digital twin model, thereby ensuring that the digital twin model can quickly respond to the dynamically changeable environment in the industrial intelligent manufacturing scenario, enhance the data consistency of the virtual and real space of industrial intelligent manufacturing, and improve the modeling accuracy of the digital twin model of the industrial heterogeneous terminal.

[0027] In the above step S101, the operating characteristic data of several industrial heterogeneous terminals in the industrial intelligent manufacturing scenario are obtained, and the operating characteristic data of each industrial heterogeneous terminal can characterize the heterogeneity of each industrial heterogeneous terminal.

[0028] The above-mentioned heterogeneous terminals refer to industrial heterogeneous terminals in industrial intelligent manufacturing scenarios, such as CNC machine tool equipment, industrial robot equipment, intelligent sensing detection equipment, intelligent logistics equipment, intelligent casting equipment and intelligent stamping equipment, but are not limited to these.

[0029] The operating characteristic data of the above-mentioned industrial heterogeneous terminals can characterize the heterogeneity of the industrial heterogeneous terminals.

[0030] In the above step S102, after obtaining the operating characteristic data of each industrial heterogeneous terminal, based on the operating characteristic data of each industrial heterogeneous terminal, multiple industrial heterogeneous terminals participating in local training are selected from a number of industrial heterogeneous terminals as training terminals to ensure the degree of adaptation between the industrial heterogeneous terminals and the digital twin model modeling, thereby improving the modeling efficiency of the digital twin model of the industrial heterogeneous terminals.

[0031] The above-mentioned training terminal refers to an industrial heterogeneous terminal used to perform local training on its local model.

[0032] The above-mentioned selecting multiple heterogeneous terminals participating in local training from a number of heterogeneous terminals based on the operating characteristic data of each heterogeneous terminal may include selecting heterogeneous terminals whose operating characteristic data is higher than a preset operating characteristic threshold from a number of heterogeneous terminals as heterogeneous terminals participating in local training, but is not limited to this.

[0033] In the above step S103, each training terminal is controlled to perform local training on its configured local model using the local data set, the local model parameters of the local model and the global model parameters of the digital twin model. During the local training, the global model parameters of the digital twin model are updated, and the update method may include an asynchronous global update method or a synchronous global update method, so as to balance the training cost and model loss of the digital twin model, ensure that the digital twin model can quickly respond to the dynamically changeable environment in the industrial intelligent manufacturing scenario, enhance the data consistency of the virtual and real space of the industrial intelligent manufacturing, and improve the modeling accuracy of the digital twin model of the industrial heterogeneous terminal.

[0034] In the above-mentioned local training, the local data set of the training terminal may include the operation data of the industrial heterogeneous terminal, the equipment parameters and the operation data of the virtual entity corresponding to the industrial heterogeneous terminal.

[0035] The operating data of the above-mentioned industrial heterogeneous terminals may include temperature data, pressure data, vibration data and hardware load data of the industrial heterogeneous terminals, and the equipment parameters of the industrial heterogeneous terminals may include equipment models, specifications, service life and equipment maintenance time values ​​of the industrial heterogeneous terminals; the operating data of the virtual entity corresponding to the industrial heterogeneous terminal can be obtained by performing virtual-to-real conversion processing on the operating data of the industrial heterogeneous terminal. The virtual-to-real conversion processing is a prior art and will not be described in detail here; or, the operating data of the virtual entity corresponding to the industrial heterogeneous terminal is obtained through preset virtual-to-real mapping data. Specifically, the virtual entity operating data corresponding to the operating data of the industrial heterogeneous terminal is searched in the preset virtual-to-real mapping data as the operating data of the virtual entity corresponding to the industrial heterogeneous terminal. The virtual-to-real mapping data may include multiple preset physical entity operating data and virtual entity operating data corresponding to each preset physical entity operating data, but is not limited to this.

[0036] In the above-mentioned local training, the local model configured by the training terminal may be a digital twin model of the training terminal. The local model is essentially the same as the global digital twin model obtained subsequently, and the local model parameters of the local model may be pre-set hyperparameters.

[0037] In the above local training, for k The global model parameters of the digital twin model in the round of training involve asynchronous global updates and synchronous global updates, which will be explained in detail later.

[0038] The above-mentioned control of each training terminal to perform local training may include controlling the training terminal to perform local training on the local model configured therein based on the preset local model parameters and the global model parameters of the digital twin model; when the training terminal completes the local training of the current training round, if the current training round is less than the preset total training rounds, the current training round is increased by one, and the local model parameters are updated, and then the process returns to the step of performing local training on the local model configured therein to achieve iterative training, otherwise it is determined that the training terminal completes the local training, but is not limited to this.

[0039] The above-mentioned asynchronous global update or synchronous global update of the global model parameters of the digital twin model may include, when there are training terminals that have completed local training at the current moment, if the number of training terminals that have completed local training is greater than a preset number threshold, then synchronously updating the global model parameters of the digital twin model to shorten the modeling time of the digital twin model; otherwise, asynchronously updating the global model parameters of the digital twin model to improve the modeling accuracy of the digital twin model, but is not limited to this.

[0040] In the above step S104, when all training terminals have completed local training, a global digital twin model is constructed according to the global model parameters of the digital twin model. The global digital twin model refers to the digital twin model of industrial heterogeneous terminals in the industrial intelligent manufacturing scenario, and the global digital twin model is universal, that is, the global digital twin model is applicable to all industrial heterogeneous terminals.

[0041] The above-mentioned constructing a digital twin model according to the global model parameters of the digital twin model may include constructing a digital twin model according to the global model parameters of the digital twin model in combination with a preset modeling tool, but is not limited to this.

[0042] The above steps will be further described below.

[0043] In some embodiments, the above-mentioned operation characteristic data may include operation efficiency, communication efficiency and historical loss function, where the historical loss function is a loss function value of a local model of a heterogeneous terminal in a training round before a current training round; The above-mentioned selecting a plurality of heterogeneous terminals participating in local training as training terminals from a plurality of heterogeneous terminals according to the operation characteristic data of each heterogeneous terminal may include: According to the selection times of each heterogeneous terminal, combined with the total number of training terminals and the historical selection times, the confidence radius of each heterogeneous terminal relative to the current training round is obtained; wherein, the selection times of the heterogeneous terminal is the number of times the heterogeneous terminal is selected to participate in local training before the current training round; the total number of training terminals is the number of heterogeneous terminals selected to participate in local training; the historical selection times is the number of times the historical heterogeneous terminals are selected to participate in local training before the current training round; the historical heterogeneous terminals are the heterogeneous terminals selected to participate in local training in the training round before the current training round; According to the computing efficiency and the number of selections of each heterogeneous terminal, combined with the confidence radius of each heterogeneous terminal relative to the current training round, the computing performance data of each heterogeneous terminal relative to the current training round is obtained; the computing performance data is used to characterize the computing performance of the heterogeneous terminal when performing local training on its local model in the current training round; According to the communication efficiency and the number of selections of each heterogeneous terminal, combined with the confidence radius of each heterogeneous terminal relative to the current training round, the communication performance data of each heterogeneous terminal relative to the current training round is obtained; the communication performance data is used to characterize the communication performance of the heterogeneous terminal when it performs local training on its local model in the current training round; According to the historical loss function of each heterogeneous terminal, the gradient norm value of the local model of each heterogeneous terminal relative to the current training round is obtained; According to the computing performance data and communication performance data of each heterogeneous terminal relative to the current training round, combined with the gradient norm value of the local model of each heterogeneous terminal relative to the current training round, the training characteristic parameters of each heterogeneous terminal are obtained; the training characteristic parameters are used to characterize the performance of the heterogeneous terminal when performing local training on its local model; According to the training characteristic parameters of each heterogeneous terminal, a plurality of heterogeneous terminals participating in local training are selected from a plurality of heterogeneous terminals as training terminals.

[0044] In this implementation, in order to improve the modeling efficiency of the digital twin model of industrial heterogeneous terminals, the training characteristic parameters of each industrial heterogeneous terminal are determined through the operating characteristic data of each industrial heterogeneous terminal. The training characteristic parameters characterize the performance of the industrial heterogeneous terminal when locally training its local model. Then, the industrial heterogeneous terminal that ultimately participates in the local training is selected from a number of industrial heterogeneous terminals through the training characteristic parameters of each industrial heterogeneous terminal, and the selected industrial heterogeneous terminal is determined as the training terminal.

[0045] Specifically, first, the confidence radius of each industrial heterogeneous terminal relative to the current training round is determined by using the selection times of each industrial heterogeneous terminal, combined with the total number of training terminals and the historical selection times. The confidence radius is mainly used to correct the computing performance data and communication performance data of each industrial heterogeneous terminal relative to the current training round to ensure the accuracy of each performance data.

[0046] Then, for each industrial heterogeneous terminal, the following are performed: using the computing efficiency and number of selections of the industrial heterogeneous terminal and the confidence radius of the industrial heterogeneous terminal relative to the current training round, the computing performance data of the industrial heterogeneous terminal relative to the current training round is determined, and the computing performance data can characterize the computing performance of the industrial heterogeneous terminal when performing local training on its local model in the current training round; and using the communication efficiency and number of selections of the industrial heterogeneous terminal and the confidence radius of the industrial heterogeneous terminal relative to the current training round, the communication performance data of the industrial heterogeneous terminal relative to the current training round is determined, and the communication performance data can characterize the communication performance of the industrial heterogeneous terminal when performing local training on its local model in the current training round; and using the historical loss function of the industrial heterogeneous terminal, the gradient norm value of the local model of the industrial heterogeneous terminal relative to the current training round is determined, and the gradient norm value can characterize the training degree of the local model of the industrial heterogeneous terminal; thereafter, using the computing performance data and communication performance data of the industrial heterogeneous terminal relative to the current training round and the gradient norm value of the local model of the industrial heterogeneous terminal relative to the current training round, the training characteristic parameters of the industrial heterogeneous terminal are determined.

[0047] Afterwards, the training characteristic parameters of each industrial heterogeneous terminal are used to select a heterogeneous terminal that ultimately participates in the local training from among a number of industrial heterogeneous terminals, and the selected industrial heterogeneous terminal is determined as a training terminal.

[0048] The above-mentioned operating characteristic data may include computing efficiency, communication efficiency and historical loss function. The historical loss function is the loss function value of the local model of the heterogeneous terminal in the previous training round of the current training round. The computing efficiency refers to the training efficiency of the heterogeneous terminal during local training. The communication efficiency refers to the efficiency of the heterogeneous terminal communicating with the edge server during local training, but is not limited to this.

[0049] The above-mentioned operation efficiency may be a ratio of the amount of training data of the heterogeneous terminal in a single local training to the training time consumed by the heterogeneous terminal in a single local training, but is not limited thereto.

[0050] The communication efficiency may be a ratio of the amount of data transmitted by the heterogeneous terminal to the edge server when completing a single local training to the transmission time consumed by the heterogeneous terminal, but is not limited thereto.

[0051] The number of selections of the heterogeneous terminal refers to the number of times the heterogeneous terminal is selected to participate in the local training before the current training round.

[0052] The total number of training terminals mentioned above refers to the number of heterogeneous terminals selected to participate in local training, which is a preset value.

[0053] The above-mentioned number of historical selections refers to the number of times that the historical heterogeneous terminals were selected to participate in the local training before the current training round, and the historical heterogeneous terminals refer to the heterogeneous terminals selected to participate in the local training in the previous training round of the current training round.

[0054] The above-mentioned method of obtaining the confidence radius of each heterogeneous terminal relative to the current training round based on the number of selections of each heterogeneous terminal, combined with the total number of training terminals and the historical number of selections may include obtaining the confidence radius of each heterogeneous terminal relative to the current training round based on the total number of training terminals, the historical number of selections and the number of selections of each heterogeneous terminal, combined with a machine learning method, but is not limited to this.

[0055] The above-mentioned machine learning method can be set according to actual conditions, and this embodiment does not specifically limit this.

[0056] For example, the above-mentioned machine learning method may be a support vector machine; or, the above-mentioned machine learning method may be a logistic regression, but is not limited thereto.

[0057] Alternatively, the above-mentioned obtaining the confidence radius of each heterogeneous terminal relative to the current training round based on the number of selections of each heterogeneous terminal, combined with the total number of training terminals and the number of historical selections, may include obtaining the confidence radius of each heterogeneous terminal relative to the current training round through the following formula (1) based on the total number of training terminals, the number of historical selections and the number of selections of each heterogeneous terminal: (1); In formula (1), Represents heterogeneous terminals Relative to the The confidence radius of the training round, which can be understood as Heterogeneous terminals in a training round Confidence radius of heterogeneous terminals It refers to Heterogeneous terminals; Represents heterogeneous terminals Relative to the The confidence radius of the training round, which can be understood as Heterogeneous terminals in a training round The confidence radius of Represents the total number of training terminals, that is, the number of heterogeneous terminals selected to participate in local training; Indicates the number of historical selections, that is, The number of times the historical heterogeneous terminal is selected to participate in local training in the training round, and the historical heterogeneous terminal is the number of times the historical heterogeneous terminal is selected to participate in local training in the training round. Heterogeneous terminals selected to participate in local training in the training round; Represents a set of historically heterogeneous terminals; Represents heterogeneous terminals In the Selected to participate in local training during the training rounds; Represents heterogeneous terminals In the Not selected to participate in local training during the training rounds; Represents heterogeneous terminals The number of selections, that is, in the previous Heterogeneous terminals in a training round The number of times selected to participate in local training, Indicates the round number; Represents heterogeneous terminals Whether or not you are selected to participate in local training, ,when When it is 1, it indicates a heterogeneous terminal. Selected to participate in local training, When it is 0, it indicates a heterogeneous terminal. Not selected for local training.

[0058] The above computing performance data is used to characterize the computing performance of the heterogeneous terminal when performing local training on its local model in the current training round.

[0059] The above-mentioned obtaining the computing performance data of each heterogeneous terminal relative to the current training round based on the computing efficiency and the number of selections of each heterogeneous terminal in combination with the confidence radius of each heterogeneous terminal relative to the current training round may include, for each heterogeneous terminal, obtaining the initial computing performance data of the heterogeneous terminal relative to the current training round based on the computing efficiency and the number of selections of the heterogeneous terminal in combination with a machine learning method; and calculating the sum of the initial computing performance data of the heterogeneous terminal relative to the current training round and the confidence radius as the computing performance data of the heterogeneous terminal relative to the current training round.

[0060] Alternatively, the above-mentioned obtaining the computing performance data of each heterogeneous terminal relative to the current training round based on the computing efficiency and the number of selections of each heterogeneous terminal in combination with the confidence radius of each heterogeneous terminal relative to the current training round may include obtaining the computing performance data of each heterogeneous terminal relative to the current training round through the following formula (2) based on the computing efficiency and the number of selections of each heterogeneous terminal and the confidence radius of each heterogeneous terminal relative to the current training round: (2); In formula (2), Represents heterogeneous terminals Relative to the The computing performance data of the training round can be understood as Heterogeneous terminals in a training round Computational performance data; Represents heterogeneous terminals Relative to the The initial computing performance data of the training round can be understood as Heterogeneous terminals in a training round The initial computing performance data satisfies the following formula (3): (3); In formula (3), Represents heterogeneous terminals operational efficiency.

[0061] The above communication performance data is used to characterize the communication performance when the heterogeneous terminal performs local training on its local model in the current training round.

[0062] The above-mentioned obtaining the communication performance data of each heterogeneous terminal relative to the current training round based on the communication efficiency and the number of selections of each heterogeneous terminal in combination with the confidence radius of each heterogeneous terminal relative to the current training round may include, for each heterogeneous terminal, obtaining the initial communication performance data of the heterogeneous terminal relative to the current training round based on the communication efficiency and the number of selections of the heterogeneous terminal in combination with a machine learning method; and calculating the sum of the initial communication performance data of the heterogeneous terminal relative to the current training round and the confidence radius as the communication performance data of the heterogeneous terminal relative to the current training round.

[0063] Alternatively, the communication performance data of each heterogeneous terminal relative to the current training round is obtained according to the communication efficiency and the number of selections of each heterogeneous terminal, combined with the confidence radius of each heterogeneous terminal relative to the current training round, which may include obtaining the communication performance data of each heterogeneous terminal relative to the current training round according to the communication efficiency and the number of selections of each heterogeneous terminal and the confidence radius of each heterogeneous terminal relative to the current training round through the following formula (4): (4); In formula (4), Represents heterogeneous terminals Relative to the The communication performance data of the training round can be understood as Heterogeneous terminals in a training round Communication performance data; Represents heterogeneous terminals Relative to the The initial communication performance data of the training round can be understood as Heterogeneous terminals in a training round The initial communication performance data satisfies the following formula (5): (5); In formula (5), Represents heterogeneous terminals communication efficiency.

[0064] The above historical loss function is the loss function value of the local model of the heterogeneous terminal in the previous training round of the current training round.

[0065] The loss function of the above-mentioned local model can be set according to actual conditions, and this embodiment does not make any specific limitation to this.

[0066] For example, the loss function may be a loss function of logistic regression, that is, a logarithmic loss function, but is not limited thereto.

[0067] The above-mentioned method of obtaining the gradient norm value of the local model of each heterogeneous terminal relative to the current training round based on the historical loss function of each heterogeneous terminal may include directly calculating the gradient norm value of the historical loss function of the heterogeneous terminal as the gradient norm value of the local model of the heterogeneous terminal relative to the current training round.

[0068] Alternatively, the above-mentioned obtaining the gradient norm value of the local model of each heterogeneous terminal relative to the current training round according to the historical loss function of each heterogeneous terminal may include obtaining the gradient norm value of the local model of each heterogeneous terminal relative to the current training round according to the historical loss function of each heterogeneous terminal through the following formula (6): (6); In formula (6), Represents heterogeneous terminals The local model is relative to the The gradient norm value of the training round can be understood as Heterogeneous terminals in a training round The gradient norm value of the local model; Represents heterogeneous terminals The gradient value of the historical loss function; Represents the historical loss function, that is, In the training round, when heterogeneous terminals The local model parameters are When heterogeneous terminals The loss function value of the local model.

[0069] The above training characteristic parameters are used to characterize the performance of a heterogeneous terminal when performing local training on its local model.

[0070] The above-mentioned training characteristic parameters of each heterogeneous terminal are obtained based on the computing performance data and communication performance data of each heterogeneous terminal relative to the current training round, combined with the gradient norm value of the local model of each heterogeneous terminal relative to the current training round. It can include weighting the computing performance data, communication performance data and gradient norm value for each heterogeneous terminal to obtain the training characteristic parameters of the heterogeneous terminal; wherein the sum of the weight of the computing performance data, the weight of the communication performance data and the weight of the gradient norm value is one.

[0071] Alternatively, the training characteristic parameters of each heterogeneous terminal are obtained according to the computing performance data and communication performance data of each heterogeneous terminal relative to the current training round, combined with the gradient norm value of the local model of each heterogeneous terminal relative to the current training round, which may include obtaining the training characteristic parameters of each heterogeneous terminal according to the computing performance data and communication performance data of each heterogeneous terminal relative to the current training round and the gradient norm value of the local model of each heterogeneous terminal relative to the current training round through the following formula (7): (7); In formula (7), Represents heterogeneous terminals Training feature parameters of Represents heterogeneous terminals Relative to the The weight of the computational performance data of each training round; Represents heterogeneous terminals Relative to the The weight of the communication performance data of each training round; Represents heterogeneous terminals The local model is relative to the The weight of the gradient norm value of the training round; Indicates the total number of heterogeneous terminals.

[0072] The selecting a plurality of heterogeneous terminals participating in local training from among several heterogeneous terminals according to the training characteristic parameters of each heterogeneous terminal may include selecting a heterogeneous terminal having a training characteristic parameter greater than a preset characteristic threshold from among several heterogeneous terminals as the heterogeneous terminal participating in local training.

[0073] Alternatively, the above-mentioned selecting multiple heterogeneous terminals participating in local training from a number of heterogeneous terminals according to the training characteristic parameters of each heterogeneous terminal may include sorting the multiple heterogeneous terminals according to the training characteristic parameters of each heterogeneous terminal to obtain a sorting sequence number of each heterogeneous terminal, wherein the sorting sequence number of the heterogeneous terminal is positively correlated with the training characteristic parameter of the heterogeneous terminal; and selecting, from the multiple heterogeneous terminals, heterogeneous terminals whose sorting sequence numbers are greater than or equal to a preset sequence number threshold as heterogeneous terminals participating in local training, and the number of the selected heterogeneous terminals is equal to the total number of the above-mentioned training terminals.

[0074] The acquisition of the training characteristic parameters of the above-mentioned heterogeneous terminals can be performed in parallel, for example, the training characteristic parameters of each heterogeneous terminal are determined simultaneously; or, the acquisition of the training characteristic parameters of the above-mentioned heterogeneous terminals can be performed serially, for example, the training characteristic parameters of each heterogeneous terminal are determined in sequence, but it is not limited to this.

[0075] The acquisition of the above-mentioned computing performance data, the above-mentioned communication performance data and the above-mentioned gradient norm value can be carried out in parallel, for example, the computing performance data of the above-mentioned heterogeneous terminal relative to the current training round, the communication performance data of the above-mentioned heterogeneous terminal relative to the current training round and the gradient norm value of the local model of the above-mentioned heterogeneous terminal relative to the current training round are simultaneously determined.

[0076] Alternatively, the acquisition of the above-mentioned computing performance data, the above-mentioned communication performance data and the above-mentioned gradient norm value can be performed serially, for example, first obtaining the computing performance data of the above-mentioned heterogeneous terminal relative to the current training round, then obtaining the communication performance data of the above-mentioned heterogeneous terminal relative to the current training round, and finally obtaining the gradient norm value of the local model of the above-mentioned heterogeneous terminal relative to the current training round, but not limited to this.

[0077] In some implementations, before controlling each training terminal to perform local training, the method may further include: According to the computing efficiency and data volume of the training terminal, combined with the total number of training terminals, the total terminal data volume and the total computing efficiency, the initial training rounds of the training terminal are updated to obtain the total training rounds of the training terminal; Among them, the total terminal data volume is the sum of the data volume of all training terminals, and the total computing efficiency is the sum of the computing efficiency of all training terminals.

[0078] In this embodiment, before controlling each training terminal to perform local training, the total training rounds of each training terminal need to be determined to improve the accuracy of the training rounds and the local training accuracy of each training terminal, and reduce the phenomenon of insufficient local training or excessive local training. Specifically, for each training terminal, first, the initial training rounds of the training terminal are obtained; then, according to the computing efficiency and data volume of the training terminal, combined with the total number of training terminals, the total terminal data volume and the total computing efficiency, the initial training rounds of the training terminal are updated to obtain the total training rounds of the training terminal.

[0079] The initial training rounds of the training terminal can be set according to actual conditions, and this embodiment does not specifically limit this.

[0080] The above-mentioned total terminal data volume refers to the sum of the data volumes of all training terminals, wherein the data volume of the training terminal refers to the training data volume of the training terminal in a single local training.

[0081] The above total computing efficiency refers to the sum of the computing efficiencies of all training terminals.

[0082] The above-mentioned updating of the initial training rounds of the training terminal according to the computing efficiency and data volume of the training terminal in combination with the total number of training terminals, the total terminal data volume and the total computing efficiency to obtain the total training rounds of the training terminal may include determining the round correction value of the training terminal according to the total number of training terminals, the total terminal data volume and the total computing efficiency as well as the computing efficiency and data volume of the training terminal in combination with a machine learning method; calculating the sum of the round correction value of the training terminal and the initial training round as the total training rounds of the training terminal, but is not limited to this.

[0083] Alternatively, the above-mentioned updating of the initial training rounds of the training terminal according to the computing efficiency and data volume of the training terminal in combination with the total number of training terminals, the total terminal data volume and the total computing efficiency to obtain the total training rounds of the training terminal may include obtaining the total training rounds of the training terminal according to the total number of training terminals, the total terminal data volume and the total computing efficiency as well as the computing efficiency and data volume of the training terminal through the following formula (8): (8); In formula (8), Indicates The total number of training rounds for training terminals; Represents the total number of training terminals, that is, the number of heterogeneous terminals selected to participate in local training; Indicates Initial training rounds for training terminals; Indicates The amount of data per training terminal; Represents the total terminal data volume, that is, the sum of the data volumes of all training terminals; Indicates training terminals are selected to participate in local training; represents the training terminal set; represents the average computing efficiency of all training terminals, which satisfies the following formula (9): (9); In formula (9), Indicates The computational efficiency of each training terminal; Represents the total computing efficiency, which is the sum of the computing efficiencies of all training terminals.

[0084] In some implementations, the controlling each training terminal to perform local training may include: Controlling the training terminal to perform local training on the local model of the training terminal based on the local model parameters of the training terminal and the global model parameters of the digital twin model; When the training terminal completes the local training of the current training round, the local model parameters of the training terminal are updated to obtain the updated local model parameters as the local model parameters of the training terminal; Compare the current training round with the total training rounds of the training terminal; If the current training round is less than the total training round, the current training round is incremented by one, and the process returns to the step of controlling the training terminal to perform local training on the local model of the training terminal based on the local model parameters of the training terminal and the global model parameters of the digital twin model; If the current training round is greater than or equal to the total training round, it is determined that the training terminal completes the local training, and the local model parameters of the training terminal are corrected to obtain the corrected local model parameters.

[0085] In this embodiment, for each training terminal, there are: first, the training terminal is controlled to perform local training on the local model of the training terminal based on the local data set of the training terminal, the local model parameters of the training terminal and the global model parameters of the digital twin model. Among them, the local data set of the training terminal may include the operation data, equipment parameters and operation data of the virtual entity corresponding to the industrial heterogeneous terminal of the industrial heterogeneous terminal, wherein the operation data of the industrial heterogeneous terminal may indicate the operation status of the industrial heterogeneous terminal, and it may include the temperature data, pressure data, vibration data and hardware load data of the industrial heterogeneous terminal, etc. The equipment parameters of the industrial heterogeneous terminal may include the equipment model, specification, service life and equipment maintenance time value of the industrial heterogeneous terminal. The local model configured by the training terminal may be the digital twin model of the training terminal, and the local model is essentially the same as the global digital twin model obtained subsequently, and the local model parameters of the local model may be pre-set hyperparameters. For the global model parameters of the digital twin model in the current training round, since asynchronous global update and synchronous global update are involved, this part will be described in detail later. When the training terminal completes the local training of the current training round, the local model parameters of the training terminal are updated to obtain the updated local model parameters as the local model parameters of the training terminal; then, it is determined whether the current training round is less than the total training rounds of the training terminal; if so, it means that the local training has not been completed. At this time, the current training round is increased by one, and the step of locally training the local model of the training terminal is returned to realize iterative training; if not, it is determined that the training terminal has completed the local training, and the local model parameters of the training terminal are corrected to obtain the corrected local model parameters.

[0086] It can be seen that this embodiment implements the training of the local model of the training terminal based on the local model parameters of the training terminal and the global model parameters of the digital twin model. The local model parameters of the training terminal are updated when each round of local training is completed, and the local model parameters of the training terminal are corrected when all rounds of local training are completed. In this way, the accuracy of the local model parameters of the training terminal can be improved, the local training accuracy of the training terminal can be ensured, and the occurrence of insufficient local training or excessive local training can be reduced.

[0087] The above-mentioned updating of the local model parameters of the training terminal to obtain the updated local model parameters may include updating the local model parameters of the training terminal through a preset parameter optimization algorithm to obtain the updated local model parameters, but is not limited thereto.

[0088] The above parameter optimization algorithm can be set according to actual conditions, and this embodiment does not specifically limit this.

[0089] For example, the parameter optimization algorithm may be a gradient descent method; or, the parameter optimization algorithm may be a stochastic gradient descent method, but is not limited thereto.

[0090] Alternatively, the above-mentioned updating of the local model parameters of the training terminal to obtain the updated local model parameters may include updating the local model parameters of the training terminal based on the average local gradient to obtain the updated local model parameters, as shown in the following formula (10): (10); In formula (10), Indicates The updated local model parameters of the training terminal, that is, In the training round, The local model parameters of the training terminal; Indicates The local model parameters of the training terminal, that is, In the training round The local model parameters of the training terminal; Indicates the preset learning rate; Indicated in In the training round The average local gradient of the training terminals satisfies the following formula (11): (11); In formula (11), Indicated in In the training round, The training terminal The updated local model parameters; Indicates The total number of training rounds for training terminals; Indicated in In the training round, when The local model parameters of the training terminal are At that time, The loss function value of the local model of the training terminal; Represents the gradient value of the loss function.

[0091] The above-mentioned correction processing of the local model parameters of the training terminal to obtain the corrected local model parameters may include correcting the local model parameters of the training terminal to obtain the corrected local model parameters if the local model parameters of the training terminal exceed the preset parameter range, and the corrected local model parameters are within the parameter range, but is not limited to this.

[0092] In some implementations, the above-mentioned correction processing of the local model parameters of the training terminal to obtain the corrected local model parameters may include: According to the model version number data of the training terminal, the local model parameters of the training terminal are corrected to obtain the corrected local model parameters; wherein the model version number data of the training terminal is the difference between the global model version number of the digital twin model and the local model version number of the training terminal.

[0093] In this implementation, the local model parameters of the training terminal are corrected based on the difference between the global model version number of the digital twin model and the local model version number of the training terminal to obtain corrected local model parameters. In this way, the accuracy of the local model parameters of the training terminal is improved and the local training accuracy of the training terminal is ensured.

[0094] The model version number data of the above training terminal is used to measure the degree of obsolescence or training lag of the local model of the training terminal.

[0095] The model version number data of the above-mentioned training terminal refers to the difference between the global model version number of the digital twin model and the local model version number of the training terminal.

[0096] The above-mentioned correction processing of the local model parameters of the training terminal according to the model version number data of the training terminal to obtain the corrected local model parameters may include searching for the model parameter correction value corresponding to the model version number data of the training terminal from the first version mapping data, and calculating the sum of the model parameter correction value and the local model parameter of the training terminal as the corrected local model parameter, wherein the first version mapping data may include multiple preset model version number data and the model parameter correction value corresponding to each preset model version number data, but is not limited to this.

[0097] The first version mapping data may be chart data or table data, but is not limited thereto.

[0098] Alternatively, the above-mentioned correcting the local model parameters of the training terminal according to the model version number data of the training terminal to obtain the corrected local model parameters may include obtaining the corrected local model parameters according to the model version number data of the training terminal and the local model parameters through the following formula (12): (12); In formula (12), Indicates The corrected local model parameters of the training terminal; Indicated in The global model parameters of the digital twin model in each training round; Indicates The aging degree of a training terminal satisfies the following formula (13): (13); In formula (13), Indicates The model version number data of the training terminal, that is, the global model version number of the digital twin model Relative to the The local model version number of the training terminal The difference.

[0099] In some embodiments, the asynchronous global update or synchronous global update of the global model parameters of the digital twin model may include: If it is detected that there is a training terminal that has completed local training at the current moment, the training terminal that has completed local training at the current moment is determined as an idle terminal, and the training terminal that has not completed local training at the current moment is determined as an occupied terminal; When the model version number data of the idle terminal is less than the preset version threshold, the global model parameters of the digital twin model are asynchronously updated globally according to the local model parameters and data volume of each training terminal and the total terminal data volume to obtain updated global model parameters; among which, the local model parameters of the training terminals that have completed local training before the current moment and the local model parameters of the idle terminals are both corrected local model parameters, and the local model parameters of the occupied terminals are the local model parameters of the previous training round before the current training round; The global model version number of the digital twin model is updated to obtain an updated global model version number.

[0100] In this embodiment, while each training terminal is performing local training, the training terminal that has completed local training at the current moment is detected; if it is detected that there is a training terminal that has completed local training at the current moment, the training terminal that has completed local training at the current moment is first determined as an idle terminal, and the training terminal that has not completed local training at the current moment is determined as an occupied terminal, and then the model version number data of the idle terminal is compared with the preset version threshold. If the model version number data of the idle terminal is less than the preset version threshold, it means that the difference between the global model version number of the digital twin model and the local model version number of the idle terminal is small. In this way, it can be determined that the local models of other training terminals are less outdated or the training lag is less, and an asynchronous global update is performed at this time.

[0101] Specifically, during the asynchronous global update, the global model parameters of the digital twin model are asynchronously globally updated according to the local model parameters and data volume of each training terminal and the total terminal data volume to obtain updated global model parameters, wherein the local model parameters of the training terminals that have completed local training before the current moment and the local model parameters of the idle terminals are both revised local model parameters, and the local model parameters of the occupied terminals are the local model parameters of the previous training round of the current training round; and, during the asynchronous global update, the global model version number of the digital twin model is updated.

[0102] In this way, when the model version number data of the idle terminal is less than the preset version threshold, this embodiment implements the global model parameter update of the digital twin model of the industrial heterogeneous terminal through the above-mentioned asynchronous global update method. It can not only effectively reduce the inherent lag effect and model obsolescence problems, but also balance the training cost and model loss of the digital twin model of the industrial heterogeneous terminal, so that the digital twin model of the industrial heterogeneous terminal can quickly respond to the dynamically changeable environment in the industrial intelligent manufacturing scenario, and ensure the data consistency of the virtual and real space of industrial intelligent manufacturing, thereby improving the modeling accuracy of the digital twin model of the industrial heterogeneous terminal.

[0103] The above version threshold can be set according to actual conditions, and this implementation does not specifically limit this.

[0104] The above-mentioned asynchronous global update of the global model parameters of the digital twin model based on the local model parameters and data volume of each training terminal in combination with the total terminal data volume to obtain the updated global model parameters can include calculating the mean of the data volume of each training terminal as the average data volume, weighting the average data volume and the total terminal data volume to obtain the overall data volume, wherein the sum of the weight of the average data volume and the weight of the total terminal data volume is one; calculating the mean of the local model parameters of each training terminal as the average model parameter; obtaining the revised value of the global model parameter based on the overall data volume and the average model parameter in combination with the machine learning method; calculating the sum of the revised value of the global model parameter and the global model parameters of the digital twin model as the updated global model parameters.

[0105] Alternatively, the above-mentioned asynchronous global update of the global model parameters of the digital twin model based on the local model parameters and data volume of each training terminal in combination with the total terminal data volume to obtain the updated global model parameters may include asynchronously globally updating the global model parameters of the digital twin model based on the total terminal data volume and the local model parameters and data volume of each training terminal through the following formula (14) to obtain the updated global model parameters: (14); In formula (14), represents the updated global model parameters; Indicates The amount of data per training terminal; Represents the total terminal data volume, that is, the sum of the data volumes of all training terminals; Represents the total number of training terminals, that is, the number of heterogeneous terminals selected to participate in local training.

[0106] It should be noted that, in this embodiment, for the parameters in the above formula (14) In the asynchronous global update, the local model parameters of the training terminals that have completed local training before the current time and the local model parameters of the idle terminals are all corrected local model parameters, which are shown in the above formula (12) ; The local model parameters of the occupied terminal are the local model parameters of the previous training round before the current training round. For example, if the current training round is round of training, the local model parameters of the occupied terminal are The local model parameters of the training round are as shown in the above formula (10) .

[0107] The above-mentioned updating of the global model version number of the digital twin model to obtain the updated global model version number may include searching from the second version mapping data to obtain the version number corresponding to the global model parameters after the asynchronous global update as the updated global model version number, so as to realize the update of the global model version number of the digital twin model, wherein the second version mapping data includes multiple preset global model parameters and the version number corresponding to each preset global model parameter.

[0108] The second version mapping data may be chart data or table data, but is not limited thereto.

[0109] Alternatively, the above-mentioned updating of the global model version number of the digital twin model to obtain the updated global model version number may include calculating the sum of the global model version number of the digital twin model and one as the updated global model version number, that is, , is the global model version number of the digital twin model, but is not limited to this.

[0110] In some implementations, the asynchronous global update or synchronous global update of the global model parameters of the digital twin model may further include: When the model version number data of the idle terminal is greater than or equal to the preset version threshold, control each occupied terminal to stop local training; According to the local model parameters and data volume of each training terminal, combined with the total terminal data volume, the global model parameters of the digital twin model are synchronously updated globally to obtain updated global model parameters; wherein the local model parameters of the training terminals that have completed local training before the current moment and the local model parameters of the idle terminals are both corrected local model parameters, and the local model parameters of the occupied terminals are the local model parameters of the current training round; The global model version number of the digital twin model is updated to obtain an updated global model version number.

[0111] In this implementation, if the model version number data of the idle terminal is greater than or equal to the preset version threshold, it means that the difference between the global model version number of the digital twin model and the local model version number of the idle terminal is large. In this way, it can be determined that the local models of other training terminals are outdated or the training lag is large. Other training terminals may have inherent lag effects and model obsolescence problems. At this time, a synchronous global update is performed.

[0112] Specifically, during the synchronous global update, first, the training terminals that have not completed local training at the current moment are controlled to stop local training; then, according to the local model parameters and data volume of each of the training terminals, combined with the total terminal data volume, the global model parameters of the digital twin model are synchronously globally updated to obtain updated global model parameters, wherein the local model parameters of the training terminals that have completed local training before the current moment and the local model parameters of the idle terminals are both corrected local model parameters, and the local model parameters of the occupied terminals are the local model parameters of the current training round; and, during the synchronous global update, the global model version number of the digital twin model is updated.

[0113] In this way, in this embodiment, when the model version number data of the idle terminal is greater than or equal to the preset version threshold, the global model parameter update of the digital twin model of the industrial heterogeneous terminal is realized by synchronous global update. This can not only effectively reduce the inherent lag effect and model obsolescence problems, but also balance the training cost and model loss of the digital twin model of the industrial heterogeneous terminal, so that the digital twin model of the industrial heterogeneous terminal can quickly respond to the dynamically changeable environment in the industrial intelligent manufacturing scenario, and ensure the data consistency of the virtual and real space of industrial intelligent manufacturing, thereby improving the modeling accuracy of the digital twin model of the industrial heterogeneous terminal.

[0114] The above-mentioned implementation method of synchronously updating the global model parameters of the digital twin model based on the local model parameters and data volume of each training terminal and the total terminal data volume to obtain the updated global model parameters can refer to the content recorded in the above embodiment, which will not be repeated here. It should be noted that in this implementation, for the parameters in the above formula (14), In the synchronous global update, the local model parameters of the training terminals that have completed local training before the current time and the local model parameters of the idle terminals are all corrected local model parameters, which are shown in the above formula (12) ; The local model parameters of the occupied terminal are the local model parameters of the current training round. For example, if the current training round is round of training, the local model parameters of the occupied terminal are The local model parameters of the training round are as shown in the above formula (10) .

[0115] The above-mentioned updating of the global model version number of the digital twin model to obtain the updated global model version number may include searching from the third version mapping data to obtain the version number corresponding to the global model parameters after the synchronous global update as the updated global model version number, so as to realize the update of the global model version number of the digital twin model, wherein the third version mapping data includes multiple preset global model parameters and the version number corresponding to each preset global model parameter.

[0116] The third version mapping data may be chart data or table data, but is not limited thereto.

[0117] Alternatively, the above-mentioned updating of the global model version number of the digital twin model to obtain the updated global model version number may include calculating the sum of the global model version number of the digital twin model and one as the updated global model version number, that is, , but not limited to this.

[0118] In some embodiments, after synchronously updating the global model parameters of the digital twin model according to the local model parameters and data volume of each training terminal in combination with the total terminal data volume to obtain the updated global model parameters, the method may further include: Local training of idle terminals is prohibited; For each occupied terminal, the updated global model version number is used as the local model version number of the occupied terminal. The local model of the occupied terminal is locally trained based on the total training rounds and local model parameters of the occupied terminal and the global model parameters of the digital twin model after synchronous global update.

[0119] In this embodiment, after the synchronous global update, two operations are performed, one of which is to prohibit idle terminals from performing local training, that is, training terminals that have completed local training will no longer participate in subsequent local training; the other operation is to control training terminals that have not yet completed local training to continue training. Specifically, for each occupied terminal, the updated global model version number is used as the local model version number of the occupied terminal, and the local model of the occupied terminal is locally trained based on the total training rounds and local model parameters of the occupied terminal and the global model parameters of the digital twin model after the synchronous global update. In this way, after the global model parameter update of the digital twin model of the industrial heterogeneous terminal is realized, the embodiment controls each occupied terminal to restart local training based on the total training rounds and local model parameters of each occupied terminal and the global model parameters of the digital twin model after the synchronous global update, and updates the local model version number of each occupied terminal, which can balance the training cost and model loss of the digital twin model of the industrial heterogeneous terminal, and reduce the lagging effect and model obsolescence problem of each occupied terminal, thereby improving the modeling accuracy of the digital twin model of the industrial heterogeneous terminal.

[0120] In some implementations, the asynchronous global update or synchronous global update of the global model parameters of the digital twin model may further include: During the asynchronous global update, the idle terminal is prohibited from performing local training, and for each occupied terminal, the local model of the occupied terminal is locally trained based on the total training rounds and local model parameters of the occupied terminal and the global model parameters of the digital twin model before the asynchronous global update.

[0121] In this embodiment, during the asynchronous global update, two operations are performed simultaneously, one of which is to update the global model parameters and global model version number of the digital twin model, and the other is to control the training terminals that have not yet completed local training to continue training. For the other operation mentioned above, idle terminals are prohibited from performing local training, that is, training terminals that have completed local training will no longer participate in subsequent local training; at the same time, each occupied terminal is controlled to continue training. Specifically, for each occupied terminal, the local model of the occupied terminal is locally trained based on the total training rounds and local model parameters of the occupied terminal and the global model parameters of the digital twin model before the asynchronous global update. In this way, this embodiment promotes the local training of other training terminals while realizing the global model parameter update of the digital twin model of the industrial heterogeneous terminal, thereby ensuring the modeling efficiency of the digital twin model of the industrial heterogeneous terminal.

[0122] In some implementations, constructing a digital twin model according to the global model parameters of the digital twin model may include: According to the global model parameters after each asynchronous global update or the global model parameters after each synchronous global update, the digital twin model is mapped in sequence at the digital twin layer.

[0123] In this embodiment, each asynchronous global update or synchronous global update will update the global model parameters of the digital twin model, so multiple updated global model parameters can be obtained. Based on this, according to the global model parameters after each update, a complete, global digital twin model is sequentially constructed in the digital twin layer, and further, based on the global model parameters after the last update, a final, global digital twin model is formed in the digital twin layer, thereby improving the modeling accuracy of the digital twin model. The global digital twin model refers to the digital twin model of industrial heterogeneous terminals in the industrial intelligent manufacturing scenario, and the global digital twin model is applicable to all industrial heterogeneous terminals, and it has universality.

[0124] The above digital twin model can be expressed as the following formula (15): (15); In formula (15), Indicates Digital twin models of industrial heterogeneous terminals in rounds of training; represents the generation function used to generate the digital twin model at the digital twin layer; Indicates The global model parameters of the digital twin model in each training round; Indicates The connection between the physical entity and the virtual entity in the training round, i.e. The connection between the industrial heterogeneous terminals and the virtual entities corresponding to the industrial heterogeneous terminals in the training rounds can be set according to the actual situation. For example, it can be a pre-calibrated mapping table that records the connection between each industrial heterogeneous terminal and the corresponding virtual entity; Indicates The running state of the physical entity in the training round, that is, Operation data of industrial heterogeneous terminals in the training rounds, such as temperature data, pressure data, vibration data, and hardware load data; Indicates The running status of the virtual entity corresponding to the physical entity in the training round, that is, The operation data of the virtual entity corresponding to the industrial heterogeneous terminal in the round of training is obtained in the aforementioned embodiment and will not be described in detail. Indicated in The digital twin model data in the training rounds, including the The operating data and equipment parameters of industrial heterogeneous terminals in rounds of training.

[0125] It is understandable that in actual industrial intelligent manufacturing scenarios, after obtaining the global digital twin model, for any industrial heterogeneous terminal, the digital twin model of the industrial heterogeneous terminal can be obtained based on the global model parameters after the last update, the real-time data of the industrial heterogeneous terminal, and the historical data of the industrial heterogeneous terminal, combined with the above formula (15). Among them, the real-time data of the industrial heterogeneous terminal may include the connection between the industrial heterogeneous terminal and the virtual entity corresponding to the industrial heterogeneous terminal at the current moment, the operation data of the industrial heterogeneous terminal at the current moment, and the operation data of the virtual entity corresponding to the industrial heterogeneous terminal at the current moment, and the historical data of the industrial heterogeneous terminal may include the operation data and equipment parameters of the industrial heterogeneous terminal at the historical moment.

[0126] To facilitate the understanding of the modeling method of the digital twin model mentioned above in the present application, the actual application scenario of the modeling method of the digital twin model mentioned above in the present application is used as an example.

[0127] In the industrial intelligent manufacturing scenario, several industrial heterogeneous terminals and an edge server are configured. The edge server and all industrial heterogeneous terminals are connected to the industrial Internet of Things. Each industrial heterogeneous terminal is built with a corresponding local model, which is the initial digital twin model. The edge server is built with a digital twin layer. The digital twin layer is used to store the global model parameters of the digital twin model and build a global digital twin model. The global digital twin model is a digital twin model suitable for each industrial heterogeneous terminal. Among them, the industrial heterogeneous terminal selected to participate in local training is called a training terminal. Each training terminal is used to perform local training on its local model, and update its local model parameters when completing the current training round or synchronizing the global update, obtain the updated local model parameters and upload them to the edge server; the edge server is used to update the global model parameters of the digital twin model based on the local model parameters uploaded by each heterogeneous terminal, and after all heterogeneous terminals have completed training, build the digital twin model of the industrial heterogeneous terminal according to the global model parameters of the digital twin model. Reference Figure 2 ,The process of building a digital twin model of an industrial heterogeneous terminal in this example is as follows steps S201-S204.

[0128] S201, the edge server determines the computing power and communication capacity of each industrial heterogeneous terminal through the confidence upper bound identification method of the optimistic estimation strategy, and obtains the gradient norm value of the local model of each industrial heterogeneous terminal, and then selects the front from a number of industrial heterogeneous terminals based on the computing power, communication capacity and gradient norm value of the local model of each industrial heterogeneous terminal. Industrial heterogeneous terminals participate in local training.

[0129] Specifically, first, the confidence radius of each industrial heterogeneous terminal relative to the current training round is determined by the above formula (1); the computing efficiency and confidence radius of each industrial heterogeneous terminal are combined with the above formulas (2)-(3) to obtain the computing performance data of each industrial heterogeneous terminal relative to the current training round; the communication efficiency and confidence radius of each industrial heterogeneous terminal are combined with the above formulas (4)-(5) to obtain the communication performance data of each industrial heterogeneous terminal relative to the current training round; the historical loss function of each industrial heterogeneous terminal is combined with the above formula (6) to obtain the gradient norm value of the local model of each industrial heterogeneous terminal relative to the current training round; then, the computing performance data, communication performance data and gradient norm value of each industrial heterogeneous terminal are combined with the above formula (7) to obtain the training characteristic parameters of each industrial heterogeneous terminal; then, according to the training characteristic parameters of each industrial heterogeneous terminal, a number of industrial heterogeneous terminals are sorted from large to small, and the top ones are selected. Industrial heterogeneous terminals serve as industrial heterogeneous terminals participating in local training.

[0130] S202: For each training terminal, the edge server adaptively adjusts the local training rounds of the training terminal based on the computing power of the training terminal and the amount of local data owned by the training terminal through the above formulas (8)-(9), obtains the total training rounds of the training terminal and sends it to the training terminal.

[0131] S203, for each training terminal, the training terminal performs local training on its local model based on its local model parameters and the global model parameters of the digital twin model; if the training terminal completes the local training of the current training round, the training terminal updates its local model parameters once by averaging local gradients, and obtains the updated local model parameters as the local model parameters of the training terminal. The updating method is shown in the above formulas (10)-(11), and it is determined whether the current training round is less than the total training rounds of the training terminal. If so, it means that the local training has not been completed. At this time, the current training round is increased by one, and the step of performing local training on the local model of the training terminal is returned to realize iterative training. Otherwise, the training terminal determines that its local training is completed, and corrects the local model parameters of the training terminal based on the aging principle, as shown in the above formulas (12)-(13), and then obtains the corrected local model parameters and uploads them to the edge server.

[0132] If the edge server receives the local model parameters uploaded by the training terminal that has completed the local training at the current moment, it means that there is a training terminal that has completed the local training at the current moment. At this time, the edge server will determine the training terminal that has completed the local training at the current moment as an idle terminal, and determine the training terminal that has not completed the local training at the current moment as an occupied terminal, and determine whether the model version number data of the idle terminal is less than the preset version threshold. If so, perform asynchronous global update, otherwise perform synchronous global update.

[0133] More specifically, idle terminals When local training is completed, the corrected local model parameters are uploaded to the edge server. The idle terminal is then calculated based on the local model parameters Model version number data , and compare it with the preset version threshold Make a comparison.

[0134] like , the edge server first aggregates idle terminals The local model parameters and global model parameters of the current training round are used to update the global model parameters to form new global model parameters. In the asynchronous global update, the local model parameters of the training terminals that have completed local training before the current time and the local model parameters of the idle terminals are the revised local model parameters, while the local model parameters of the occupied terminals are the local model parameters of the previous training round of the current training round. Then let (that is, the global model version number is increased by one), (Idle terminal Add to collection ), is the global model version number of the digital twin model, represents the set of training terminals that perform asynchronous global updates and do not participate in local training. During this period, the remaining occupied terminals Still using the unupdated global model parameters for local training, represents the set of training terminals, Represents the set of training terminals that perform synchronous global updates.

[0135] like , then all occupied terminals The edge server aggregates the global model parameters and the local model parameters of all training terminals, and updates the global model parameters through formula (14) to form new global model parameters. In the synchronous global update, the local model parameters of the training terminals that have completed local training before the current moment and the local model parameters of the idle terminals are both the corrected local model parameters, and the local model parameters of the occupied terminals are the local model parameters of the current training round. Then let (that is, the global model version number is increased by one), (Idle terminal Add to collection ), which does not participate in local training.

[0136] After the global update is synchronized, the edge server sends the updated global model version number and global model parameters to each occupied terminal; for each occupied terminal, set (That is, the local model version number of each occupied terminal is the updated global model version number), and the updated global model parameters are used to continue local training.

[0137] S204, the edge server maps the weighted global model parameters at the digital twin layer to form a digital twin model applicable to any industrial heterogeneous terminal, as shown in the above formula (15).

[0138] In this example, for any industrial heterogeneous terminal, the digital twin model of the industrial heterogeneous terminal can be obtained based on the global model parameters after the last update, the real-time data of the industrial heterogeneous terminal, and the historical data of the industrial heterogeneous terminal, combined with the above formula (15). Among them, the real-time data of the industrial heterogeneous terminal can include the connection between the industrial heterogeneous terminal and the virtual entity corresponding to the industrial heterogeneous terminal at the current moment, the operating data of the industrial heterogeneous terminal at the current moment, and the operating data of the virtual entity corresponding to the industrial heterogeneous terminal at the current moment, and the historical data of the industrial heterogeneous terminal can include the operating data and equipment parameters of the industrial heterogeneous terminal at the historical moment.

[0139] As an expandable application, after obtaining the digital twin model of each industrial heterogeneous terminal, each industrial heterogeneous terminal is subjected to fault detection. Specifically, the operation status of each industrial heterogeneous terminal can be obtained through the digital twin model of each industrial heterogeneous terminal, and then the operation status of each industrial heterogeneous terminal can be subjected to fault detection using a preset fault detection model to obtain a result that characterizes whether each industrial heterogeneous terminal is faulty. This is conducive to reducing the failure risk of industrial heterogeneous terminals and facilitating preventive equipment maintenance.

[0140] In addition, refer to Figure 3 , the embodiment of the present application also provides a modeling device for a digital twin model, which may include: An acquisition module 301 is used to acquire operation characteristic data of a plurality of heterogeneous terminals; A first processing module 302 is used to select a plurality of heterogeneous terminals participating in local training as training terminals from among a plurality of heterogeneous terminals according to the operation characteristic data of each heterogeneous terminal; The second processing module 303 is used to control each training terminal to perform local training, and perform asynchronous global update or synchronous global update on the global model parameters of the digital twin model during the local training; The third processing module 304 is used to build a digital twin model according to the global model parameters of the digital twin model if all training terminals have completed local training.

[0141] The contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0142] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the application is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Optional embodiments are expected, wherein the order of various operations is changed and the sub-operation described as a part of a larger operation is performed independently.

[0143] In addition, although the present application is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present application. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional techniques of the engineer. Therefore, those skilled in the art can implement the present application set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present application, which is determined by the full scope of the attached claims and their equivalents.

[0144] If the 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 application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several programs to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0145] The logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable programs for implementing logical functions, and may be embodied in any computer-readable medium for use by a program execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch and execute a program from a program execution system, device or apparatus), or in conjunction with such program execution system, device or apparatus. For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by a program execution system, device or apparatus, or in conjunction with such program execution system, device or apparatus.

[0146] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0147] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable program execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0148] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0149] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the claims and their equivalents.

[0150] The above is a specific description of the preferred implementation of the present application, but the present application is not limited to the described embodiments. Technical personnel familiar with the field may make various equivalent modifications or substitutions without violating the spirit of the present application. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present application.

Claims

1. A modeling method for a digital twin model, characterized in that: The method comprises the following steps: Obtaining operational characteristic data of several heterogeneous terminals; According to the operation characteristic data of each of the heterogeneous terminals, selecting a plurality of the heterogeneous terminals participating in the local training as training terminals from among the plurality of the heterogeneous terminals; Controlling each of the training terminals to perform local training, and performing asynchronous global update or synchronous global update on the global model parameters of the digital twin model during the local training; If all the training terminals have completed local training, the digital twin model is constructed according to the global model parameters of the digital twin model.

2. The modeling method of the digital twin model according to claim 1, characterized in that: The operation characteristic data includes operation efficiency, communication efficiency and historical loss function; the historical loss function is the loss function value of the local model of the heterogeneous terminal in the previous training round of the current training round; The selecting, according to the operation characteristic data of each of the heterogeneous terminals, a plurality of the heterogeneous terminals participating in the local training as training terminals from among the plurality of the heterogeneous terminals comprises: According to the number of selections of each of the heterogeneous terminals, combined with the total number of training terminals and the number of historical selections, the confidence radius of each of the heterogeneous terminals relative to the current training round is obtained; wherein the number of selections of the heterogeneous terminals is the number of times the heterogeneous terminals are selected to participate in local training before the current training round; the total number of training terminals is the number of the heterogeneous terminals selected to participate in local training; the number of historical selections is the number of times the historical heterogeneous terminals are selected to participate in local training before the current training round; the historical heterogeneous terminals are the heterogeneous terminals selected to participate in local training in the previous training round of the current training round; Obtaining computing performance data of each heterogeneous terminal relative to the current training round according to the computing efficiency and the number of selections of each heterogeneous terminal and the confidence radius of each heterogeneous terminal relative to the current training round; the computing performance data is used to characterize the computing performance of the heterogeneous terminal when performing local training on its local model in the current training round; According to the communication efficiency and the number of selections of each of the heterogeneous terminals, combined with the confidence radius of each of the heterogeneous terminals relative to the current training round, the communication performance data of each of the heterogeneous terminals relative to the current training round is obtained; the communication performance data is used to characterize the communication performance of the heterogeneous terminal when performing local training on its local model in the current training round; Obtaining, according to the historical loss function of each of the heterogeneous terminals, a gradient norm value of the local model of each of the heterogeneous terminals relative to the current training round; According to the computing performance data and communication performance data of each of the heterogeneous terminals relative to the current training round, combined with the gradient norm value of the local model of each of the heterogeneous terminals relative to the current training round, a training characteristic parameter of each of the heterogeneous terminals is obtained; the training characteristic parameter is used to characterize the performance of the heterogeneous terminal when performing local training on its local model; According to the training characteristic parameters of each of the heterogeneous terminals, a plurality of the heterogeneous terminals participating in the local training are selected from among a plurality of the heterogeneous terminals as the training terminals.

3. The modeling method of the digital twin model according to claim 1, characterized in that: Before controlling each of the training terminals to perform local training, the method further includes the following steps: According to the computing efficiency and data volume of the training terminal, combined with the total number of training terminals, the total terminal data volume and the total computing efficiency, the initial training rounds of the training terminal are updated to obtain the total training rounds of the training terminal; The total terminal data volume is the sum of the data volumes of all the training terminals, and the total computing efficiency is the sum of the computing efficiencies of all the training terminals.

4. The modeling method of the digital twin model according to claim 1, characterized in that: The controlling each training terminal to perform local training includes: Controlling the training terminal to perform local training on the local model of the training terminal based on the local model parameters of the training terminal and the global model parameters of the digital twin model; When the training terminal completes the local training of the current training round, the local model parameters of the training terminal are updated to obtain the updated local model parameters as the local model parameters of the training terminal; Comparing the current training round with the total training rounds of the training terminal; If the current training round is less than the total training round, the current training round is incremented by one, and the step of controlling the training terminal to perform local training on the local model of the training terminal based on the local model parameters of the training terminal and the global model parameters of the digital twin model is returned; If the current training round is greater than or equal to the total training round, it is determined that the training terminal completes local training, and the local model parameters of the training terminal are corrected to obtain corrected local model parameters.

5. The modeling method of the digital twin model according to claim 4, characterized in that: The correcting the local model parameters of the training terminal to obtain corrected local model parameters includes: According to the model version number data of the training terminal, the local model parameters of the training terminal are corrected to obtain corrected local model parameters; wherein the model version number data of the training terminal is the difference between the global model version number of the digital twin model and the local model version number of the training terminal.

6. The modeling method of the digital twin model according to claim 1, characterized in that: The asynchronous global update or synchronous global update of the global model parameters of the digital twin model includes: If it is detected that there is the training terminal that has completed the local training at the current moment, the training terminal that has completed the local training at the current moment is determined as an idle terminal, and the training terminal that has not completed the local training at the current moment is determined as an occupied terminal; When the model version number data of the idle terminal is less than a preset version threshold, the global model parameters of the digital twin model are asynchronously updated globally according to the local model parameters and data volume of each training terminal, combined with the total terminal data volume, to obtain updated global model parameters; wherein the local model parameters of the training terminal that completed local training before the current moment and the local model parameters of the idle terminal are both revised local model parameters, and the local model parameters of the occupied terminal are the local model parameters of the previous training round of the current training round; The global model version number of the digital twin model is updated to obtain an updated global model version number.

7. The modeling method of the digital twin model according to claim 6, characterized in that: The asynchronous global update or synchronous global update of the global model parameters of the digital twin model further includes: When the model version number data of the idle terminal is greater than or equal to the preset version threshold, controlling each of the occupied terminals to stop local training; According to the local model parameters and data volume of each training terminal, combined with the total terminal data volume, the global model parameters of the digital twin model are synchronously updated globally to obtain updated global model parameters; wherein the local model parameters of the training terminals that have completed local training before the current moment and the local model parameters of the idle terminals are both revised local model parameters, and the local model parameters of the occupied terminals are the local model parameters of the current training round; The global model version number of the digital twin model is updated to obtain an updated global model version number.

8. The modeling method of the digital twin model according to claim 7, characterized in that: After synchronously updating the global model parameters of the digital twin model according to the local model parameters and data volume of each training terminal in combination with the total terminal data volume to obtain the updated global model parameters, the method further includes the following steps: prohibiting the idle terminal from performing local training; For each occupied terminal, the updated global model version number is used as the local model version number of the occupied terminal, and the local model of the occupied terminal is locally trained based on the total training rounds and local model parameters of the occupied terminal and the global model parameters of the digital twin model after synchronous global update.

9. The modeling method of the digital twin model according to claim 6, characterized in that: The asynchronous global update or synchronous global update of the global model parameters of the digital twin model further includes: During the asynchronous global update, the idle terminal is prohibited from performing local training, and for each occupied terminal, the local model of the occupied terminal is locally trained based on the total training rounds and local model parameters of the occupied terminal and the global model parameters of the digital twin model before the asynchronous global update.

10. A modeling device for a digital twin model, characterized in that: include: An acquisition module, used to acquire operation characteristic data of a number of heterogeneous terminals; A first processing module, configured to select, from among a plurality of the heterogeneous terminals, a plurality of the heterogeneous terminals participating in the local training as training terminals according to the operation characteristic data of each of the heterogeneous terminals; A second processing module is used to control each of the training terminals to perform local training, and to perform asynchronous global update or synchronous global update on the global model parameters of the digital twin model during the local training; The third processing module is used to construct the digital twin model according to the global model parameters of the digital twin model if all the training terminals have completed local training.

Images