Intelligent manufacturing digital twin modeling method and system

Abstract

The application provides a smart manufacturing digital twin modeling method and system, which comprises the following steps: according to the physical layout of the production equipment and the processing sequence of the materials, a digital twin model is decomposed into multiple functional nodes; at least two functional nodes along the process route are divided into one process correlation section; for each process correlation section, a node-level error is generated by comparing the simulation value and the measured value; for adjacent process correlation sections, a section-to-section transmission error is obtained by analyzing the difference between the output of the previous section and the input of the next section at the coupling boundary; a joint fidelity coefficient is constructed by fusing the node-level error and the section-to-section transmission error; when the joint fidelity coefficient exceeds a threshold value, the model parameters of the functional nodes or the state coupling mapping rules between the process correlation sections are iteratively corrected until the fidelity coefficient is lower than the threshold value. By using the scheme of the application, double-layer fusion correction of the node-level and section-to-section transmission errors can be realized, thereby improving the dynamic synchronization fidelity of the digital twin model and the physical production line.

Description

Technical Field

[0001] This application relates to the field of digital twin technology, and more specifically, to a digital twin modeling method and system for intelligent manufacturing. Background Technology

[0002] Digital twin technology is increasingly widely used in the field of intelligent manufacturing. By constructing a virtual model corresponding to the physical production line, it enables real-time monitoring, prediction and optimization of the production process. Currently, digital twin modeling methods mainly focus on improving the simulation accuracy of individual equipment or workstations, and using sensor data to drive model parameter calibration in order to reduce the deviation between simulation and actual measurement.

[0003] However, existing technologies typically treat each workstation as an independent unit during modeling, focusing only on the simulation accuracy of individual nodes and neglecting the coupling effect of state transfer between processes. In actual production, the output state of the preceding process becomes the input of the following process, and errors propagate step by step along the process route and may be amplified. Traditional methods lack quantitative assessment and correction mechanisms for the consistency of state transfer between adjacent process segments, resulting in significant deviations from the physical production line even if the local simulation accuracy of each workstation meets the standards, due to boundary transfer distortion. Therefore, how to achieve dual-layer fusion correction of node-level and inter-segment transfer errors to improve the dynamic synchronization fidelity between the digital twin model and the physical production line has become a challenge for the industry. Summary of the Invention

[0004] This application provides a digital twin modeling method and system for intelligent manufacturing, which can achieve dual-layer fusion correction of node-level and inter-segment transmission errors, thereby improving the dynamic synchronization fidelity between the digital twin model and the physical production line.

[0005] In a first aspect, this application provides a digital twin modeling method for intelligent manufacturing, comprising the following steps: Based on the physical layout of the production equipment and the processing sequence of the materials, the digital twin model is deconstructed into multiple functional nodes that have a preset mapping relationship with the physical workstations. Divide at least two functional nodes connected in series along at least one process route into a process-related segment; For each process-related segment, the simulated values ​​of each functional node within the process-related segment are compared with the actual sensor values ​​of the corresponding physical workstation to generate node-level errors. For two sequentially adjacent process segments, analyze the degree of difference between the output state of the previous segment and the input state of the next segment at the coupling boundary to obtain the inter-segment propagation error; By integrating the node-level error and the inter-segment propagation error, a joint fidelity coefficient is constructed. When the joint fidelity coefficient exceeds a preset threshold, the model parameters of the functional nodes and / or the state coupling mapping rules between process-related segments are iteratively corrected until the joint fidelity coefficient is lower than the preset threshold, so that the digital twin model and the physical production line state remain dynamically consistent.

[0006] In some embodiments, according to the physical layout of the production equipment and the processing sequence of materials, the digital twin model is deconstructed into multiple functional nodes that have a preset mapping relationship with the physical workstations, specifically including: Obtain the physical layout diagram and material processing flow chart of the target production line; Determine the spatial location and process attributes of each physical workstation based on the physical layout diagram; According to the material processing sequence specified in the material processing process flow chart, each physical workstation is mapped to an independent functional node, and a process attribute label is set for each functional node. Based on the process attribute labels and spatial locations, configure corresponding simulation model interfaces and data interaction protocols for each functional node to complete the deconstruction of the functional node.

[0007] In some embodiments, dividing at least two functional nodes connected in series along at least one process route into a process association segment specifically includes: Extract the process connection relationships between all functional nodes and construct a directed graph with functional nodes as vertices and material flow direction as edges; In the directed graph, all branchless paths from the starting point to the ending point are traversed, and a continuous and unbranched sequence of functional nodes on each path is marked as a candidate process route. For each candidate process route, the functional nodes on the route are divided into several groups based on the coupling strength between nodes. The functional nodes in each group are spatially and temporally continuous and the process correlation is higher than a set threshold. Define each group of functional nodes as a process association segment, and record the input status interface and output status interface of the process association segment.

[0008] In some embodiments, for each process-related segment, comparing the simulated values ​​of each functional node within that process-related segment with the actual sensor measurements of the corresponding physical workstation to generate node-level errors specifically includes: Read the simulation output state vectors of each functional node at the current moment from the digital twin model; The measured state vectors at the same moment are collected by sensors deployed at the corresponding physical workstations; The simulated state vector of each functional node is compared with the measured state vector component by component, and the absolute deviation of each component is calculated. The absolute deviations of all components are fused to obtain the scalar error value of the functional node; The scalar error values ​​of all functional nodes within the process-related segment are summarized to obtain the node-level error.

[0009] In some embodiments, for two sequentially adjacent process segments, the degree of difference between the output state of the preceding segment and the input state of the following segment at the coupling boundary is analyzed to obtain the inter-segment propagation error, which specifically includes: Identify the set of state variables and their numerical range defined by the output state interface of the previous process segment to obtain the actual output state vector of the previous segment. Identify the set of state variables and their numerical ranges expected by the input state interface of the next process segment, and obtain the state vector expected to be received by the next process segment. At the coupling boundary, the actual output state vector of the previous segment is aligned and matched with the expected received state vector of the next segment, and the difference metric between the two aligned state vectors is calculated. Based on the difference metric, the inter-segment propagation error is calculated, and the position information is marked at the coupling boundary.

[0010] In some embodiments, constructing a joint fidelity coefficient by fusing the node-level error and the inter-segment propagation error specifically includes: Arrange the node-level errors of each process-related segment in the process sequence to construct a node error diagonal matrix. The diagonal elements of this matrix correspond to the node-level errors of the corresponding segments, and the off-diagonal elements are zero. Based on the inter-segment transmission error between adjacent process segments, construct the upper diagonal error matrix; The joint error matrix is ​​obtained by adding the nodal error diagonal matrix and the upper diagonal error matrix. Calculate the spectral norm of the joint error matrix and use the result as the joint fidelity coefficient.

[0011] In some embodiments, constructing the upper diagonal error matrix based on the inter-segment transfer error between adjacent process segments specifically includes: The process-related segments are numbered from the first to the Nth according to the process route sequence, where N is the total number of process-related segments; For each pair of sequentially adjacent segments i and i+1, obtain their corresponding inter-segment propagation error; Construct an N x N zero matrix, assign the element in the i-th row and i+1-th column to the inter-segment propagation error between the i-th segment and the i+1-th segment, and set the remaining elements to zero. The resulting matrix is ​​the upper diagonal error matrix.

[0012] Secondly, this application provides an intelligent manufacturing digital twin modeling system, the system comprising: The deconstruction module is used to deconstruct the digital twin model into multiple functional nodes with preset mapping relationships to physical workstations, according to the physical layout of the production equipment and the processing sequence of materials. A partitioning module is used to divide at least two functional nodes connected in series along at least one process route into a process-related segment; The node error generation module is used to compare the simulation values ​​of each functional node in each process-related segment with the actual sensor values ​​of the corresponding physical workstation to generate node-level errors. The inter-segment error generation module is used to analyze the degree of difference between the output state of the previous segment and the input state of the next segment at the coupling boundary for two sequentially adjacent process segments, and to obtain the inter-segment transmission error. The fusion module is used to fuse the node-level error and the inter-segment propagation error to construct a joint fidelity coefficient; The correction module is used to iteratively correct the model parameters of the functional nodes and / or the state coupling mapping rules between process-related segments when the joint fidelity coefficient exceeds a preset threshold, until the joint fidelity coefficient is lower than the preset threshold, so that the digital twin model and the physical production line state remain dynamically consistent.

[0013] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described intelligent manufacturing digital twin modeling method.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent manufacturing digital twin modeling method.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: This application deconstructs a digital twin model into functional nodes and process-related segments, independently generating node-level errors and inter-segment propagation errors. It then constructs a joint fidelity coefficient based on the matrix spectral norm and implements targeted iterative correction based on the dominant error source, ultimately improving the dynamic synchronization fidelity between the digital twin model and the physical production line. First, the model is deconstructed into functional nodes according to equipment layout and processing sequence, and at least two nodes connected in series along the process route are divided into process-related segments, establishing a clear hierarchical structure for two-layer error assessment. Second, within each process-related segment, simulated values ​​and sensor measurements are compared to generate node-level errors, quantifying local fidelity. Third, the coupling boundary differences between the output state and the desired input state between adjacent segments are analyzed to generate inter-segment propagation errors, separating state propagation distortion from node errors. This overcomes the limitations of traditional methods that only focus on single-point accuracy. This approach ignores the shortcomings of boundary coupling. Then, node-level errors are arranged into a diagonal matrix, and inter-segment propagation errors are arranged into an upper diagonal matrix. These are summed to calculate the spectral norm as the joint fidelity coefficient. This coefficient comprehensively reflects the propagation and amplification effect of local deviations and errors along the process route through the maximum singular value. Compared to conventional aggregation methods such as weighted summation, it has higher evaluation sensitivity and provides a more accurate measurement basis for dynamic synchronization fidelity. Finally, when the joint fidelity coefficient exceeds a threshold, the mean values ​​of the two types of errors are compared to determine the dominant error source: if node-level errors dominate, gradient descent is used to iteratively update the model parameters of functional nodes; if inter-segment propagation errors dominate, state coupling mapping rules such as unit transformation, coordinate transformation, attribute inheritance, and time delay are iteratively corrected; joint correction is performed when necessary, and the joint fidelity coefficient is recalculated after each correction until it falls below the threshold. Through the above-mentioned collaborative mechanism of hierarchical error modeling, spectral norm fusion, and directional correction of dominant error sources, this application can achieve dual-layer fusion correction of node-level and inter-segment propagation errors, thereby improving the dynamic synchronization fidelity between the digital twin model and the physical production line. Attached Figure Description

[0016] Figure 1 This is an exemplary flowchart of a digital twin modeling method for intelligent manufacturing according to some embodiments of this application; Figure 2 This is a flowchart illustrating the division of process-related segments according to some embodiments of this application; Figure 3 This is a flowchart illustrating the determination of inter-segment transmission error according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a smart manufacturing digital twin modeling system according to some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device for implementing a digital twin modeling method for intelligent manufacturing, according to some embodiments of this application. Detailed Implementation

[0017] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] refer to Figure 1 The figure is an exemplary flowchart of a digital twin modeling method for intelligent manufacturing according to some embodiments of this application. The digital twin modeling method for intelligent manufacturing mainly includes the following steps: In step 101, the digital twin model is deconstructed into multiple functional nodes with preset mapping relationships with the physical workstations, according to the physical layout of the production equipment and the processing sequence of the materials.

[0019] It should be noted that the digital twin model in this application refers to a virtual mapping system constructed based on the physical layout of production equipment and the processing sequence of materials. It consists of multiple functional nodes with preset mapping relationships to physical workstations, connected in series according to the process route to form process-related segments. This system can receive real-time sensor data from the physical production line and drive simulation calculations to simulate and reflect the operating state of the physical production line. This digital twin model has the following characteristics: First, the model's structure is consistent with the spatial layout and processing sequence of the physical production line, with each functional node corresponding to a physical workstation. Second, the model continuously compares the simulation output with the physical measured data through a two-layer evaluation mechanism of node-level error and inter-segment transmission error. Third, the model has self-correcting capabilities; when the joint fidelity coefficient exceeds a preset threshold, it can iteratively adjust the model parameters or state coupling mapping rules to ensure that the model dynamically maintains consistency with the state of the physical production line.

[0020] In some embodiments, deconstructing a digital twin model into multiple functional nodes with preset mapping relationships to physical workstations, according to the physical layout of production equipment and the processing sequence of materials, can be achieved through the following steps: Obtain the physical layout diagram and material processing flow chart of the target production line; Determine the spatial location and process attributes of each physical workstation based on the physical layout diagram; According to the material processing sequence specified in the material processing process flow chart, each physical workstation is mapped to an independent functional node, and a process attribute label is set for each functional node. Based on the process attribute labels and spatial locations, configure corresponding simulation model interfaces and data interaction protocols for each functional node to complete the deconstruction of the functional node.

[0021] It should be noted that the physical layout diagram is an engineering drawing used to record the spatial location and relative arrangement of each physical workstation on the production line; the material processing flow chart is a flowchart used to describe the processing sequence and process parameters that materials undergo from raw materials to finished products; the functional node is an executable unit used to carry the simulation logic, state update, and external communication functions of a single physical workstation in the digital twin model; the process attribute label is a classification tag string used to identify the process type and key process parameters corresponding to the functional node; and the data interaction protocol is used to standardize the message format and communication timing rules for data exchange between the functional node and the sensors and actuators on the physical workstation.

[0022] In practice, during the process of obtaining the physical layout diagram and material processing flow diagram of the target production line, the implementer can retrieve the physical layout diagram containing the location information of all physical equipment from the computer-aided design system of the production line, and at the same time obtain the material processing flow diagram from the manufacturing execution system. This diagram clearly marks each processing step of the material flow and its sequence. By comparing the above two documents, the coordinate range of each physical workstation in space and the type of process task undertaken by that workstation can be determined.

[0023] In specific implementation, during the process of determining the spatial location and process attributes of each physical workstation based on the physical layout diagram, the outline of the equipment corresponding to each physical workstation is first read from the layout diagram, and then its center point coordinates, the length, width and height of the minimum bounding box and the orientation angle of the equipment spindle are estimated. This information is used to form a spatial location set. Based on the process node corresponding to the workstation defined in the material processing process flow diagram, the process type code, standard process parameters and tolerance requirements of the process are read. This information is encoded into a process attribute vector, and the obtained spatial location and process attributes of each physical workstation are used as the static configuration parameters of the workstation in the digital twin model.

[0024] In specific implementation, each physical workstation is mapped to an independent functional node according to the material processing sequence specified in the material processing process flow chart, and process attribute labels are set for each functional node. This can be achieved in the following way: Based on the material flow direction defined in the material processing process flow chart, the execution order of each physical workstation on the time axis is determined; for each physical workstation, a corresponding functional node instance is created in the software container of the digital twin model. This instance contains independent threads or coroutines for performing simulation calculations; then, the process type code and main process parameter values ​​in the previously acquired process attributes of the physical workstation are serialized to form a structured process attribute label, and this label is bound to the functional node in the form of metadata. The functional node with the bound process attribute label is used as the basic simulation unit in the digital twin model to simulate the behavior of the physical workstation.

[0025] In specific implementation, the deconstruction of a functional node can be achieved by configuring a corresponding simulation model interface and data interaction protocol for each functional node based on the process attribute tags and the spatial location. This can be done in the following way: Read the process attribute tags attached to the functional node, select the corresponding simulation model type according to the process type in the tag (e.g., select the finite element simulation model for cutting and the rigid body dynamics model for material handling), and then configure a standardized simulation model interface for the functional node. This interface defines the data structure of input parameters such as the initial workpiece state and output results such as the processed dimensions. At the same time, based on the spatial location of the physical workstation corresponding to the functional node and the communication capability of the equipment, select an appropriate data interaction protocol from a predefined protocol library, such as a publish-subscribe protocol based on message queue telemetry transmission or a real-time protocol based on Ethernet control automation technology, and configure the specific topic name, data sampling frequency, and retransmission strategy. Bind the configured simulation model interface and data interaction protocol to the functional node to complete the complete deconstruction of the functional node.

[0026] In step 102, at least two functional nodes connected in series along at least one process route are divided into a process association segment.

[0027] In some embodiments, reference Figure 2 As shown in the figure, this is a flowchart illustrating the division of process-related segments according to some embodiments of this application. In this embodiment, dividing at least two functional nodes connected in series along at least one process route into a process-related segment can be achieved by the following steps: In step 1021, the process connection relationships between all functional nodes are extracted, and a directed graph with functional nodes as vertices and material flow direction as edges is constructed. In step 1022, all branchless paths from the start point to the end point are traversed in the directed graph, and a continuous and unbranched sequence of functional nodes on each path is marked as a candidate process route. In step 1023, for each candidate process route, the functional nodes on the route are divided into several groups based on the coupling strength between nodes. The functional nodes in each group are spatially and temporally continuous and the process correlation is higher than a set threshold. In step 1024, each group of functional nodes is defined as a process association segment, and the input status interface and output status interface of the process association segment are recorded.

[0028] In this application, the inter-node coupling strength is a metric used to measure the degree of interdependence and close interaction between two adjacent functional nodes; the process association segment is a logical combination that includes a set of functional nodes and functions as a simulation unit with independent input and output interfaces in the digital twin model; the input state interface is a data specification for defining the set of external state variables that the process association segment needs to receive before performing simulation; and the output state interface is a data specification for defining the set of state variables that the process association segment passes out after completing simulation.

[0029] In practical implementation, extracting the process connection relationships between all functional nodes and constructing a directed graph with functional nodes as vertices and material flow directions as edges can be achieved as follows: Traverse all functional nodes. For each functional node, analyze the downstream functional nodes to which its output material flows, record this flow relationship, treat each functional node as a vertex, and treat the direction of material flow from one functional node to another as an edge, thus constructing a directed graph with functional nodes as vertices and material flow directions as edges. Use this constructed directed graph as the basic topology model for subsequent division of process connection segments; traverse all branchless paths from the starting point to the ending point in the directed graph, and record each... The following method can be used to mark a continuous and non-branched sequence of functional nodes on a path as a candidate process route: In the constructed directed graph, first identify all functional nodes with an in-degree of 0 as the starting point of the path and all functional nodes with an out-degree of 0 as the ending point of the path. Starting from each starting point, perform a depth-first traversal along the direction of the edge. When the out-degree of a functional node is greater than 1, terminate the current path; when the out-degree of a node is equal to 1, continue to move forward; until the ending point is reached, record the continuous and non-branched sequence of functional nodes passed during the traversal as a candidate process route, and use all the obtained candidate process routes as the set of alternative paths for subsequent division of process association segments.

[0030] In practice, for each candidate process route, the functional nodes on the route are divided into several groups based on the coupling strength between nodes. The functional nodes within each group are spatially and temporally continuous, and their process correlation is higher than a set threshold. This can be achieved as follows: For each candidate process route, the coupling strength between each pair of adjacent functional nodes on the route is calculated sequentially. This coupling strength can be determined comprehensively based on the material transfer time between the physical workstations corresponding to the two nodes, the frequency of shared resources, and the degree of mutual influence of process parameters. Specifically, a simple implementation method is as follows: the ratio of the material transfer time to the preset maximum transfer time... After taking the reciprocal and normalizing to the range of 0 to 1, the frequency of shared resources is divided by the preset maximum frequency and also normalized to the range of 0 to 1. The degree of mutual influence of process parameters is defined as the influence coefficient of the change of key output parameters of upstream station on the processing qualification rate of downstream station and normalized to the range of 0 to 1. Then, weights are assigned to the above three factors respectively, for example, the weight of material transfer time is 0.3, the weight of shared resource frequency is 0.3, and the weight of mutual influence of process parameters is 0.4. Finally, the weighted sum is calculated, and the resulting value is the coupling strength between two adjacent functional nodes, with a value between 0 and 1. The larger the value, the tighter the coupling. Furthermore, starting from the first functional node of the route, it is sequentially determined whether the coupling strength between the current node and the next node is higher than a preset threshold. If it is higher than the threshold, they are grouped into the same group; if it is lower than or equal to the threshold, the group division ends after the current node, and a new group begins from the next node. The functional nodes in each group are spatially adjacent, temporally continuous, and their process correlation is higher than the set threshold. Each functional node group obtained is used as the basic unit for defining subsequent process correlation segments. It should be noted that the threshold mentioned in this application can be preset according to the actual application scenario, such as based on the statistical characteristics of historical calculation data, such as the average, median, or percentile. This application does not make specific limitations on this.

[0031] In specific implementation, each group of functional nodes is defined as a process association segment, and the input and output state interfaces of the process association segment are recorded. This can be achieved in the following way: logically combine each group of functional nodes into a process association segment, obtain the definition of the original input state variable of the first functional node in the segment, and directly use this definition as the input state interface of the process association segment; obtain the definition of the original output state variable of the last functional node in the segment, and directly use this definition as the output state interface of the process association segment, while recording the expected value range of the state variables of the input state interface and the type of state variables actually generated by the output state interface.

[0032] In step 103, for each process-related segment, the simulated values ​​of each functional node within the process-related segment are compared with the actual sensor values ​​of the corresponding physical workstation to generate node-level errors.

[0033] In some embodiments, for each process-related segment, the node-level error can be generated by comparing the simulated values ​​of each functional node within the process-related segment with the actual sensor values ​​of the corresponding physical workstations using the following steps: Read the simulation output state vectors of each functional node at the current moment from the digital twin model; The measured state vectors at the same moment are collected by sensors deployed at the corresponding physical workstations; The simulated state vector of each functional node is compared with the measured state vector component by component, and the absolute deviation of each component is calculated. The absolute deviations of all components are fused to obtain the scalar error value of the functional node; The scalar error values ​​of all functional nodes within the process-related segment are summarized to obtain the node-level error.

[0034] The simulation output state vector in this application is an ordered array used to describe the values ​​of each state variable in the simulation results generated by the functional node at a certain moment; the measured state vector is an ordered array used to describe the values ​​of each state variable obtained by the physical workstation through sensor measurement at a certain moment; the node-level error is a characteristic index used to measure the local simulation accuracy within a single process-related segment.

[0035] In practical implementation, firstly, the simulation output state vectors of each functional node at the current moment are read from the digital twin model. Digital twin platforms typically provide application programming interfaces (APIs), such as through a unified process control architecture interface for object linking and embedding, to read the simulation results of each functional node according to preset node identifiers and variable names, and store the results as an ordered array format. Secondly, the measured state vectors at the same moment are collected by sensors deployed at corresponding physical workstations. In industrial settings, precise time protocols or Ethernet-based control automation technology can be used to achieve multi-sensor clock synchronization, ensuring that sensors at each workstation trigger sampling at the same timestamp. The sampled data is aggregated by the programmable logic controller and uploaded to the data acquisition and monitoring control system via manufacturing message specifications or message queue telemetry transmission protocols, also organized into an ordered array with the same dimension as the simulation state vectors. Finally, the simulation state vectors of each functional node are... The state vector is compared component by component with the measured state vector, and the absolute deviation of each component is calculated. This can be done using conventional numerical calculations, such as subtracting the corresponding elements of the vectors and taking the absolute value. That is, for the i-th component, the absolute deviation is equal to the absolute value of the difference between the simulated value and the measured value. Then, the absolute deviations of all components are fused to obtain the scalar error value of the functional node. This can be done using known vector norm fusion methods, such as calculating the arithmetic mean of the absolute deviations of all components, or calculating the square root of the sum of the squares of the absolute deviations of each component, or assigning different weights to each component according to its physical importance and then calculating a weighted average. All of these methods can yield a single value representing the overall deviation of the node. Finally, the scalar error values ​​of all functional nodes within the process-related segment are summarized to obtain the node-level error. The arithmetic mean of the scalar error values ​​of each node within the segment or the maximum value can be used as the node-level error of the segment. The specific choice is determined based on engineering experience.

[0036] In step 104, for two sequentially adjacent process segments, the degree of difference between the output state of the previous segment and the input state of the next segment at the coupling boundary is analyzed to obtain the inter-segment transmission error.

[0037] In some embodiments, reference Figure 3 As shown in the figure, this is a flowchart illustrating the determination of inter-segment propagation error according to some embodiments of this application. In this embodiment, for two sequentially adjacent process segments, the degree of difference between the output state of the preceding segment and the input state of the following segment at the coupling boundary is analyzed to obtain the inter-segment propagation error. This can be achieved through the following steps: In step 1041, the set of state variables and their numerical range defined by the output state interface of the previous process segment are identified to obtain the state vector of the actual output of the previous segment. In step 1042, the expected set of state variables and the numerical range of the input state interface of the next process segment are identified to obtain the expected state vector to be received in the next process segment. In step 1043, at the coupling boundary, the actual output state vector of the previous segment is aligned and matched with the expected received state vector of the next segment, and the difference metric between the two aligned state vectors is calculated. In step 1044, the inter-segment propagation error is calculated based on the difference metric, and position information is marked at the coupling boundary.

[0038] It should be noted that the difference measure in this application refers to the feature value used to quantify the overall deviation between the actual output state vector of the previous segment and the expected received state vector of the next segment; the inter-segment transmission error is a feature index used to measure the consistency of state transmission between adjacent process segments.

[0039] In specific implementation, firstly, the set of state variables and their numerical range defined by the output state interface of the preceding process segment are identified to obtain the actual output state vector of the preceding segment. Each process segment in the digital twin model is pre-configured with an output state interface, which records the name, unit, and upper and lower limits of the state variables in the form of a structured data table. By calling the standard interface of the unified process control architecture for object linking and embedding, the specific values ​​of each state variable generated at the current moment after the simulation calculation of this segment are read and organized into an output state vector according to the interface definition order. Secondly, the set of state variables and their numerical range expected by the input state interface of the following process segment are identified to obtain the expected state vector of the following segment. The expected state variable types and their standard target values ​​are extracted from the input state interface of the following segment, and these target values ​​are arranged into an expected received state vector according to the interface order. Thirdly, alignment and matching are performed at the coupling boundary, which can utilize the precise time protocol commonly deployed in industrial Ethernet to ensure... The output time of the previous segment is kept consistent with the expected time of the next segment. Then, the corresponding components of the two vectors are converted according to physical units, such as millimeters to meters and degrees Celsius to Kelvin. After alignment, the well-known Euclidean distance formula is used to calculate the square root of the sum of the squares of the differences of each component to obtain the difference measure. Finally, the inter-segment transmission error is calculated based on the difference measure. The difference measure can be divided by the maximum allowable difference value pre-calibrated using historical data to obtain the inter-segment transmission error normalized to the interval of 0 to 1. It should be noted that the maximum allowable difference value refers to the normalized reference value determined by statistical analysis based on historical data accumulated by the production line under normal operating conditions. It serves as the upper limit of the acceptable deviation between state vectors and can be taken as the 95th quantile or the 3σ upper limit of the difference measure in historical data. At the same time, the spatial coordinates and sequential number in the process route at the coupling boundary are read from the physical layout database of the production line and stored as location information associated with the error value.

[0040] In step 105, the node-level error and the inter-segment propagation error are fused to construct a joint fidelity coefficient.

[0041] In some embodiments, the joint fidelity coefficients can be constructed by fusing the node-level errors and the inter-segment propagation errors using the following steps: Arrange the node-level errors of each process-related segment in the process sequence to construct a node error diagonal matrix. The diagonal elements of this matrix correspond to the node-level errors of the corresponding segments, and the off-diagonal elements are zero. Based on the inter-segment transmission error between adjacent process segments, construct the upper diagonal error matrix; The joint error matrix is ​​obtained by adding the nodal error diagonal matrix and the upper diagonal error matrix. Calculate the spectral norm of the joint error matrix and use the result as the joint fidelity coefficient.

[0042] It should be noted that the joint error matrix in this application is a composite matrix used to simultaneously accommodate node-level errors and inter-segment propagation errors. It can reflect the comprehensive deviation information between local simulation accuracy and inter-segment state propagation consistency. The joint fidelity coefficient is a quantitative indicator used to comprehensively measure the overall deviation of the digital twin model in terms of local node fidelity and inter-segment state propagation consistency.

[0043] It should also be noted that since node-level errors usually have physical dimensions, while inter-segment transmission errors have been normalized to dimensionless values, directly adding the two would violate the principle of dimension consistency. Therefore, before constructing the joint error matrix, the node-level errors need to be dimensionlessly normalized: for each process-related segment, obtain the historical measured data fluctuation range (difference between the maximum and minimum values) of the physical workstation corresponding to that segment under normal operating conditions, divide the node-level error by this fluctuation range to obtain the normalized node-level error, with a value range of 0 to 1. Then, use the normalized node-level error to construct the node error diagonal matrix, and add it to the upper diagonal error matrix to obtain the dimensionally consistent joint error matrix. For ease of description, the node-level errors referred to in this embodiment are all assumed to be normalized values.

[0044] In practical implementation, the node-level errors of each process-related segment are arranged according to the process sequence to construct a node error diagonal matrix. The diagonal elements of this matrix correspond to the node-level errors of the corresponding segments, and the off-diagonal elements are zero. This can be achieved as follows: First, count the total number of process-related segments, denoted as N. Then, arrange the node-level errors of each process-related segment in the process route from front to back, constructing an N x N square matrix. Initialize all elements in this square matrix to zero, and then fill the node-level error of the first process-related segment into the first row and first column. The node-level error of the second process-related segment is filled into the second row and second column, and so on, until the node-level error of the Nth process-related segment is filled into the Nth row and Nth column. This matrix is ​​the node error diagonal matrix. Then, according to the method mentioned above, each diagonal element is normalized without dimension: obtain the historical measured data fluctuation range Δ_i of the physical workstation corresponding to the i-th process-related segment, replace the node-level error e_i with e_i / Δ_i, and obtain the normalized node error diagonal matrix.

[0045] It should be noted that the node error diagonal matrix and the upper diagonal error matrix in this scheme have the same dimension. Specifically, if the total number of process-related segments is N, then the node error diagonal matrix is ​​an N-row N-column square matrix, and its diagonal elements are the node-level errors of each segment. The upper diagonal error matrix is ​​also constructed as an N-row N-column square matrix, where the inter-segment transmission error between the i-th segment and the i+1-th segment is placed only at the position of the i-th row and i+1-th column (i ranges from 1 to N-1), and all other elements are zero. Although there are only N-1 inter-segment transmission errors, placing them on the upper diagonal of an N-order square matrix is ​​a necessary condition for matrix addition. If the dimension of the upper diagonal error matrix is ​​not N×N, it is impossible to perform addition with the node error diagonal matrix. Therefore, the two matrices have the same dimension, both N×N, thus ensuring the correct construction of the joint error matrix.

[0046] In specific implementation, the joint error matrix can be obtained by adding the node error diagonal matrix and the upper diagonal error matrix. This can be achieved by performing matrix addition on the node error diagonal matrix and the upper diagonal error matrix, that is, adding the elements at corresponding positions. Since the two matrices will not have non-zero elements at the same position, the matrix after addition retains the normalized node-level error of each process-related segment at the diagonal position and retains the inter-segment transmission error between adjacent segments at the upper diagonal position. This result is the joint error matrix.

[0047] In specific implementation, the calculation of the spectral norm of the joint error matrix and the use of the calculation result as the joint fidelity coefficient can be achieved in the following way: The spectral norm of the joint error matrix is ​​calculated, that is, the maximum singular value of the matrix is ​​obtained. In numerical calculation, the power iteration method or solving the eigenvalues ​​of the matrix and then taking the square root is usually used to obtain the maximum singular value. The calculated spectral norm value is then used as the joint fidelity coefficient. It should be noted that the reason why this application calculates the spectral norm of the joint error matrix instead of using a simple weighted sum or average is that the spectral norm can simultaneously reflect the comprehensive influence of all elements in the matrix, especially the coupling effect between the diagonal elements and the upper diagonal elements. When node-level errors propagate backward through inter-segment propagation errors, the spectral norm can capture the worst case of this error being amplified step by step along the process route, while the traditional summation and averaging method cannot reflect the amplification effect along the propagation path. The advantages of using the spectral norm are: on the one hand, it makes the joint fidelity coefficient more sensitive to error accumulation in long-process production lines, and can detect the overall model inaccuracy caused by inter-segment coupling distortion in a timely manner; on the other hand, the spectral norm has rotation invariance and scale scaling predictability, which makes it easy to set the threshold uniformly across production lines of different lengths and dimensions, thereby improving the versatility and robustness of this method.

[0048] Preferably, in some embodiments, constructing the upper diagonal error matrix based on the inter-segment transfer error between adjacent process segments can be achieved using the following steps: The process-related segments are numbered from the first to the Nth according to the process route sequence, where N is the total number of process-related segments; For each pair of sequentially adjacent segments i and i+1, obtain their corresponding inter-segment propagation error; Construct an N x N zero matrix, assign the element in the i-th row and i+1-th column to the inter-segment propagation error between the i-th segment and the i+1-th segment, and set the remaining elements to zero. The resulting matrix is ​​the upper diagonal error matrix.

[0049] In practice, firstly, the pre-divided process-related segments are numbered consecutively from the first to the Nth according to the process route sequence, where N is the total number of process-related segments. The numbering result uniquely determines the sequential position of each segment in the material flow direction. Then, for each pair of sequentially adjacent segments i and i+1, where i ranges from 1 to N-1, the pre-calculated inter-segment transmission error value corresponding to the two segments is read from the data storage area of ​​the digital twin model. Finally, an N-row N-column all-zero square matrix is ​​constructed in the computer memory. For each i, the obtained inter-segment transmission error between the i-th segment and the i+1th segment is assigned to the element in the i-th row and i+1-th column of the matrix, while all other elements remain zero. The matrix obtained after the assignment is completed is the upper diagonal error matrix.

[0050] In step 106, when the joint fidelity coefficient exceeds a preset threshold, the model parameters of the functional nodes and / or the state coupling mapping rules between process-related segments are iteratively corrected until the joint fidelity coefficient is lower than the preset threshold, so that the digital twin model and the physical production line state remain dynamically consistent.

[0051] It should be noted that when the joint fidelity coefficient exceeds a preset threshold, it indicates that the overall synchronization deviation between the digital twin model and the physical production line has exceeded the allowable range, and the model parameter correction process must be initiated. The preset threshold is determined based on the statistical distribution of the joint fidelity coefficient in historical data collected under normal production line operating conditions. For example, it can be two or three times the standard deviation of the historical mean as the threshold, or it can be directly set by the process engineer based on the allowable error range of product quality. This application does not make specific limitations on this.

[0052] In some embodiments, when the joint fidelity coefficient exceeds a preset threshold, iteratively correcting the model parameters of the functional nodes and / or the state coupling mapping rules between process-related segments until the joint fidelity coefficient is lower than the preset threshold can be achieved through the following steps: By comparing the mean node-level error with the mean inter-segment propagation error, the dominant error source can be determined. If node-level error is dominant, then the gradient descent algorithm is used to iteratively update the model parameters of the functional nodes with the goal of minimizing node-level error. If the inter-segment propagation error is dominant, the state coupling mapping rule between process-related segments is iteratively corrected with the goal of minimizing the difference between the output of the previous segment and the expected input of the next segment. If the average node-level error and the average inter-segment propagation error both exceed the preset corresponding thresholds, then the parameters and rules are jointly corrected. After each correction, the joint fidelity coefficient is recalculated until it falls below a preset threshold, at which point the final model configuration parameters are output.

[0053] It should be noted that the model parameters of the functional nodes in this application refer to the internal adjustable variables used to determine the simulation behavior and output characteristics of the functional nodes, including but not limited to kinematic parameters, dynamic parameters and process parameters; the state coupling mapping rules between process-related segments in this application refer to the functions or parameterized transformation relationships used to transform the output state vector of the previous segment into the expected input state vector of the next segment, including but not limited to unit transformation functions, coordinate transformation matrices, attribute inheritance functions and time delay compensation functions.

[0054] In practical implementation, comparing the mean of node-level errors with the mean of inter-segment propagation errors to determine the dominant error source can be achieved as follows: First, calculate the mean of node-level errors and the mean of inter-segment propagation errors separately, then compare the two means. If the mean of node-level errors is greater than the mean of inter-segment propagation errors, then the node-level errors are determined to be the dominant error source; otherwise, the inter-segment propagation errors are determined to be the dominant error source. If node-level errors are dominant, the gradient descent algorithm is used to iteratively update the model parameters of functional nodes with the goal of minimizing node-level errors. This can be achieved as follows: When node-level errors are determined to be the dominant error source, a correction procedure for the model parameters of functional nodes is initiated. The sum of the squares of the normalized node-level errors of all functional nodes is used as the objective function. The gradient descent algorithm is used iteratively. In each iteration, the partial derivatives of the objective function with respect to each model parameter of each functional node are calculated, and then the parameter values ​​are updated along the negative gradient direction with a set step size. For example, for a functional node simulating cutting machining, its model parameters... The parameters include the tool feed rate coefficient and the spindle speed correction factor. The gradient descent algorithm calculates the magnitude and direction of these two parameters to be adjusted based on the current simulation error, gradually reducing the deviation between the simulation and the actual measurement of this node. For example, for a functional node simulating heat treatment, its model parameters include the thermal conductivity coefficient and specific heat capacity. The algorithm will repeatedly fine-tune these two coefficients based on the difference between the simulated temperature value and the measured value until the deviation is reduced. The iteration continues until the average node-level error drops below the preset sub-threshold or reaches the maximum number of iterations, and the updated model parameters are used as the internal configuration of the corrected digital twin model.If inter-segment transmission error is dominant, and the goal is to minimize the difference between the output of the previous segment and the expected input of the next segment, the iterative correction of the state coupling mapping rules between process-related segments can be achieved as follows: When the inter-segment transmission error is determined to be the dominant error source, a correction procedure for the state coupling mapping rules between adjacent process-related segments is initiated. The sum of the squares of the transmission errors between each pair of adjacent segments is used as the objective function, and numerical optimization methods are employed for iterative correction. Specifically, the state coupling mapping rules can include various types. The following examples illustrate the correction process. For cases requiring unit conversion, the mapping rule is a linear transformation function. For instance, if the workpiece length output in the previous segment is in millimeters, and the expected input in the next segment is in meters, the mapping rule is a linear coefficient divided by one thousand. The correction process adjusts the precise value of this linear coefficient. For cases requiring coordinate transformation, the mapping rule is a rotation and translation matrix. For instance, the position coordinates of the workpiece output in its own coordinate system in the previous segment need to be transformed in the next segment. In the global coordinate system, the correction process involves adjusting the rotation angle and translation parameters in the matrix. For cases requiring material property inheritance, the mapping rule is an exponential decay function. For example, when the workpiece temperature output from the previous segment is transferred to the initial temperature of the next segment, the decay factor is 0.95 due to environmental heat dissipation, and the correction process adjusts this decay factor. For cases with time delays, the mapping rule is a delay compensation function. For example, if the previous segment completes processing and then arrives at the next segment after a 5-second delay via the conveyor belt, the correction process adjusts the delay time parameter. In each iteration, the actual output state vector of the previous segment is transformed according to the current mapping rule to obtain the transformed state vector. Then, the difference between the transformed state vector and the expected input state vector of the next segment is calculated. The parameters of the mapping rule are then adjusted according to the gradient of the difference measure to gradually reduce the difference. The iteration continues until the average value of the inter-segment transmission error drops below the preset sub-threshold, and the corrected state coupling mapping rule is used as the new data transmission standard between adjacent process segments.

[0055] In specific implementation, if both the mean node-level error and the mean inter-segment propagation error exceed their respective preset thresholds, the joint correction parameters and rules can be implemented as follows: If the mean node-level error is greater than its corresponding threshold, and the mean inter-segment propagation error is also greater than its corresponding threshold, a joint correction strategy is adopted. Joint correction refers to executing model parameter updates and state coupling mapping rule corrections in parallel or alternately. Specifically, an alternating optimization strategy is used to reduce mutual interference: first, a model parameter update is executed, then a mapping rule correction is executed. After one round of alternation, the two mean error values ​​are recalculated. Then, based on the comparison results of the new round of mean error values, the focus of subsequent corrections is dynamically adjusted. For example, after several rounds of alternation, if the mean node-level error still exceeds the threshold while the mean inter-segment propagation error is lower than the threshold, only model parameter updates are executed subsequently; otherwise, only mapping rule corrections are executed. Joint correction continues until both mean error values ​​are lower than their respective corresponding thresholds. The simultaneously optimized model parameters and state coupling mapping rules are then used as the dual correction result of the digital twin model.

[0056] In practice, after each correction, the joint fidelity coefficient is recalculated until it falls below a preset threshold. Outputting the final model configuration parameters means that after any correction operation, the digital twin simulation is immediately rerun based on the updated functional node model parameters and state coupling mapping rules. The joint fidelity coefficient is recalculated according to the aforementioned fusion steps, and compared with the preset threshold. If the coefficient is still greater than or equal to the preset threshold, the process returns to the dominant error source determination step to continue iterative correction. If the coefficient is less than the preset threshold, the iteration stops, and the model parameters of all currently active functional nodes and the state coupling mapping rules between all adjacent process segments are combined as the optimal configuration parameters output to ensure dynamic consistency between the digital twin model and the physical production line. Specifically, an iteration counter is set in the digital twin platform. After each correction, the counter is incremented by 1, and the joint fidelity coefficient calculation module is called to obtain the current value. This value is then compared with the preset threshold until the condition is met. Finally, the model parameter snapshot and mapping rule snapshot stored in memory are serialized into a configuration file or written to a database for real-time synchronization by the production line.

[0057] On the other hand, in some embodiments, this application provides a smart manufacturing digital twin modeling system, with reference to Figure 4 The figure is a schematic diagram of the structure of a smart manufacturing digital twin modeling system according to some embodiments of this application. The smart manufacturing digital twin modeling system 400 includes: a deconstruction module 401, a partitioning module 402, a node error generation module 403, an inter-segment error generation module 404, a fusion module 405, and a correction module 406, which are described below: The deconstruction module 401 is used to deconstruct the digital twin model into multiple functional nodes with preset mapping relationships to physical workstations according to the physical layout of the production equipment and the processing sequence of materials. The partitioning module 402 is used to partition at least two functional nodes connected in series along at least one process route into a process-related segment; The node error generation module 403 is used to compare the simulation values ​​of each functional node in each process-related segment with the actual sensor values ​​of the corresponding physical workstation to generate node-level errors for each process-related segment. The inter-segment error generation module 404 is used to analyze the degree of difference between the output state of the previous segment and the input state of the next segment at the coupling boundary for two sequentially adjacent process segments, and to obtain the inter-segment transmission error. The fusion module 405 is used to fuse the node-level error and the inter-segment propagation error to construct a joint fidelity coefficient. The correction module 406 is used to iteratively correct the model parameters of the functional nodes and / or the state coupling mapping rules between process-related segments when the joint fidelity coefficient exceeds a preset threshold, until the joint fidelity coefficient is lower than the preset threshold, so that the digital twin model dynamically maintains consistency with the physical production line state. Additionally, this application also provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to acquire the code and execute the above-described intelligent manufacturing digital twin modeling method.

[0058] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device implementing a digital twin modeling method for intelligent manufacturing, according to some embodiments of this application. The intelligent manufacturing digital twin modeling method in the above embodiments can... Figure 5 The computer device 500 shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0059] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0060] The communication bus 502 can be used to transmit information between the aforementioned components.

[0061] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CDROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0062] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. In the above embodiments, the intelligent manufacturing digital twin modeling method can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0063] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0064] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single CPU) processor or a multi-core (multi CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0065] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0066] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent manufacturing digital twin modeling method.

[0067] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0068] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A digital twin modeling method for intelligent manufacturing, characterized in that, Includes the following steps: Based on the physical layout of the production equipment and the processing sequence of the materials, the digital twin model is deconstructed into multiple functional nodes that have a preset mapping relationship with the physical workstations. Divide at least two functional nodes connected in series along at least one process route into a process-related segment; For each process-related segment, the simulated values ​​of each functional node within the process-related segment are compared with the actual sensor values ​​of the corresponding physical workstation to generate node-level errors. For two sequentially adjacent process segments, analyze the degree of difference between the output state of the previous segment and the input state of the next segment at the coupling boundary to obtain the inter-segment propagation error; By integrating the node-level error and the inter-segment propagation error, a joint fidelity coefficient is constructed. When the joint fidelity coefficient exceeds a preset threshold, the model parameters of the functional nodes and / or the state coupling mapping rules between process-related segments are iteratively corrected until the joint fidelity coefficient is lower than the preset threshold, so that the digital twin model and the physical production line state remain dynamically consistent.

2. The method as described in claim 1, characterized in that, Based on the physical layout of production equipment and the processing sequence of materials, the digital twin model is deconstructed into multiple functional nodes with preset mapping relationships to physical workstations, specifically including: Obtain the physical layout diagram and material processing flow chart of the target production line; Determine the spatial location and process attributes of each physical workstation based on the physical layout diagram; According to the material processing sequence specified in the material processing process flow chart, each physical workstation is mapped to an independent functional node, and a process attribute label is set for each functional node. Based on the process attribute labels and spatial locations, configure corresponding simulation model interfaces and data interaction protocols for each functional node to complete the deconstruction of the functional node.

3. The method as described in claim 1, characterized in that, Dividing at least two functional nodes connected in series along at least one process route into a process association segment specifically includes: Extract the process connection relationships between all functional nodes and construct a directed graph with functional nodes as vertices and material flow direction as edges; In the directed graph, all branchless paths from the starting point to the ending point are traversed, and a continuous and unbranched sequence of functional nodes on each path is marked as a candidate process route. For each candidate process route, the functional nodes on the route are divided into several groups based on the coupling strength between nodes. The functional nodes in each group are spatially and temporally continuous and the process correlation is higher than a set threshold. Define each group of functional nodes as a process association segment, and record the input status interface and output status interface of the process association segment.

4. The method as described in claim 1, characterized in that, For each process-related segment, the simulated values ​​of each functional node within that segment are compared with the actual sensor values ​​of the corresponding physical workstations to generate node-level errors, specifically including: Read the simulation output state vectors of each functional node at the current moment from the digital twin model; The measured state vectors at the same moment are collected by sensors deployed at the corresponding physical workstations; The simulated state vector of each functional node is compared with the measured state vector component by component, and the absolute deviation of each component is calculated. The absolute deviations of all components are fused to obtain the scalar error value of the functional node; The scalar error values ​​of all functional nodes within the process-related segment are summarized to obtain the node-level error.

5. The method as described in claim 1, characterized in that, For two sequentially adjacent process segments, the degree of difference between the output state of the preceding segment and the input state of the following segment at the coupling boundary is analyzed to obtain the specific inter-segment propagation error, which includes: Identify the set of state variables and their numerical range defined by the output state interface of the previous process segment to obtain the actual output state vector of the previous segment. Identify the set of state variables and their numerical ranges expected by the input state interface of the next process segment, and obtain the state vector expected to be received by the next process segment. At the coupling boundary, the actual output state vector of the previous segment is aligned and matched with the expected received state vector of the next segment, and the difference metric between the two aligned state vectors is calculated. Based on the difference metric, the inter-segment propagation error is calculated, and the position information is marked at the coupling boundary.

6. The method as described in claim 1, characterized in that, The construction of the joint fidelity coefficient by integrating the node-level error and the inter-segment propagation error specifically includes: Arrange the node-level errors of each process-related segment in the process sequence to construct a node error diagonal matrix. The diagonal elements of this matrix correspond to the node-level errors of the corresponding segments, and the off-diagonal elements are zero. Based on the inter-segment transmission error between adjacent process segments, construct the upper diagonal error matrix; The joint error matrix is ​​obtained by adding the nodal error diagonal matrix and the upper diagonal error matrix. Calculate the spectral norm of the joint error matrix and use the result as the joint fidelity coefficient.

7. The method as described in claim 6, characterized in that, Based on the inter-segment transmission error between adjacent process segments, the upper diagonal error matrix is ​​constructed, specifically including: The process-related segments are numbered from the first to the Nth according to the process route sequence, where N is the total number of process-related segments; For each pair of sequentially adjacent segments i and i+1, obtain their corresponding inter-segment propagation error; Construct an N x N zero matrix, assign the element in the i-th row and i+1-th column to the inter-segment propagation error between the i-th segment and the i+1-th segment, and set the remaining elements to zero. The resulting matrix is ​​the upper diagonal error matrix.

8. A digital twin modeling system for intelligent manufacturing, characterized in that, The system includes: The deconstruction module is used to deconstruct the digital twin model into multiple functional nodes with preset mapping relationships to physical workstations, according to the physical layout of the production equipment and the processing sequence of materials. A partitioning module is used to divide at least two functional nodes connected in series along at least one process route into a process-related segment; The node error generation module is used to compare the simulation values ​​of each functional node in each process-related segment with the actual sensor values ​​of the corresponding physical workstation to generate node-level errors. The inter-segment error generation module is used to analyze the degree of difference between the output state of the previous segment and the input state of the next segment at the coupling boundary for two sequentially adjacent process segments, and to obtain the inter-segment transmission error. The fusion module is used to fuse the node-level error and the inter-segment propagation error to construct a joint fidelity coefficient; The correction module is used to iteratively correct the model parameters of the functional nodes and / or the state coupling mapping rules between process-related segments when the joint fidelity coefficient exceeds a preset threshold, until the joint fidelity coefficient is lower than the preset threshold, so that the digital twin model and the physical production line state remain dynamically consistent.

9. A computer device comprising a memory and a processor, the memory storing code, characterized in that, The processor is configured to acquire the code and execute the intelligent manufacturing digital twin modeling method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent manufacturing digital twin modeling method as described in any one of claims 1 to 7.

Images