Business model dynamic configuration and online updating method and system based on digital twinning

Abstract

The invention provides a dynamic configuration and online updating method and system for a business model based on digital twinning. The method comprises the following steps: constructing a digital twinning 3D model; collecting data of a physical system in real time, configuring point locations and simulating and driving the digital twinborn body; configuring and executing a business model calculation graph; a user-defined algorithm registration mechanism is introduced, and dynamic task mapping and calculation graph updating are achieved; carrying out online incremental learning and a soft update strategy of parameters; through calculation results and monitoring data, the state and behavior of the 3D digital twin are driven in real time, and the state change of a physical system is accurately reflected. Dynamic configuration and parameter updating of the digital twinning service model are completed in the form of the computational graph, the adaptive capacity of the digital twinning system to the real environment is improved, and therefore intelligent management and control are achieved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing technology, and in particular to a method and system for dynamic configuration and online updating of business models based on digital twins. Background Art

[0002] Digital Twin technology has received extensive attention and application in recent years. Its core idea is to achieve real-time monitoring, analysis and optimization of the system by building a virtual model of the physical system. However, traditional digital twin models are usually static and cannot adapt to dynamic changes in the environment and system status.

[0003] Dynamic model update mechanisms require that models evolve over time to reflect the actual state of the device. This requires a complex model management system that can continuously update model parameters based on real-time data while maintaining model consistency and interpretability. When designing such a dynamic update mechanism, we must consider both computational efficiency and ensure that updates do not cause model drift or instability.

[0004] Therefore, the present invention proposes a method for dynamic configuration and updating of business models for digital twins, which enables the digital twin model to continuously optimize and update itself through the introduction of real-time data feedback and online learning. Summary of the invention

[0005] In view of the defects in the prior art, the purpose of the present invention is to provide a method and system for dynamic configuration and online update of business models based on digital twins.

[0006] A method for dynamic configuration and online update of a business model based on digital twins provided by the present invention includes: Step S1: construct a digital twin 3D model of the target object; Step S2: collecting target object point data at a fixed frequency and saving it in the IHD real-time database; Step S3: construct a business model calculation graph; based on the collected target object point data, use the business model calculation graph to perform logical calculations to obtain calculation results; Step S4: Transmit the calculation results to the digital twin 3D model to dynamically drive the state and behavior of the digital twin.

[0007] Preferably, the step S3 comprises: Step S3.1: define the business logic as a Node class and encapsulate it into calculation blocks. Each calculation block includes: multiple input interfaces, output interfaces, configuration parameters, calculation methods, and status information;

[0008] in, Indicates the data received by the input interface. Represents the configuration parameters bound to the computational block, Indicates the calculation method, Indicates status information. Indicates the data received by the output interface; Step S3.2: Build a business model based on multiple computing blocks and manage it through a computing graph; ; Among them, G represents the business model; It is a collection of nodes, representing a single business logic. It is a set of edges, which represents the dependency relationship between business logics. After the business models are connected by directed edges, a composite chain of functions is formed: ; Step S3.3: Use the depth-first search DFS method to perform topological sorting of the computation graph, and save the sorting results in a list ; Execute each business logic in sorted order; where, Represents business logic; Step S3.4: Update the status of each computation block by traversing the computation graph and executing the business logic according to the topological sorting until global convergence is reached.

[0009] Preferably, the method further comprises: Step S3.5: Obtain the uploaded algorithm file through the graphical editor, obtain the defined function based on the uploaded algorithm file; and construct a temporary function space respectively. and the main function space ; The main function space Used to store the initial function set; the temporary function space Used to store uploaded algorithm files; functions in temporary function space When , it is added to the main function space; when , it is not added to the main function space; Step S3.6: extracting the function's meta information based on the obtained algorithm file, including parameters, default values, and type annotations; converting and filtering the extracted function's meta information according to preset requirements; Step S3.7: Use the AST abstract syntax tree to parse the function object and obtain the function's meta information, including name, parameters, default values, function body, return type, comments, etc.; pass the function meta information to the graphical editor in json format, automatically generate the corresponding calculation block, and realize dynamic registration of business logic functions.

[0010] Preferably, step S3.6 includes: assuming that the registration function is , whose parameters are expressed as a set , where each parameter Each has its own type and default values ; For each parameter , the input string Convert to , ;in, Is the converted value, ensuring that the converted parameter conforms to the function Signature.

[0011] Preferably, the method further comprises: implementing self-learning of the business model based on the evaluation network, the target network and the parameter migration to implement updating of the business model; The network structures of the evaluation network and the target network are consistent; The evaluation network receives new data and continuously optimizes to achieve online learning, and then migrates the learned parameters to the target network through parameter migration, thereby achieving business model update.

[0012] Preferably, the evaluation network dynamically updates parameters using an online incremental learning method; Assume that the parameters of the evaluation network are , the goal of evaluating the network at time t is to minimize a loss function , and its update rule is expressed as:

[0013] in, is the learning rate, is the gradient under the current data.

[0014] Preferably, migrating the learned parameters to the target network by means of parameter migration includes: Assume the evaluation network parameters are , the target network parameters are , introducing the parameter soft update strategy, the parameter update formula of the target network is:

[0015] in, is a small constant whose value is between [0,1].

[0016] Preferably, the output nodes of the business model calculation graph include display nodes, storage nodes, and transmission nodes; The display node can directly view the calculation results, including scalars, matrices, tables, and graphs; The storage node stores the calculation result in a database; The transmission node configures an interface to transmit the calculation results.

[0017] Preferably, step S4 includes: configuring an http interface on the output node of the computation graph, transmitting data to the digital twin 3D model, and dynamically driving the state and behavior of the digital twin to reflect changes in the physical system.

[0018] A business model dynamic configuration and online update system based on digital twins provided by the present invention includes: Module M1: Build a digital twin 3D model of the target object; Module M2: collects target object point data at a fixed frequency and saves it in the IHD real-time database; Module M3: Construct a business model calculation graph; based on the collected target object point data, use the business model calculation graph to perform logical calculations to obtain calculation results; Module M4: Transmits the calculation results to the digital twin 3D model and dynamically drives the state and behavior of the digital twin.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention provides a unified framework for dynamic configuration and online update of digital twin business models. By building and executing business model calculation graphs, real-time data monitoring and online update of model parameters are performed, and finally the state and behavior of the digital twin are driven by simulation; 2. The present invention introduces a parameter soft update strategy so that parameter update will not cause model drift or instability; 3. The whole process improves the digital twin system's ability to adapt to the real environment. Through continuous optimization and dynamic updating, the system has high flexibility and responsiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 This is a flow chart of the model dynamic configuration and online update method based on digital twins in the present invention.

[0021] Figure 2 This is a point configuration diagram for driving the digital twin in the present invention.

[0022] Figure 3 It is the business logic calculation graph editor in the present invention.

[0023] Figure 4 This is a diagram of the dynamic function registration process in the present invention.

[0024] Figure 5 This is a sample diagram of the digital twin in the present invention. DETAILED DESCRIPTION

[0025] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0026] Example 1 The purpose of the present invention is to overcome the shortcomings of the prior art and propose a method for dynamic configuration and online update of business models based on digital twins, which combines the configuration and parameter update of business models into digital twins, simplifies the configuration process of business models in complex industrial environments, and is also conducive to the optimization and update of digital twin systems.

[0027] To achieve the above-mentioned purpose, the present invention provides a model dynamic configuration and online update method based on digital twins, such as Figure 1 As shown, the following steps are included: Step 1: Build a digital twin 3D model, simulate the operating status and behavior of the system according to the structure and operating principle of the physical system, and provide an interface to receive data in real time; Step 2: Collect data of the physical system in real time, configure points and simulate the driving of the digital twin; in this embodiment, the collected data is saved in a real-time database, such as Figure 2 As shown, the points can be configured online, and then the required point data, such as temperature, pressure, flow, etc., can be transmitted to the 3D model through the web interface.

[0028] Step 3: Configuration and execution of business model calculation graph; Step 3.1: Creation of business logic; The business logic defines a Node class, which encapsulates a computing module. Each computing block can have multiple input interfaces, output interfaces, configuration parameters, computing methods, and status information. The process is represented as:

[0029] in, Indicates the data received by the input interface. Represents the configuration parameters bound to the computational block, Indicates the calculation method, Indicates status information. Indicates the data received by the output interface.

[0030] Step 3.2: Creation of business model The computational graph is used to manage and execute multiple business logics. A single business logic is essentially a computational unit, and multiple computational units constitute a complete business model. The business model can be represented as a directed acyclic graph. ,in It is a collection of nodes, representing a single business logic. It is a set of edges, representing the dependencies between business logics. After the business models are connected by directed edges, a composite chain of functions is formed: .

[0031] Step 3.3: Execution of the computation graph Use the depth-first search DFS method to perform topological sorting of the computational graph, and save the sorting results in a list . Execute each business logic in sorted order. If an error occurs in a business logic, the calculation unit and all subsequent related calculation units will be skipped, and other branches can continue to execute, ensuring that local errors do not affect global calculations. In this embodiment, L is a list that contains the execution order of each node V after background parsing. We organize it in a graph way. A graph may have one business model or multiple business models without interference.

[0032] Step 3.4: State transfer and update: By traversing the computation graph and executing the business logic in topological order, the state of each computation unit is updated until global convergence is reached, that is, all logic is calculated. Its status includes: whether the computation unit is executed and the execution status. The execution status is divided into three types: success, failure, and error. The computation graph is serialized into a file for easy persistence and recovery operations.

[0033] Step 3.5: Graphical configuration like Figure 3 As shown, a graphical editor is constructed. Users can build complex business models by dragging and connecting. The constructed business model diagram is saved in XML format for subsequent loading. In the editor, when adding a new node, a unique calculation unit name needs to be generated to avoid name conflicts.

[0034] Step 4: Introduce a custom algorithm registration mechanism to implement dynamic task mapping and computation graph updates; Step 4.1: Registration of custom algorithm: You can upload the algorithm file through the graphical editor in step 3.5, such as Figure 4 As shown, functions are read and registered from the file, and each function is a computing unit. All defined functions are dynamically registered into the function space, and temporary function spaces are constructed respectively. and the main function space . Main function space Store the initial set of functions; when the algorithm file is imported, it is stored in the temporary function space The functions in the temporary function space are When , add it to the main function space; When adding a function, it needs to be confirmed by the user before it is added to the main function space to avoid conflicts with existing functions in the main function space, thereby changing the business logic of the existing calculation graph. In this embodiment, functions can be dynamically registered, including uploading files, editing and managing existing functions, and saving to complete the registration process.

[0035] Step 4.2: Parameter type parsing and conversion: After uploading the algorithm file, you need to extract the function's meta information from the string, including parameters, default values, type annotations, and convert and filter them through rules. Assume that the registered function is , whose parameters are expressed as a set , where each parameter Each has its own type and default values For each parameter , the input string Convert to ,Right now ,in Is the converted value, ensuring that the converted parameter conforms to the function If the conversion fails, an exception is thrown.

[0036] Step 4.3: Function extraction Use the AST abstract syntax tree to parse the function object and obtain the function's meta information, including name, parameters, default values, function body, return type, comments, etc.; pass the function meta information in json format to the graphical editor in step 3.5 to automatically generate the corresponding calculation unit to realize dynamic registration of business logic functions.

[0037] Step 4.4: Dynamic task mapping and computation graph update; After building the business model calculation graph, you can schedule tasks to execute the process calculation of the graph. After the task is started, the graph will be reloaded and parsed each time it is scheduled to run, that is, the calculation units in the calculation graph can be dynamically updated after changes occur, without the need to manually pause or rebuild the calculation graph.

[0038] Step 5: Online incremental learning and soft update strategy of parameters; Step 5.1: Access real-time data The calculation graph is used to represent the dependencies of the entire data processing and model update process, and the complete business model calculation Figure 1Generally, it includes nodes such as data input, preprocessing, algorithm, post-processing and output. In the industrial field, the data input node obtains data from the real-time database through the configuration interface and point position, and then passes it to the subsequent nodes.

[0039] Step 5.2: Online incremental learning of algorithm nodes Algorithm nodes include evaluation network nodes, target network nodes, and parameter migration nodes. The evaluation network node uses new data to train the model in real time, and the target network node is used to stabilize parameters to prevent training instability caused by too frequent parameter updates. The network structures of the two need to be consistent.

[0040] The evaluation network and the target network are deep learning algorithms with built-in neural network structures. The network structures of the two are the same.

[0041] The evaluation network is used for online learning, while the target network is not learned.

[0042] The evaluation network can receive new data to continuously optimize the model and adjust internal weight parameters, while the target network directly calculates the output results.

[0043] Because the network structure is consistent, the parameters learned by the evaluation network can be migrated to the target network. However, if the parameters are directly replaced, the target network will be unstable. Therefore, a parameter migration node is introduced. This node will perform new processing on the evaluation network parameters and replace the target network parameters after processing.

[0044] The evaluation network nodes use online incremental learning methods to dynamically update parameters. Assume that the parameters of the model are , at time t the goal of the model is to minimize a loss function , its update rule can be expressed as:

[0045] in, is the learning rate, is the gradient under the current data.

[0046] Step 5.3: Parameter soft update strategy The evaluation network nodes and target network nodes in the above steps update the evaluation network parameters to the target network through the parameter migration node. Suppose the evaluation network parameters are , the target network parameters are , introducing the parameter soft update strategy, the parameter update formula of the target network is:

[0047] in, It is a very small constant with a value between [0,1], which prevents the model from being unstable due to too fast update of the target network parameters.

[0048] Step 5.3: Output Node The output nodes of the business model calculation graph are divided into multiple types, including display nodes, save nodes, and transmission nodes. Display nodes can directly view the calculation results, including scalars, matrices, tables, and graphs; save nodes save the calculation results in the database for further analysis, historical queries, or long-term data tracking; transmission nodes configure interfaces to transmit calculation results to other functions.

[0049] Step 6: Configure the http interface on the output node of the computation graph to transfer data to the digital twin 3D model. Figure 5 As shown, the digital twin constructed for this embodiment drives the state and behavior of the 3D digital twin in real time through calculation results and monitoring data, accurately reflecting the state changes of the physical system.

[0050] The present invention also provides a business model dynamic configuration and online update system based on digital twins. The business model dynamic configuration and online update system based on digital twins can be implemented by executing the process steps of the business model dynamic configuration and online update method based on digital twins, that is, technical personnel in this field can understand the business model dynamic configuration and online update method based on digital twins as the preferred implementation mode of the business model dynamic configuration and online update system based on digital twins.

[0051] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0052] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for dynamic configuration and online update of business models based on digital twins, characterized in that: include: Step S1: construct a digital twin 3D model of the target object; Step S2: collecting target object point data at a fixed frequency and saving it in the IHD real-time database; Step S3: construct a business model calculation graph; based on the collected target object point data, use the business model calculation graph to perform logical calculations to obtain calculation results; Step S4: Transmit the calculation results to the digital twin 3D model to dynamically drive the state and behavior of the digital twin.

2. The method for dynamic configuration and online update of business models based on digital twins according to claim 1 is characterized in that: The step S3 comprises: Step S3.1: define the business logic as a Node class and encapsulate it into calculation blocks. Each calculation block includes: multiple input interfaces, output interfaces, configuration parameters, calculation methods, and status information; in, Indicates the data received by the input interface. Represents the configuration parameters bound to the computational block, Indicates the calculation method, Indicates status information. Indicates the data received by the output interface; Step S3.2: Build a business model based on multiple computing blocks and manage it through a computing graph; ； Among them, G represents the business model; It is a collection of nodes, representing a single business logic. It is a set of edges, which represents the dependency relationship between business logics. After the business models are connected by directed edges, a composite chain of functions is formed: ; Step S3.3: Use the depth-first search DFS method to perform topological sorting of the computation graph, and save the sorting results in a list ; Execute each business logic in sorted order; where, Represents business logic; Step S3.4: Update the status of each computation block by traversing the computation graph and executing the business logic according to the topological sorting until global convergence is reached.

3. The method for dynamic configuration and online update of business models based on digital twins according to claim 1 is characterized in that: The method further comprises: Step S3.5: Obtain the uploaded algorithm file through the graphical editor, obtain the defined function based on the uploaded algorithm file; and construct a temporary function space respectively. and the main function space ; The main function space Used to store the initial function set; the temporary function space Used to store uploaded algorithm files; functions in temporary function space When , it is added to the main function space; when , it is not added to the main function space; Step S3.6: extracting the function's meta information based on the obtained algorithm file, including parameters, default values, and type annotations; converting and filtering the extracted function's meta information according to preset requirements; Step S3.7: Use the AST abstract syntax tree to parse the function object and obtain the function's meta information, including name, parameters, default values, function body, return type, comments, etc.; pass the function meta information to the graphical editor in json format, automatically generate the corresponding calculation block, and realize dynamic registration of business logic functions.

4. The method for dynamic configuration and online update of business models based on digital twins according to claim 3 is characterized in that: The step S3.6 includes: assuming that the registration function is , whose parameters are expressed as a set , where each parameter Each has its own type and default values ; For each parameter , the input string Convert to , ;in, Is the converted value, ensuring that the converted parameter conforms to the function Signature.

5. The method for dynamic configuration and online update of business models based on digital twins according to claim 2 is characterized in that: The method further includes: implementing self-learning of the business model based on the evaluation network, the target network and the parameter migration to implement updating of the business model; The network structures of the evaluation network and the target network are consistent; The evaluation network receives new data and continuously optimizes to achieve online learning, and then migrates the learned parameters to the target network through parameter migration, thereby achieving business model update.

6. The method for dynamic configuration and online update of business models based on digital twins according to claim 5 is characterized in that: The evaluation network dynamically updates parameters using an online incremental learning method; Assume that the parameters of the evaluation network are , the goal of evaluating the network at time t is to minimize a loss function , and its update rule is expressed as: in, is the learning rate, is the gradient under the current data.

7. The method for dynamic configuration and online update of business models based on digital twins according to claim 5 is characterized in that: The method of migrating the learned parameters to the target network by means of parameter migration includes: Assume the evaluation network parameters are , the target network parameters are , introducing the parameter soft update strategy, the parameter update formula of the target network is: in, is a small constant whose value is between [0,1].

8. The method for dynamic configuration and online update of business models based on digital twins according to claim 1 is characterized in that: The output nodes of the business model calculation graph include display nodes, storage nodes, and transmission nodes; The display node can directly view the calculation results, including scalars, matrices, tables, and graphs; The storage node stores the calculation result in a database; The transmission node configures an interface to transmit the calculation results.

9. The method for dynamic configuration and online update of business models based on digital twins according to claim 1 is characterized in that: The step S4 includes: configuring the http interface of the output node of the calculation graph, transmitting the data to the digital twin 3D model, dynamically driving the state and behavior of the digital twin, so as to reflect the changes of the physical system.

10. A business model dynamic configuration and online update system based on digital twins, characterized in that: include: Module M1: Build a digital twin 3D model of the target object; Module M2: collects target object point data at a fixed frequency and saves it in the IHD real-time database; Module M3: Construct a business model calculation graph; based on the collected target object point data, use the business model calculation graph to perform logical calculations to obtain calculation results; Module M4: Transmits the calculation results to the digital twin 3D model and dynamically drives the state and behavior of the digital twin.

Images