Industrial digital twinning real-time cooperative control method and system based on fog computing

Abstract

The invention relates to the technical field of industrial automation and intelligent manufacturing, in particular to an industrial digital twinning real-time cooperative control method and system based on fog computing. The system adopts a cloud, fog and edge three-layer architecture, wherein the cloud layer is responsible for a global slow-varying model; the fog layer is responsible for local real-time model and collaboration; and the edge layer is responsible for an equipment-level super real-time model and direct control. The method comprises the following steps: initializing a hierarchical model; based on self-adaptive data flow management of data value density, an edge layer only uploads key data and events; according to cooperative control in the fog layer, when one fog node predicts a fault, the fog node directly communicates with the associated fog node and performs joint simulation, a cooperative strategy is rapidly formulated, a control instruction is distributed, and reporting to the cloud layer is not needed. According to the invention, millisecond-level real-time control of industrial digital twinning is realized, the network bandwidth dependence is obviously reduced, the autonomy and reliability of the system under the conditions of network disconnection and the like are improved, and the method is particularly suitable for intelligent manufacturing scenes.

Description

Technical Field

[0001] This invention relates to the fields of industrial automation and intelligent manufacturing technology, and in particular to an industrial digital twin system based on fog computing architecture and its real-time collaborative control method. Background Technology

[0002] Digital twins, as a core technology for realizing cyber-physical systems (CPS), have been widely used in the industrial manufacturing field. By constructing virtual mappings of physical entities, they enable status monitoring, fault diagnosis, and process optimization.

[0003] Most mainstream industrial digital twin systems currently use cloud computing centers for model building and data analysis. This centralized processing model has the following inherent drawbacks:

[0004] High control latency: The massive amount of sensor data on the production site needs to be transmitted to a remote cloud data center via the network, processed, and then the control commands are sent to the actuators. This loop results in an end-to-end latency of hundreds of milliseconds or even seconds, which cannot meet the millisecond-level real-time control requirements of scenarios such as precision machining and robot collaboration.

[0005] High network bandwidth pressure: Continuously uploading full amounts of raw data (such as high-frequency vibrations and high-definition videos) poses a huge challenge to the factory's network bandwidth, resulting in high communication costs and easily creating network bottlenecks.

[0006] Poor system reliability: The cloud computing center is a single point of failure. Once the cloud center or network connection fails, the digital twin and control functions of the entire production system will be paralyzed, seriously affecting the continuity and safety of production.

[0007] Data privacy risks: Uploading all production process and status data to a third-party cloud platform poses a risk of sensitive data leakage.

[0008] Fog computing, as a computing paradigm situated between cloud computing and terminal devices, is characterized by deploying computing, storage, and network services near the data source, offering a potential solution to the aforementioned problems. However, how to deeply integrate fog computing with digital twins, design a hierarchical and collaborative real-time control architecture, and achieve efficient data management and task allocation remain pressing technical challenges. Summary of the Invention

[0009] The purpose of this invention is to overcome the problems of high latency and poor reliability of digital twin systems under existing cloud computing architecture, and to provide a real-time collaborative control method and system for industrial digital twins based on fog computing, which can achieve millisecond-level real-time control, significantly reduce bandwidth consumption, and improve overall robustness and autonomy.

[0010] On one hand, embodiments of the present invention provide a real-time collaborative control method for industrial digital twins based on fog computing, comprising the following steps:

[0011] S100: Acquire physical device operation data collected at the edge layer;

[0012] S200, the edge layer processes the running data according to preset rules and decides on the content and frequency of data to be uploaded to the fog layer;

[0013] S300, when the first fog node in the fog layer determines an anomaly or an impending anomaly based on the received data, it initiates a collaborative control process with the second fog node to jointly agree on a control strategy;

[0014] S400, the first fog node and the second fog node respectively send the agreed control commands to the edge computing devices under their jurisdiction, and the edge computing devices drive the physical devices to perform corresponding actions.

[0015] In some embodiments, in S200, the preset rule is an adaptive rule based on data value density, including:

[0016] S210, for the operating data characterizing the normal state of the device, the edge layer only uploads the processed feature data or statistical summary;

[0017] S220, for the operational data characterizing device malfunctions or specific events, the edge layer uploads high-precision raw data fragments or event details.

[0018] In some embodiments, S300, the initiation of the collaborative control process with the second fog node includes:

[0019] S310, the first fog node sends the information and impact prediction of the anomaly or impending anomaly to the second fog node;

[0020] S320, the first fog node and the second fog node perform joint simulation based on their respective local real-time digital twin models to generate a new collaborative production plan;

[0021] S330, the first fog node and the second fog node confirm the new collaborative production plan.

[0022] In some embodiments, the method further includes:

[0023] S500, the cloud layer periodically obtains model parameter increments and key event logs from the fog layer, updates the global slowly changing digital twin model, and sends the updated model parameters to the fog layer.

[0024] On the other hand, embodiments of the present invention provide an industrial digital twin real-time collaborative control system based on fog computing, comprising:

[0025] Cloud layer, which constructs and maintains a global, slowly changing digital twin model;

[0026] A fog layer, which consists of at least one fog computing node, is used to construct and maintain a local real-time digital twin model corresponding to a physical production unit.

[0027] An edge layer, comprising at least one edge computing device, which builds and maintains a device-level ultra-real-time digital twin model corresponding to a single physical device and interacts directly with the physical device;

[0028] The cloud layer, the fog layer, and the edge layer communicate with each other via a network and form a collaborative control closed loop.

[0029] In some embodiments, a lateral communication link is provided between fog computing nodes in the fog layer for real-time data exchange and collaborative decision-making across production units within the fog layer, wherein the collaborative decision-making does not need to pass through the cloud layer.

[0030] The beneficial effects of this invention are:

[0031] This invention introduces a three-layer architecture of cloud, fog, and edge layers. It retains raw operational data locally at the edge layer and achieves privacy-preserving collaborative training through hierarchical aggregation of model parameters. By deploying lateral communication links in the fog layer, real-time collaborative decision-making across production units in different regions is achieved without traversing the cloud layer. Through adaptive data flow management based on data value density, key control loops are enclosed in the edge and fog layers, achieving millisecond-level response from perception to decision-making to control. This invention effectively solves the latency, bandwidth, reliability, and privacy issues of industrial digital twin systems and is suitable for smart manufacturing scenarios. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart illustrating the steps of a real-time collaborative control method for industrial digital twins based on fog computing, provided in an embodiment of the present invention.

[0034] Figure 2 This is an architecture diagram of the real-time collaborative control method for industrial digital twins based on fog computing described in this invention.

[0035] Figure 3 This is a flowchart of the adaptive data flow management in the fog computing-based industrial digital twin real-time collaborative control method described in this invention.

[0036] Figure 4 This is a timing diagram of the collaborative control within the fog layer in this invention. Detailed Implementation

[0037] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of this invention; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this invention as detailed in the appended claims.

[0039] It is understood that the terms "first," "second," etc., used in this invention may be used to describe various concepts, but unless specifically stated otherwise, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, first information may also be referred to as second information without departing from the scope of embodiments of this invention, and similarly, second information may also be referred to as first information. Depending on the context, the words "if" or "when" as used herein may be interpreted as "when," "in response to a determination," or "in the event of a determination."

[0040] The terms "at least one", "multiple", "each", "any", etc., used in this invention, "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiple, and "any" refers to any one of the multiple.

[0041] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this invention is for descriptive purposes only and is not intended to limit the invention.

[0042] This invention discloses an industrial digital twin real-time collaborative control method based on fog computing. The core of the method is to construct a hierarchical digital twin architecture and achieve millisecond-level real-time control and efficient resource utilization through adaptive data flow management and fog layer horizontal collaboration mechanism.

[0043] refer to Figure 1 This invention provides a real-time collaborative control method for industrial digital twins based on fog computing, the method comprising the following steps:

[0044] S100: Acquire physical device operation data collected at the edge layer;

[0045] S200, the edge layer processes the running data according to preset rules and decides on the content and frequency of data to be uploaded to the fog layer;

[0046] S300, when the first fog node in the fog layer determines an anomaly or an impending anomaly based on the received data, it initiates a collaborative control process with the second fog node to jointly agree on a control strategy;

[0047] S400, the first fog node and the second fog node respectively send the agreed control commands to the edge computing devices under their jurisdiction, and the edge computing devices drive the physical devices to perform corresponding actions.

[0048] In steps S100 to S400 of this embodiment of the invention, industrial control tasks are transformed into hierarchical collaborative problems, and data value density adaptive filtering is used to reduce communication load. Step S100 is the foundational step for data acquisition, where edge computing devices interact directly with physical devices to collect high-frequency operational data, providing raw data input for subsequent local analysis and anomaly detection. Step S200 performs data value assessment at the edge layer, compressing normal state data and maintaining high-precision transmission of abnormal event data, balancing bandwidth consumption and information integrity. Step S300 implements cross-unit collaborative decision-making at the fog layer; when a single fog node predicts an anomaly, it negotiates directly with related fog nodes through a horizontal communication link, bypassing the cloud layer and significantly reducing decision latency. Step S400 translates the collaborative decision into specific control commands, which are then executed by physical devices via edge computing devices, forming a closed-loop control.

[0049] In some embodiments, in S200, the preset rule is an adaptive rule based on data value density, including:

[0050] S210, for the operating data characterizing the normal state of the device, the edge layer only uploads the processed feature data or statistical summary;

[0051] S220, for the operational data characterizing device malfunctions or specific events, the edge layer uploads high-precision raw data fragments or event details.

[0052] This embodiment clarifies the criteria and processing rules for determining data value density. Edge computing devices perform real-time analysis of operational data, calculating statistical characteristics and anomaly indicators. When all monitored parameters are within the normal range, the device is considered to be in normal condition, and only compressed feature data is uploaded. When any parameter exceeds a preset threshold or exhibits an abnormal trend, it is determined to be an anomaly or a specific event, triggering the upload of high-precision data. The feature data includes statistical quantities such as mean, variance, and peak factor, with a data compression ratio of 100:1. The original data segment contains complete time-series data for one second before and after the anomaly, with the sampling frequency remaining constant at 1kHz to ensure the accuracy of fault diagnosis. The event details include structured information such as event type code, timestamp, anomaly parameter values, and duration.

[0053] In some embodiments, S300, the initiation of the collaborative control process with the second fog node includes:

[0054] S310, the first fog node sends the information and impact prediction of the anomaly or impending anomaly to the second fog node;

[0055] S320, the first fog node and the second fog node perform joint simulation based on their respective local real-time digital twin models to generate a new collaborative production plan;

[0056] S330, the first fog node and the second fog node confirm the new collaborative production plan.

[0057] This embodiment implements an autonomous collaborative mechanism within the fog layer. Upon predicting an anomaly, the first fog node immediately broadcasts a warning message to the associated second fog node via a lateral communication link. The warning message includes the abnormal device ID, anomaly type, predicted failure time, and an assessment of the impact range. Upon receiving the warning message, the second fog node establishes a collaborative session with the first fog node, and both nodes run their respective local real-time digital twin models for parallel simulation. The first fog node simulates the output changes of its production line under different control strategies, while the second fog node simulates the response capabilities of related systems (such as material scheduling and energy supply). The two nodes exchange simulation results through multiple iterations, ultimately generating a consistent collaborative production plan. The confirmation process employs a three-way handshake protocol to ensure that both parties have a consistent understanding of the collaborative plan and avoid execution deviations.

[0058] In some embodiments, the method further includes:

[0059] S500, the cloud layer periodically obtains model parameter increments and key event logs from the fog layer, updates the global slowly changing digital twin model, and sends the updated model parameters to the fog layer.

[0060] This embodiment achieves continuous optimization of the hierarchical model. The cloud layer acquires data from each fog node during daily production off-peak periods (e.g., 2:00-4:00 AM). The model parameter increment represents the weight change of the local real-time digital twin model of the fog node compared to the previous version, using differential encoding compression, with a data volume approximately 10% of the original model. The key event log contains all warning events, collaborative decision-making records, and execution effect evaluations for the day, with a data volume of approximately 5MB. The cloud layer uses a federated averaging algorithm to aggregate the parameter increments of each fog node and combines this with the key event log to incrementally train the global slowly changing digital twin model, optimizing the process degradation model and energy efficiency assessment model. After training, the cloud layer distributes the updated global model parameters to each fog node during the nighttime maintenance period. The fog nodes extract and update their local models, achieving hot model updates.

[0061] refer to Figures 2 to 4 The system's operating method includes the following steps:

[0062] Step S1: Model Initialization and Synchronization

[0063] The cloud center distributes the initial parameters of the global slow-changing digital twin model to the relevant fog nodes. The fog nodes extract and instantiate local real-time digital twin models from the global model according to their jurisdiction, and then distribute them to the corresponding edge devices to complete the construction of the device-level ultra-real-time digital twin model. The edge devices load the ultra-real-time model locally and establish communication connections with the physical devices.

[0064] Step S2: Adaptive Data Flow Management Based on Data Value Density

[0065] Edge nodes perform local processing and judgment on the collected raw runtime data:

[0066] For normal equipment operation data, edge nodes perform trend calculations and feature extraction locally and periodically upload feature summaries to fog nodes. When data anomalies or specific events are detected, edge nodes will trigger the synchronous upload of high-precision data fragments, event types, and local diagnostic results to fog nodes. Fog nodes only upload cross-unit abnormal events, optimization results, and model parameter increments to the cloud center.

[0067] Step S3: Cooperative control within the fog layer

[0068] When the first fog node predicts, through its local real-time digital twin model, that a device under its jurisdiction is about to fail, it executes the following collaborative process:

[0069] The first fog node immediately notifies the associated second fog node of the early warning information and the possible scope of impact through a low-latency network;

[0070] The first fog node and the second fog node perform joint simulation based on their respective local real-time digital twin models to quickly negotiate new production rhythms or alternative process paths.

[0071] The first fog node sends the adjusted control commands to the relevant local edge nodes; the second fog node also sends the adjustment commands to the relevant edge nodes under its jurisdiction, realizing seamless collaboration across production lines. The entire process is completed autonomously within the fog layer without reporting to the cloud center.

[0072] The following section uses electrode wear early warning and collaborative scheduling in a robotic welding production line as an example to explain the implementation process of this invention in detail:

[0073] Step 1: System Initialization and Model Deployment

[0074] The cloud center pre-trains a global, slowly varying digital twin model based on historical production data. The model architecture is a hybrid of LSTM and physical mechanism models, with input dimensions of [batch_size, time_steps, features], where time_steps = 1440 (corresponding to 24 hours of production data), and features include 128 dimensions such as equipment utilization, energy consumption, and failure rate. The model parameters are 50MB. The cloud center distributes the global model parameters to the workshop fog nodes via an encrypted channel.

[0075] Fog node A is responsible for welding production line 1. It extracts and instantiates local real-time digital twin models from the global model. The model structure is RNN, with input dimensions of [batch_size, 60, 64] and a parameter size of 5MB. Fog node A distributes the local model to the edge devices corresponding to the 6 welding robots on production line 1.

[0076] Edge device A1 is deployed next to the controller of robot A, constructing a device-level ultra-real-time digital twin model. The model uses a state-space model combined with a Kalman filter, with a state vector dimension of 12, including joint position, velocity, acceleration, torque, etc. The model update frequency is 1kHz, and the control closed-loop latency is less than 1ms. Edge device A1 interacts directly with robot A via the EtherCAT bus to collect operational data.

[0077] Step 2: Adaptive Data Flow Management

[0078] Edge device A1 collects operational data of robot A at a frequency of 1kHz, including joint torque, welding current, voltage, end effector position, temperature, etc.

[0079] Under normal production conditions, edge device A1 calculates feature data every second:

[0080] The 1-second mean, variance, and peak factor of the joint torque;

[0081] The 1-second mean, standard deviation, and harmonic content of the welding current;

[0082] The rate of temperature rise over 10 seconds.

[0083] Edge device A1 packages the feature data into a JSON format data packet (approximately 500 bytes) every 10 seconds and uploads it to fog node A via the MQTT protocol, with a QoS level of 1.

[0084] When electrode wear causes fluctuations in welding current, edge device A1 detects a current deviation exceeding 5% of the set value and determines it as a current anomaly event. Edge device A1 immediately extracts raw data segments (containing raw sample values ​​of all operational data, approximately 200KB) one second before and after the anomaly, generates event details (including event type, timestamp, anomaly parameter values, and duration), and immediately uploads them to fog node A via TCP protocol.

[0085] Step 3: Cooperative control within the fog layer

[0086] After receiving abnormal data uploaded by edge device A1, fog node A inputs it into a local real-time digital twin model for inference. The model output shows a predicted electrode wear rate of 0.3 mm / min. According to the process model calculation, the electrode will wear to the critical value in 30 seconds, resulting in substandard welding quality. Fog node A determines that an anomaly is about to occur.

[0087] Fog node A immediately generates an early warning message:

[0088] Device ID: Robot_A_01;

[0089] Warning type: Accelerated electrode wear;

[0090] Predicted failure time: 30 seconds;

[0091] Impact: Subsequent workstations Robot_B_01 will be forced to wait, resulting in a 5-second delay in the production line cycle time per unit.

[0092] Fog node A sends the warning information to fog node B (responsible for AGV scheduling) via a horizontal communication link using the UDP protocol, with a transmission delay of less than 5ms.

[0093] After receiving the warning information, fog node B initiates a joint simulation with fog node A:

[0094] Simulation of different cycle time adjustment strategies for fog node A: If the cycle time is reduced from 60 seconds / unit to 65 seconds / unit, a 5-second time window can be gained for electrode replacement, with an output impact of -8%;

[0095] Fog node B simulation AGV scheduling: If the AGV departs from the spare parts warehouse and travels at a speed of 1m / s, it can deliver the spare electrode to the Robot_A_01 station within 25 seconds;

[0096] The two nodes exchange simulation results through three iterations and finally generate a collaborative production plan: the cycle time is adjusted to 65 seconds / unit, and the AGV scheduling starts 25 seconds in advance.

[0097] Fog node A and fog node B confirm their collaborative production plan through a three-way handshake protocol:

[0098] Fog node A sends a plan proposal (65 seconds per node, effective time current + 5 seconds);

[0099] Fog node B returns a confirmation message (AGV will arrive in 25 seconds).

[0100] Fog node A sends final confirmation, and the plan takes effect.

[0101] Step 4: Issuance and execution of control commands

[0102] Fog node A generates a control command packet containing a clock speed adjustment command, a target clock speed of 65 seconds per device, an effective timestamp, and a list of target devices (Robot_A_01, Robot_B_01, and all devices within the line). Fog node A sends the command packet to the corresponding edge device via the MQTT protocol, with a QoS level of 2.

[0103] Fog node B generates AGV scheduling instructions, including task ID, start point, destination, delivery timestamp, and spare electrode model. Fog node B sends the instructions to the AGV control system via the Modbus TCP protocol.

[0104] After receiving the control command, edge device A1 automatically calibrates the welding parameters after electrode replacement and sends a completion confirmation to fog node A. The edge device of Robot_B_01 automatically adjusts its waiting logic to adapt to the new cycle time. The AGV control system drives the AGV to the target station to deliver the spare electrode.

[0105] Step 5: Cloud Model Update

[0106] Every day at 2:00 AM, fog node A uploads the day's model parameter increments (using differential encoding compression, approximately 500KB) and key event logs (including early warning events, collaborative decision records, and execution effect evaluations, approximately 5MB) to the cloud center via HTTPS. Fog node B simultaneously uploads its data.

[0107] The cloud center uses a federated averaging algorithm to aggregate parameter increments from each fog node and update the global slowly changing digital twin model. The cloud center inputs key event logs into the global model for incremental training, optimizing the process degradation model and energy efficiency assessment model. After training, new global model parameters are generated, and the model version number increments from v1.0.0 to v1.0.1.

[0108] During the nighttime maintenance period from 4:00 AM to 6:00 AM, the cloud center distributes the updated global model parameters v1.0.1 to fog node A and fog node B via a secure channel. Fog node A and fog node B extract and update their local real-time digital twin models. The hot update process does not affect current production, and a confirmation message is returned to the cloud center after the update is completed.

[0109] Compared with related technologies, the present invention has the following improvements.

[0110] The hierarchical digital twin architecture enables industrial data to be "usable but not visible," with raw operational data always remaining at the edge layer. Model parameters and feature summaries are only exchanged between the fog and cloud layers, reducing the risk of sensitive data leakage.

[0111] By implementing adaptive data flow management based on data value density, bandwidth consumption is reduced under normal operating conditions, with raw data fragments only uploaded during abnormal events, significantly alleviating network bandwidth pressure. This results in a reduction in network bandwidth requirements in a typical welding workshop.

[0112] Cross-production line collaborative decision-making is achieved through a lateral communication link within the fog layer, enclosing critical control loops at the edge and fog layers, enabling millisecond-level response from perception to decision-making to execution. End-to-end latency is reduced, meeting the real-time requirements of precision welding.

[0113] Through a layered architecture design, the system can maintain autonomous operation for extended periods even when the cloud layer is disconnected. The fog layer and edge layer can independently perform local control and coordination, improving system reliability. The risk of a single point of failure causing a plant-wide shutdown is reduced.

[0114] The embodiments of the present invention achieve millisecond-level real-time control, efficient bandwidth utilization, high reliability and data privacy protection of industrial digital twins, and are particularly suitable for intelligent manufacturing scenarios such as automobile manufacturing, precision machining and semiconductor manufacturing.

[0115] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they are fundamentally similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0116] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application is not limited to any single aspect, any single embodiment, or any single claim.

[0117] The method step numbers in this application are for convenience of reference only and do not indicate the execution order. The execution order of each step should be determined by its function and internal logic.

[0118] The scope of protection claimed in this application is defined by the claims. The contents of the specification and drawings are used to support the scope of protection of the claims and shall not be construed as limiting the scope of protection of the claims.

[0119] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application.

[0120] All publications, patents and patent applications mentioned in this application are incorporated herein by reference in their entirety, as if each individual publication, patent or patent application were specifically and individually indicated to be incorporated herein by reference in its entirety.

[0121] Embodiments of this application may be implemented using software, hardware, firmware, or any combination thereof. When implemented in software, the method may be stored in a computer-readable medium and executed by a computer or processor. A computer-readable medium includes any medium or apparatus that can be used to store or transmit information in a computer-accessible form.

[0122] Those skilled in the art will recognize that this application is not limited to the details of the embodiments described above, and that it may be implemented in other specific forms without departing from the spirit and essential characteristics of this application. Therefore, the embodiments should be considered exemplary rather than restrictive in all respects. The scope of this application is defined by the appended claims rather than the foregoing description, and therefore all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within this application. Any reference numerals in the claims should not be construed as limiting the claims.

Claims

1. A real-time collaborative control method for industrial digital twins based on fog computing, characterized in that, The method includes the following steps: S100: Acquire physical device operation data collected at the edge layer; S200, the edge layer processes the running data according to preset rules and decides on the content and frequency of data to be uploaded to the fog layer; S300, when the first fog node in the fog layer determines an anomaly or an impending anomaly based on the received data, it initiates a collaborative control process with the second fog node to jointly agree on a control strategy; S400, the first fog node and the second fog node respectively send the agreed control commands to the edge computing devices under their jurisdiction, and the edge computing devices drive the physical devices to perform corresponding actions.

2. The method according to claim 1, characterized in that, In S200, the preset rule is an adaptive rule based on data value density, including: S210, for the operating data characterizing the normal state of the device, the edge layer only uploads the processed feature data or statistical summary; S220, for the operational data characterizing device malfunctions or specific events, the edge layer uploads high-precision raw data fragments or event details.

3. The method according to claim 1, characterized in that, In S300, the coordinated control process for initiating the second fog node includes: S310, the first fog node sends the information and impact prediction of the anomaly or impending anomaly to the second fog node; S320, the first fog node and the second fog node perform joint simulation based on their respective local real-time digital twin models to generate a new collaborative production plan; S330, the first fog node and the second fog node confirm the new collaborative production plan.

4. The method according to claim 1, characterized in that, The method further includes: S500, the cloud layer periodically obtains model parameter increments and key event logs from the fog layer, updates the global slowly changing digital twin model, and sends the updated model parameters to the fog layer.

5. A real-time collaborative control system for industrial digital twins based on fog computing, characterized in that, include: Cloud layer, which constructs and maintains a global, slowly changing digital twin model; A fog layer, which consists of at least one fog computing node, is used to construct and maintain a local real-time digital twin model corresponding to a physical production unit. An edge layer, comprising at least one edge computing device, which builds and maintains a device-level ultra-real-time digital twin model corresponding to a single physical device and interacts directly with the physical device; The cloud layer, the fog layer, and the edge layer communicate with each other via a network and form a collaborative control closed loop.

6. The system according to claim 5, characterized in that, The fog computing nodes are connected by a horizontal communication link, which is used for real-time data exchange and collaborative decision-making across production units within the fog layer. The collaborative decision-making does not need to pass through the cloud layer.

Images